surface contamination and cleaning.pdf

Surface Contamination and Cleaning,

Volume 1

K.L. Mittal,Editor

VSP

Surface Contamination and Cleaning, Volume 1

This page intentionally left blank

SURFACE

CONTAMINATION

AND CLEANING

VOLUME 1

Editor:

K.L. Mittal

UTRECHT BOSTON 2003

VSP BV Tel: +31 30 692 5790 P.O. Box 346 Fax: +31 30 693 2081 3700 AH Zeist [email protected] The Netherlands www.vsppub.com © VSP BV 2003 First published in 2003 ISBN 90-6764-376-9

All rights reserved. No part of this publication may be reproduced, stored in a retrieval

system, or transmitted in any form or by any means, electronic, mechanical, photocopy-

ing, recording or otherwise, without the prior permission of the copyright owner.

Printed in The Netherlands by Ridderprint bv, Ridderkerk

Contents

Preface vii

Mapping of surface contaminants by tunable infrared-laser imaging

D. Ottesen, S. Sickafoose, H. Johnsen, T. Kulp, K. Armstrong,

S. Allendorf and T. Hoffard 1

Monitoring cleanliness and defining acceptable cleanliness levels

M.K. Chawla 23

Tracking surface ionic contamination by ion chromatography

B. Newton 43

A new method using MESERAN technique for measuring surface

contamination after solvent extraction

M.G. Benkovich and J.L. Anderson 49

Methods for pharmaceutical cleaning validations

H.J. Kaiser 75

Influence of cleaning on the surface of model glasses and their

sensitivity to organic contamination

W. Birch, S. Mechken and A. Carré 85

Decontamination of sensitive equipment

R. Kaiser and K. Haraldsen 109

The fundamentals of no-chemistry process cleaning

J.B. Durkee II 129

Development of a technology for generation of ice particles

D.V. Shishkin, E.S. Geskin and B. Goldenberg 137

Cleaning with solid carbon dioxide pellet blasting

F.C. Young 151

Development of a generic procedure for modeling of waterjet

cleaning

K. Babets and E.S. Geskin 159

Contents vi

Experimental and numerical investigation of waterjet derusting

technology

K. Babets, E.S. Geskin and B. Goldenberg 173

Practical applications of icejet technology in surface processing

D.V. Shishkin, E.S. Geskin and B. Goldenberg 193

Correlating cleanliness to electrical performance

T. Munson 213

Qualifying a cleaning system for space flight printed wiring assemblies

J.K. “Kirk” Bonner and A. Mehta 225

Investigation of modified SC-1 solutions for silicon wafer cleaning

C. Beaudry and S. Verhaverbeke 241

Performance qualification of post-CMP cleaning equipment

in a semiconductor fabrication environment

M.T. Andreas 249

Spatial and temporal scales in wet processing of deep

submicrometer features

M. Olim 261

Microdenier fabrics for cleanroom wipers

J. Skoufis and D.W. Cooper 267

Fine particle detachment studied by reflectometry and atomic

force microscopy

A. Feiler and J. Ralston 279

Dust removal from solar panels and spacecraft on Mars

S. Trigwell, M.K. Mazumder, A.S. Biris, S. Anderson and C.U. Yurteri 293

Laser cleaning of silicon wafers: Prospects and problems

M. Mosbacher, V. Dobler, M. Bertsch, H.-J. Münzer, J. Boneberg and

P. Leiderer 311

Particle removal using resonant laser detachment

K. Kearney and P. Hammond 335

The future of industrial cleaning and related public policy-making

C. LeBlanc 345

Surface Contamination and Cleaning, Vol. 1, pp. vii–viii

Ed. K.L. Mittal

© VSP 2003

Preface

This volume chronicles the proceedings of the International Symposium on Sur-

fase Contamination and Cleaning held under the aegis of MST Conferences in

Newark, New Jersey, May 23–25, 2001.

Even a cursory look at the literature will evince that there has been tremendous

interest and R&D activity in the arena of surface contamination and cleaning, so

we decided to organize this symposium. Because of the importance of this topic

in many technological areas, tremendous efforts have been devoted to devise

novel and more efficient ways to monitor, analyse and characterize contamination

on surfaces as well as ways to remove such contamination from a wide variety of

surfaces.

The ubiquitous nature of surface contamination causes concern to everyone

dealing with surfaces, and the world of surfaces is wide and open-ended. A con-

taminant is defined as “unwanted matter or energy” or “material or energy in the

wrong place”. Also contaminants can by broadly classified as: film-type, particu-

lates; ionic, and biological or microbial. The technological areas where surface

contamination has always been a bete noire and thus surface cleaning is of cardi-

nal importance are too many and range from aerospace to microelectronics to

biomedical. Here a few eclectic examples will suffice to underscore the impor-

tance of surface contamination and cleaning. In the world of ever-shrinking de-

vice dimensions in the microelectronics, the need to remove ever smaller particles

(of nanosize dimension) is quite patent. On the other hand, film-type (organic)

contamination is of crucial importance in the area of adhesive bonding, as even a

very thin layer of contamination can be very detrimental in attaining good bond

strength. In operation theaters, the concern about microbial contamination is all

too obvious. So in light of the great concern about surface contamination, people

dealing with surfaces are rightfully afflicted with molysmophobia.*

The technical program for this symposium comprised 45 papers dealing with

all kinds of contaminations on a host of surfaces, and many ramifications of sur-

face contamination and cleaning were addressed. There were brisk and illuminat-

ing (not exothermic) discussions, both formally and informally, throughout the

symposium. Also if comments from the participants are a barometer for the suc-

cess of a symposium then this event was quite successful.

Now coming to this volume, it contains a total of 24 papers (others are not in-

cluded for a variety of reasons). It must be recorded that all manuscripts were rig-

orously peer reviewed and suitably revised (some twice or thrice) before inclusion

in this volume. So this volume is not a mere collection of unreviewed papers −

Preface viii

which is generally the case with many symposia proceedings − rather it reflects

information which has passed peer scrutiny. The topics covered include: mapping

of surface contaminants; various techniques for cleaning surfaces; various tech-

niques for monitoring level of cleanliness; acceptable cleanliness levels, ionic

contamination; pharmaceutical cleaning validations; cleaning of glass surfaces;

decontamination of sensitive equipment; no-chemistry process cleaning; waterjet

cleaning; cleaning with solid carbon dioxide pellet blasting; cleanroom wipers;

dust removal from solar panels and spacecraft on Mars; laser cleaning of silicon

surfaces; particle removal; implications of surface contamination and cleaning;

and future of industrial cleaning and related public policy-making.

I sincerely hope that this volume addressing many aspects and recent develop-

ments in the domain of surface contamination and cleaning will be of interest to a

wide range of people working in many different industries.

Acknowledgements

It is always a pleasure to write this particular segment of a book as it offers the

opportunity to thank those who helped in many ways. First, my sincere thanks are

extended to my colleague and friend, Dr. Robert H. Lacombe, for taking care of

the organizational aspects of this symposium. The comments from the peers are a

sine qua non to maintain the highest standard of a publication, so I am most ap-

preciative of the time and efforts of the unsung heroes (reviewers) in providing

many valuable comments. I am profusely thankful to the authors for their interest,

enthusiasm and contribution without which this book would not have seen the

light of day. In closing, my thanks go to the staff of VSP (publisher) for giving

this book a body form.

K.L. Mittal

P.O. Box 1280

Hopewell Jct., NY 12533

*Molysmophobia means fear of dirt or contamination, from Mrs. Byrne’s Dictionary of Unusual,

Obscure, and Preposterous Words, University Books, Secaucus, NJ (1974).

Surface Contamination and Cleaning, Vol. 1, pp. 1–22

Ed. K.L. Mittal

© VSP 2003

Mapping of surface contaminants by tunable

infrared-laser imaging

DAVID OTTESEN, SHANE SICKAFOOSE,∗ HOWARD JOHNSEN,

TOM KULP, KARLA ARMSTRONG, SARAH ALLENDORF and

THERESA HOFFARD1

Sandia National Laboratories, P.O. Box 969, MS 9403, Livermore, CA 94551-0969 1Naval Facilities Engineering Service Center, 1100 23rd Avenue, Port Hueneme, CA 93043-4370

Abstract—We report the development of a new, real-time non-contacting monitor for cleanliness verification based on tunable infrared-laser methods. New analytical capabilities are required to maximize the efficiency of cleaning operations at a variety of federal (Department of Defense [DoD] and Department of Energy [DOE]) and industrial facilities. These methods will lead to a re-duction in the generation of waste streams while improving the quality of subsequent processes and the long-term reliability of manufactured, repaired or refurbished parts.

We have demonstrated the feasibility of tunable infrared-laser imaging for the detection of con-taminant residues common to DoD and DOE components. The approach relies on the technique of infrared reflection spectroscopy for the detection of residues.

An optical interface for the laser-imaging method was constructed, and a series of test surfaces was prepared with known amounts of contaminants. Independent calibration of the laser reflectance images was performed with Fourier transform infrared (FTIR) spectroscopy. The performance of both optical techniques was evaluated as a function of several variables, including the amount of contaminant, surface roughness of the panel, and the presence of possible interfering species (such as water). FTIR spectra demonstrated that a water film up to 7 µm thick would not interfere with the effectiveness of the laser-imaging instrument. The instrumental detection limit for the laser reflec-tance imager was determined to be on the order of a 10-20 nm thick film of a general hydrocarbon contaminant.

Keywords: Infrared; tunable-laser; imaging; cleaning; surface contamination.

1. INTRODUCTION

Real-time techniques to provide both qualitative and quantitative assessments of

surface cleanliness are needed for a wide variety of governmental and industrial

applications. The range of potential applications include aircraft, shipboard, vehi-

cle, and weapon component surfaces to be coated, plated, or bonded. The avail-

∗To whom all correspondence should be addressed. Phone: (925) 294-3526,

Fax: (925) 294-3410, E-mail: [email protected]

D. Ottesen et al. 2

ability of a convenient analysis technology for on-site, post-cleaning determina-

tion of surface contamination will allow more rapid and accurate assessments of

the efficiency of chosen cleaning techniques. By developing an on-line technique,

processed parts or extracted samples will not have to be sent to a separate labora-

tory for analysis, thereby eliminating processing delays. The information provided

by the optical method will assist the process operator in distinguishing between

specific contaminants and determining subsequent actions to be taken.

In this paper we report the development of an infrared laser-based imaging ap-

proach that will reduce the use, emission, and handling of waste-stream materials

in cleaning operations. This work is supported by the separate development of a

hardened, portable Fourier transform infrared (FTIR) reflectance instrument at the

Naval Facilities Engineering Service Center (NFESC), Port Hueneme, CA in co-

operation with the Surface Optics Corporation. Simultaneous development of an

FTIR instrument is complementary in nature to the laser-imaging technique and is

described in detail elsewhere [1]. Both instruments will be used primarily for the

real-time on-line or nearly on-line detection of contaminant residues on reflective

surfaces. In each case, surface contamination is detected by its absorption of a

grazing-incidence infrared beam reflected from the surface.

The instruments differ in the nature of the information they provide. The laser-

based instrument produces images that directly indicate the spatial extent and lo-

cation of infrared-absorbing surface hydrocarbon contaminants. In contrast, FTIR

instrumentation provides a wide-band spectral measurement of the surface reflec-

tance averaged over a small area for nearly all organic materials, and many inor-

ganic components. Thus, the laser-imaging system allows the rapid determination

of surface cleanliness for organic residues over a large area, while the spectrally-

resolved FTIR method is useful in identifying the specific molecular composition

of a surface contaminant at a particular location.

The imaging system under development employs a widely tunable infrared-

laser illumination source in conjunction with an infrared camera. This approach

provides an on-line technique for surveying contamination levels over large sur-

face areas in a real-time imaging mode. The laser is broadly-tunable over the 1.3-

4.5 µm wavelength range, thus allowing the detection of many hydrocarbon con-

taminants via absorption bands associated with CH-, OH-, and NH-stretching vi-

brations.

Currently, the detection and identification of surface contaminants on reflective

surfaces is conveniently and rapidly done by FTIR reflectance methods. These

non-destructive, non-contacting optical techniques identify the chemical constitu-

ents of the contaminants, and can yield quantitative measurements with appropri-

ate calibration. Infrared optical methods are particularly useful for cleanliness

verification since the surface is probed under ambient conditions. More sensitive

high-vacuum electron and ion spectroscopic techniques (X-ray photoelectron

spectroscopy, Auger electron spectroscopy, and secondary-ion mass spectrome-

try) are not suited for on-line application.

Tunable IR-laser mapping of surface contaminants 3

Commercial instruments that employ infrared reflectance spectroscopy are

available for surface analysis and provide both quantitative and qualitative infor-

mation on surface coatings. These instruments are limited in their ultimate sensi-

tivity to surface contaminants by the nature of their optical design. Infrared radia-

tion is focused onto the surface to be analyzed at a near-normal angle of

incidence, resulting in a compact hand-held apparatus. The infrared light is col-

lected as either specularly or diffusely reflected radiation depending on the

roughness and scattering properties of the surface [2, 3]. The resulting sensitivity

to very thin layers of surface species is limited by poor coupling of the incident

electromagnetic field with the vibrating dipoles of the surface molecular species

[4-6] in layers less than 0.1 µm thick.

In order to maximize the sensitivity of infrared reflectance measurements for

absorption bands of thin layers of contaminants on metallic surfaces, theoretical

and experimental studies [7-9] have shown that the angle of incidence of infrared

radiation on the surface should be increased to at least 60° from the surface nor-

mal. This is also true for many thin-film residues on the surface of non-metals,

such as dielectrics and semiconductors (although the detectability of contaminant

absorption bands under these circumstances depends strongly on the optical con-

stants of both surface and substrate, and any absorption features intrinsic to the

non-metallic substrate). Additional sensitivity in the reflectance measurement is

obtained by measuring only the component of the reflected infrared radiation po-

larized parallel to the plane of incidence. This experimental method is variously

referred to as, “grazing-angle” reflectance spectroscopy or infrared reflection-

absorption spectroscopy (IRRAS). We have adapted the technique of “grazing-

angle” reflectance spectroscopy to utilize the newly developed tunable-laser

source.

2. EXPERIMENTAL

The laser-based instrument described in this report offers the capability to rapidly

survey large surface areas and to determine the location and extent of residual hy-

drocarbon contaminants following cleaning operations. In contrast, a spectro-

scopic analysis by an FTIR-based infrared reflectance instrument is able to char-

acterize a very broad range of organic constituents and many inorganic species.

However, a surface-probing FTIR instrument measures a spectrum at only a sin-

gle small area on a sample, thus requiring broad area surveys to be done by se-

quentially probing many points. Even at a rate of ~ 10 seconds per measurement

point, this can be a time-consuming process. The rate of measurement by FTIR

spectroscopy is constrained by the relatively low spectral brightness (compared to

a laser) of the incandescent illumination sources. This makes it necessary to use

relatively long integration times to achieve an acceptable signal-to-noise ratio.

The tunable-laser-based instrument overcomes these limitations by illuminating

a broad surface area with a high-brightness infrared laser. This approach allows a

D. Ottesen et al. 4

single-wavelength reflectance measurement over an area of several square centi-

meters to be made on a timescale of less than a second. In order to acquire meas-

urements at multiple wavelengths, the laser is tuned and an image is collected at

each of the desired wavelengths. While a detailed spectral map of a surface can be

generated over the laser tuning range, the primary use of the system is to provide

rapid areal surveys at a few key wavelengths that are indicative of hydrocarbon

contaminants. The detection sensitivity for several hydrocarbon species at various

illumination wavelengths was evaluated in this work, as well as a method to sup-

press image noise due to laser speckle while maintaining high illumination inten-

sity.

2.1. Quasi-phasematching tunable infrared laser

The broadly-tunable infrared laser illuminator is based on a technology called

quasi-phasematching (QPM) [10]. This approach has been exploited to increase

the tuning range and power of the infrared light source while reducing its size. For

example, continuous-wave (cw) optical parametric oscillators (OPOs) that employ

the QPM material, periodically-poled lithium niobate (PPLN), are capable of tun-

ing over the 1.3-4.5 µm spectral region while emitting more than 0.5 W of power.

This technique has been used to generate tunable infrared laser light for imaging

natural gas emissions, and developing laser-based spectroscopic gas sensors [10].

In this work we are extending it to the analysis of hydrocarbon residues on mate-

rial surfaces.

The limit of the current tuning range of the PPLN-based laser at long wave-

lengths is about 4.5 µm (2222 cm-1

) due to the transmission characteristics of lith-

ium niobate. This property restricts the sensitivity of the chemical imaging system

to functional groups containing hydrogen atoms (C-H, N-H, O-H). Extension of

the laser tuning wavelength range beyond 5 µm (2000 cm-1

) is desirable to pro-

vide specific identification of hydrocarbon and some inorganic molecular species.

The light source assembled for the IR imaging sensor is an OPO pumped by a

continuous-wave (cw) Nd:YAG laser, as shown in Figure 1 [10]. An electric field

is induced in the OPO’s PPLN crystal by the electric field of the pump laser; these

fields interact to form two new laser beams whose frequencies sum to the fre-

quency of the pump laser. The reflectivities of the mirrors in the optical cavity are

selected to resonate one of the generated waves, while the other wave is simply

generated and released from the cavity. The resonated wave is called the signal;

the non-resonated wave is called the idler. The exact frequencies of the signal and

the idler are determined by the phasematching properties of the crystal (described

below), the reflectivity of the cavity, and by any spectrally-selective optics that

may be added to the laser cavity (e.g. an étalon). While either the signal or the

idler beam can be used for measurements, only the idler is used in the experi-

ments reported here.

As shown in Figure 1, the OPO used in the imaging sensor is of the “bowtie-

ring” design. A diode-pumped, cw, multimode Nd:YAG laser (Lightwave Elec-


tronics, Mountain View, CA) that is capable of generating at least 6 W of output

power at a wavelength of 1064 nm is used as the OPO pump source. Two flat mir-

rors (M3 and M4) and two curved mirrors (M1 and M2, 50-mm radius of curva-

ture), all coated to be highly reflective at the signal and highly transmissive at the

pump and idler wavelengths, form the bow-tie-shaped, single-wavelength reso-

nant ring oscillator cavity designed to resonate the signal wave. An anti-

reflection-coated lens, positioned between the pump laser and the OPO cavity,

serves to image the Gaussian pump beam into the PPLN crystal. In this way, a

beam waist (E-field radius) of 70 µm is created in the center of the crystal, which

itself is centered between the two curved cavity mirrors. During normal operation,

the OPO resonates on a single signal mode for minutes at a time, whereupon it

hops to another cavity mode. The idler bandwidth is, however, determined by that

of the pump beam, which is 10-15 GHz.

The use of the QPM material, PPLN, makes cw OPO operation more tunable

and efficient than it would be for a conventional birefringently phasematched

crystal. Simply stated, phasematching is a condition in which all of the interacting

waves (i.e., signal, pump, and idler) maintain a specified relative phase relation-

ship as they propagate through a nonlinear medium, and is a necessary condition

for efficient nonlinear generation. In birefringent materials, phasematching is

Figure 1. Diagram of the PPLN OPO and projection optics.

D. Ottesen et al. 6

achieved by careful selection and/or control of the crystal birefringence, tempera-

ture, and beam propagation angles.

In a QPM medium, phasematching is designed into the medium during the

crystal growth process. Phasematching is achieved by causing the crystal to have

a periodically inverting optical axis. The engineering process used to create these

crystals increases conversion efficiency by allowing the use of much stronger

nonlinear coefficients of the crystal, and frees the system from reliance on bire-

fringence thereby increasing tunability. As the light beams cross the crystal-axis-

inverting boundaries, any relative dephasing of the waves is corrected. For a crys-

tal of a given periodicity, the rephasing is effective for a particular set of pump,

signal, and idler frequencies. Some degree of tuning of these waves can be

achieved within the crystal phasematching bandwidth (typically 10-20 cm-1

).

Broader tuning is achieved by accessing a portion of the same crystal having a

different periodicity, or by changing the temperature of the crystal.

In the present system, two 50-mm-long PPLN crystals (Crystal Technology,

Palo Alto, CA) with an aperture of 11.5 mm × 0.5 mm are used as the active me-

dium. Each crystal contains eight poled regions with different periodicities. One

crystal’s periodicities range from 28.5 to 29.9 µm, and of the other crystal from

30.0 to 31.2 µm. When operating at a crystal temperature of 148°C, these periods

collectively allow tuning of the idler from 2720 to 3702 cm-1

. The crystals are

mounted in a stacked fashion within a temperature-stabilized copper oven that is

attached to a vertical translation stage. Each crystal is tuned by selecting a period

using the vertical motion of the stage; horizontal motion of the oven is used to se-

lect between the two crystals.

The raw output of the OPO contains the idler beam as well as portions of the

signal and pump beams and some higher-order (red, green) beams created spuri-

ously in the PPLN crystal. Spectral filtering is used to dump all but the idler

beam. Prior to illumination of the sample, the idler is passed through a set of pro-

jection optics, also shown in Figure 1. The first of these is a ZnSe diffuser (mean

roughness of ~ 3-4 µm) that is mounted on a motor-driven spindle. The diffuser

serves to reduce the phase coherence of the idler in order to minimize laser

speckle noise in the transmitted beam and viewed by the IR camera in the light re-

flected from the sample surface. The cone of radiation leaving the diffuser is col-

lected by a ZnSe faceted lens (Laser Power Optics, Murrieta, CA). The faceted

lens is formed to contain the equivalent of 16 6.4 mm facets and 16 partial facets

around the edge of the lens on a 3.8 cm diameter with an effective f-number of

1.7. It operates as a prism array – the expanded beam is segmented into 32 differ-

ent square beamlets that are subsequently overlapped at a distance of 5 cm from

the surface of the lens. A ZnSe wire grid polarizer (not shown in Figure 1) is lo-

cated at the overlap point, and serves to produce a p-polarized beam for the infra-

red reflectance measurement. The square-shaped overlap region is then imaged

onto the target using an f/1.7, 8.4 cm focal-length ZnSe lens. As a unit, the system

converts the Gaussian profile of the idler beam into a uniform square illumination

on the sample surface.


The infrared laser light is incident on the sample surface at an angle of 60°

from the surface normal, and the specularly reflected component is detected by an

InSb focal-plane array (FPA) camera with an infrared macro-lens assembly and

an array size of 256 x 256 pixels. The FPA camera is located approximately 0.3 m

from the sample surface, and the resulting field of view is 20 x 35 mm.

FTIR instruments at both Sandia and NFESC were used to characterize the

mid-infrared spectra of contaminated surfaces via optical interfaces for grazing-

angle reflectance spectroscopy. The system at NFESC uses a commercially avail-

able sampling accessory that permits a variable angle of incidence from 30 to 80°,

which is convenient for evaluating detection limits for contaminants on a variety

of surfaces. The optical interface used by the Sandia National Laboratories FTIR

instrument was constructed with a fixed 60° angle of incidence with optics exter-

nal to the spectrometer. It also differs from the NFESC system in the large solid-

angle used both to illuminate the surface and collect reflected light. This feature is

particularly useful in the examination of rougher surfaces that cause significant

scattering of the infrared beam, with a consequent degradation in both sig-

nal/noise ratio and detection limits. Both systems use infrared polarizers to en-

hance the sensitivity of the measurements by restricting the surface illumination

to p-polarization [4]. Unless otherwise noted, all reflectance spectra presented in

this paper are for p-polarized measurements.

2.2. Test sample preparation for calibration

In order to evaluate the usefulness of the laser-imaging technique as a cleaning

verification method, we prepared a number of test surfaces with well-

characterized levels of contamination. These were used to determine detection

limits as a function of contaminant species, level of contamination, degree of sur-

face roughness, effect of spectral interference, and instrumental parameters such

as angle-of-incidence. Seven candidate materials were chosen as contaminant

species for evaluation as shown in Table 1. These materials have proven to be

particularly difficult to remove during cleaning operations, and are representative

of many other organic contaminants encountered in government and industrial

cleaning processes. Detailed measurements on the first four materials have been

made in the course of this work and preliminary measurements have been made

on the remaining three.

A number of metals were chosen as substrates for the target contaminants,

based on usage information obtained from military and contractor facilities. These

were Aluminum-7075-T6, Titanium 6Al-4V, Steel Alloy 4340, Stainless Steel

304, and Magnesium AZ31B. The metals were fabricated into 3.8 x 12.7 cm flat

coupons for laboratory testing and method demonstration.

Six surface roughness finishes of the Aluminum 7075-T6 test coupons were

obtained, ranging from 80 to 600 grit (600 grit being the smoothest). A profilome-

ter instrument was used to examine the surface roughness profiles and provide

average Ra values. A Ra value is an arithmetic average of the absolute deviations

D. Ottesen et al. 8

from the mean surface level, in millionths of an inch; therefore, a Ra value of 1.5

= 0.00000015 inches (3.8 µm). Due to the nature of metal-shop finishing proc-

esses, surface roughness values vary considerably across a given surface area.

Finishing operations also result in a directional “grain” parallel to the sample

coupons’ longitudinal direction. Surface roughness measurements, therefore, ex-

hibit large variations between measurements taken along an orientation longitudi-

nal or transverse to the polishing axis. Two surface roughness levels, 600 and 220

grit, were obtained for the other metal alloys.

Prior to contaminant application, the aluminum alloy coupons were cleaned

with acetone and underwent sonication with a clean-rinsing aqueous cleaner.

They were then thoroughly rinsed in distilled water and dried in an oven at 50°C.

Once cooled, they were weighed on a microbalance with a precision of 0.01 mg.

Two or three weighings were averaged.

Both drawing agent and lubricant contaminated Al-7075 coupons were pro-

duced by two primary deposition methods – airbrushing and manual brushing.

Several other techniques were attempted, including “wire-cator” drawing, coupon

spinning, and “manual drop and spread.” These techniques were not used to pro-

duce test samples for calibration for these particular contaminants due to the supe-

rior results obtained from airbrushing and manual brushing. Three levels of draw-

ing agent were applied by airbrushing to three Al test coupons for each of the six

surface finishes, creating a suite of 18 panels. Varying concentrations of drawing

agent in water were prepared for the airbrush solutions. Similarly, four levels of

lubricant were applied to four Al test coupons for each of six surface finishes,

creating a suite of 24 panels. Manual brushing was used for all but the least con-

taminated samples, which were airbrushed. Lubricant solutions for both tech-

Table 1.

Contaminant materials used for preparation of test coupon for calibration

Material Description Usage

Drawing Agent White soft solid – ester grease Metal drawing, cutting, and lubricating agent

Lubricant Brown liquid – paraffin hydrocarbons Rust preventative, cleaner, lubricant, protectant for metals

Silicone Silicone Lubricant

Mold Release 1 Green liquid – ethanol homopolymer Mold release agent

Mold Release 2 Clear liquid – proprietary polymeric resins

Mold release agent

Solder Flux Yellow liquid – abietic acid or anhydride

Soldering flux for electrical and electronic applications

Hydraulic Oil

MIL-H-5606A AM2

Blue liquid – castor oil base Hydraulic systems, shock and strut lubricant


niques were prepared using pentane as the solvent. Similar methods were used in

preparing calibration samples with the mold release, solder flux, and hydraulic oil

samples.

All contaminated coupons were gently heated in an oven at 50°C for several

days to remove both semi-volatile and volatile components. This served to stabi-

lize the contaminants, allowing for quantification by weighing. Once the weights

became stable, the coupons were cooled and weighed to determine the amount of

contaminant present on the surface. When not being weighed or examined, the

coupons were kept in a desiccator.

3. RESULTS AND DISCUSSION

Grazing-angle incidence reflectance spectroscopy acts to enhance the detection

sensitivity for thin layers of residue predominantly through improved coupling of

the electric field intensity of the incident beam with the vibrating dipoles of the

surface contaminant layer perpendicular to the metallic surface. Some additional

enhancement of the infrared absorption spectrum will also occur due to a length-

ening of the effective path length through the absorbing thin film layer [4-6].

If the optical properties of both thin film and substrate are known (or can be de-

termined), the reflection-absorption spectrum can be calculated as a function of

film thickness and angle of incidence. This capability is particularly useful for in-

terpreting experimental data and designing optical instrumentation. Computer

codes written at Sandia [7] performed these calculations for a variety of materials.

3.1. FTIR measurements

FTIR reflectance data for the full drawing-agent sample set were obtained at

NFESC and Sandia using angles of incidence of 75 and 60° for average film

thickness ranging from 0.1 to 1 µm, and aluminum substrates with surface finish

ranging from 600 to 80 grit. Since the surface finishing operation produced a

highly directional roughness, measurements were made both longitudinally and

transversely with respect to the polishing grooves. Ra values were determined at

NFESC using profilometer measurements, and resulted in surface roughness val-

ues of 0.3 to 1.5 µm for the longitudinal direction, and 0.5 to 6 µm for the trans-

verse direction.

The FTIR reflectance spectra were normalized using the uncoated back of a

panel as a clean reference standard, and the intensity data are presented as either

reflectance or –log reflectance in the following discussion. The C-H stretching vi-

brations near 2900 cm-1

proved to be generally useful in quantifying instrument

response since these frequencies are well isolated from atmospheric interference

due to water vapor and carbon dioxide. However, the baseline for these reflec-

tance data was often non-linear. A simple single-point measurement of intensity

was therefore not sufficient to determine the instrument response function.

D. Ottesen et al. 10

Optical constants (n and k) were derived for the contaminant C-H stretching

vibrations using the Sandia reflectance code and a dispersion model to calculate a

fit to the experimental data for one of the test coupons [7]. Reflectance-absorption

spectra for the 2800-3000 cm-1

range were calculated for 1-µm thick films of a

specific hydrocarbon contaminant on an aluminum surface at either 60 or 75° an-

gle of incidence. This function was then used as a linear variable in conjunction

with a second-order polynomial to produce a least-squares fit of the experimental

reflectance data for the test coupons. An example is shown in Figure 2 for the

longitudinal measurements of three thicknesses of drawing-agent contaminant at

Figure 2. Linear least-squares fit of experimental reflectance data for drawing-agent contaminant on

600 grit polished aluminum surfaces. Average film thickness: (Top) 0.9 µm, (Middle) 0.4 µm, (Bot-

tom) 0.1 µm.


75° angle-of-incidence. This procedure produces extremely rapid, robust analyses

of the FTIR reflectance data, even for very thin films in the presence of noise, and

accounts for baseline shifts and curvature due to interference fringes.

Fitting coefficients for the linear spectral function (which are proportional to the

integrated intensity) are plotted against the average calculated film thickness, and

these results are shown in Figures 3 and 4 for longitudinal and transverse reflec-

tance measurements at 75° and 60° angle-of-incidence, respectively. Results for

Figure 3. Integrated reflection-absorption intensity at 60° angle-of-incidence for C-H stretching bands of drawing-agent films deposited on aluminum test coupons with varying degree of surface roughness (longitudinal, top; transverse, bottom).


the longitudinal, 60° angle-of-incidence follow a linear relationship with film

thickness except for the roughest surface finish (80 grit, Ra = 1.5 µm). The instru-

ment response functions for transverse measurements at 60° angle-of-incidence are

also reasonably linear, with the same average slope as seen in Figure 3.

In contrast, analysis of the FTIR reflectance data at 75° angle-of-incidence for

both longitudinal and transverse sample orientations shows a marked departure

from linearity at the highest values of film thickness (Figure 4). The initial slopes

Figure 4. Integrated reflection-absorption intensity at 75° angle-of-incidence for C-H stretching bands of drawing-agent films deposited on aluminum test coupons with varying degree of surfaceroughness (longitudinal, top; transverse, bottom).


of the spectral response, the integrated reflection-absorption intensity, of these

samples are slightly greater than the intensity of the spectral response of the same

samples measured via a 60° angle of incidence data (Figure 3). This behavior is

expected due to the increase in reflection-absorption sensitivity with increasing

angle of incidence. Here, too, the average initial slope (and hence instrument sen-

sitivity) is the same for both transverse and longitudinal orientations.

The pronounced non-linearity in slope for the thickest films at 75° angle-of-

incidence was unexpected. An increasingly non-linear response may be observed

for thicker absorbing films, and this effect will become more pronounced as the

angle of incidence is also increased. The interpretation of the data implying that

measurement of a thicker film, sampled at a steeper angle, generated the observed

non-linearity in the data is not substantiated by the calculated spectra for the pre-

sent measurement conditions due to the small change from 60 to 75° in the angle

of incidence. Furthermore, such a non-linear effect would be most pronounced for

measurements on the smoothest substrate (Figure 4, filled circles) where the ef-

fective local orientation of the surface is most constant with respect to the illumi-

nation beam. Instead of observing such non-linear behavior the measurements

made on the smoothest surface are by far the most linear sample series for the 75°

data.

We attribute the pronounced non-linearity of the 75° data for the thickest draw-

ing-agent films to the morphological characteristics of the material as deposited

on the aluminum test panel surface. As described above, the drawing-agent mate-

rial is highly viscous and forms a visibly heterogeneous white film at 1-µm thick-

ness. Variations in the deposition process produce relatively thick local areas of

drawing-agent film and result in accretion of solid residue along the polishing

grooves and ridges of the aluminum substrate. Under these circumstances, illumi-

nation of the surface with the FTIR beam at an angle of 75° may result in shadow-

ing by contaminant material on ridge structures for all except the smoothest (600

grit polish) surface. The 12-mm diameter focal area of the infrared beam is elon-

gated by a factor of four for this angle of incidence. In contrast, reflectance meas-

urements at 60° result in only a factor of 2 elongation, and minimize the shadow-

ing effect of thick films except for ridges on the roughest (80 grit polish) surfaces.

This interpretation is substantiated by reflectance data for the second test set

(lubricant material) as shown in Figure 5. FTIR reflectance measurements have

been made at 75° angle-of-incidence for a test series similar to that of the draw-

ing-agent set. An analysis of the C-H stretching frequencies shows a strikingly

more linear dependence of instrument response with film thickness (with the ex-

ception of a single point for one of the panels with a 220 grit surface finish). We

believe that this is due to the more fluid characteristic of the lubricant material,

which allows the deposited film to conform much more closely to the surface to-

pography of the test coupons. This behavior may also account for the stronger de-

pendence of the integrated intensity slope with surface roughness, when compared

to the nearly constant results for the drawing-agent contaminant examined above.


Figure 5. Integrated reflection-absorption intensities of C-H stretching bands for lubricant films de-posited on aluminum test coupons with varying degree of surface roughness for longitudinal illumi-nation.

Even though excellent sensitivity was demonstrated for common hydrocarbon

contaminants using grazing-angle infrared reflectance spectroscopy, concerns re-

main due to potential interference from other molecular species that may be pre-

sent in the measurement environment. Chief among these is water, resulting either

from cleaning operations or the local environment. Water is a very strong infrared

absorber, and its presence on the surface to be measured may cause distortion or

obscuration of the characteristic contaminant reflection spectrum.

We performed an evaluation of this interference using lubricant-contaminated

test panels with an average hydrocarbon thickness of 0.7 µm on aluminum. A wa-

ter film was created on the surface of the test coupon using an airbrush, and re-

flection-absorption measurements were acquired at a 75° angle of incidence for

several conditions. The thickness of the water film was difficult to determine due

to continuous evaporation during the reflectance measurements. We estimated the

thickness by measuring coupon weight gain immediately prior to and following

the infrared measurements. Film thickness was calculated based on the average

weight gain.

Reflection-absorption spectra are presented in Figure 6 for three water films on

the lubricant-contaminated test panel. These water films range in thickness from 1

µm (not visible to the eye) to 7 µm (clearly visible to the eye). Substantial inter-

ference is present in the 1700 cm-1

spectral range (not shown) due to the strong H-

O-H bending mode. This strong absorption obscures carbonyl absorption features

that may be present in some, but not all, hydrocarbon contaminant species. The


broad H-OH stretching bands centered near 3400 cm-1

, however, do not obscure

the C-H stretching bands near 2900 cm-1

. This is particularly important for the ef-

fective and accurate use of the tunable infrared-laser imaging instrument, since

images are acquired for only a small number of frequencies near 3000 cm-1

, in

contrast to the broad-band spectral data collected by the FTIR instrument.

3.2. Tunable infrared-laser imaging

Initial images of test panel surfaces were acquired at two frequencies (2915 and

3000 cm-1

) that correspond to highly absorbing and non-absorbing portions, re-

spectively, of the hydrocarbon infrared spectrum (see above, Figures 2 and 6). We

used an acquisition time of 0.5 ms per frame, and averaged a minimum of 20

frames for each frequency in order to reduce noise (shot noise and laser speckle

noise). Although the InSb FPA camera is square (256 x 256 pixels), the aspect ra-

tio of the surface area scanned by the spectrometer and the resulting images in

this work are elongated by a factor of two due to the trigonometric effects of the

60° angle of incidence and reflectance.

Images were acquired for illumination transverse to the polishing direction.

They have been corrected for thermal background emission and normalized for

system spectral response at the measurement frequencies. The normalization fac-

Figure 6. Potential interference effects of water on C-H stretching bands of hydrocarbon lubricant

film (0.7 µm) on aluminum. Three thicknesses of water film were examined (1 µm, top; 3 µm, mid-

dle; and 7 µm, bottom).


tor was determined by the average intensity ratio of a clean surface (the uncon-

taminated back surface of the test panel) for the two measurement frequencies.

The ratios of successive images using the PPLN-based laser system showed a

noise level of 0.44% for the entire 65,536-pixel image under our current operating

conditions. This noise level corresponds to a hydrocarbon film thickness of ap-

proximately 10-20 nm for the species examined in this report, and is the primary

factor in determining the present instrumental detection limit.

Gray-scale images at these two frequencies for the hydrocarbon drawing-agent

(thickness of 0.9 µm on aluminum) are shown in Figure 7. Structure in the images

is primarily in the form of vertical lines that represent ridges in the aluminum

substrate formed during surface polishing operations. A darker vertical band near

the center of the image manifests the presence of an absorbing hydrocarbon in the

Figure 7. Gray-scale on-resonance (2915 cm-1, top) and off-resonance (3000 cm-1, bottom) images

for an aluminum test panel contaminated with hydrocarbon drawing agent of 0.9-µm thickness.


2915 cm–1

image. However, it is difficult to differentiate the absorbing organic

film from the high contrast presented by the surface polishing marks in images at

a single wavelength.

The image created from the ratio of the two images, corrected for thermal

background and normalized for the average image intensity, is a relative reflec-

tance image, as shown in Figure 8 (A), assuming that the reflectance of the sub-

strate remains constant at these two frequencies. Unprocessed image ratios such

as these show a periodic grid pattern due to coherent interference effects that tend

to obscure the hydrocarbon image, and we have investigated several image en-

hancement procedures to reduce noise while maintaining spatial resolution and

contrast in the reflectance ratio images. Weighted Gaussian smoothing in a 7 x 7

pixel neighborhood and Fourier filtering have both been successful in suppressing

this noise without significant degradation in spatial resolution, as shown in Figure

8 (B). The image ratios presented in this report have all been Gaussian smoothed.

Reflectance intensity profiles along the horizontal line in each image ratio are also

shown in Figure 8 (C) and (D) to demonstrate the magnitude of laser coherence

noise and the effects of the smoothing procedure.

Figure 8. Reflectance images and line-intensity profiles for an aluminum test panel contaminated

with a hydrocarbon drawing-agent of 0.9-µm thickness. Laser coherence noise (A) and results of Gaussian smoothing (C) are illustrated with corresponding intensity profiles (B and D, respectively) sampled along the horizontal lines superimposed on the images.


Examples of reflectance ratio images for several test surfaces are shown in Fig-

ures 9 and 11 in false color. A calibrated color-table (“Rainbow”) for these false-

color images is shown in Figure 10. Images for a series of 600-grit polished alu-

minum substrates contaminated with drawing agent are presented in Figure 9.

These are the same specimens whose FTIR spectra are shown in Figure 2. Aver-

age film thicknesses for the three samples are 0.9 µm (top, left), 0.4 µm (middle,

left), and 0.1 µm (bottom, left).

The images are presented in false color format with identical dynamic range to

help visualize the location of contaminants. Hydrocarbon material was manually

deposited along the orientation of the surface polishing grooves, which is oriented

vertically in these images. Heavy deposits of the hydrocarbon residue are easily

Figure 9. False-color reflectance images and thickness profiles for three aluminum test panels con-

taminated with a hydrocarbon drawing agent (thicknesses are: 0.9 µm, top-left; 0.4 µm, middle-left;

0.1 µm, bottom-left). Corresponding line thickness profiles are shown to the right of each false-color image.


visible in the top reflectance image (red and yellow indicating the lowest reflec-

tance, hence the thickest deposit, locations), with a particularly thick vertical band

near the center. Very few areas in this image possess high reflectance values (dark

blue) characteristic of low contamination. A horizontal line across the center of

the image indicates the thickness profile, shown in Figure 9 (top, right) for this

sample. Reflectance values have been converted to thickness of the drawing-agent

hydrocarbon contaminant using the FTIR data analysis discussed above. The data

shown here indicate the thickness averaging about 0.7 µm along the profile line,

with heavier deposits up to 2 µm.

False color images of the test surfaces contaminated with lower amounts of hy-

drocarbon (Fig. 9, middle and bottom) show much less spatial variation in the dis-

tribution of hydrocarbon residue. Hydrocarbon residues are thinner and appear as

predominantly green and light blue in the false-color images while the line pro-

files show quantitatively the thickness of lubricant in these images. The average

thickness values of the three profiles presented here are consistent with the weight

change and thickness values determined by FTIR.

The potential value of the infrared-laser imaging method for cleanliness verifi-

cation is clearly demonstrated for these test panels. For these samples distribution

of the residual hydrocarbon contaminant is quite variable. In the case of the

heaviest contaminated sample, a localized cleaning to effect substantial removal

can be profitably applied to the most heavily contaminated areas.

Figure 10. Color bar for false-color images presented in Figures 9 and 11. Film thickness was cali-brated by weight-gain measurements during sample preparation and by comparison with FTIR re-flectance data.


We also acquired reflectance ratio images for test surfaces with rougher fin-

ishes for average hydrocarbon thicknesses of 0.9 µm, again using transverse illu-

mination. False-color images and corresponding thickness profiles for these two

samples are compared to the 0.9-µm thick hydrocarbon residue deposited on the

smoothest, 600-grit polished surface in Figure 11. Average thickness values from

the three profiles are in reasonable agreement for all three test panels, demonstrat-

ing that large changes in surface roughness (0.5, 2.1, and 6.1 µm) do not substan-

tially affect the measured thickness of hydrocarbon residue.

We observe a qualitative change in the false-color images in Figure 11. In-

creasingly rough test surfaces (middle and bottom) exhibit a grainier image qual-

Figure 11. False-color reflectance images and thickness profiles for three aluminum test panels with a hydrocarbon drawing-agent contaminant (surface polishes are: 600-grit, top-left; 220-grit, middle-left; 80-grit, bottom-left). Corresponding line thickness profiles are shown to the right of each false-color image.


ity due to the large diversity of surface orientations relative to the infrared laser il-

lumination beam. Distribution of the hydrocarbon residue on the 220-grit surface,

however, is much more even (Fig. 11, middle, left) than for the smoothest surface

(Fig. 11, top, left). The drawing-agent material shows a strong thickness gradient

toward the right-hand side of the image for the roughest, 80-grit, surface (Fig. 11,

bottom, left) that is clearly visible despite the grainy image appearance.

4. CONCLUSIONS

The work presented in this report has shown tunable infrared-laser imaging to be

an extremely attractive method for on-line detection of hydrocarbon contaminants

and determination of their spatial distribution for efficient cleaning operations.

Calibrated test panels of hydrocarbon contaminants on metallic substrates were

prepared and characterized with FTIR grazing-angle reflectance spectroscopy.

Measurements were made over a range of film thicknesses and surface roughness,

and the derived instrument sensitivity was quite robust with respect to the degree

of surface roughness and the orientation of the reflectance unit to the direction of

polishing grooves.

Tunable infrared-laser images were acquired at both absorbing and non-

absorbing frequencies for hydrocarbon contaminants on aluminum test panels.

The thickness of the contaminant layers calculated from the laser images showed

good agreement with the measured film thickness determined by spatially aver-

aged FTIR spectroscopic results. The laser images clearly reveal the heterogene-

ous distribution of the contaminant species on the component surfaces for a vari-

ety of film thicknesses and degree of surface roughness.

Primarily, the effects of laser-coherence noise determine the current detection

limits of the laser-imaging method. The noise is introduced when an image ratio

is formed from images taken at absorbing and non-absorbing wavelengths. For

typical hydrocarbon species, the detection limit appears to be on the order of 10-

20 nm for film thickness. Improvements in the system despeckling and projection

optics may substantially decrease this noise level with an attendant increase in

sensitivity.

The configuration of a future prototype imaging system instrument will be

strongly determined by system formats that employ either a pulsed or continuous-

wave laser, and staring focal-plane array (FPA) cameras or raster-scanned imag-

ers. The design of an imaging system will include a consideration of the ultimate

instrument cost. At the present time, it appears that a continuous-wave system

with a scanned imager offers the system with the lowest cost. However, the per-

formance of some newly developed inexpensive infrared microbolometer arrays

will also be evaluated as a possible component of a low-cost pulsed imager. Fu-

ture work will enlarge both the laser illumination area and image field of view in

order to develop a prototype instrument capable of rapid large-area surveys during

cleaning verification.


Acknowledgments

We gratefully acknowledge the financial support for these investigations by the

Department of Defense through the Strategic Environmental Research and Devel-

opment Program.

REFERENCES

1. T.A. Hoffard, C.A. Kodres and D.R. Polly, Technical Memorandum, NFESC-TM-2335-SHR

(2000).∗ 2. C.A. Kodres, D.R. Polly and T.A. Hoffard, Technical Report, NFESC-TR-2067-ENV (1997).* 3. C.A. Kodres, D.R. Polly and T.A. Hoffard, Metal Finishing 95, 48-53 (1997). 4. R.G. Greenler, J. Chem. Phys. 44, 310-315 (1966). 5. D.L. Allara, in: Characterization of Metal and Polymer Surfaces, L.H. Lee (Ed.), Vol. 2, pp.

193-206, Academic Press, New York (1977). 6. W.G. Golden, in: Fourier Transform Infrared Spectroscopy-Applications to Chemical Systems,

J.R. Ferraro and L.J. Basile (Eds.), Vol. 4, pp. 315-344, Academic Press, New York (1985). 7. D.K. Ottesen, J. Electrochem. Soc. 132, 2250-2257 (1985). 8. D.K. Ottesen, L.R. Thorne and R.W. Bradshaw, Sandia Report, SAND86-8789 (1986).* 9. R.W. Bradshaw, D.K. Ottesen, L.R. Thorne, A.L. Newman and L.N. Tallerico, Sandia Report,

SAND87-8241 (1987).* 10. P.E. Powers, T.J. Kulp and S.E. Bisson, Optics Letters 23, 159-169 (1998).

∗NFESC technical reports may be ordered from the web at www.dtic.mil. Reports from Sandia

National Laboratories may be ordered by contacting Sandia National Laboratories’ Technical Li-braries at (505) 845-8287 or the National Technical Information Service (NTIS) at www.ntis.gov.


Ed. K.L. Mittal

© VSP 2003

Monitoring cleanliness and defining acceptable

cleanliness levels

MANTOSH K. CHAWLA∗

Photo Emission Tech., Inc., 3255 Grande Vista Drive, Newbury Park, CA 91320

Abstract—Defining and maintaining a “proper” level of surface cleanliness is, at best, subjective. Often the failure of surface preparation processes is not discovered until problems, such as poor ad-hesion, occur down stream. Surface cleanliness is critical for good surface finish or success of sub-sequent operations that depend on surface cleanliness. To assure consistent quality of surface cleanliness, it is important to: understand the types of contaminants that need to be monitored, most common cleanliness monitoring methods and their strengths and limitations, factors to be consid-ered in choosing appropriate cleanliness monitoring method(s), and cost impact of various cleanli-ness levels.

The selection of a cleanliness monitoring method should take into account several factors, such as the type of substrate and the types of contaminants to be monitored, etc.

In order to define “Acceptable” level of cleanliness, a total cost approach is needed. Total cost is defined as the cost of cleaning added to the cost of non-conformance related to a particular level of surface cleanliness. An acceptable level of cleanliness is the one that minimizes or optimizes this “total cost”.

Keywords: Acceptable cleanliness levels; optimum cleanliness level; total cost of cleaning; cleanliness monitoring methods.

1. INTRODUCTION

Defining and maintaining the surface preparation at “proper” levels is the key to

good surface finish. However defining a “proper” level of surface cleanliness is,

at best, subjective. For consistent results, it is important to define “how clean is

clean”. Often the inadequacy of surface preparation processes is not discovered

until problems, such as poor adhesion, occur downstream resulting in non-

conformance due to poor surface cleanliness. To assure consistent quality of sur-

face cleanliness, it is important to: understand the types of contaminants to be

monitored; most common cleanliness monitoring techniques and their strengths

and limitations; factors that affect the choice of cleanliness monitoring tech-

nique(s); select an appropriate cleanliness monitoring method; specify a desirable

∗Phone: (805) 499-7667, Fax: (805) 499-6854, E-mail: [email protected]

M.K. Chawla 24

level of surface cleanliness; and monitor the surface cleanliness to an established

level on an on-going basis.

The selection of a cleanliness verification technique, as a minimum, should

take into account the type of substrate and the types of contaminants to be moni-

tored, desired level of cleanliness, speed of measurement, operator skill level re-

quired, and acquisition and operating costs. In addition, it is very important that

the cleanliness monitoring technique be quantitative, non-destructive and readily

usable.

For every level of cleanliness, there is a corresponding level of product per-

formance (i.e. failure / non-conformance rate). Each level of cleanliness has a cost

associated with achieving that level, just as there is a cost associated with the fail-

ure / non-conformance rate corresponding to each level of cleanliness. These two

cost components can be combined to assess “total cost” of cleaning. A minimum

“total cost” can only be achieved by balancing the cost of incremental cleaning

with the reduced cost of corresponding failure / non-conformance rate. The “op-

timum” level of cleanliness is the one that minimizes the “total cost”. Since all

processes have some variation, there is bound to be some variation in the level of

cleanliness achieved. An acceptable variation around the “optimum” level of

cleanliness, where the total cost is minimum, would define the “Acceptable

cleanliness level”. Some suggested approaches to defining acceptable levels of

surface cleanliness are also discussed.

2. TYPES OF CONTAMINATION

A contamination is defined as any undesirable foreign matter that is present on a

surface. Contaminations can be classified into three different categories: 1) Par-

ticulate, 2) Thin Film (Both Organic and Inorganic), and 3) Microbial or biologi-

cal contamination.

(1) Particulate contamination can be defined as any foreign matter present on the

surface as a physical object. Some examples of particulate contaminants are

dust, hair, micro-fragments and fibers.

(2) Thin film contamination, also called Molecular contamination, is present on

the surface in the form of a thin film covering the whole surface or some areas

of the surface. This type of contamination can be organic or inorganic. Some

examples of thin film contaminants are skin oil, grease, surfactant/chemical

residues, oxides and other unwanted films.

(3) Microbial contamination can be present on the surface in the form of particles

or thin films or a combination of both and refers to generally unwanted living

organisms present on the surface. Some examples of microbial contaminants

are spore, bacilli and organic cultures. This type of contamination generally

occurs from the environment or residues from processes.

Monitoring cleanliness and defining acceptable cleanliness levels 25

3. TYPES OF CLEANLINESS MONITORING METHODS

Cleanliness monitoring methods can also be generally classified into three differ-

ent categories: 1) Indirect Methods, 2) Direct Methods, and 3) Analytical Meth-

ods. All of these methods have certain strengths and limitations, which will be

discussed later; hence, it is important to select the method that will be most ap-

propriate for a particular application. Most of these methods are appropriate for

thin film or molecular contamination.

(1) Indirect methods – Any technique that does not take a measurement on the

actual surface or area of interest would be classified as an indirect method.

See Table 1 for some of the most common indirect methods along with their

features.

(2) Direct methods – Any technique that takes a measurement directly from the

actual surface or area of interest but does not directly identify the species of

contamination present would be classified as a direct method. Some of the

most common direct methods along with their features are listed in Table 1.

(3) Analytical methods – Any technique that identifies the species of, and meas-

ures the amount of contamination would be classified as an analytical tech-

nique. Analytical techniques can be direct or indirect; however all of them

usually determine the amount of and the species of contamination. Some of

the most common analytical methods along with their features are listed in

Table 2.

4. MOST COMMON VERIFICATION / MEASUREMENT METHODS

Some of the most common indirect, direct and analytical methods, with a brief

discussion of their principles of operation, are presented below.

4.1. Indirect methods

4.1.1. Determination of non-volatile residue (NVR) [1]

Also known as gravimetric measurement. This method requires a highly sensitive

scale that can weigh parts to an accuracy of plus or minus one milligram, or bet-

ter. A container is weighed before collecting fluid that flushes the part of interest.

After the collected fluid has evaporated, the container is weighed again. The dif-

ference in the weight of the container before and after flushing and evaporation is

the weight of the contamination removed by flushing.

4.1.2. Ultraviolet (UV) spectroscopy

It involves the use of a spectrometer to analyze solvent extract from the parts of

interest. Only contaminants that have an absorption wavelength in the UV region

can be detected and analyzed. Calibration curves, utilizing samples with known

concentration of contamination, can be developed and used to determine actual

amount of contamination.

M.K

. Ch

aw

la

26

Mo

nito

ring clea

nlin

ess an

d d

efinin

g a

ccepta

ble clea

nlin

ess levels

27

M.K

. Ch

aw

la

28

Mo

nito

ring clea

nlin

ess an

d d

efinin

g a

ccepta

ble clea

nlin

ess levels

29

M.K. Chawla 30

4.1.3. Use of an optical particle counter (OPC)

As the name implies, this method is used for detecting particulate contamination.

Typically the part or surface of interest is flushed with some fluid. The fluid is

then analyzed using a particle counter. OPC gives both the count and size of par-

ticles in the suspension measured.

4.2. Direct methods

4.2.1. Magnified visual inspection

It is a step above visual inspection with the naked eye. Using some means of mag-

nification, gross contamination that may not be visible to the naked eye can be

observed. Due to its nature it is only effective with smaller parts that can be han-

dled by an operator. The method also limits the surface area that can be checked.

4.2.2. Black light

Using a black-light, i.e., UV light it is possible to visually detect gross level of

contamination. For this technique to work, however, the contaminant of interest

must fluoresce under black light. This method is somewhat similar to magnified

visual inspection, except that since the contaminants fluoresce, if present, they are

easier to see. Typically the level of contamination that can be detected with this

method is too high for most precision cleaning applications. Experiments have

shown that a skilled operator can, at best, detect 1 mg/cm2 [2].

4.2.3. Water break test

This technique utilizes the difference in surface tension of water and organic con-

taminants to detect contamination. This test will detect the presence of hydropho-

bic films on surfaces. When water is applied to the surface to be checked for con-

tamination, water covers the areas of the surface that are clean. The presence of

organic contamination on the surface prevents water from forming a film over it.

This test can be used for checking small parts as well as large surfaces. It is very

cost effective and will enable detection of molecular layers of hydrophobic or-

ganic contaminants. The sensitivity of the test may be questionable for rough or

porous surfaces.

4.2.4. Contact angle

A drop of water resting on a solid surface forms a shape that is influenced by the

solid surface tension. The shape is influenced by presence of organic contami-

nants on the surface. If a tangent is drawn from the droplet to the solid surface,

the angle formed is called “Contact Angle”. Contact angle measurements can be

used to detect organic films, coatings or contaminants on the surface. “A con-

taminated metal part would have a high contact angle, such as 90° or more. Some

parts, such as plastics, have positive contact angles even when “clean” so the

method is not typically used for cleanliness analysis for these materials. While a

number is obtained from this test, the test is still non-quantitative in terms of the

contaminants on the part [3]”. Because of its simplicity, contact angle measure-


ments have been broadly accepted for material surface analysis related to wetting,

adhesion, and absorption.

4.2.5. Optically stimulated electron emission (OSEE) [4]

A probe illuminates the surface to be tested with ultraviolet light of a particular

wavelength. This illumination stimulates the emission of electrons from the metal

surface. The emitted electrons are collected and measured as current by the in-

strument. Contamination reduces the electron emission and, therefore, the current

measured. The equipment may be connected to a computerized scanning system

that can scan a flat or cylindrical surface for cleanliness. The results can be pre-

sented as a color map or 3-D map. The user can define the level of cleanliness

each color represents in the graphic presentation of the results. This feature makes

it easy to compare “before” and “after” effect of a cleaning process or side-by-

side comparison of two pieces cleaned in alternative cleaners. OSEE is simple to

operate, fast, and relatively inexpensive. In addition, it is quantitative, non-

destructive, and non-contact. This technique detects both organic and inorganic

contamination, such as oxides, and can be used on any shape of parts as long as

the geometry of the part is presented to the sensor in a consistent manner. This

system lends itself to scan small parts or large surface areas very quickly. This

test can be used in the production line as well as for on-line real time measure-

ment of surface cleanliness. The surface of interest must emit electrons for the

technique to work. Nearly all materials of engineering importance emit electrons

when exposed to UV light.

4.2.6. MESERAN surface analyzer – (measurement and evaluation of surfaces by

evaporative rate analysis) [5]

A measurement begins by depositing onto the test surface a small volume of test

solution. A thin- end-window Geiger Müller detector is positioned above the

droplet and a metered flow of gaseous nitrogen is passed between the detector and

the test surface. To sense the volatile compound, organic compounds are used in

which one or more of the carbon atoms are Carbon-14. The β-particles given off

by the C-14 molecules at the surface are counted. Specifically measurements are

made of how many molecules there are, how many are evaporating away, how

fast they are evaporating away and, how many remain retained on the surface.

Measuring molecules provides a high degree of sensitivity and the opportunity to

analyze surfaces on a molecular scale with observations and results available in

only a few minutes. The choice of volatile chemical compounds determines

whether they react with the surface material, evaporate, or are retained by the

various physical/chemical molecular forces acting at the surface.

Chemical compounds can be found which tend to both volatilize (evaporate)

and yet tend to be retained by the surface upon which they are placed. The bal-

ance of these tendencies determines just how long the volatile compound remains

on the surface, or just how much remains. In fact, it is possible to choose a com-

pound that reacts with specific properties of the surface, or a compound where the

evaporation and/or retention are affected by certain characteristics of the surface

M.K. Chawla 32

material. By using only a monolayer equivalent of the radiochemical, the ob-

served rate of evaporation becomes a function of the residual concentration of the

non-evaporated molecules of radiochemical compound.

4.2.7. Total organic carbon (TOC) analysis [6]

This method uses oxygen gas in a combustion chamber at a set temperature to

combust carbon-based contaminants into carbon dioxide which is then detected

by CO2 coulometer. Coulometer detection uses electricity to electrochemically

measure the weight of carbon combusted in the combustion chamber. The method

is very sensitive and can detect as little as one microgram of carbon. The TOC

method works on a variety of materials and is surface-geometry independent. The

method works only on small parts or pieces of larger parts. Due to the high tem-

perature in the combustion chamber (more than 400°C) the method is not suitable

to parts sensitive to high temperature. In addition, the TOC method detects only

carbon-based contaminants, although this is generally not an issue since the ma-

jority of contaminants encountered in a manufacturing environment are carbon

based. The TOC method can be used in a laboratory but is adaptable to production

environment. It is a technique that works by oxidizing the sample to convert the

carbon into carbon dioxide, and detecting and measuring carbon dioxide. The de-

tection of carbon implies that there was some contamination that had carbon as its

constituent. The level of TOC detected determines the level of cleanliness of a

part. Since a TOC Analyzer detects only carbon, the compound of interest must

contain some carbon in a detectable quantity, in order for the analysis to be car-

ried out.

4.3. Analytical methods

Any technique that identifies the species of, and measures the amount of contami-

nation would be classified as an analytical technique. Analytical techniques can

be direct or indirect; however all of them usually determine the amount of and the

species of contamination. All of the analytical techniques involve “Probing the

surface, near-surface region, or interior of a material with electrons, ions, or pho-

tons produced radiation that has been altered depending on the number, energy, or

type of particles emitted. Changes can also occur in the frequency or absorbance

of the radiation transmitted through or reflected from the material. Each type of

analytical instrument looks at these emissions in a different way to provide infor-

mation about certain aspects of the sample, such as structure, composition, or

chemistry, and electronic or optical properties” [9]. Most of the analytical tech-

niques test the specimen in vacuum, are expensive and require high skill level to

operate and interpret the results. Testing takes time and rarely provides real-time

information. Because of the cost of analytical testing, it is recommended that its

use be limited to applications where identification of the species of contamination

is required to enhance or improve the process.

Analytical techniques can be divided into two groups; 1) Chemical/elemental

surface analysis, and 2) bulk analysis techniques. There are many techniques that


are capable of performing these analyses, some of the most common analytical

techniques are summarized below. For a more complete list of most common ana-

lytical techniques, visit www.cea.com/table/htm, website of Charles Evans & As-

sociates. For a more comprehensive list of analytical techniques visit the website

of ESCA users group in England – www.ukesca.org/tech/list/html.

4.3.1. Chemical/elemental surface analysis techniques

4.3.1.1. Auger electron spectroscopy (AES) / scanning Auger microscopy (SAM)

[7–9]

They are used to obtain elemental composition information (and some chemical

information) from the top two to five atomic layers of a material; identify the

composition of very small features and particulates on surfaces; and provide depth

composition profiles of thin films, metals, and alloys. Micro-beam AES is also

used to study grain boundaries in high temperature alloys, and to examine fracture

surfaces to determine composition and extent of damage. The Auger electrons,

named after the discoverer of the process, are produced (among other emissions)

with discrete energies, which are specific to each element, when the surface is ir-

radiated by a finely focused electron beam. Auger electrons are collected and

measured. Auger electrons have discrete kinetic energies that are characteristic of

the emitting atoms, making this technique particularly useful for identifying ele-

mental composition. The energy level of Auger electrons is specific to a species

of contamination. The escape depth of Auger electrons (1–5 nm) makes this tech-

nique very surface sensitive.

4.3.1.2. Electron spectroscopy for chemical analysis (ESCA) [7–9]

Also known as X-ray Photoelectron Spectroscopy, or XPS, is a surface analysis

technique that provides information on both elemental identity and chemical

bonding. This information can be used to identify functional groups and molecu-

lar types. This method uses special equipment to bombard the surface of interest

with X-rays under vacuum conditions, causing electrons to be ejected from the

surface. The actual elemental composition can be quantified by measuring the en-

ergy level of ejected electrons, since each element ejects electrons at a unique en-

ergy. Its application is limited to mostly research and development, but it can be

used to calibrate and evaluate other, less sophisticated measurement methods.

4.3.1.3. Secondary ion mass spectrometry (SIMS – static) [7–9]

A surface analysis technique used for identifying molecules on a surface, as well

as for depth profiling for tracking very low concentrations of contaminants or ion-

implanted species. SIMS technique includes static SIMS (SSIMS), dynamic

SIMS, and time-of flight SIMS (TOF SIMS). SSIMS can identify organic and in-

organic species. TOF SIMS is an ultra-precise and accurate technique for measur-

ing the mass of molecules in the near-surface layers of material. A pulsed primary

ion beam is used to sputter material from the surface of the sample. Secondary

ions are collected and focused into a reflection time-of-flight (TOF) mass spec-

trometer, where they are mass analyzed. Analysis involves measuring the length

M.K. Chawla 34

of time it takes the secondary ions to reach the detector. The lighter the ion, the

less time it takes to reach the detector. From the arrival time the masses of the

species can be identified. High sensitivity depth profiling is a key feature.

4.3.1.4. Secondary ion mass spectrometry (SIMS – dynamic) [7–9]

It uses a much higher intensity bombarding beam than Static SIMS, and is a par-

ticularly sensitive (less than part-per-billion level) method for depth profiling of

dopants and trace elements in semiconductors. It can also map the X-Y distribu-

tion of atomic species with sub-micrometer spatial resolution. An energetic pri-

mary ion beam is used to sputter atoms from the sample surface. Secondary ions

emitted are mass analyzed. It is inherently a profiling technique. It uses O2 or Cs

ions to bombard a surface in high vacuum. High sensitivity depth profiling is a

key feature.

4.3.1.5. Variable-angle spectroscopic ellipsometry (VASE) [7, 8]

It is a noninvasive technique that offers information about surface composition,

layer thickness, and optical properties. Its applications include examining optical

surfaces and crystals, and measuring and analyzing band gaps in semiconductors,

optical devices, thin films, and carbon coatings on computer hard disks.

4.3.1.6. Energy dispersive X-ray (EDX) and wavelength dispersive X-ray (WDX)

analyses [7, 8]

They are often combined with a scanning electron microscope or electron micro-

probe. EDX provides simultaneous multi-element analysis and elemental mapping

capabilities for a region up to a few micrometers deep. WDX analyzes trace

amounts of one element at a time and is more quantitative than EDX. An example

of EDX application is identifying silicon nitride and titanium carbide inclusions in

stainless steel.

4.3.2. Bulk analysis techniques

The following are several analytical techniques that typically are used for chemi-

cal or elemental analysis of bulk materials, but these can also be adapted for the

characterization of surfaces and thin films. Many times these techniques are used

in industry for characterizing surfaces, sometimes without full knowledge of the

strengths and limitations of these techniques. It is hoped that information about

how these techniques work, their strengths and limitations would help the reader

in determining their usefulness and limitations for their applications.

4.3.2.1. Fourier transform infrared (FTIR) spectroscopy [7, 8]

It provides information about the chemical bonding and molecular structure of or-

ganics and some inorganic solids, liquids, gases and films. This technique is espe-

cially good for identifying unknowns when reference IR spectra are available.

When an infrared beam impinges on a surface, the molecular constituents vibrate

in the infrared regime. The identities, surrounding environments, and concentra-

tions of these oscillating chemical bonds can be determined. FTIR is a powerful

analytical tool for characterizing and identifying organic molecules. The IR spec-

trum of an organic compound serves as its fingerprint and provides information


about chemical bonding and molecular structure. This information can be used to

detect the types of organic materials present on the surface.

4.3.2.2. Raman spectroscopy (RS) [7, 8]

It is used to examine the energy levels of molecules that cannot be well character-

ized via infrared spectroscopy. The two techniques, however, are complimentary.

In the RS, a sample is irradiated with a strong monochromatic light source (usu-

ally a laser). Most of the radiation will scatter or “reflect off” the sample at the

same energy as the incoming laser radiation. However, a small amount will scat-

ter from the sample at a wavelength slightly shifted from the original wavelength.

It is possible to study the molecular structure or determine the chemical identity

of the sample. It is quite straightforward to identify compounds by spectral library

search. Due to extensive library spectral information, the unique spectral finger-

print of every compound, and the ease with which such analyses can be per-

formed, the RS is a very useful technique for various applications. An important

application of the RS is the rapid, nondestructive characterization of diamond,

diamond-like, and amorphous-carbon films.

4.3.2.3. Scanning electron microscopy (SEM) / energy dispersive X-ray analysis

(EDX) [7, 8]

The SEM produces detailed photographs that provide important information about

the surface structure and morphology of almost any kind of sample. Image analy-

sis is often the first and most important step in problem solving and failure analy-

sis. With SEM, a focused beam of high-energy electrons is scanned over the sur-

face of a material, causing a variety of signals, secondary electrons, X-rays,

photons, etc. – each of which may be used to characterize the material with re-

spect to specific properties. The signals are used to modulate the brightness on a

CRT display, thereby providing a high-resolution map of the selected material

property. It is a surface imaging technique, but with Energy Dispersive X-ray

(EDX) it can identify elements in the near-surface region. This technique is most

useful for imaging particles.

4.3.2.4. X-ray fluorescence (XRF) [7, 8]

Incident X-rays are used to excite surface atoms. The atoms relax through the

emission of an X-ray with energy characteristic of the parent atoms and the inten-

sity proportional to the amount of the element present. It is a bulk or “total mate-

rials” characterization technique for rapid, simultaneous, and nondestructive

analysis of elements having an atomic number higher than that of boron. Tradi-

tional bulk analysis applications include identifying metals and alloys, detecting

trace elements in liquids, and identifying residues and deposits.

4.3.2.5. Total-reflection X-ray fluorescence (TXRF) [7, 8]

It is a special XRF technique that provides extremely sensitive measures of the

elements present in a material’s outer surface. Applications include searching for

metal contamination in thin films on silicon wafers and detecting picogram-levels

of arsenic, lead, mercury and cadmium on hazardous, chemical fume hoods.

M.K. Chawla 36

5. CONSIDERATIONS FOR SELECTING A CLEANLINESS MONITORING

METHOD [10]

There are several factors that should be considered in selecting a method for

monitoring surface cleanliness. The factors discussed here are the ones that are

most important but by no means represent a complete list of factors that should be

considered. There may be other factors that are pertinent to a particular applica-

tion that should be considered.

(1) Type of contaminant – One of the first factors that should be considered in se-

lecting a cleanliness monitoring method is the type of contaminant that need

to be monitored. Is the contaminant particulate or thin film type? If thin film

contamination, is it organic or inorganic or both? Does the technique under

consideration monitor the type of contaminants that need to be monitored?

(2) Types of substrates – What type of substrate is going to be monitored? Are

the techniques under consideration capable of monitoring this type of sub-

strates? Are the techniques likely to damage the substrate to be monitored?

(3) Level of cleanliness to be monitored – It is important that the level of contami-

nation that is expected or tolerable can be monitored by the technique under

consideration. It is recommended that samples with different levels of contami-

nation be monitored with the technique(s) under consideration. In evaluating

the technique for suitability, prepared samples should have levels of contamina-

tion spanning a range from 0% (i.e. clean surfaces) to maybe 200% of the ex-

pected level of contamination on the surface. The technique(s) should not have

any problem in distinguishing between different levels of contamination.

(4) Features of monitoring method – It is important to consider various features

of the method under consideration. For example, is the technique non-contact

and/or non-destructive? Does the technique require deposit of some medium

on the surface? For example, the contact angle measurement requires that a

droplet of water be placed on the surface of interest. How large an area can

the technique measure? Is it sensitive to surface roughness? Can the technique

check parts of different geometries? Can the technique be used on-line? Is the

technique suitable for the environment it is going to be used in? Does the

technique cause any permanent changes to the surface? All of these questions

should be considered to determine the most appropriate monitoring technique

for a particular application.

(5) Measurement speed – Is the measurement speed critical for the application

under consideration? If so, how fast can the technique make a measurement?

Is the speed sufficient to keep up with the production flow?

(6) Acquisition and operating cost – How does the acquisition cost compare

among the techniques that meet other requirements for the application? Are

there any expendable items that would have to be purchased for continued use

of the equipment? How much does that add to the operating cost? What are

the maintenance and calibration requirements and how much these require-


ments will add to the operating cost? All these questions should be answered

to truly compare the total cost of any cleanliness monitoring system.

(7) Skill level required – The operator skill level can be a key factor in the use of

some techniques, particularly the analytical techniques. Some techniques may

involve interpretation of the data to determine the quality of surface cleanliness.

These factors should also be considered in the selection of a cleanliness measur-

ing technique. A high operator skill level will result in higher operating cost. In

the event of personnel turnover, higher training costs may also be incurred.

6. COST OF CLEANLING [10]

For every level of cleanliness, there is a cost to achieve that level of cleanliness.

There is corresponding level of failure/non-conformance for each cleanliness

level, hence cost of failures/non-conformance. “Total Cost” of achieving a certain

level of cleanliness is the sum of these two costs.

As the achieved level of surface cleanliness increases, the cost of cleaning also

increases. Eventually the incremental cost of cleaning rises exponentially. Hence

the cost of surface cleaning is directly proportional to the surface cleanliness level.

Intuitively, we know that the higher the cleanliness level the lower the fail-

ure/non-conformance rate, hence cost, due to surface cleanliness. The incremental

drop in costs due to lower failure/non-conformance also exhibits exponential rela-

tionship. Hence the cost of failures/non-conformance is inversely proportional to

the surface cleanliness level. If both of these costs were plotted on a graph, the

typical result would be like the one shown in Figure 1.

Figure 1. Total cost vs. cleanliness level.

M.K. Chawla 38

An optimum level of cleanliness is the one that minimizes the total cost. Even-

tually one can arrive at a cleanliness level where the savings in the failure/non-

conformance costs will not be offset by incremental cost of achieving cleanliness

beyond the optimum level. A small range around the optimum level of cleanliness

can be established as the “Acceptable Level” of cleanliness.

7. DEFINING ACCEPTABLE (“OPTIMUM”) LEVEL OF CLEANLINESS

It is expected that the non-conformance levels will increase as the level of

cleanliness decreases or vice versa. It is important to understand the relationship

between the level of cleanliness and non-conformance rate in order to establish

the “acceptable level of cleanliness”. For example, if the failure/non-conformance

rate is too high due to the surface cleanliness level, then the surface cleanliness

level will have to be improved in order to reduce the failure rate. On the other

hand, no failures or a very low failure rate due to the surface cleanliness level im-

plies that the surface may be “over-cleaned.”

It may be desirable to optimize the cleaning process by comparing the cost of

failures/non-conformance with the cost of cleaning the surface. Generally, in-

creasing the level of surface cleanliness will result in increased cleaning cost. An

increased level of cleanliness should lower the rate of non-conformance, which, in

turn, reduces the non-conformance cost. As long as the reduction in non-

conformance cost more than offsets the increased cost of cleaning, it would be

cost effective to increase the achieved level of surface cleanliness. When the de-

crease in non-conformance cost fails to offset the increase in the cleaning cost,

then an optimum or “acceptable” level of cleanliness has been achieved.

To establish the optimum level of surface cleanliness, two approaches are out-

lined here. One approach utilizes the success of the subsequent operation that de-

pends on surface cleanliness level. The other approach is to start monitoring the

cleanliness levels achieved and corresponding level of failure/non-conformance

rate. Once an acceptable level of cleanliness is established using one of the two

approaches, cleaning process can be monitored in production to assure ongoing

product quality.

7.1. Controlled experiment

This approach requires that the measure of success be defined for the subsequent

operation that depends on surface cleanliness. For example, if the parts are to be

bonded, then the adhesion strength of the bond will be the measure of success. If

the parts are to be coated after cleaning, then the adhesion strength of the coating

should be correlated to surface cleanliness. The acceptable level of surface

cleanliness is the one that results in the desired level of bond/adhesion strength.

One simple approach is to start monitoring and recording the cleanliness level

of each part. A statistically significant sample must be monitored to assure valid


results. These parts then must be followed through the process to measure the

level of success for each part at the subsequent operation. The level of cleanliness

that results in the desired minimum level of success is the minimum level of

cleanliness that must be achieved in production. This approach has its limitations.

For example, the results depend on what level of cleanliness is being achieved in

production. If the surface is “too clean” there may not be enough variation in the

cleanliness level to identify the point where minimum success is achieved. On the

other hand, if the surface is not clean enough the desirable success may not occur.

A more proactive approach is to prepare parts with different levels of surface

cleanliness, measure and record the cleanliness level and follow through with the

subsequent operation to correlate the success level with cleanliness level. It is

recommended that the range of cleanliness should be as wide as possible to help

identify the minimum level of cleanliness. Once again it is important that a statis-

tically significant sample be used. It is also recommended that, if possible, several

cleanliness measurements should be taken from each part and the mean cleanli-

ness level be correlated to the mean success level. Figure 2 [10] graphically de-

picts the typical result of correlating the success level to surface cleanliness level.

A minimum level of cleanliness is the one that corresponds to the target minimum

level of success.

7.2. “Benchmark” testing

Once a cleanliness monitoring method has been selected, it can be used to estab-

lish the cleanliness level achieved by current cleaning process (“Benchmark”).

The production can then be monitored to assure that benchmark cleanliness level

is being achieved. In addition, the product can be followed through the manufac-

turing process to assure that no problems occur downstream as a result of inade-

quate surface cleanliness. The level of non-conformance related to the level of

Figure 2. Peel strength vs. surface cleanliness.

M.K. Chawla 40

cleanliness achieved should be monitored. The cost associated with a given level

of cleanliness and the cost of non-conformance associated with that given level of

surface cleanliness should be combined to determine the “total cost”. Changes

should be made to the cleaning process to minimize the “total cost”, i.e. optimize

the “total cost”. The level of cleanliness associated with the “optimum total cost”

should be considered the optimum or “acceptable” level of cleanliness.

8. ON-GOING, IN-PROCESS SURFACE CLEANLINESS MONITORING

Surface cleanliness monitoring system must be used to monitor the process and

assure that the desired cleanliness level is being achieved on an on-going basis.

Surface cleanliness monitoring system can be very useful in assessing how the

surface cleanliness level is affected by making changes to the cleaning process or

for evaluating alternative cleaning processes for their ability to achieve the de-

sired cleanliness level.

The required level of cleaning agent concentration in the cleaning solution can

also be objectively determined and maintained by using a surface cleanliness

monitoring system. Measuring the effect of varying the concentration level of the

cleaning agent on surface cleanliness can help determine the “optimum” concen-

tration level.

In most industries, the chemical or cleaning agent replenishment schedule is

usually time-dependent. The success of this approach relies on the level of con-

tamination on each part and the number of parts processed in a given time interval

being relatively constant. In real life, the amount of contamination can vary con-

siderably from part to part. In addition, the number of parts cleaned during a

given time frame can also vary considerably. A time-dependent replenishment

schedule is not the ideal way of assuring product quality. On-going, in-process

monitoring of surface cleanliness helps in replenishment of chemicals or cleaning

agents only when needed, and not based on a pre-determined, somewhat arbitrary

schedule.

9. SUMMARY

In order to define an acceptable level of cleanliness, it is important to minimize

the total cost of cleaning. The total cost of cleaning is the sum of the cost of

achieving a certain level of surface cleanliness and the cost of failure/non-

conformance associated with that level of surface cleanliness. Selecting a method

for monitoring cleanliness is the first step in establishing an acceptable level of

cleanliness or defining “how clean is clean”. Several factors need to be considered

in selecting an appropriate surface cleanliness method, which include, but are not

limited to, type of contaminant to be detected, level of cleanliness to be moni-

tored, acquisition and operating cost of the monitoring method, and the skill level


required to operate the system. Surface cleanliness monitoring method may be di-

rect, indirect or analytical. A monitoring method can be used to optimize the

cleaning process by varying different parameters of the cleaning process while

monitoring surface cleanliness to see how it is affected by the change. It can also

help in ongoing monitoring of the cleaning process to assure that the desirable

level of cleanliness is being achieved.

REFERENCES

1. B. Kanegsberg and M. Chawla, “Non Volatile Residue”, A2C2 Magazine, 5, No. 3, 41 (2002) and 5, No. 4, 45 (2002).

2. R.L. Gause, “A Noncontacting Scanning Non Contact Photoelectron Emission Technique for Bonding Surface Cleanliness Inspection”, Marshall Space Flight Center, Huntsville, AL, pre-sented at Fifth Annual NASA Workshop, Cocoa Beach, Florida (1987).

3. B. Kanegsberg and M. Chawla, “Contact Angle”, A2C2 Magazine, 4, No. 8, 41 (2001). 4. Surface Quality Monitors Brochure, Photo Emission Tech., Inc. 5. B. Kanegsberg and M. Chawla, “MESERAN”, A2C2 Magazine, 4, No. 9, 49 (2001). 6. B. Kanegsberg and M. Chawla, “Total Organic Carbon”, A2C2 Magazine, 4, No. 10, 37 (2001). 7. Charles Evans Associates Website – www.cea.com 8. R.D. Cormia, “Problem-Solving SURFACE ANALYSIS Techniques”, Surface Sciences Labo-

ratories, Mountain View, CA: Advanced Materials & Processes, 16-23 (Dec. 1992). 9. Measurement and Characterization Website – www.nrel.gov/measurements/surface/html

10. M. Chawla, “How Clean is Clean? Measuring Surface Cleanliness and Defining Acceptable Level of Cleanliness”, in Handbook for Critical Cleaning, B. Kanegsberg and E. Kanegsberg (Eds.), pp. 415-430, CRC Press, Boca Raton, FL (2001).


Ed. K.L. Mittal

© VSP 2003

Tracking surface ionic contamination by ion

chromatography

BEVERLY NEWTON∗

Dionex Corporation, 500 Mercury Drive, Sunnyvale, CA 95032, USA

Abstract—Surface ionic contamination can cause device failures. In order to find the source of the contamination many questions must be answered first. Are the failures due to incoming materials that are not clean? Has there been a change in the process that is introducing contamination? What is the exact nature of the contaminant, ionic, particulate, metallic, etc? Is there a training issue that needs to be addressed. Can the failure be tested for or is it a long term reliability problem? These are just a few of the questions that must be answered as part of the troubleshooting process. This paper addresses how ion chromatography can be used to troubleshoot a manufacturing or cleaning process and to assure the quality and reliability of electronic devices. Topics covered include: 1. What is ion chromatography. 2. How does it differ from other cleanliness testing methods. 3. How can ion chromatography be used to troubleshoot a cleaning process. 4. Real life examples showing how the use of ion chromatography has improved cleaning processes.

Keywords: Ionic contamination; ion chromatography; electronic devices.

1. INTRODUCTION

As electronic devices and assemblies become smaller and more complex, the re-

quirements for improved quality control of product cleanliness have begun to es-

calate. Surface contamination from ions such as chloride, bromide, sodium, and

organic acids has been shown to cause failures in electronic devices [1]. Ionic

residues can cause corrosion, metal migration and electrical leakage. The failures

cased by these residues may be hard or soft failures and may occur several

months after the product has been manufactured and shipped to customers. Upon

re-testing the returned product, the failures can be intermittent or “no trouble

found” making troubleshooting the device for a root cause of the failure difficult.

These residues may be on the exposed surface of an electronic device, they may

be encapsulated in flux or resin deposits, they may be trapped under surface

mounted devices or they may be encapsulated in polymer finishes (Figure 1).

∗Phone: (408)4814272, Fax: (408)7372470, E-mail: [email protected]

B. Newton 44

Figure 1. Chromatogram of a board extract using IPA/water mixture.

Figure 2. Analysis of a cassette used to transport disk drive components during manufacture.

The manufacture of electronic devices typically involves a series of chemical

and mechanical operations such as plating, masking, soldering, rinsing, etching,

cleaning, etc. Each of these operations along with the environment in which they

occur leaves some effect on the device or assembly. The processes and manufac-

turing environment leave chemical “fingerprints” on the device that are unique to

the manufacturing process. In the same way that a forensic scientist would use

fingerprints to trace a criminal, analytical techniques can be used to troubleshoot a

manufacturing process or field failure to understand and correct the root cause.

Tracking surface ionic contamination by ion chromatography 45

Ionic contamination can also be found on materials that come in contact with

electrical devices during manufacture, e.g. gloves, cassettes, etc. (Figure 2). These

manufacturing consumables can transfer contamination to the manufactured

products and need to be examined for contamination in the same way that the fi-

nal product is evaluated.

2. TYPES OF IONIC CONTAMINATION

Potentially corrosive ions found on printed circuit boards and electronic devices

include:

Bromide - commonly found in solder masks

Sulfate - comes from a variety of materials such as oils and release agents

Chloride - commonly found in fluxes

Organic acids such as adipic or succinic acid - found in fluxes

Typically, the higher the concentration of corrosive ions on a particular assem-

bly, the higher the risk of electrochemical failure.

3. TEST METHODS FOR IONIC CONTAMINATION

In the past, electronic component manufacturers, board manufacturers and elec-

tronic assemblers have relied on resistivity of solvent extract (ROSE) type test

methods to assure ionic cleanliness. Several studies reported by Contamination

Studies Laboratory (CSL, Kokomo, IN) have shown that the ROSE method is in-

adequate for true quantification of ionic contamination. Recently, a modified

ROSE method has been proposed as an IPC (Association Connecting Electronics

Industries) Standard Method IPC-TM-650 2.3.25.1. Although this new technique

is an improvement for reporting overall ionic contamination, it too provides insuf-

ficient information to troubleshoot the root cause of electronic failures caused by

ionic contamination.

The technique of ion chromatography is uniquely qualified for troubleshooting

the root cause of failures due to ionic contamination on electronic devices and

printed circuit boards. Ion chromatography can provide information on the chemi-

cal nature of the residue causing the failure. The output of the ion chromatograph

is called a “chromatogram” and gives the identity and quantity of each ion found

in a sample of a rinse extract of the device of interest.

Ion chromatography is a form of liquid chromatography. The technique is

based on the use of specialized column packings for analytical separation of ions

found in a chemical mixture. The main advantages of ion chromatography for

residue analysis are:

Multi-component ion analysis

Most sensitive detection technique available for many ionic compounds

Method versatility

B. Newton 46

Figure 3. Ion chromatography system configuration.

Ion chromatography is an analytical technique used to separate, identify and

quantify ions in a sample matrix such as a water extract of a printed circuit board.

The simplest ion chromatography system is composed of a sampling device, a

pump, an analytical column, a suppressor and a detector (Figure 3).

The analysis begins with a sample, typically a water extract containing ions of

interest such as chloride, sulfate, or nitrate. A portion of the sample is injected

into the ion chromatography system and combined with an eluent stream com-

posed of sodium hydroxide or bicarbonate solution. The eluent stream carries the

sample through the ion chromatography system to the analytical column. The ana-

lytical column separates the ions of interest in the sample into narrow bands

within the stream of the eluent. Thus, by the time the sample leaves the analytical

column, all of the chloride ions are grouped together, then all of the nitrate ions

and then all of the sulfate ions. The eluent then sweeps these groups of ions into

the suppressor device. This device electrolytically transforms the eluent into pure

water leaving just the ions of interest in pure water to be swept along to the con-

ductivity detector. The detector detects the ions based on their conductivity rela-

tive to the water eluent. At this point all interfering ions have been removed and

the detector’s sensitivity has been maximized allowing for detection of very low

(part per billion) levels of ions [2].

This is a very simplified explanation of ion chromatography but it is important

to note that more complex samples and analytes can also be analyzed using this

technique (for instance, cations such as sodium and magnesium, transition metals

such as iron and copper and even certain biological analytes such as amines and

nucleic acids).

Tracking surface ionic contamination by ion chromatography 47

4. TRACKING SURFACE IONIC CONTAMINATION IN MANUFACTURING

AND ASSEMBLY OPERATIONS

There has been a growing interest in the analysis of ionic contamination on elec-

tronic components. Absolute contamination level requirements and guidelines

have not been determined; however, Contamination Studies Laboratory (CSL,

Kokomo, IN) recommends maximum levels of chloride ion in the range 1.0 µg/sq.

in for assembled boards with sensitive components such as microBGAs. The level

recommended for bare boards is less than 2.0 µg/sq. in [3].

Ion chromatography provides the unique capability of identifying the individ-

ual ions for a given contamination issue. Since the source for chloride contamina-

tion can be much different than the source for organic acid contamination it is im-

portant to know which ions the manufacturer is dealing with in order to

understand and correct the root cause of the problem. This is not possible with re-

sistivity of solvent extract (ROSE) measurements. The capability to identify and

quantify individual ions makes ion chromatography a valuable troubleshooting

tool for process contamination issues and process monitoring programs. In addi-

tion to being the most economical analytical technique for monitoring multiple

ions, ion chromatography also provides the ability to distinguish between noncor-

rosive and corrosive ions, something that ROSE testing is unable to do.

A number of studies have been published to show the use of ion chromatogra-

phy to troubleshoot reliability issues with electronic products. One of the best

sources of case study information can be found on the web site for Contamination

Studies Laboratory (CSL) at www.residues.com. CSL regularly publishes case

studies showing the hazards of ionic contamination to electronic device reliability

on their web site and in each issue of Circuits Assembly magazine.

A good explanation of how ion chromatography has been used to identify

sources of CAF (conductive anodic filament) failures can be found in a study

completed by Ready et al. [4].

Several studies [5-7] have been completed on the analysis of ionic contamina-

tion on failed disk drive components.

As mentioned earlier, manufacturing materials and packaging can be an impor-

tant source of ionic contamination. Two recent studies by Lin and Graves [8] and

Bahten and McMullen [9] provide information on the use of ion chromatography

for the analysis of ionic contamination on materials such as pink poly film (a

common packaging material) and cleaning brushes.

5. STANDARD TEST METHODS FOR TRACKING IONIC CONTAMINATION

IPC (Association Connecting Electronics Industries) has standard test methods

documented for the ROSE, Modified ROSE and Ion Chromatography analysis

techniques. These are:

B. Newton 48

IPC-TM-650, TM 2.3.25 Detection and Measurement of Ionizable Surface

Contaminants by Resistivity of Solvent Extract (ROSE).

IPC-TM-650, TM 2.3.25.1 Ionic Cleanliness Testing of Bare PWBs (modi-

fied ROSE Test Method).

IPC-TM-650, TM 2.3.28 Ionic Analysis of Circuit Boards, Ion Chromatogra-

phy Method.

IDEMA (International Disk Drive Equipment and Materials Association) has

developed the following standard test method for ionic cleanliness testing.

M13-99, Measurement of Extractable/Leachable Anion Contamination on

Drive Components by Ion Chromatography.

6. CONCLUSION

The ion chromatography, ROSE and modified ROSE test methods have been de-

veloped to allow electronics manufacturers to identify and control ionic contami-

nation before it evolves into a failed component or board. Tracking ionic con-

tamination requires systematic troubleshooting and improved cleanliness of the

product as it is manufactured. This means cleaner raw materials and processes

which are controlled by systematic analysis using standard methods such as those

documented by IPC, IDEMA, IEST (Institute of Environmental Sciences and

Technology), and SEMI (Semiconductor Equipment and Materials International).

REFERENCES

1. D. Yang, C. Lee, Y. Yang, E. Kaiser, S. Heberling and B. Newton, Precision Cleaning, 17-23 (May 1998).

2. B. Newton, Precision Cleaning, 38-39 (March 2000). 3. D. Pauls and T. Munson, Circuits Assembly, 110-112 (September 1998). 4. W.J. Ready, B.A. Smith, L.J. Turbini and S.R. Stock, Mater. Res. Soc. Symp. Proc. 515, 45-54

(1998). 5. A. Toxen, A2C2, 13-16 (September 1998). 6. P. Mee, M. Smallen and D. Vickers, IDEMA Insight, 1 (March/April 1997). 7. J. Thompson, T. Prommanuwat, A. Siriraks and S. Heberling, IDEMA Insight, 24-29 (May/June

1999). 8. S. Lin and S. Graves, Micro, 95-106 (October 1998). 9. K. Bahten and D. McMullen, Proc. Semiconductor Pure Water and Chemicals Conference, 355-

364 (March 1999).


Ed. K.L. Mittal

© VSP 2003

A new method using MESERAN technique for

measuring surface contamination after solvent

extraction

MARK G. BENKOVICH∗,1 and JOHN L. ANDERSON2 1Honeywell Federal Manufacturing & Technologies,† PO Box 419159, D/833 MS-2C43,

Kansas City, MO 64141-6159 2ERA Systems, Inc., The MESERAN Company, PO Box 3609, Chattanooga, TN 37404-0609

Abstract—The precision analytical technique known as MESERAN Analysis permits, in 2 minutes, quantitative measurement of the level of pre-existing nonvolatile organic residue (NVOR) on a sub-

strate from <1 ng/cm2 to >100 µg/cm2. MESERAN Analysis is also applicable for determining NVOR deposited from solvents and solvent extracts. The MESERAN method is able to quantify or-

ganic contamination levels down to and below 1 ng by depositing as little as 10 µL of solvent con-taining a known amount of contamination on a clean substrate, allowing it to evaporate, and measur-ing the resultant residue. The method is described in detail. In addition, NVOR measurements determined from MESERAN data are presented for a specific project conducted at Honeywell Fed-eral Manufacturing & Technologies (FM&T), Kansas City Plant (KCP).

Keywords: MESERAN; surface contamination; solvent extraction; non-volatile organic residue.

DEFINITIONS

In this paper a number of abbreviations, special terms, and trademarks are em-ployed:

(1) µCi means microCurie, a unit of radiation which corresponds to 3.7 E 4 (37,000) disintegrations per second.

(2) Carbon-14 (C-14) refers to the radioactive isotope of the element Carbon, an isotope which emits only soft or low energy beta particles; most C-14 beta particles are stopped by a sheet of paper.

(3) USNRC EXEMPT means the very low level of Carbon-14 that is not regu-lated by the U S Nuclear Regulatory Commission. No license is required for

∗To whom all correspondence should be addressed. Phone: (816) 997-3529,

Fax: (816) 997-2049, E-mail: [email protected] †Operated for the United States Department of Energy under prime contract

DE-AC04-01AL66850. Copyright Honeywell LLC, 2002.


possession or use. Only USNRC licensed companies are permitted to distrib-

ute EXEMPT quantities not to exceed ten 100 µCi of C-14 (or combinations

of smaller quantities that added up to 100 µCi of C-14) at one time. Multiple quantities may be stored by the user. Shipments may be made to anyone in the US. Foreign shipments under IATA regulations must conform to the regula-tions of the country of final destination.

(4) ng means nanogram (1 E - 9 grams or 0.000000001 grams); µg means micro-

gram (1 E - 6 grams or 0.000001 grams), mg means milligram (1 E - 3 grams or 0.001 grams).

(5) µL means microliter (1 E - 6 liter or 0.000001 liter).

(6) GM detector refers to a thin end-window Geiger Müller detector tube which detects the C-14 beta emissions which penetrate through the 1.4–2.0 mg/cm2 mica window.

(7) One nanomole (nmole) is 1 E - 9 moles which equals approximately. 6 E 14 molecules (from Avogadro’s ~ 6 E 23 molecules per gram mole).

(8) 1 square centimeter (sq cm) with a roughness factor of 3 is equal to 3 E 16 square Angstroms.

(9) Monolayer refers to the number of molecules of a material which covers 1 sq cm in a conventional non-close-packed configuration. For example, on a smooth, flat surface with a roughness factor of 3, each molecule of n-tridecane occupies about 50 sq. Angstroms – which equates to ~ 6 E 14 mole-cules per sq. cm – i.e. one nanomole.

(10) NVR means non-volatile residue; NVOR means non-volatile organic residue.

(11) MESERAN is an acronym for Measurement and Evaluation of Surfaces by Evaporative Rate ANalysis.

(12) MESERAN, MicroSolventEvaporator (MSE), MicroOrganicResidue, and MOR are trademarks licensed to ERA Systems, Inc.

(13) Ln or ln is the natural logarithm.

(14) 1 mg/ft2 is equivalent to 1.0764 µg/cm2 or 1 µg/cm2 is equivalent to 0.929 mg/ft2.

1. INTRODUCTION

The principle of the MESERAN technique was discovered by one of us (JLA) in 1960. This analytical technique is used in a number of industrial and governmen-tal facilities (within the United States and abroad) for research and development purposes as well as for quality and production control. The characterization of the surface being analyzed is carried out by depositing a chemical detector onto the test surface and observing the rate at which the chemical detector disappears from the surface. The MESERAN technique is routinely used for quantifying organic

A new method using MESERAN technique 51

contamination on surfaces and the crosslink density (or degree of cure) in poly-mers. In addition, the MESERAN technique can be used for quantifying chemi-cally active sites on surfaces [1-3].

Honeywell FM&T, KCP (henceforth KCP) has been using MESERAN Ana-lyzers for approximately 30 years to detect and quantify organic contamination on parts and evaluate various cleaning processes for removing organic contamina-tion. KCP has used MESERAN Analyzers extensively to evaluate the ability of alternate solvents and processes for removing specific organic contaminants to eliminate the use of chlorinated and fluorinated solvents [4-10]. In recent years, KCP has been working on several projects with The MESERAN Company to im-prove data analysis and develop new methods for using the MESERAN technol-ogy [11-14].

2. PRINCIPLE OF MESERAN TECHNIQUE [15]

The standard microcomputer-based MESERAN technology involves deposition,

using a “clean” precision microsyringe, onto a flat or concave surface of 18 µL of a test solution consisting of a low boiling solvent or solvent combination (for these evaluations – cyclopentane) and a high-boiling-but-volatile Carbon-14 la-beled compound (in a ratio of approximately 60,000:1). Figure 1 shows the appli-cation of test solution. For example, the amount of tridecane-C14 radiochemical

per single test (< 0.06 µCi) corresponds to approximately 6 E 14 (6 x 1014) mole-cules which equates to one nanomole, the equivalent of approximately one mo-lecular layer over one square centimeter. Metered air or nitrogen gas is permitted to flow across the surface and between the surface and a Geiger Müller detector positioned directly above the surface. The evaporation of the low boiling solvent and then the radiochemical is observed as a function of time by recording the de-tected emissions per second arising from the radiochemical molecules remaining on, or retained by, the surface – the vapor-phase, already-evaporated molecules having been swept out from under the detector by the metered gas (see Figure 2).

Figure 1. Application of test solution. Figure 2. Measurement of emissions.


Each test takes less than 3 minutes and the amount of radiochemical employed is EXEMPT from U S Nuclear Regulatory Commission and/or ‘Agreement State’ licensing regulations due to the very low level of C-14 involved.

For the measurement of microorganic residues, the MESERAN method may be used:

(1) Directly on a flat or concave surface and any microorganic residue thereon which is chemically compatible with the particular radiochemical employed, or

(2) Indirectly using an extracting solvent followed by depositing and evaporating an aliquot amount onto a “clean” reference surface. Subsequent deposition and evaporation of the radiochemical solution permits measurement of the amount of deposited residue by comparing the results with previously ob-tained standards similarly deposited from volumetric dilutions.

For non-polar and/or hydrocarbon type residues, tridecane-C14 in cyclopentane (designated BK) is employed. For more polar residues, tetrabromoethane-C14 in cyclopentane (designated AK) is used. In order to provide a high number of de-tected emissions for the minimal amount of radiochemical deposited, the tridecane-C14 has a specific activity of approximately 57 µCi/µmole (one carbon atom of tridecane is essentially pure C14 isotope) while the tetrabromoethane has

both carbon atoms labeled (approximately 114 µCi/µmole). Approximately 200 ng of radiochemical are deposited in each test with similar levels of radioactivity.

The MESERAN method assumes that the particular radiochemical employed is chemically compatible with the residue, that the test solution droplet covers all of the residue, and that the test solution solvent substantially dissolves the residue within the time period of the solvent evaporation. Attention to the avoidance of

inadvertent contamination and the maintenance of reasonably constant tempera-

ture and pressure are required for optimal reproducibility from test to test.

2.1. Mechanism of the MESERAN technique for quantifying organic residues

[1-3, 11-14]

When a homogeneous chemical is permitted to evaporate, the classical mecha-nism of the process (normally measured by monitoring the already evaporated portion) follows first order kinetics, i.e., the plot of log concentration vs. time is a straight line. This mechanism applies to pure materials as well as to solutions of chemicals in which the components are chemically compatible and in which the second component is non-volatile under the conditions of the process. In the pres-ence of the second component, the rate of evaporation is slowed.

In the MESERAN technology, however, the amount of radiochemical retained by the surface as a function of time is measured by counting the emissions arising from the radiochemical molecules remaining on the surface. In this discussion, the temperature and pressure are assumed constant and the concentration of already evaporated molecules in the adjacent gaseous phase approaches zero due to the flowing air or nitrogen referred to above. The molecular weight of each evaporat-ing molecule and the intermolecular forces among the near-neighbor molecules


are thus the primary factors in determining the tendency of each molecule to re-main in solution or conversely to escape from the liquid portion of the air/liquid (or semisolid) interface. In the MESERAN technology, which employs only a monolayer equivalent of the radiochemical, the observed rate of evaporation is thus a function of the residual concentration of the non-evaporated molecules of the Carbon-14 radiochemical. Figure 3 illustrates the typical evaporation of the radiochemical solution from a clean surface. The A-B line represents the evapora-tion of the low boiling solvent (e.g., cyclopentane). The rationale for the initial in-crease in counts/second is that the C-14 soft beta emissions are partially absorbed by the solvent molecules. B represents the point at which substantially all of the low boiling solvent has evaporated and the maximal amount of residual radiation reaches the GM detector. The B-C line represents the evaporation of the radio-chemical from the surface under the conditions of the test. C represents a level where the GM detector can no longer adequately differentiate the residual radia-tion from background.

A solution of the high-boiling-but-volatile tridecane-C14 in higher boiling hy-drocarbons (i.e., contamination) follows a similar but slower path than does the evaporation of the labeled tridecane itself since the non-volatile “residue” mole-cules occupy increasing portions of the liquid (or semi-solid) interface. The rate at which the solvent evaporates is slowed somewhat and the rate at which the radio-chemical evaporates is slowed considerably with the observed rate of evaporation being a function of the amount of residue on the surface. The observed rate of evaporation of the radiochemical (the slope expressed as a positive integer) thus is an inverse measure of the amount of non-evaporating residue. The lower the slope, the more the residue and vice versa. Figure 4 illustrates typical evapora-tions of the radiochemical solution with increasing amounts of residue. ABC is repeated from Figure 3 and illustrates a typical evaporation of the radiochemical solution with no interactions from residue (i.e., a clean substrate). A*B*C* illus-

Figure 3. Typical evaporation of radiochemical Figure 4. Typical evaporations of radiochemical solution from a clean surface. solution with increasing amounts of residue.


trates a typical evaporation of the radiochemical solution with some contamina-tion present. A**B**C** illustrates a typical evaporation of the radiochemical so-lution with a larger amount of contamination present.

2.2. Methods of analyzing MESERAN data [1-14]

There are two general methods for analyzing the MESERAN data: (1) Total Counts (total area under each curve based on counts minus background) and (2) Slope of the evaporation of radiochemical (the post-peak portion of the curve). Based upon raw data minus background, Figure 5 illustrates three typical experi-mentally derived curves of natural logarithm (Ln or ln) counts per second minus background vs. time in seconds. Figure 5 is similar to Figure 4 except raw data from actual tests are shown.

In Figure 5, the upper curve represents a high level of organic residue, the mid-dle curve represents a medium level of organic residue, and the lower curve repre-sents a low level of organic residue. The scatter, particularly at the lower values, is due to the inherent randomness of radiation (the Poisson distribution in which the square root of each count total is the best estimate of one standard deviation).

Figure 5. Plot of raw data showing low, medium, and high levels of organic contamination.


In an effort to reduce the effect of the randomness of radiation, the data (ln (counts – background)) are “smoothed” from six seconds through 115 seconds (for 120 second length tests) and replotted. The smoothing is accomplished by summing the ln (counts – background) at the sixth second with the previous five seconds and the subsequent five seconds and dividing this number by 10. A divi-sor of 10 is used instead of 11 because it is statistically sound to take the number of items being smoothed and subtract one from it because a degree of freedom is lost. This process is carried out through the 115th second and the subsequent data are replotted as smoothed (ln (counts – background)) versus time. Figure 6 repre-sents the same data as in Figure 5 except that the data in Figure 6 are logarithmi-cally smoothed to increase the reliability of the individual points. The plotted smoothed curve is then analyzed via linear regression to determine the slope of the post-peak line (down to near background) which best fits the data representing the evaporative process. The determined slope is multiplied by –10,000 to convert it to a positive integer; this becomes the reported MESERAN slope value with units of smoothed (ln (counts – background))/sec x (–10,0000).

Figure 6. Logarithmic plot of smoothed data showing low, medium, and high levels of organic con-tamination.


The slope method is more sensitive, especially at low levels of contamination. For measuring microorganic residues, the total counts method of analysis (i.e., the area under each overall curve based on actual counts) is valid from somewhat less

than 200 ng to approximately 100,000 ng (100 µg). Higher total counts are indica-tive of higher organic residue levels and vice versa. The slope method of data analysis, normally based on statistically smoothed data and based on the log count vs. time relationship, increases the sensitivity of the lower limit markedly (to less than 1 ng) since the total counts method (the total area under each curve) ap-proaches statistical insignificance somewhat below 200 ng. Expressed as a posi-tive integer, the higher (or steeper) the slope, the cleaner the surface and con-versely, the lower (or more flat) the slope, the higher the residue.

Both the total counts and slope methods of analysis can be used qualitatively or quantitatively. The total counts method has been used for approximately 30 years at KCP by testing a surface with the MESERAN Analyzer and comparing the re-sults to those obtained from known clean standards for that particular surface. The total counts of the clean standard are subtracted from the total counts obtained on the surface being tested to give a net total counts representing the contamination amount. This result can be compared to previously performed calibrations of con-tamination to obtain a quantitative result for the contamination amount. Similarly, the slope method can be used to compare the slope obtained on the surface being tested to the slope obtained from known clean standards for that particular sur-face. The slope can also be compared to previously performed calibrations of con-tamination to obtain a quantitative result for the contamination amount.

In many cases, quantitative data are not needed. For instance, if one is perform-ing process control work to determine if the cleaning process is performing as de-signed, quantitative data on the actual amount of contamination may not be neces-sary. Often times, as long as the parts being cleaned are less than a certain level of contamination, they are clean enough. Therefore, one only has to establish the MESERAN total counts or MESERAN slopes that correspond to that level of contamination and relate the tests as being in compliance or not. KCP has used this technique for years to control cleanliness and compare the abilities of differ-ent cleaners and cleaning processes to remove various contaminants. Net total count values were established that corresponded to electrical failures and catas-trophic adhesion failures. As long as the MESERAN net total counts were below these levels, no cleaning related failures occurred [4-10].

In recent years, KCP has been incorporating the use of the slope technique to give more quantifiable data for lower amounts of contamination. Calibrations of various contaminants have been performed by KCP to develop calibration curves for these contaminants on substrates of interest. MESERAN slope results obtained can now be compared to the calibration curves to determine quantitative amounts of contamination detected [11-14].

The volumetric dilution process for making calibration solutions is shown in Figure 7.


Figure 7. Calibration solutions formulations.

Volumetric dilutions were used to make contamination solutions for depositing known amounts of the organic contaminant on reference substrates (e.g., alumi-num panels, stainless steel disks, glass cones, etc.). These calibrations were per-formed in the following manner. A master calibration solution was prepared in a 10-mL volumetric flask by dissolving 100 mg of the organic contaminant in 10 mL of solvent (e.g., cyclopentane, methylene chloride, or hexane that has been double distilled in an all-glass still with no grease in the joints – NVOR of these solvents are approximately 10 ppb). The master calibration solution was thor-oughly mixed and 1 mL of this solution was placed in another 10-mL volumetric flask. The second volumetric flask was then diluted with the double distilled sol-vent until the solution level was at 10 mL and this solution was thoroughly mixed.

Subsequent dilutions were carried out in a similar fashion. Ten microliters (µL) of each calibration solution were deposited on the precleaned substrates and allowed to evaporate. This resulted in the following amounts of contamination on the sub-

strates: 1 ng, 10 ng, 100 ng, 1 µg, 10 µg, and 100 µg. Some intermediate levels

were obtained by depositing 3 µL and 5 µL of the calibration solutions. The substrates that were contaminated were then tested using the MESERAN

Analyzer to develop a calibration curve for the contaminant. For example, calibra-tion curves for Dioctyl Phthalate (DOP) using radiochemical test solution AK (tetrabromoethane-C14 in cyclopentane) on aluminum panels are shown in Fig-ures 8 and 9. Figure 8 shows the calibration curve for DOP using the total counts method of analysis. Figure 9 shows the calibration curve for DOP using the slope method of analysis. As can be seen from examining Figure 8, the total counts method of analysis loses its ability to differentiate contamination amounts (i.e., loses its statistical significance) below a few hundred nanograms of contamina-tion (approximately a monolayer). However, the slope method of analysis shown in Figure 9 is able to differentiate contamination amounts down to 1 ng. In gen-eral practice, total counts can be used to quantify contamination amounts greater

than a monolayer (a few hundred nanograms) up towards the 100 µg range. The slope method can be used to quantify contamination levels well below the monolayer (down to a nanogram) as well as up to approximately 100 µg.


Figure 8. Calibration curve for DOP on aluminum panels using MESERAN total counts.

Figure 9. Calibration curve for DOP on aluminum panels using MESERAN low variance slopes.


Whenever possible it is advantageous to develop calibration curves for the con-taminants of interest. KCP has developed calibration curves for numerous con-taminants typically encountered in KCP operations such as oils, greases, mold re-leases, solder fluxes, resins, etc. However, since one does not always know all of the contaminants that may be present on a part, KCP developed a general calibra-tion curve to use for unknown samples based upon hydrocarbon residues. Thus far, most hydrocarbon residues tested have similar calibration curves for the vari-ous amounts of residue.

3. EXPERIMENTAL

3.1. Purpose

KCP conducted a cleanliness evaluation to determine the NVOR amounts on aluminum and stainless steel panels which were machined using KCP machining fluids and cleaning processes chosen for production of hardware for a particular customer. This section of the paper describes how KCP used recent advances in MESERAN technology to determine the NVOR amounts on four stainless steel panels and four aluminum panels (31 in2 each, excluding edges) by extracting the panels with methylene chloride and quantifying the extracted residues in mg/ft2. The virgin methylene chloride solvent was also evaluated so that its contribution could be subtracted from the solvent extracts. The customer specifically requested

that the results be reported in mg/ft2 as opposed to µg/cm2, therefore that is how the results are reported in this paper. It is common practice in the Aerospace in-dustry (as well as other industries) to report contamination amounts on large sur-faces in mg/ft2. The conversion factors for these units are 1 mg/ft2 is equivalent to 1.0764 µg/cm2 or 1 µg/cm2 is equivalent to 0.929 mg/ft2.

3.2. Sample details

For the NVOR evaluations, four samples each of the aluminum and stainless steel (10 cm x 10 cm x 0.7 cm) were machined at KCP using particular machining flu-ids and associated machining methods. The four KCP machining fluids evaluated were a hydrocarbon blend (mixture of 70% Pennex N 47 and 30% Hangsterfer’s Hard Cut # 511) and three aqueous-based coolants (Cimtech 200, Trimsol, and Cimstar 3700). The suppliers for these materials are: Pennex N 47 – Exxon Com-pany, Houston, TX; Hangsterfer’s Hard Cut #511 – Hangsterfer’s Laboratories, Mantua, NJ; Cimtech 200 – Cincinnati Milacron Marketing, Cincinnati, OH; Trim Sol – Master Chemical Corporation, Perrysburg, OH; and Cimstar 3700 – Cincin-nati Milacron Marketing, Cincinnati, OH. Two of the stainless steel samples were improperly labeled; therefore, the contaminant for these two panels is not known for sure. They were either contaminated with the hydrocarbon blend or Cimtech 200 and are described as such in subsequent portions of this paper (including sev-eral tables). All of the stainless steel samples were passivated by the KCP plating


group. This evaluation was conducted on these eight small samples to allow cleanliness verifications to be performed at KCP prior to cleaning large samples (25.4 cm x 25.4 cm x 1.3 cm) that would be sent to the customer for cleanliness verification.

3.3. Cleaning process

The aluminum and stainless steel panels tested in this evaluation were cleaned us-ing the following steps: (1) ultrasonic cleaned in Dirl-Lum 603 (30 g per liter con-

centration) for 5 minutes at 140°F (60°C), (2) rinsed in flowing DI water for 15–30 seconds, (3) DI water rinsed in ultrasonic cascade rinse station with 3 tanks (30

seconds in each tank) at 110–115°F (43.3–46.1°C), (4) blown dry with filtered

nitrogen, and (5) baked for 30 minutes minimum at 220°F (104.4°C )in a HEPA filtered convection oven with nitrogen flowing into the oven. The panels were then packaged in nylon bags and heat sealed.

Dirl-Lum 603, supplied by Blue Wave Ultrasonics, Davenport IA, is a pow-dered alkaline cleaner. It contains sodium metasilicate, sodium carbonate, sodium tripolyphosphate, dodecyl benzene sulfonate, polyethoxyolated phenol, and nonyl phenol.

3.4. Customer cleanliness requirements and associated problems

The customer has cleanliness level requirements for this hardware which can be ex-tremely difficult to measure. The desired cleanliness of the hardware is <0.1 mg/ft2. Cleanliness measurements performed by the customer for these parts are typically carried out using a gravimetric NVR procedure. This procedure requires that the part being measured is rinsed with a known volume of a “clean” solvent (methylene chloride) to extract contamination from the part. The solvent and extracted residues are caught in a clean receptacle and evaporated in a precleaned and preweighed dish. After all of the solvent extract has evaporated, the dish is reweighed to obtain a weight of the dish plus the residue. The dish weight is subtracted from the dish plus residue weight to determine the level of the contamination extracted from the part. Similar evaluations are performed on the virgin solvent to determine the resi-due in the solvent itself. This residue amount is then subtracted from the extracted residue amount to give a final result for the residue extracted.

Gravimetric analysis can be a difficult technique to use consistently when measuring low levels of contamination (<1 mg) because many factors can affect these small weight measurements. Customer cleanliness criterion of <0.1 mg/ft2 further complicates matters because error is prone to being introduced during sample collection and sample processing which can be significant when trying to accurately measure to <0.1 mg/ft2. Every piece of laboratory equipment (such as glassware, weighing trays, funnels, etc.) that comes in contact with the solvent will contribute a small but variable amount of NVR to the solvent. The magnitude of this contribution is not constant due to fluctuations in contact time, surface temperature, and other variables. Although these variables are controlled as well


as possible, they still will contribute some level of error to the reported results, and at the low levels of contamination being discussed, this error may be signifi-cant. In addition, gravimetric analysis is typically performed after rinsing a 1 ft2 (144 in2) sample area while the samples used in this evaluation were significantly smaller (31 in2).

3.5. KCP proposed method of evaluating NVR levels

Due to the problems discussed above, KCP proposed the following: (1) Extract the contamination from the panels with methylene chloride just as would be done in the gravimetric procedure, (2) use the newly developed MicroSolventEvapora-tor to evaporate 250-µL aliquots of the methylene chloride extract onto clean ref-erence substrates, (3) test the evaporated residues using the MESERAN Analyzer to quantify the NVOR levels extracted from the samples, and (4) perform gravim-etric analysis on the remaining extract to determine if there was enough residue to be weighed on a balance.

It should be noted that gravimetric NVR methods will weigh all contamination (organic, inorganic, and particles such as metals) whereas the MESERAN method detects only NVOR (inorganic contamination and metal particles are not de-tected). The real thrust of these evaluations is to quantify the organic contamina-tion and it is estimated that the majority of the contamination being extracted from the parts is organic residue. Whereas, the contribution of inorganic and metal particles is thought to be small.

3.6. KCP machining fluid information

The ingredients for the machining fluids (as obtained from their Material Safety Data Sheets) are as follows:

Hydrocarbon Blend

Pennex N 47 – petroleum distillates (~ 78%) and proprietary additives (~ 22%). Hangsterfer’s Hard Cut #511 – petroleum oil, chlorinated paraffin, and triglyc-erides.

Cimtech 200

Ethanolamine (10% max), caprylic acid (10% max), triethanolamine (10% max), isononanoic acid (10% max), and balance water.

Trim Sol

Petroleum oil (30–40%), petroleum sulfonate (20–30%), chlorinated alkene polymer (20–30%), nonionic surfactant (1–10%), aromatic alcohol (1–10%), pro-pylene glycol ether (1–10%), propylene glycol (<1%), substituted indole (<1%), blue-green dye (<1%), silicone defoamer (<1%), and balance water.

Cimstar 3700

Mineral Oil (10% max), diethanolamine (10% max), triethanolamine, (10% max), aminomethylpropanol (10% max), and balance water.


3.7. Calibration of hydrocarbon blend

Based upon the volatility evaluation and petroleum oils being the most likely con-taminant from the machining fluids being evaluated, it was decided to develop a calibration curve using MESERAN Analysis for the hydrocarbon blend as a refer-ence for future cleanliness evaluations.

A master calibration solution was prepared in a 10-mL volumetric flask by dis-solving 100 mg of the hydrocarbon blend in 10 mL of hexane (NVR<1 mg/L). Volumetric dilutions were used to make contamination solutions for depositing known amounts of the hydrocarbon blend on stainless steel disks. The master calibration solution was thoroughly mixed and 1 mL of this solution was placed in another 10-mL volumetric flask. The second volumetric flask was then diluted with hexane until the solution level was at 10 mL and this solution was thor-

oughly mixed. Subsequent dilutions were carried out in a similar fashion. Ten µL of each calibration solution were deposited on the precleaned stainless steel disks and allowed to evaporate. This resulted in the following amounts of contamina-

tion on the stainless steel disks: 1 ng, 10 ng, 100 ng, 1 µg, 10 µg, and 100 µg.

Some intermediate levels were obtained by depositing 3 µL and 5 µL of the cali-bration solutions. The volumetric dilution process for making calibration solu-tions was illustrated previously in Figure 7.

3.8. Overview of extractions and NVR measurement processes

One-hundred mL of methylene chloride (Optima Grade, stated residue after evaporation – 1 ppm) were used to extract both sides of the cleaned aluminum and passivated and cleaned stainless steel panels. The panels were 10 cm x 10 cm x 0.7 cm for a total of 200 cm2 (31 in2) surface area (discounting the edges). The samples were placed (one at a time) in a large cleaned stainless steel funnel. One-hundred mL of the methylene chloride were measured and poured into a cleaned Teflon squeeze bottle. Each side of the panel was rinsed with the solvent to ex-tract contamination. The methylene chloride extract drained from the funnel into precleaned glass bottles which were then capped. The caps had Teflon inserts and were screwed onto the bottles.

The funnel was ultrasonically cleaned in aqueous Dirl-Lum 603, ultrasonically rinsed in cascading DI water, blown dry with nitrogen, dried in a HEPA filtered oven, and rinsed with virgin methylene chloride after each extraction before being used for the next sample. Prior to being used, all glassware (bottles, graduated cylinders, etc.) used in the extraction process was also precleaned using the above process.

The methylene chloride extracts were thoroughly mixed using an ultrasonic cleaner and vigorous shaking of the bottle prior to taking aliquots for analysis.

Using the MicroSolventEvaporator and clean microsyringes, 250 µL of the ex-tracts were deposited onto cleaned stainless steel reference substrates. The me-thylene chloride evaporation process using the MicroSolventEvaporator takes 10–


20 minutes per 250 µL sample when performed at 30°C. The evaporated residue was then tested with the MESERAN Analyzer. A minimum of 5 replicates were run for each condition tested. The data were compared to calibrations performed using the KCP hydrocarbon blend cutting oil (mixture of 70% Pennex N 47 and 30% Hangsterfer’s Hard Cut # 511) deposited on the same stainless steel sub-strates. Interpolations between the calibration points were performed to determine how much residue was in the extract. This amount was then multiplied by the ap-propriate factor to determine the quantity of residue in the entire amount of ex-tract that was obtained from each panel. The amount of extract obtained from each sample was measured so this calculation could be performed. Recovery of 70–80% of the 100-mL of methylene chloride solvent was typical. The NVR of the methylene chloride was analyzed in a similar fashion to generate a baseline value for the solvent. Ultimately, the NVR of the methylene chloride was sub-tracted to determine the amount of residue actually extracted from the sample panels. This result was then converted to an amount per square foot for compari-son with the customer’s specification.

3.9. KCP gravimetric analysis

Only a small portion of the solvent extracts (generally about 4–5 mL out of the approximate 80 mL obtained after the extraction) was used in the evaluation con-ducted with the MicroSolventEvaporator and MESERAN Analyzer. Therefore, the remainder of each solvent extract (generally 70–80 mL) was sent to KCP’s Analytical Sciences Laboratory for complete evaporation in an attempt to quan-tify the resulting residue gravimetrically. A Mettler AE163 balance was used to make the weight measurements. The calibration label on the balance indicates the performance specification is +/– 0.005% of reading + 0.1 mg. The weights were measured to the nearest 0.01 mg. There is definitely greater error associated with weighing these small amounts. KCP Analytical Sciences personnel indicated that the residue values determined gravimetrically were at best good only to approxi-mately +/– 0.03 mg.

The virgin methylene chloride and the methylene chloride extracts were proc-essed as follows:

1. Weighing trays were baked in an oven for 1 hour @ 105°C.

2. The weighing trays were taken from the oven and placed in a desiccator and allowed to cool.

3. The weighing trays were placed on an analytical balance and massed to the nearest 0.01 mg.

4. The methylene chloride and methylene chloride extracts were poured into in-dividual weighing trays and allowed to evaporate. This process continued un-til all of the solvent evaporated for each sample.

5. The weighing trays were placed in an oven for 1 hour @ 105°C.


6. The weighing trays were removed from the oven and placed in a desiccator to cool.

7. The weighing trays were placed on an analytical balance and massed to the nearest 0.01 mg.

8. The NVR of the methylene chloride was calculated by subtracting the result of step 3 from the result of step 7.

Two samples of virgin methylene chloride were evaluated using this procedure. One sample was 100 mL of methylene chloride taken straight from the original bottle, measured in a clean graduated cylinder, poured into a precleaned glass bot-tle, and then sealed with caps containing Teflon inserts. The second methylene chloride NVR sample was obtained using the following procedure: (1) 100 mL of methylene chloride from the original bottle was measured in a clean graduated cylinder, (2) the methylene chloride was poured into a precleaned Teflon squeeze bottle, (3) the precleaned stainless steel funnel used to hold the samples that were extracted was rinsed with 100 mL of methylene chloride, (4) the methylene chlo-ride drained through the funnel and was captured in a precleaned glass bottle, and (5) the glass bottle was sealed with caps containing Teflon inserts. This process captured 89 mL of methylene chloride for analysis.

After aliquots from the extracts were analyzed using the MicroSolventEvapora-tor and MESERAN Analyzer, the remainder of each of the extracts was poured into clean graduated cylinders to measure the amount of solvent left for gravimet-ric analysis. The extracts were then poured back into their respective bottles and sent to be evaporated and measured gravimetrically.

3.10. Gravimetric NVR evaluations of large panels by the customer

Based upon the MESERAN results obtained for the small panels (31 in2), KCP was convinced that they could clean the large panels to acceptable levels for the customer. Therefore, large panels were manufactured with the same machining flu-ids, cleaned, and sent to the customer for evaluation. The customer evaluated the samples using their extraction and gravimetric NVR procedure described below.

The analytical procedure involves three steps: 1) rinsing the surface to be tested with solvent and collecting the rinse solvent; 2) concentrating the solvent to near dryness by evaporation with a clean gas; and 3) weighing the dry residue to de-termine the NVR. In the first step, the organic material is rinsed from the metal surface with methylene chloride, which must be completely captured in the sam-ple container. The concentration stage consists of evaporating the solvent first by bubbling with clean gas (helium or nitrogen) and then by blow-down after trans-ferring the residual solvent with the NVR into progressively smaller containers to minimize the container surface area and potential loss of NVR. All transfer steps must be quantitative and will require small aliquots of methylene chloride to rinse the containers. The final step is to transfer the remaining solvent containing the NVR into a tared weighing boat and weigh to a constant weight after all of the solvent has evaporated.



4.1. KCP machining fluid volatility

A quick evaluation was made to determine the volatility of the machining fluids. The fluids were weighed out in a weighing dish, allowed to air dry at room tem-perature for 20 hours, and then reweighed. The results are shown in Table 1. As would be expected from looking at their ingredients, the hydrocarbon blend is the least volatile and the aqueous machining fluids are the most volatile. Therefore, theoretically, there should be larger quantities of the hydrocarbon blend to be cleaned off the samples than the aqueous cutting oils. This should make it much easier to clean the aqueous machining fluids to acceptable levels since their resi-due amount is relatively low even if they are not removed.

4.2. Calibration results for hydrocarbon blend

The stainless steel disks that were contaminated with various amounts of the hydro-carbon blend were tested using the MESERAN Analyzer to develop a calibration curve. Example plots of the MESERAN data for these calibrations at various con-tamination levels are shown in Figure 10. The low variance slope calibration curve for the hydrocarbon blend on these stainless steel disks is shown in Figure 11.

4.3. MESERAN analysis of virgin methylene chloride and methylene chloride

extracts

The newly developed MicroSolventEvaporator (a prototype designed and devel-oped by The MESERAN Company) was used to deposit and evaporate the me-thylene chloride onto precleaned stainless steel disks. This system was used to evaporate 250 µL of the methylene chloride onto the stainless steel surfaces in se-

quential small quantities (~ 5–10 µL increments) to concentrate the residue in a small area so that it could be tested with the MESERAN Analyzer. The evapora-

tions were carried out at 30°C to slightly speed up the process. The MicroSol-ventEvaporator allowed the evaporations to be carried out in a reproducible fash-ion to help reduce the error normally associated with manual evaporations (i.e., someone using a syringe and depositing multiple small quantities of solvent in the

Table 1.

Volatility of KCP machining fluids

Cutting Oil Initial Weight (g) Weight After 20 Hour Air Drying (g)

% Volatile

Hydrocarbon Blend 0.11322 0.11292 0.26

Cimtech 200 0.10790 0.00290 97.31

Trim Sol 0.17500 0.01100 93.71

Cimstar 3700 0.11665 0.00400 96.57


Figure 10. Plot of various amounts of hydrocarbon blend (70% Pennex N 47 and Hangsterfer’s Hard Cut # 511) on stainless steel disks using smoothed Ln (counts – background). (MESERAN tests carried out using radiochemical BK).

Figure 11. MESERAN low variance slope calibration of hydrocarbon blend (70% Pennex N 47 & 30% Hangsterfer’s Hard Cut # 511) on stainless steel disks with 0.015 inch/RPM spiral grooves.


same area until a large volume of solvent has been evaporated). In addition, these stainless steel disks have grooves machined in them which act as a self-centering mechanism so that the solvent evaporates in a confined space and does not mi-grate elsewhere.

After the methylene chloride was evaporated onto the stainless steel disks with the MicroSolventEvaporator, the disks were tested with the MESERAN Analyzer using radiochemical BK (best suited for detecting nonpolar residues). Table 2

shows the average low variance MESERAN slope results for the 250-µL deposi-tions from the methylene chloride extracts.

The results were compared to those obtained previously from the calibration using known amounts of the KCP hydrocarbon blend machining fluid (mixture of 70% Pennex N 47 and 30% Hangsterfer’s Hard Cut # 511) on these same sub-strates, which were shown previously in Figure 11. The average low variance slope obtained for the methylene chloride blanks was 2412. The MESERAN slope value has units of smoothed (ln (counts – background))/sec x (–10,0000); however, the units are generally not given when the data are shown. A linear-log

Table 2.

MESERAN low variance slope results for 250-µL depositions from methylene chloride extractions

Description of Sample Tested

Average MESERAN Low Variance Slope (smoothed (ln (counts – background))/sec x (–10,0000))

Standard Deviation

Coefficient of Variation (%)

Methylene Chloride (Straight from Bottle)

2412

111.71

4.63

Hydrocarbon Blend on Aluminum

2392

186.34

7.79

Hydrocarbon Blend or Cimtech 200 on Stainless Steel

2319

196.57

8.48

Hydrocarbon Blend or Cimtech 200 on Stainless Steel

2324

176.14

7.58

Cimtech 200 on Aluminum

2049

45.35

2.21

Trim Sol on Aluminum

2205

351.69

15.95

Trim Sol on Stainless Steel

2399

55.25

2.30

Cimstar 3700 on Stainless Steel

2208

46.33

2.10

Cimstar 3700 on Aluminum

2385

141.25

5.92


interpolation using the values from the calibration in Figure 11 was performed to calculate the contamination level. The slope of 2412 corresponds to a contamina-

tion level of 3 ng (or 0.000003 mg) for the 250 µL of methylene chloride evapo-rated on the stainless steel disks. Therefore, the NVR of the 100-mL methylene

chloride blanks would be 1.2 µg (or 0.0012 mg) as calculated below:

(3 ng/250 µL) x (1000 µL/1 mL) x (100 mL) x (1 µg/1000 ng) = 1.2 µg (or 0.0012 mg)

The contamination results obtained from the MESERAN Analysis for the me-thylene chloride extracts from the panels tested are shown in Table 3. The average low variance slopes obtained for the samples were converted into equivalent con-

tamination amounts (in nanograms) for the 250 µL of extract evaporated by per-forming a linear-log interpolation using the values from the calibration curve in Figure 11. Then the amount of contamination in all of the extract collected for each sample was calculated (shown in micrograms) using similar equations as shown previously for the methylene chloride blanks. Next, the methylene chloride NVR was subtracted from the contamination amount in all of the extract collected to determine the NVR amount extracted from each sample (this result is shown as

µg/31 in2). Finally, this contamination was converted to an amount per square foot to compare to the customer requirement of <0.1 mg/ft2. The results of all samples tested passed the requirement of <0.1 mg/ft2 and most of them were sig-nificantly lower.

4.4. KCP gravimetric analysis results

The results obtained in the gravimetric analysis on the methylene chloride blanks and the methylene chloride extracts are shown in Table 4. Since the extract amount analyzed gravimetrically was less than the original extract amount (due to aliquots being taken for MESERAN Analysis), the gravimetrically determined NVRs were factored up to account for the lost contamination. In all cases except two, the contamination determined gravimetrically was less than the gravimetri-cally determined methylene chloride NVR, which by definition makes the ex-tracted residue less than the customer limit of <0.1 mg/ft2. One of the hydrocar-bon blend or Cimtech 200 samples on stainless steel had a positive result after the methylene chloride NVR was subtracted. After converting the contamination from mg/31 in2 to mg/ft2, the contamination determined is higher than the cus-tomer limit of <0.1 mg/ft2. The Cimstar 3700 on stainless steel sample also had a positive result after the methylene chloride NVR was subtracted; however, the re-sulting contamination was still less than the customer limit of <0.1 mg/ft2. The er-ror associated with measuring these small amounts of contamination makes all of these gravimetric results doubtful and it is recommended that they be dismissed.

4.5. Results of gravimetric NVR evaluations of large panels by the customer

The gravimetric NVR results obtained by the customer on the large aluminum and stainless steel panels (100 in2) machined, cleaned, and packaged at KCP are

A n

ew m

etho

d u

sing

ME

SE

RA

N tech

niq

ue

69

M.G

. Ben

kovich

and

J.L. A

nd

erson

7

0


shown in Table 5. The customer determined that 100 mL of virgin methylene chloride had a NVR of 0.004 mg. They did not subtract this amount from the final results of the extracts when they reported their results. The results of every panel evaluated by the customer indicated that the contamination level was less than their requirement of <0.1 mg/ft2.

Table 5.

Gravimetric analysis of methylene chloride extracts from 100 in2 aluminum and stainless steel pan-els performed by the customer

Description of Sample Tested

NVR Level (mg/ft2)

Hydrocarbon Blend on Aluminum 0.03

Hydrocarbon Blend on Stainless Steel 0.04

Cimtech 200 on Aluminum 0.07

Cimtech 200 on Stainless Steel 0.04

Trimsol on Aluminum 0.04

Trimsol on Stainless Steel 0.02

Cimstar 3700 on Aluminum 0.02

Cimstar 3700 on Stainless Steel 0.06

5. CONCLUSIONS

Analysis of the solvent extracts with the MicroSolventEvaporator and MESERAN Analyzer indicated that all of the samples tested passed the customer requirement of <0.1 mg/ft2 with most of them significantly lower. Thus, the results indicated that the cleaning process used sufficiently cleaned the cutting oils from the panels tested.

In short, the amounts of contamination being determined are too small to be ac-curately measured gravimetrically for the small samples extracted (31 in2) be-cause the error associated with these measurements is extremely high at these lev-els. These results provide further proof that for gravimetric NVR values to be accurate, it is best to extract at least a one square foot area so there is plenty of contamination in the extract to measure properly. Therefore, when quantifying ex-tremely small residues using solvent extracts from small parts, the only viable method currently available is the MicroSolventEvaporator/MESERAN Analysis method.

The large panels (100 in2) evaluated gravimetrically by the customer indicated that all of the panels were cleaned to acceptable levels (i.e., with residual con-tamination less than the customer requirement of <0.1 mg/ft2). Even though pan-els the customer evaluated were significantly larger than ones evaluated by KCP


and different methods were used for determining the NVR levels, the results ob-tained were remarkably consistent.

The NVR data determined by KCP with the MicroSolventEvaporator and ME-SERAN technique and the NVR data determined gravimetrically by the customer follow a similar pattern. Tables 3 and 5 indicate that the results of both test meth-ods exhibited the same cleanliness level relationship between the aluminum and stainless steel panels for each specific contaminant evaluated. Hydrocarbon blend-contaminated samples had more contamination after cleaning on the stainless steel panels than on the aluminum panels. Cimtech 200-contaminated samples had more contamination after cleaning on the aluminum panels than on the stainless steel panels. Trimsol-contaminated samples had more contamination after cleaning on the aluminum panels than on the stainless steel panels. Cimstar 3700-contaminated samples had more contamination after cleaning on the stainless steel panels than on the aluminum panels.

Acknowledgements

The work of Garth Christoff, Tom Hand, and Ed Fuller was instrumental in the performance of these evaluations.

REFERENCES

1. J.L. Anderson, “Quantitative Detection of Surface Contaminants,” Journal of the American As-sociation of Contamination Control, II (6),9 (1963).

2. J.L. Anderson et al., “Measurement and Evaluation of Surfaces and Surface Phenomena by Evaporative Rate Analyses,” Journal of Paint Technology, 40, No. 523, 320-327 (August 1968).

3. J.L. Anderson, “Evaporative Rate Analysis: Its First Decade”, in: Characterization of Metal and

Polymer Surfaces, L.H. Lee (Ed.), Vol 2, pp. 409-427, Academic Press, New York (1977). This paper summarizes all known references prior to 1975.

4. L.C. Jackson, “Solubility Parameters and Evaporative Rate Analyses in Organic Residue Char-acterization” (Topical Report), UNCLASSIFIED, Bendix-Kansas City Division: BDX-613-1099, March 1974. (Available from NTIS)

5. L.C. Jackson, “Contaminant Removal Based on Solubility Parameter and Evaporative Rate Analysis Technologies” (Topical Report), UNCLASSIFIED, Bendix-Kansas City Division: BDX-613-1128, August 1974. (Available from NTIS)

6. L.C. Jackson, “How to Select a Substrate Cleaning Solvent,” Adhesives Age, 22-31 (December 1974).

7. L.C. Jackson, “Solvent Cleaning Process Efficiency,” Adhesives Age, 31-34 (July 1976). 8. L.C. Jackson, “Removal of Silicone Grease and Oil Contaminants,” Adhesives Age, 29-32

(April 1977). 9. L.C. Jackson, “Contaminant Cleaning for Critical Electrical Assembly Areas” (Final Report),

UNCLASSIFIED, Bendix-Kansas City Division: BDX-613-1695, February 1978. (Available from NTIS)

10. M.G. Benkovich, “Solvent Substitution for Electronic Products,” International Journal of Envi-ronmentally Conscious Manufacturing, 1, No. 1, 27-32 (1992).

11. M.G. Benkovich and J.L. Anderson, “Measurement of Organic Residues on Surfaces to a Low Fraction of a Monolayer,” Precision Cleaning, 16-28 (May 1996). This paper includes many of the more current references to MESERAN technology.


12. M.G. Benkovich and J.L. Anderson, “Quantification of MicroOrganic Residues to Low Nanogram Levels,” Precision Cleaning ’96 Proceedings, 115-122 (1996).

13. J.L. Anderson, R.F. Russell and M.G. Benkovich, “Quantitative Measurement of Extremely Low Levels of Non-Volatile Residues (NVR) on Surfaces and in Liquids,” Precision Cleaning

’97 Proceedings, 96-108 (1997). 14. J.L. Anderson, R.F. Russell and M.G. Benkovich, “Solvent NVR: A Problem and a Solution,”

CleanTech ’98 Proceedings, 331-340 (1998). 15. MESERAN Analyzer Literature, ERA Systems Inc., The MESERAN Company, Chattanooga,

TN.


Ed. K.L. Mittal

© VSP 2003

Methods for pharmaceutical cleaning validations

HERBERT J. KAISER∗

STERIS Corporation, P.O. Box 147, St. Louis, MO 63166

Abstract—Cleaning validations are very difficult to perform. They can be made easier if an appro-

priate method for analyzing the samples is used. The method used should be based on the previously

established residue limits of both the active and cleaning agent. There are many choices of analyti-

cal techniques that can potentially be used. This article describes various analytical techniques that

are available for use, particularly for cleaning agent residues.

Keywords: Pharmaceutical cleaning; cleaning validation; total organic carbon; high performance

liquid chromatography; Fourier transform infrared spectroscopy; ultraviolet detection; evaporative

light scattering detection; mass spectrometry; photoelectron emission; residue; method validation.

1. INTRODUCTION

In the pharmaceutical industry cleaning is an important component of the manu-

facturing process. Regulatory agencies both in Europe and North America require

that the cleaning process be validated [1]. It must be clearly shown that product

residues are removed to an acceptable level. In addition, not only do the manufac-

turers have to show that they have cleaned their product to an acceptable level but

they have to demonstrate that they have removed the cleaning agent also to an ac-

ceptable level. Cleaning validations require tools to measure the residue levels left

on surfaces. These tools are required to be validated. This poses a challenge to

analytical chemists in that they must first choose the appropriate method to meas-

ure the residues and then they must validate that method. If the method cannot be

validated it cannot be used to measure residue levels. There are many types of

analytical techniques that can be used for these analyses. The methods can be

specific or non-specific. The chemist can also utilize complementary techniques

such as total organic carbon (TOC) determination or high performance liquid

chromatography (HPLC). The analytical chemist must examine the pros and cons

of each technique in order to choose the appropriate technique. Again, whichever

method or methods are chosen they must be validatable.


H.J. Kaiser 76

2. CHOOSING THE ANALYTE

The analytical chemist must determine what is being measured. Is the residue a

drug active, formulation component, or the cleaning agent? Is the residue organic

or inorganic? Is the residue water soluble or water insoluble? Is the residue par-

ticulate, microbial or is it an endotoxin? The answers to these questions will nar-

row the choice of analytical techniques available. For example, if the residue is a

drug active, HPLC or gas chromatography (GC) could be utilized. If the residue is

a formulation component or a cleaning agent, TOC determination may be the

method of choice.

An important piece of information that the analytical chemist will require is the

target residue limit. The target residue limit should always be established prior to

the selection of the analytical technique. Once the residue limit has been estab-

lished the analytical technique can be chosen and a method developed that can de-

termine residue levels below the acceptance limit [2, 3].

3. SPECIFIC VERSUS NON-SPECIFIC TECHNIQUES

Once the type of residue has been established the analytical chemist can chose be-

tween either specific or non-specific techniques. A specific technique detects a

unique compound. A non-specific technique detects any compound that produces

a certain response. Examples of specific techniques include HPLC, Fourier trans-

form infrared spectroscopy (FTIR), ion chromatography (IC), atomic absorption

(AA), inductively coupled plasma atomic emission (ICP), capillary electrophore-

sis (CE) and various protein methods. Examples of non-specific techniques in-

clude TOC determination, pH determination, acid/base titrations and conductivity.

4. TECHNIQUES FOR SAMPLING

In addition to choosing an appropriate technique a method of sampling must also

be established. There are two acceptable forms of sampling. One form is a direct

surface swab. This is where a swab is utilized to rub a known surface area to re-

cover the residue. The other acceptable form of sampling is rinse water sampling.

In this technique the equipment is rinsed with a known volume of water and then

the water is analyzed for its residue content. The test must be a direct measure of

potential contaminants and not just compendial tests for water (e.g., a U.S. Phar-

macopeial test). In other words, if rinse water is used as the sample a method must

be utilized that can measure the contaminants coming off the surface, not just the

quality of the rinse water. The analysis cannot just be based on standard water

quality parameters. Also, if rinse water is utilized it must be demonstrated that the

rinse water does indeed remove the residues from the surface of the materials be-

ing cleaned. A questionable sampling technique is the use of placebos. This is a

Methods for pharmaceutical cleaning validations 77

technique where, after the manufacture of a normal drug product, the product, less

the drug active, is manufactured again. A sample is then taken of this material and

analyzed for the drug active in the previous batch. This is a questionable sampling

technique because it assumes that the residue is distributed uniformly throughout

the placebo. Another problem with placebo sampling is that the residue may be so

dilute in the placebo that the analytical technique may not be able to detect it.

This is an acceptable technique only if it is used along with swab or rinse water

data. Therefore, since swabbing or rinse water analysis must be done in conjunc-

tion with placebo sampling, this technique is rarely utilized. Placebo sampling is

important when validating the cleaning of filling equipment or small pieces of

equipment where direct rinse water measurements are not practical.

5. ANALYZING THE SAMPLE

5.1. Interferences

If the detection technique is non-specific a good strategy is to first identify possi-

ble sources of interference. These interferences could be either positive or nega-

tive. For example, in TOC determinations, if the person performing the sample

collection accidentally coughs onto the surface being analyzed this would cause a

positive interference. Since TOC measures a non-specific property, the residue

amount would be calculated as if all of the measured property were due to that

residue. For example, if the target residue was aspirin and the technique utilized

was TOC determination, the TOC measured would be calculated as if all of the

TOC came from the aspirin even though there may be excipients and/or deter-

gents present. This provides an upper limit value and is a worst-case situation.

The identification of possible interferences is important to both specific and

non-specific methods. In the case of specific methods there should not be any

interferences if the method was properly developed. It is important that the

method is able to detect the analyte after exposure to the cleaning environment.

Studies should be conducted to demonstrate that the analyte does not change after

exposure to alkaline or acidic cleaners. This can be simply done by exposing the

drug active to a dilution of the cleaning agent at the temperature that will be

utilized during the cleaning process. This sample can then be analyzed to

determine if the analyte can still be detected and quantitated. If this is not done

the analyte may not be detected even though it is present in the sample. If the

cleaning environment does change the nature of the analyte a new method would

be required or modifications would have to be made to the existing method. When

utilizing a non-specific method to measure a non-specific property, any

compound that displays that property and is introduced into the sample will

interfere. Possible sources of interferences could be from the environment,

sampling technique, and/or people.

H.J. Kaiser 78

5.2. High performance liquid chromatography

The most common analytical technique available to pharmaceutical analytical

laboratories is the HPLC. There are many different detectors that are used in

HPLC analyses. These include ultraviolet (UV), fluorescence, electrochemical, re-

fractive index, conductivity, evaporative light scattering and many others. The

most common detection technique is ultraviolet. However, evaporative light scat-

tering detection (ELSD) is a very good method for the detection of surfactants.

5.2.1. UV detection

There are many advantages in using a UV detector. A large number of com-

pounds can be detected because they contain chromophores. UV detection also al-

lows for diode array spectral possibilities. This feature is very useful when exam-

ining peaks for coeluting compounds. UV detectors are easy to use since they

require no special reagents. In addition, temperature is not an important consid-

eration since it typically does not affect molar absorptivities. This means that no

special heating or cooling are required for the detector.

There are disadvantages in using UV detection. UV detectors are not universal.

This means that not all compounds have chromophores and, therefore, are ex-

tremely difficult, if not impossible, to detect using UV detection. Limits of detec-

tion can be higher than other detector types due to background interferences.

These interferences are especially predominant when low wavelengths are used in

combination with a gradient elution scheme.

HPLC analysis along with UV detection is most commonly used for the analy-

ses of drug actives. However, this technique is also sometimes utilized for the de-

tection of surfactants that may have been left behind by the cleaning agent. There

are many problems associated with surfactant analyses using chromatographic

methods. High levels of detection result unless special sample preparations are

taken. There are many examples in the literature where surfactants are analyzed in

environmental samples, such as river water. However, this is typically accom-

plished by pre-concentrating the sample up to 1000-fold. Due to the higher level

of detection, the quantitation levels are also higher. Again, this problem can be

avoided by pre-concentrating the sample. Also, the peaks present in a chroma-

togram of a surfactant must be summed to obtain an appropriate level of quantita-

tion. This is due to the fact that a surfactant material is not a single compound.

Surfactants are a mixture of compounds containing various chain lengths and per-

haps various degrees of ethoxylation. Another potential problem is deciding

which surfactant to analyze. A cleaning agent may contain more than one surfac-

tant. It may contain an anionic surfactant along with an amphoteric surfactant.

The cleaner may contain a combination of anionic surfactants with nonionic sur-

factants or any number of combinations. Figure 1 shows the chromatogram of a

typical ethoxylated surfactant. The different peaks represent different degrees of

ethoxylation. All of the peaks must be summed in order to generate the calibration

curve. The smaller peaks will disappear into the baseline as the dilution increases.

This will produce less than desirable results.


Figure 1. Chromatogram of an ethoxylated surfactant using UV detection.

A significant disadvantage to the use of UV detection is that many surfactants

used in formulated cleaners do not contain chromophores. ELSD can be used to

overcome this.

5.2.2. Evaporative light scattering detection

An ELSD is a mass detector and does not depend on the presence of chromopho-

res. The eluent from a typical HPLC column enters the detector through a nebu-

lizer and is carried along by a gas stream through a heated column. The mobile

phase is evaporated in the column leaving small particles of the analyte. The

small particles are then passed through a detector containing a laser. The light

from the laser is scattered by the small particles. The detector then measures the

degree of scattering.

There are many advantages associated with the use of an ELSD. This type of

detector is simple (in theory), versatile and rugged in use. Their manufacturers

describe these detectors as “universal detectors”. This is because, in theory, they

should be able to detect any non-volatile compound since they solely rely on

mass. Since they do rely on mass all compounds should produce a similar re-

sponse as opposed to UV detection where the extinction coefficient is important.

There should also not be any baseline drifts due to mobile phase effects. This is

due to the fact that the mobile phase is evaporated prior to entering the detector.

There are disadvantages to the use of ELSD. There is a very limited choice of

buffer salts that can be utilized. This is because the buffer has to be volatile. If the

buffer is not volatile it will remain in the system and flow through the detector.

The detector will then detect the buffer salts instead of the analyte. Another dis-

H.J. Kaiser 80

advantage is that the ELSD requires special care and cleaning of the nebulizer.

The nebulizer must produce consistent particle sizes so that an accurate detection

can occur. Therefore, careful cleaning and maintenance must be performed.

5.3. Capillary electrophoresis

Capillary electrophoresis (CE) is a very good method to use for cleaning valida-

tions. CE can be used in different modes. These modes include capillary zone

electrophoresis, micellar electrokinetic chromatography, and isoelectric focusing.

CE can be used for both drug actives and detergent residues. The ability of CE to

pre-concentrate samples makes it very useful in the detection and quantitation of

surfactants. UV detection is the most common detection technique used in CE.

Inverse detection can be readily used with CE. Inverse detection is a procedure

where a strongly absorbing compound is utilized in the buffer solutions and pro-

duces a strong signal in the detector. When the analyte, which may not be

strongly absorbing, passes through the detection window a “negative” peak is

produced. Software accompanying CE units can typically invert this “negative”

peak producing a normal looking “chromatogram”.

5.4. Ion chromatography

Ion chromatography (IC) is another useful technique that can be utilized in resi-

due analyses. This is because IC can be used to quantitate both organics and inor-

ganics. The inorganics may include sulfates, phosphates, chlorides, sodium, po-

tassium and other inorganics that may be present in a cleaning agent. IC can also

be used to determine organic acids such as citric or glycolic acids and even sur-

factants. There are several modes of detection that can be used in IC but the most

common are UV and conductivity. IC can analyze multiple analytes in the same

sample in many cases. Figure 2 is a typical chromatogram showing sodium and

potassium ions that are present in a cleaning agent. Low levels of detection can be

achieved especially with the conductivity detector.

5.5. Total organic carbon determination

TOC determination is the most commonly used analytical technique in cleaning

validations. Instrument manufacturers use several different ways to determine the

TOC in a sample. In general, the sample is oxidized and the carbon dioxide pro-

duced is measured. The TOC detection can occur in a variety of ways including

infrared and conductivity. There are several advantages to using TOC determina-

tion as the analytical technique for cleaning validations. TOC is a non-specific

measurement technique for carbon. Therefore, it can be used to quantitate any

residue that contains carbon. Low limits of quantitation/detection can be achieved

with TOC. Manufacturers of TOC units claim a detection limit of <1 ppb. How-

ever, quantitation and detection limits are primarily based on the quality of the

water used in the analysis. There is typically a high recovery from samples in


TOC analyses. There are also minimal chemical interferences. TOC determination

is easy to carry out and involves a minimal amount of method development.

Typically, samples are simply diluted and analyzed. No separation procedures

need to be developed unlike HPLC. As with other instrumentations, there are

automated sampling accessories that can be used. These accessories include on-

line monitors. On-line monitoring should only be used for monitoring water qual-

ity. On-line monitoring should not be used for monitoring a cleaning validation

study. The instrumentation for TOC determination is also very cost effective.

There are two major disadvantages to TOC determination. The first is that the

sample must be water-soluble. TOC determination cannot be used as the analyti-

cal technique if the residue cannot be dissolved in water. It should be noted that

solubility in the TOC determination sense means in the part per billion range

rather than in the part per million range. If a compound is reported to be water in-

soluble it does not automatically mean that it is not soluble enough to be analyzed

by TOC determination. The second disadvantage is that samples cannot be pre-

pared in organic solvents. If the samples were prepared in organic solvents the

TOC of the solvents and not of the residue would be determined.

TOC determination is an excellent complimentary technique to HPLC. TOC

can be used to monitor water-soluble ingredients in a formulation and HPLC can

be used to monitor the water insoluble ingredients. An example of this is a formu-

lated ibuprofen. The ibuprofen itself is not soluble even at TOC determination

levels, but HPLC can be utilized to monitor the ibuprofen. The formulation com-

ponents are soluble in water and TOC determination can be used to monitor their

contribution to the residue.

Figure 2. Chromatogram showing sodium and potassium ions in a cleaning agent using conductiv-

ity detection.

H.J. Kaiser 82

5.6. Fourier transform infrared spectroscopy

FTIR spectroscopy can be used to conduct cleaning evaluations. Fiber optics for

mid-IR use is not developed far enough where it could be used in a cleaning vali-

dation situation. Therefore, FTIR spectroscopy is used in the lab to evaluate the

rinsibility of various residues including detergents. The use of a grazing angle ac-

cessory allows for the semi-quantitation of residues on coupons. This can be used

in a method where a coupon is spiked with a known amount of residue and al-

lowed to dry. The coupons are then subjected to a simulated rinse procedure and

the concentration on the surface of the coupon is monitored using FTIR spectros-

copy. The data generated from these studies can be used in establishing rinsing

procedures and for choosing appropriate cleaners for the manufacturing environ-

ment.

5.7. Photoelectron emission

There is a very sensitive direct surface measurement technique that is based on

photoelectron emission [4]. This relatively portable instrument has a probe that is

placed onto the surface to be analyzed. The instrument uses a UV source to bom-

bard the surface that excites the atoms on the surface. Radiation is emitted when

the atoms return to the ground state. The instrument detects the radiation. Clean

surfaces generate more radiation than do soiled surfaces. This is an extremely

sensitive technique and any change to the surface will affect the amount of radia-

tion detected. The normal wear and tear that occurs to equipment surfaces in the

manufacturing environment will cause changes to occur which will affect the

readings.

5.8. Mass spectrometry

A “portable” mass spectrometer has been developed at Lawrence Livermore Na-

tional Laboratories [5]. It consists of a probe connected via a cable to a vacuum

pump, the electronics, and controller. The tip of the probe forms a tight seal

against the surface of the substrate and a vacuum is generated. A heating element

heats the surface and vaporizes the residues that may be present on the surface.

This instrument not only can quantitate what is on a surface but it can also be

used to identify the residues. The only drawback with this instrument is that it re-

quires a relatively flat surface to be effective.

6. METHOD VALIDATION

It is time now to validate the methods once the residues have been identified, lim-

its set and method(s) chosen. Worldwide regulatory agencies require that these

methods be validated [6]. In order to be validated the method(s) must be shown to

have linearity, precision, accuracy and robustness. Ideally, there would be at least

two different analysts performing analyses on two different instruments using two


different columns (if it is a chromatographic method) on two different days with

two different sets of prepared standards and samples. It is important to note that

not only does the analytical method need to be validated but the sampling method

needs to be validated too. For example, if a swabbing technique is utilized it must

be demonstrated that the technique recovers the residues being analyzed. The per-

cent recovery whether 90% or 50% must be established.

Figure 3 shows one type of swabbing technique. It really does not matter which

technique is utilized as long as it is repeated the same way each time and the en-

tire sampling area is covered. If an analyst swabs in one manner and another ana-

lyst swabs in another manner different results will be obtained. It is very impor-

tant to train whoever is doing the swabbing in the appropriate technique. A

diagram of the swabbing technique should be placed in the procedure document.

As was mentioned earlier, it is important to perform recovery studies and vali-

date the sampling technique. Both swabs and rinse sampling methods must be

validated. A typical procedure would be to spike a coupon with a known amount

of residue and remove the residue with either a swab or a simulated rinse proce-

dure. If a swabbing technique was used the residue from the swab should be de-

sorbed in a suitable solvent. If a TOC analysis were being performed the solvent

should be low TOC water and if an HPLC method was being preformed the sol-

vent should most likely be the mobile phase. These studies are done below the

residue acceptance limits that had been previously established. This is done to en-

sure that the method can at least recover the actual residue limit amount.

What is an acceptable recovery? A recovery of >80% is very good. A recovery

of >50% may be sufficient but a recovery of <50% is questionable. Additional

work should be done to improve the <50% recovery. Factors that could cause low

recoveries are adhesion of the residue to the swab or the volatility of the residue.

Figure 3. Example of one method for swabbing a surface to recover residues for subsequent analysis.

H.J. Kaiser 84

Applying a known amount of the test residue directly onto a swab can be used to

perform swab recoveries. The swab is then extracted and the extract analyzed us-

ing the appropriate method. The extraction solvent spiked with the same amount

should be used as the control. If the swab is retaining the residue, an alternate

swab material should be sought. The recovery factor obtained should either be in-

cluded in the analytical calculations or the acceptance limit calculations but not in

both. If the recovery factor is included in both, too tight a limit will have been set.

7. SUMMARY

In summary, when choosing an analytical technique to determine residues, resi-

due limits should be established first. This limit would then be used to calculate

the limit in the analyzed sample. The next step would be to select, develop and

validate a method to determine residue levels at and below the acceptance limit.

Interferences should also be investigated and addressed. Developing and validat-

ing a sampling procedure should be the next step. The validation of the sampling

procedure should include recovery studies. There are many different techniques

that can be used to analyze residues. In practice, the simplest technique that can

quantitate at the residue limit and below should be utilized. Less complicated

techniques normally have less potential problems associated with them.

REFERENCES

1. “Points to Consider for Cleaning Validation”, Technical Report No. 29, Parenteral Drug Asso-

ciation (1998).

2. H.J. Kaiser and M. Minowitz, J. Validation Technol., 7(3), 226-236 (2001).

3. H.J. Kaiser, J.F. Tirey and D.A. LeBlanc, J. Validation Technol., 6(1), 424-436 (1999).

4. M.K. Chawla, Precision Cleaning, 8(6), 36, 38 (2000).

5. M. Meltzer, C. Koester and C. Steffani, “Criteria Evaluation for Cleanliness Phase 0”, Lawrence

Livermore National Laboratory, UCRL-CR-133199 (1999).

6. USP 23, United States Pharmacopeial Convention, Rockville, MD, 1982-1984 (1995).


Ed. K.L. Mittal

© VSP 2003

Influence of cleaning on the surface of model glasses and

their sensitivity to organic contamination

W. BIRCH,∗ S. MECHKEN and A. CARRÉ Corning SA, Fontainebleau Research Center, 7 bis, Avenue de Valvins, 77210 Avon, France

Abstract—Both wet and dry cleaning processes impact the chemical composition of glass surfaces. The alteration of glass surface chemistry is probed by wettability measurements as a function of pH. Wettability measurements are performed using the two-liquid method. This method consists of de-positing sessile water drops on a substrate immersed in liquid octane. The measured contact angle allows an estimation of the number of hydroxyl groups exposed at the glass/water interface. Meas-urements with varying pH of the water drops probe the surface charge at the glass surface and indi-cate the isoelectric points of predominant functional groups. Variations in glass surface composition correlate with sensitivity to organic contaminant adsorption. Three typical species of glass were probed: a silica surface, the surface of sodalime glass, and the surface of a Corning aluminoborosili-cate glass. The three glass species show different contamination behavior, depending on wet or dry cleaning of the glass. The surface compositions, as probed by wettability measurements, correlate to grazing incidence XPS data, indicating the alteration of surface chemical sites by leaching during the wet cleaning process. It appears that sensitivity to organic contamination is controlled by the surface chemistry and the hydration of the exposed substrate.

Keywords: Glass; cleaning; contamination; surface composition; wettability; surface charge; hydroxyl group density; self-assembled monolayers; isoelectric point.

1. INTRODUCTION

The cleaning of glass surfaces [1] precedes a variety of coating processes, ranging from silane-based sol-gel coatings, to adhesion promoting films, to conductive coatings such as ITO (indium tin oxide). These cleaning processes are generally designed to remove particles from the glass surface and to render it uniformly wetting. For most applications, it is desirable that this wettability be associated with the exposure of silanol groups at the glass surface. These exposed chemical functions form part of the glass substrate and they may be used to covalently graft silane-based coatings to the glass substrate [2].

∗To whom all correspondence should be addressed. Phone: +33-1-64 69 74 14,

Fax: +33-1-64 69 74 54, E-mail: [email protected]

W. Birch et al. 86

Rendering a glass surface hydrophilic may be achieved by a variety of tech-niques [1]. These may be broadly classed into two categories: solution cleaning and dry cleaning. Solution cleaning is generally easy to set up for a small number of samples. It often has the disadvantage of using toxic products and generating hazardous waste, while being effective for removing particles from the glass sur-face [1]. Dry cleaning requires a larger initial investment. It is efficient for clean-ing large numbers of samples while generating a minor amount of waste products. An advantage of dry cleaning techniques is that they generate a clean, dry surface. This should be compared to the blow-drying that follows solution cleaning, where streaks and surface inhomogeneities are almost unavoidable.

This article will focus on the impact on glass surfaces of both solution cleaning and dry cleaning. Glass surfaces are not inert and their exposure to aqueous solu-tions generally causes a modification of their surface chemical composition. Dry cleaning processes have little or no impact on the chemical composition of the glass surface. Solution cleaning uses chromic acid, while dry cleaning uses pyro-lysis.

Three types of glass were analyzed. They were chosen to span a representative range of silica-based glasses, from sodalime glass to aluminoborosilicate glass, to silica. Sodalime glass is the most common and cheaply available glass substrate. It is the material used for almost all microscope slides, as well as being the stan-dard “window glass”. Corning code 1737 aluminoborosilicate glass provides a substrate with increased chemical durability, improved optical properties, and higher thermal resistance. It is primarily used in flat panel display applications and has been used to make microscope slide substrates. Silica provides the refer-ence substrate, bearing a surface composed of silicon oxide chemical sites. These three glass species are of increasing cost and are commercialized for different ap-plications. Sodalime glass is a low cost substrate with a low melting point and minimal chemical durability in aqueous solution. Aluminoborosilicate glass is a more expensive glass, providing an increased thermal resistance and improved chemical durability. Its high melting point and the absence of sodium in its com-position make it suitable for use in making flat panel displays. Finally, silica pro-vides a substrate with good chemical resistance and a very high melting point. Its high cost limits its use to critical applications, where it is uniquely suited to meet the required specifications.

The surface properties of glass substrates were directly probed by wettability measurements. Since clean glass surfaces are completely wetted by water, water drops deposited in air spread to give a contact angle of less than five degrees, which is below the measuring threshold of our contact angle goniometer. Thus, the glass surface wettability was measured under octane, where sessile water drops give a finite contact angle value. The wettability as a function of pH reveals features arising from the chemical functions exposed at the glass surface.

Influence of cleaning on the surface of model glasses 87

2. EXPERIMENTAL

2.1. Substrates and cleaning processes

The glass materials examined were: sodalime (SDL) glass, in the form of micro-scope slides from Erie Scientific (Portsmouth, NH, USA); aluminoborosilicate (ABS) glass, in the form of Corning code 1737F glass (Corning, NY, USA); and silica surfaces, in the form of the native oxide layer of a silicon wafer. Silicon wa-fers were supplied by Unisil Corporation (Santa Clara, CA, USA) and were p-doped or n-doped four-inch wafers from standard production. All the glass sur-faces were believed to have a surface roughness of 3 Å rms or less. This has been measured for silicon wafers [3, 4] and for float glass. The sodalime and alumino-borosilicate glasses bear their as-formed (“fire-polished”) surfaces, with a low surface roughness. Unfrosted microscope slides were cut from sodalime glass, made by a draw process that generates two identical surfaces. The sodalime glass made with a draw process contains no iron. This is visible by looking at the glass slide from the side, where it appears white. A lower cost process for manufactur-ing these slides is by floating the glass over a bath of molten tin, whereby the glass is referred to as “float glass”. The float glass has two faces: one in contact with a reducing atmosphere, referred to as the “air” side, and one in contact with the molten tin, referred to as the “tin” side, where a small amount of tin diffuses into the glass surface. This is different from the glass drawing process that gener-ates identical glass faces. The presence of trace amounts of iron in the sodalime glass composition gives it a mild green color. This can be seen when observing a microscope slide from the side. In one experiment, frosted glass slides, made from float glass, were also used. The frosted side of these slides corresponds to the air side of the float glass. Primary chemical compositions of the three glass species used are given in Table 1.

The chromic acid cleaning solution (CHR) is also known as Chromerge. It con-sists of a strong oxidizing agent in concentrated sulfuric acid. The solution was prepared by completely dissolving 20 grams of potassium dichromate (K2Cr2O7, from Prolabo, Fontenay sous Bois, France) in 90 grams of water. To this, 900 grams of concentrated sulfuric acid (Normapur grade, from Prolabo) was slowly added. It was then allowed to cool to room temperature. This solution could be used for cleaning while it remained yellow or brown. It was discarded when the color changed to green, indicating a change in the oxidation state of the chro-mium. As the concentrated sulfuric acid is hygroscopic, so the solution was stored in a closed container, avoiding dilution by ambient moisture. Dipping in chromic acid solution at room temperature for 20 minutes cleaned the glass samples. They were then rinsed in pure water and dipped for 20 minutes in a 1:1 concentrated hydrochloric acid/water solution (designed to remove chromium ions from the surface). They were then finally rinsed in pure water and dried under a stream of pure nitrogen.

W. Birch et al. 88

Table 1.

Composition of glass substrates. Principal components of the glass composition are given in weight percentage. For the silica substrate, this is the composition of the native oxide layer at the surface. The bulk glass composition is given for the other two glasses. The first column recalls the process used to make the glass

Substrate Glass composition (by weight)

Silica substrate:

– 1-2 nm thick silica layer

– polished silicon wafer with native oxide

100% silica

Corning code 1737F glass (ABS)

– aluminoborosilicate

– made by fusion draw process

70% SiO2

10% Al2O3

10% B2O3

Sodalime glass (SDL)

– microscope slides

– made by draw process

70% SiO2

13% Na2O

10% CaO

The water used in these experiments was purified using an Elga UHP unit. It

filters deionized (18 MΩ.cm resistivity) water through an activated carbon filter and a UV lamp. This water was used to rinse samples, as well as in the prepara-tion of acidic or basic solutions to measure surface wettability.

The dry cleaning process used was pyrolysis (PYR). The glass substrates were heated to 500°C, according to the following cycle: a rise in temperature from room temperature to 500°C over four hours, followed by five hours at 500°C and a return to room temperature over five hours. The samples were placed in a glass rack and the rack was enclosed in a glass container, closed with a folded alumi-num foil sheet. The containers were removed from the oven after cooling to room temperature. Provided the lid was not opened, the glass samples remained clean for about one week. The samples were used within 5 minutes after removing the aluminum foil lid.

The glass samples showed no visible organic contamination on their surfaces before cleaning by CHR or PYR. Neither technique is suitable for cleaning a glass surface bearing macroscopic contamination. At most, the samples had ambient contamination, in the form of a monolayer or sub-monolayer coverage of organic contaminants, expected to have a thickness of order 1 nm or less.

Following cleaning, one glass sample per cleaning batch was verified as being wettable by water. To achieve this, a one-microliter drop of pure water was placed on the sample. This drop was expected to spread. When the sample was tilted, its receding edge was not expected to dewet the glass surface. The glass substrates were used if the sample passed these criteria of glass surface wettability. The sample tested was not used for further experiments.


For surface composition analyses by XPS and monolayer adsorption and depo-sition experiments, PYR was replaced by UV/ozone dry-cleaning (details are pre-sented in Sections 2.3 and 2.4). Since the cleaning of silicon wafers with pyrolysis led to rapid subsequent contamination on exposure to air and the XPS equipment was far from the site where the samples were cleaned, UV/ozone cleaning was used to improve sample cleanliness during transport. The two dry cleaning tech-niques were expected to have a similar impact on the glass surfaces composition, which is measured by XPS to a depth of about 5 nm.

2.2. Contact angle measurements, SiOH group density and dissociation

The wettability of the cleaned glass surfaces was measured under liquid octane. Octane was chosen due to its dispersive interactions with glass being almost equal to those between glass and water. This results in the measured wettability giving information on the dispersive interactions between glass and water, as described below. The clean glass substrate was immersed in liquid octane and its wettability measured with a water drop deposited on the glass surface under octane. The wa-ter drop displaces the liquid octane that was in contact with the clean glass. When the drop ceases to spread, the measured contact angle, θ, may be interpreted as shown in Figure 1.

The force balance between the solid and liquid 1 (glass/octane), liquid 2 and liquid 1 (water/octane), and solid and liquid 2 (glass/water) interfacial tensions can be expressed in the form of Young’s equation:

. (1)

This equation can be reformulated as follows to give an expression for cosθ:

. (2)

Figure 1. Schematic diagram of a sessile water drop under liquid octane. The interfacial tensions

used in equation (1) are indicated and the definition of the contact angle, θ, is illustrated.

γ

γ γ cosθ

newater/octa

rglass/wateneglass/octa −=

rglass/watenewater/octaneglass/octa γ cosθ γ γ +=

W. Birch et al. 90

Octane is a non-polar liquid. Its interaction with the glass substrate is purely dispersive giving rise to the octane/water interfacial tension. Water is a polar liq-uid. Its interactions with the glass substrate may be broken down into two compo-nents: dispersive and non-dispersive. The dispersive (London) interactions are those generated by instantaneous (non-permanent) dipole interactions. The non-dispersive interactions include the permanent dipole interactions (Keesom and Debye interactions), hydrogen bonding and acid-base interactions. The latter two generate a surface charge on the glass substrate. The dissociation of silanol (SiOH) groups at the glass surface may be represented by the following reaction:

The surface charge generated by these acid-base interactions increases the glass/water interfacial energy. The dispersive component of the surface tension of water is almost equal to the surface tension of octane: γWD = 21.8 mJ/m2 [5] and γO = 21.6 mJ/m2 [6], where the subscript “w” stands for water and the subscript “o” stands for octane. Hence, the interaction energy between glass and octane may be considered as being equal to the dispersive interaction between glass and water:

(3)

The non-dispersive interactions for the octane/glass and octane/water interfaces are considered as negligible. Using

, (4)

and

(5)

in equation (1) gives

(6)

which directly relates the non-dispersive interaction of the glass/water interface to the measured contact angle and known constants (γwater = 72.8 mJ/m2). The inter-facial tension between water and octane is given by equation (7) below [7]. It can be measured using a standard procedure, such as a Wilhelmy plate set-up.

. (7)

Finally, equation (6) may be written as:

. (8)

2/1

octane

D

wateroctanewaternewater/octa)γγ(2γγγ −+=

neglass/octaoctaneglassneglass/octa Iγγγ −+=

ND

rglass/wate

D

rglass/watewaterglassrglass/wate IIγγγ −−+=

D

rglass/wateneglass/octa II =

cosθγγ-γI eroctane/watoctanewater

ND

rglass/wate +=

NDglass/waterI (1+cos θ)*51 mJ/m²≈

– +

3 2SiO +H O SiOH + H O .↔


The non-dispersive (dipolar, hydrogen bond, and acid-base) interactions be-tween a clean glass surface and water can give a simple estimate of the hydroxyl group density at the glass surface, providing a measure of the density of SiOH re-active sites at the glass surface. If we assume the non-dispersive interaction be-tween glass and water to be dominated by hydrogen bonding and acid-base inter-actions, we may consider that at the point of zero charge of the glass substrate the non-dispersive interaction energy to be dominated by hydrogen bonding. At this point, the surface energy per unit area can be divided by the energy per hydrogen bond to obtain the number of hydrogen bonds per unit area. Taking the -OH bond energy between water and the glass surface as 24 kJ/mol [8] gives the following expression for the -OH group density, expressed as the number of hydroxyl bonds per square nanometer:

(9)

where θ is the contact angle of sessile water drop under octane at the point of zero charge. The above equations relating wettability measurements to the non-dispersive interaction between the glass surface and water are valid if organic contamination is absent from the glass surface.

The wettability measurements of water drops under liquid octane were per-formed in a cube-shaped glass cell (from Hellma, in Müllheim, Germany). This glass cell was first cleaned with a 2% solution of Hellmanex detergent (Hellma) in pure water for 20 minutes. It was then rinsed with pure water and blow-dried under a flow of pure nitrogen. During rinsing, the cell was verified to be uni-formly wetting. Immediately after cleaning, octane liquid was poured into the cell to a depth of about two centimeters. The cleaned glass substrate was then im-mersed in octane. Following a waiting period, as given below, three two-microliter sessile water drops were deposited on the sample. The contact angles on both sides of the drops were measured using a Ramé-Hart contact angle go-niometer (Mountain Lakes, NJ, USA). These contact angle values were then aver-aged.

The first measurements were made using unpurified octane. The immersion of the substrate in octane was noted and the sessile water drop contact angles were plotted as a function of the elapsed time between substrate immersion in octane and deposition of the water drop. The second set of experiments was performed using purified octane, cleaned by passing over silica and alumina chroma-tographic columns. Octane (purissimum grade from Fluka, Ref. 74821) was passed through a chromatography column filled with silica (silica gel, 60, Ref. 1.09385.1000, from Merck, Darmstadt, Germany), followed by passage through a chromatography column filled with alumina (basic alumina, 60, activity stage I, Ref. 1067.2000, from Merck). The silica and alumina were used as received. The chromatography columns were cleaned with Hellmanex surfactant, as described above for the glass cell. They were rinsed with pure water and dried in a clean oven with aluminum foil covering their ends. The first few milliliters of octane

-OH/nm² 1.3 (1+cos θ)≈

W. Birch et al. 92

passed through the columns were used to rinse the original octane bottle and then discarded. The purified octane was then stored in this rinsed bottle. The reasoning behind this purification of the liquid octane is that the high active surface area of the silica and alumina particles will adsorb polar impurities in the octane. Given the similarity of the surface sites on silica and alumina particles to those found on clean glass surfaces, we expect the impurities adsorbed in the columns to be those that would have adsorbed onto a clean glass surface.

To vary the pH of the water, aqueous solutions of hydrochloric acid or sodium hydroxide were used. These solutions were made using calculated volumes and verified with a pH sensitive paper. In the preliminary experiments, cleaned and dried glass substrates were immersed into octane and the wettability of water drops at different pH values was measured. These experiments showed trends in the wettability as a function of pH. To reduce the error bars on the measured con-tact angle values, the cleaned substrates were first immersed for twenty minutes in an aqueous solution of the same pH as that of the water drops to be measured. This was found to give the same trends in wettability as a function of pH, while decreasing the error bar on the measured contact angle values. The increased ac-curacy of the measurement may be due to the fact that the deposited water drop has insufficient time for it to interact with the cleaned glass surface before settling at the measured contact angle. Further, contact angle hysteresis may prevent the contact angle value from changing if the surface charge of the cleaned glass changes with its exposure the water drop.

The wettability of the cleaned glass substrates was measured after their immer-sion in the same acidic or basic solution as used in the measuring water drops. Af-ter cleaning, each glass substrate was cut into five pieces and each piece placed in a solution of different pH. Following the 20-30 minute soaking time, the glass substrates were rinsed with water and blow-dried with pure nitrogen. Sessile drop contact angles were measured for water drops deposited five minutes after immer-sion. The deposited water drops came from the same solution as the one in which the substrates were immersed. For each glass species, three samples were meas-ured at each pH and three sessile water drops were measured for each sample. The data points used for the graphs in Figures 4 and 5 represent an average of these 18 contact angle values.

2.3. Surface composition

The glass surfaces were examined by normal angle x-ray photoelectron spectros-copy (XPS). This technique provides quantitative information on the atomic com-position of the first 5 nm of the glass surface. The measured percentage of inor-ganic elements in this region provides information on changes in the glass surface composition following CHR or UV/ozone cleaning. UV/ozone cleaning was used to replace pyrolysis due to rapid adsorption of organic contaminants on the pyro-lyzed silicon wafers. Placing the samples 1 cm away from the surface of a 4” x 4” low-pressure mercury lamp (BHK Inc., Claremont, CA, USA) for 30 minutes


achieved the UV/ozone cleaning. The UV lamp emits radiation at the 184.9 and 253.7 nm wavelengths, where the first generates ozone from oxygen while the second combines with ozone to oxidize hydrocarbons [10, 11].

2.4. Monolayer adsorption and deposition

Practical applications for bare glass surfaces may utilize their surface charge or surface silanol sites to adsorb or graft a layer of molecules for a thin film coating. Charge-adsorbed surfactant monolayers were deposited to probe the glass surface charge. The deposition of a grafted self-assembled monolayer of octadecylsilane was used to probe the quality of coatings grafted to the silanol groups on the glass surface.

Monolayers of anionic and cationic surfactants were deposited on the cleaned glass substrates from aqueous solutions. For the anionic surfactant, sodium dode-cyl sulfate (SDS, purissimum grade, from Fluka Chemie AG, Buchs, Switzerland) was used. The deposition was made by pulling the clean glass out of a solution at one half of the critical micellar concentration (cmc) for SDS, corresponding to a 4 mM solution of SDS. No deposition occurs on a negatively charged surface in so-lution. A monolayer of surfactant is deposited onto the glass surface from the thinning film of SDS solution as the sample is slowly pulled vertically from the surfactant solution [3]. The cationic surfactant used was hexadecyltrimethylam-monium bromide (CTAB, purum grade, from Fluka). Its deposition on the nega-tively charged surface occurred in solution, resulting in a substrate that was pulled out “dry” (with no liquid film on its surface) from solution. The solution concen-tration was 0.4 cmc, corresponding to 0.4 mM of CTAB. The glass substrate was left in the surfactant solutions for at least thirty seconds and pulled out vertically at a speed of about one millimeter per second.

The quality of coatings grafted to the glass via their surface silanol groups was tested by depositing a self-assembled monolayer of octadecyltriethoxysilane (OTES). For this experiment, the glass slides were cleaned using the CHR and UV/ozone cleaning processes. UV/ozone cleaning replaced the PYR dry cleaning to avoid rapid contamination of the silicon wafers following PYR cleaning. The OTES molecules were hydrolyzed before deposition for two purposes. By con-verting the ethoxy groups surrounding the silicon atom to hydroxyl groups, hy-drolysis reduces the size of the silicon functional group at the end of the aliphatic hydrocarbon chain. This allows a dense packing of the aliphatic hydrocarbon chains into a structure perpendicular to the substrate, somewhat reminiscent of the hair on a carpet. The second purpose is an increased reactivity of the SiOH groups, as compared with the unhydrolyzed Si-ethoxy groups. This reactivity al-lows for cross-linking of the silanol groups and grafting of the aliphatic molecules to the silanol surface sites on the glass, resulting in a durable grafted coating on the glass surface. To form a densely-packed structure, the grafting reaction should take place after the self-assembly of the aliphatic hydrocarbon chains [12]. This results in a high quality monolayer structure with only a few defects, resulting in

W. Birch et al. 94

an optimal non-wetting behavior towards water. The OTES was pre-hydrolyzed using the recipe given in the paper by Peanasky et al. [13]. A 50-ml volumetric flask was filled three quarters with inhibitor-free THF (Ref. 24,288-8, from Al-drich). To this were added 0.25 g of a 1.31 N aqueous solution of hydrochloric acid. The hydrochloric acid solution was made by dissolving 13.27 g of 36% (specific gravity of 1.19) hydrochloric acid (ref. A 7405201, from Fisher Scien-tific, Chicago, IL, USA) in water, making 100 ml of aqueous solution. The mix-ture was stirred and 0.42 grams of OTES were slowly added (ref. SIO6642.0, from ABCR, Karlsruhe, Germany). The flask was then filled up to the 50-ml mark with THF. This solution was allowed to sit at ambient temperature for four to five hours before use. It could be refrigerated for up to one week at 4°C. To make the OTES coating solution, 18.6 grams of cyclohexane (HPLC grade, Ref. 27,062-8, from Aldrich, purified in the same manner as the octane above) were poured into a pyrolyzed glass container. While stirring, 1.11 grams of pre-hydrolyzed OTES solution were slowly added. Cleaned glass slides were placed in a pyrolyzed glass rack and the rack placed in the glass container for coating. The container was then closed with a sheet of folded aluminum foil taped down to its sides. The glass slides were left to incubate for 24 hours at room temperature. They were then rinsed three times with stabilized THF (Ref. 87368, from Fluka) in a clean glass container. Ultrasonic agitation was applied to each rinse by plac-ing the glass container for five minutes in a Crest ultrasonic bath (68 kHz with modulated frequency, from Crest Ultrasonics, Trenton, NJ, USA). For optimal transmission of ultrasonic vibration energy, the ultrasonic bath contained a 3% aqueous solution of Chemcrest 14 detergent (from Crest Ultrasonics). The deter-gent solution was degassed and heated to 45°C to improve the uniformity and transmission efficiency of ultrasonic vibration energy in the bath. The wettability to water of three samples of each glass species and cleaning procedure was then measured. Each sample was measured using three sessile water drops and one wa-ter drop for the advancing and receding contact angles. The contact angle meas-urements were made as described above, using the Ramé-Hart goniometer. The sessile drop values quoted were averaged over 18 values, corresponding to both sides of three drops on each of three samples. The advancing and receding contact angle values quoted were averaged over three values, corresponding to one drop on each sample.


3.1. Contact angle measurements: surface contamination of cleaned glass in octane

In the first experiments, the contact angles of sessile water drops deposited under unpurified octane were measured. Figure 2 shows the evolution of the contact an-gle of the deposited sessile water drop as a function of the elapsed time between immersion of the cleaned substrate in octane and the water drop deposition. The


contact angle of the sessile water drops did not change after their deposition. The observed increase in contact angle suggests an evolution of the cleaned surface with time over the first five minutes after drop deposition. The increasing contact angle values also suggest a contamination of the substrate. Other possibilities, such as a hydration or dehydration of the cleaned surface were considered. Hydra-tion of the cleaned surface is expected to lead to a decrease in the measured con-tact angle value. To check for dehydration, a drop of water was placed in the oc-tane and the cleaned substrates immersed after 48 hours. The contact angle evolution was identical to that in Figure 2, suggesting that dehydration of the glass surface in octane is not what causes the observed increase in contact angle with immersion time. To verify contamination by exposure to octane, the sub-strates were blow-dried with pure nitrogen after immersion for at least five min-utes in the liquid octane. The wettability of sessile water drops was then measured in air. The measured contact angles are given in Table 2. They range from 32 to 46 degrees, indicating a significant level of organic contamination adsorbed on the glass surfaces. Further, the large error bar of about 8 degrees indicates a non-uniform surface wettability. This behavior is typically found for random contami-nation deposited on a clean surface. The contact angle values measured after ex-posure to octane, when compared with the complete spreading of water drops on the freshly cleaned glass substrates, suggest that the liquid octane contaminates the cleaned glass substrates.

The sessile drops deposited following longer immersion time of the cleaned substrate in octane showed a higher contact angle than those on a freshly im-mersed substrate. In fact, a water drop deposited immediately after immersion of the cleaned glass into octane spread to form a low contact angle, generally of 5°

Figure 2. Water sessile drop contact angles measured after exposure of cleaned glass substrates to unpurified octane. The legend is the same as for Table 2. The quoted contact angles are averages ofsix values for three sessile drops at times of 1, 3, and 5 minutes. The contact angle values quoted for“silica PYR” at less than one minute are for one drop at each data point.

W. Birch et al. 96

or less. The drops, once deposited, did not change their shape, indicating that the contact angles of the sessile drops did not evolve. Young’ s equation invokes the notion of a force equilibrium at the three-phase line, where the two fluid phases, namely octane and water, meet the solid surface. The absence of any increase in the contact angle of the deposited drops, when compared to the increase in contact angle for freshly deposited drops, suggests that an equilibrium contact angle value is not reached for the sessile water drops under octane. The notion of a force bal-ance at the three-phase line is valid. A force balance can be considered as a mini-mum in the energy of a system with respect to a small displacement of the system from this minimum. This concept is valid for systems that are free to move, where the influence of frictional forces is negligible. For contact angles, the concept of a contact angle hysteresis is analogous to a frictional force, preventing the contact angle from reaching a value corresponding to the overall minimum in the free en-ergy of the system. This is the case for water drops deposited under octane, where the observed contact angle is not reached by the drop retracting to form a contact angle equal to that of sessile drops deposited at later times.

Following a five minute exposure to octane, the contact angles of three or more freshly deposited sessile water drops were measured on each of the glass sub-strates. At least three substrates were investigated for each glass species and cleaning procedure. Table 3 gives the averaged contact angle values for each type of glass using CHR or PYR cleaning. For the silica surface, we find a slightly higher contact angle following PYR cleaning than for the CHR cleaned substrate. This is compatible with a loss of about 10% of the surface hydroxyl groups during glass pyrolysis [14]. However, for the sodalime glass, the wettability of the sur-face following PYR cleaning remains high, even when immersed in the liquid oc-tane that has contaminated the other glass substrates. To obtain more information on this behavior, the advancing and receding contact angles were measured for the substrates exposed to liquid octane for more than five minutes. The measured values are shown in Table 4. If these contact angle values are interpreted in terms of surface contamination of the cleaned glass substrates by immersion in the liq-

Table 2.

Water sessile drop contact angles (in degrees) measured on glass substrates exposed to unpurified octane for more than five minutes. The substrates were removed from the octane and blow-dried with pure nitrogen before wettability measurements. The error bar over several sessile drops and several substrates is nominally 8°. The superscript number indicates the number of substrates over which the quoted wettability was measured. Each substrate was measured using between three and eight sessile drops. In the legend, silica refers to the native oxide layer on a silicon wafer, ABS re-fers to Corning code 1737 aluminoborosilicate glass, and SDL refers to sodalime glass. CHR and PYR refer to chromic acid and pyrolysis cleaning, respectively

Cleaning process

silica ABS SDL

CHR 46 1 43 1 40 1

PYR 46 5 35 7 32 2


uid octane, they suggest that the pyrolysed silica is more susceptible to adsorbing contamination than the chromic acid cleaned silica. Further, the pyrolysed so-dalime glass does not strongly adsorb organic contaminants from the octane, given that the water drop spreads easily. The pyrolysed aluminoborosilicate glass shows a behavior intermediate between the silica surface and the sodalime glass, showing a finite advancing contact angle and zero receding contact angle. The ab-sence of a finite receding contact angle value suggests that few or no organic con-taminants remain adsorbed on the glass surface after passage of the water drop. On the other hand, following chromic acid cleaning, the three glass species show similar wettability, suggesting that they have almost the same affinities for ad-sorbing organic contaminants in liquid octane. This suggests that the sensitivity to adsorbing contaminants depends on the composition of the glass surface and the procedure used to clean the glass surface. Considering that PYR cleaned silica surface is more dehydrated than CHR cleaned silica surface, we may attribute the reduction in the affinity for adsorbing organic contaminants to the presence of a layer of water on the glass. PYR cleaned sodalime glass shows the lowest level of adsorbed contaminants. Both the soluble alkaline oxide film and the ambient moisture it adsorbs may enhance the durability of this clean glass surface. It is in-teresting to note that the surface behavior of the glass species is similar to that of the silica following cleaning with chromic acid.

Table 3.

Water sessile drop contact angles (in degrees) measured on cleaned glass substrates under octane, following a five minute exposure to the octane. The error bar over several sessile drops and several substrates is nominally 4°. The legend is the same as for Table 2

Cleaning process

silica ABS SDL

CHR 29 44 26

PYR 59 29 3

Table 4.

Advancing and receding contact angles (in degrees) of water drops under unpurified octane. The substrates were immersed in the unpurified octane for at least five minutes before measuring the contact angle hysteresis of one water drop per substrate. The superscript number indicates the num-ber of substrates over which the quoted wettability was measured. The legend is the same as for Ta-ble 2

Cleaning process

silica ABS SDL

θa 38 5 50 1 35 1 CHR

θr 10 9 15 13 13 3

θa 68 37 spreading PYR

θr 45 0 0

W. Birch et al. 98

Figure 3. Schematic side-view of a drop on an inclined surface, indicating the advancing contact

angle, θa, and the receding contact angle, θr.

3.2. Use of purified octane to avoid surface contamination

To avoid the surface contamination described in the preceding section, subsequent measurements of water drops under octane were made in purified liquid octane, prepared as described in Section 2.2. Following exposure to the purified octane, substrates dried under a stream of pure nitrogen gave low contact angles for de-posited water drops, indicating a low level of contamination from strongly ad-sorbing organic molecules. This suggests that the purified octane did not signifi-cantly contaminate the cleaned glass surfaces, unlike the behavior cited for the unpurified octane in Section 3.1.

The contact angles for sessile water drops measured under purified octane showed a different behavior from those measured under unpurified octane. The contact angles of water under octane for silicon wafers cleaned with chromic acid were in the range 8° to 16° and showed time-dependent increases of only 2 to 6°. This result was obtained again when a freshly cleaned substrate was immersed in purified octane that had been left in the measuring cell for two days, indicating no significant increase of contaminants in the octane over two days. For Corning code 1737F glass, the contact angles of deposited water drops showed a different behavior: the water drops gave a finite contact angle immediately after deposition, later spreading to wet the glass surface within a few minutes. This behavior re-mains unexplained. For CHR cleaned sodalime glass, the contact angle of freshly deposited water drop increased from about 6° to 14° over the first ten minutes fol-lowing immersion of the substrate. The contact angles are for freshly deposited water drops. There is no increase in the contact angle after drop deposition.

For the pyrolysis-cleaned substrates, water completely wets the substrates im-mersed in octane, regardless of their immersion time. The spreading of water drops under octane on pyrolyzed silicon wafers was immediate. The final contact angle was below the measurement capability of the Ramé-Hart goniometer (less than about 4°). For the Corning code 1737F glass, the water drop spreads, leaving only a slightly thicker film. The contact angle is still too small to be measured. Drops of water deposited on sodalime glass under octane spread rapidly, forming dendrimers. A simple classification would rate the affinity to water of the PYR cleaned glass surfaces in the order: sodalime glass > Corning code 1737F glass > silica.


The kinetics of water drop spreading described for water drops on glass under purified octane remains unexplained. The increase in contact angle observed for glass surfaces cleaned with chromic acid is not attributable to contamination, given the high wettability of the cleaned glass following exposure to the octane. The unusual spreading behavior of water drops on acid cleaned Corning code 1737F glass also remains unexplained. Finally, the complete wetting of pyrolyzed glass substrates by water drops under octane suggests a high affinity of these glass surfaces for water. Further, the presence of some components of the glass compo-sition that are soluble in water, such as sodium oxide and calcium oxide (see Ta-ble 1), appears to enhance the spreading of the water drops under octane.

3.3. Surface wettability as a function of pH, surface charge and SiOH group

density

While the contact angle values for sessile water drops at neutral pH provide a first information on the surface energy of the cleaned glass substrates, further informa-tion can be provided by measuring the surface wettability as a function of the pH. These experiments were conducted using purified octane to avoid the substrate contamination described in Section 3.1. The absence of contamination of the cleaned glass substrates and the behavior under octane of sessile water drops at neutral pH is described in Section 3.2.

The pH of water drops deposited on the cleaned glass surfaces was varied in the range from 0 to 13, in order to probe the surface charge variations from acidic and basic surface groups. The surface wettability was measured at pH intervals of 1 unit for acidic pH and two units for alkaline pH values. The standard deviation of the measurements was generally 1° to 2°. About 10% of the data points showed standard deviations ranging from 3° to 5°.

Figure 4 shows the non-dispersive interaction energy of glass substrates cleaned with chromic acid as a function of pH, in the range from 0 to 13. The in-terval of one pH unit for acidic solutions and two units for alkaline solutions was chosen arbitrarily. The contact angle values, ranging from 0 to 21°, have been converted into the interaction energy, INDglass/water, using equation (8). Figure 5 shows data from the same measurements on glass surfaces cleaned by pyrolysis. The two graphs in Figure 5 show the same data. The upper graph was plotted to show differences in the behavior of Corning code 1737 glass and sodalime glass, while the lower graph shows data for all three glasses, with silica showing a sig-nificantly lower non-dispersive interaction energy with water than the other two glass species.

The hydroxyl (SiOH) surface groups on the clean glass may be positively or negatively charged, depending on the solution pH in contact with the surface. The isoelectric point, also known as the point of zero charge, or p.z.c., defines the pH at which the surface densities of positive and negative charges are equal, associ-ated with an equal surface density of negatively charged SiOH groups (SiO–) and positively charged SiOH (SiOH2

+) [9]. The isoelectric point of silica has been

W. Birch et al. 100

measured to be in the range of pH 2 to 3 [9]. Figure 4 shows a minimum in the in-teraction energy near the quoted isoelectric point of silica. While the quoted value for the isoelectric point of silica is close to pH 2, the minimum in Figure 4 is close to pH 3.

At the p.z.c., we may consider the glass surface as being uncharged. If we as-sume the non-dispersive interactions between glass and water to be dominated by hydrogen bonding, the contact angles measured at the p.z.c. provide an estimate of the number of hydroxyl groups per unit area on the cleaned glass. Table 5 gives the measured contact angle values at pH 3 for the cleaned glass surfaces. Table 6 shows the corresponding calculated hydroxyl group densities, converted from the contact angles in Table 5 using equation (9). The values in Table 6 compare fa-vorably to Iler’s quoted value of 2-5 -OH/nm2 (between 2 and 5 hydroxyl groups per square nanometer) for silica surfaces [14]. The measured contact angles range from 10° to 39°, while the corresponding hydroxyl group densities vary only by about 10%, indicating similar density of hydroxyl groups on the glass surfaces for all glass species and cleaning procedures. For comparison, sessile water drops measured under octane on polypropylene or polyethylene, which are not expected to show any hydrogen bonding, give contact angles of 171° [15], yielding a calcu-lated -OH density of 0.003 (or less) per nm2.

Figure 4. Non-dispersive interaction energy between sessile water drops and Chromerge cleanedglass substrates as a function of pH. The water drops were at the pH indicated and each point repre-sents an average of 18 contact angle measurements. The standard deviation of the data points istypically 1-2°, with about 10% of the data points giving a standard deviation of 3-5°.


Figure 5. Non-dispersive interaction energy between sessile water drops and pyrolysis cleaned glass substrates as a function of pH. The upper graph shows the data from sodalime and Corning code 1737 glass, comparable to that from Figure 4. The lower graph shows the full data for the silica sur-face, which shows an extremely different behavior, resulting in far lower interaction energy and higher measured contact angles. Data points are averaged over 18 values with standard deviations as described in the legend for Figure 4.

Table 5.

Water sessile drop contact angles (in degrees) measured on cleaned glass substrates under octane. The measured water drops are aqueous hydrochloric acid at pH 3, corresponding to the point of zero charge of the data in Figures 4 and 5. The legend is the same as for Table 2

Cleaning process

silica ABS SDL

CHR 15 21 16.2

PYR 38.5 19.5 10.2

W. Birch et al. 102

Table 6.

Hydroxyl group density (-OH/nm2) for glass surfaces, calculated from the contact angles given in Table 5. These contact angles were measured on the cleaned glass surfaces, for sessile water drops under octane at the point of zero charge. The legend is the same as for Table 2

Cleaning process

silica ABS SDL

CHR 2.56 2.51 2.55

PYR 2.32 2.53 2.58

In terms of estimated hydroxyl group density on the cleaned glass surfaces, py-

rolyzed silica shows a lower hydroxyl group density than chromic acid cleaned silica. This is compatible with a minor amount of dehydroxylation during the py-rolysis cycle [14]. Unexpectedly, pyrolyzed sodalime glass appears to have a higher surface SiOH concentration than silica. This may be an artifact caused by the soluble alkali salts (see Table 1) present on the surface of the pyrolyzed so-dalime glass. These salts may increase the wettability to water when compared to the same glass cleaned by chromic acid, where alkaline oxides are leached during the cleaning procedure.

Returning to the graphs in Figures 4 and 5, the data show two minima in the non-dispersive interaction energy between glass and water. The first and more prominent minimum, found at pH 3, may be associated with the isoelectric point of the SiOH groups. Recalling the acid-base reaction responsible for generating the surface charge from the SiOH groups:

– +

3 2SiO +H O SiOH + H O,↔

the hydrogen ion (H3O+) concentration at this pH is sufficient to drive the equa-

tion towards the right. At higher pH values, corresponding to a lower hydrogen ion concentration, the glass surface is negatively charged from its surface SiOH groups. At lower pH values, the surface hydroxyl groups become positively charged, according to the following equation:

+ +

3 2 2SiOH+H O SiOH + H O.↔

Thus, only at the isoelectric point does the glass surface bear no overall charge from its SiOH groups. A second, shallower, minimum in the interaction energy may be seen at pH 9. This may correspond to the sodium ions (Na+), from the so-dium hydroxide used to increase the pH, being present in sufficient quantity to neutralize the SiO– ions from the surface hydroxyl groups.

For all the glass samples, the measured contact angles fall sharply above pH 10 and are zero above pH 12. This is probably due to the glass surface being de-graded by dissolving in the alkaline solution, leading to its being fully wettable by water.


The non-dispersive interaction energy between glass and water as a function of pH is expected to reflect the surface charge generated by the exposed chemical functions on the clean glass surface. The variations in surface charge, generated by the exposed SiOH and aluminum oxide groups, is expected to give rise to fea-tures representing the surface chemistry of the clean glass. The scatter in the data shown in Figures 4 and 5 allows only general trends to be discerned. The p.z.c.’s at pH 3 and 9 have been described in the preceding paragraphs. It is interesting to note that the chromic acid cleaned glass surfaces behave in a similar manner, showing virtually identical trends. The pyrolysis cleaned glass surfaces show dif-ferences in their behavior across the different glass compositions. These trends correlate with those observed for organic contamination of these surfaces, as de-scribed in Section 3.1, where the chromic acid cleaned glass surfaces all showed similar behavior, while the pyrolyzed glass showed significant differences in its sensitivity to contamination. In particular, the pyrolyzed silica surface shows far lower non-dispersive interaction energy with water than the pyrolyzed Corning code 1737 or sodalime glasses. This features correlates with the high degree of adsorbed contamination, described in Section 3.1, for the pyrolyzed silica surface.

The datum in Figure 5 for the non-dispersive interaction energy between a py-rolyzed silica surface and water at pH 7 corresponds to a contact angle of 31°. This is significantly higher than the contact angle of water on a pyrolyzed silica surface freshly immersed into liquid octane. While the surface cleanliness was measured after cleaning, it was not measured after substrate immersion in the acidic or alkaline solutions. It is possible that the comparatively low non-dispersive interaction energy observed for pyrolyzed silica is partially an artifact caused by contamination of the cleaned silica before immersion into liquid oc-tane.

Figure 4 shows similar behavior for the glass surfaces, suggesting that the alu-minoborosilicate and sodalime glasses show behavior similar to that of a silica surface. This phenomenon may be due to the leaching of soluble alkaline oxides from the glass surfaces during chromic acid cleaning, leaving a surface enriched in silica that behaves essentially in the same way as a chromic acid cleaned silica surface.

In Figure 5, the minimum in the non-dispersive interaction energy between glass and water at pH 9 is not present for pyrolyzed sodalime glass. This mini-mum was presumed to be associated with a high sodium ion concentration in solution, neutralizing the SiO– groups at the glass surface. The presence of sodium oxide (see Table 1) in the sodalime glass composition may generate a high so-dium environment for the the silanol groups at the glass surface. The high sodium concentration in the glass may thus be equivalent to a high sodium concentration in solution, neutralizing the p.z.c.

W. Birch et al. 104

3.4. Glass surface composition

The glass surface composition was measured using XPS following chromic acid and UV/ozone cleaning. UV/ozone cleaning replaced the PYR dry cleaning to avoid the contamination described in Section 3.1 for pyrolyzed silica surfaces. The XPS probed a 5 nm depth from the glass surface. Table 7 shows the percent-age atomic composition of the glass surfaces.

The silica surface shows a slight increase in the oxidation of its native surface silicon oxide film following UV/ozone cleaning. The aluminoborosilicate shows little change in its surface composition between the chromic acid or pyrolysis cleaning. The sodalime glass shows a loss of sodium following chromic acid cleaning. This reflects a leaching of the sodium oxide from the glass surface layer by exposure to acid. The residual sodium atoms may lie at a sufficient depth in the glass so as not to be removed by contact with the chromic acid. This meas-urement of glass surface composition indicates leaching of soluble alkaline oxides from the glass surface. However, the changes in the atomic composition of the glass surfaces appear small when compared to the observed differences in glass surface behavior.

Table 7.

Surface composition of cleaned glass measured by XPS. The penetration depth is estimated at 5 nm. The percentage concentrations of selected elements are given, resulting in a total of less than 100%. The carbon signal ranged from 4-7%, indicating a layer of ambient contamination adsorbed during transfer of the samples from cleaning to the measurement location. For the silicon wafer, only the signal attributed to the native oxide layer is given. Since the native oxide layer is 1-2 nm thick and the beam penetrates 5 nm into the sample, the total of the percentage compositions is much less than 100

Element % Silica UV/O3

Silica CHR

1737 ABS UV/O3

1737 ABS CHR

SDL UV/O3

SDL CHR

Si 13 12 27 29 26 29

O 34 30 59 59 47 59

Al 0 0 5 6 1 0

Na 0 0 1 0 8 3

3.5. Monolayer adsorption and deposition

Probing the practical behavior of the cleaned glass surfaces by charge-adsorbed surfactant monolayers gave results indicating a negatively charged surface on all the cleaned glass substrates. The anionic SDS surfactant was not deposited from solution, resulting in a wettable surface when the substrate was pulled from the surfactant solution. The cationic CTAB surfactant was deposited from solution, resulting in a non-wettable surface as the glass was pulled out of the solution. A negative surface charge is predicted for silicate surfaces in water at neutral pH.


The presence of surfactant monolayers coating the glass surfaces was con-firmed by measuring the wettability of non-polar tricresyl phosphate sessile drops in air. The contact angles on the SDS monolayers were in the range 61-66 de-grees; those on CTAB monolayers were in the range 40-41 degrees. This wettabil-ity is compatible with a lower packing density for the CTAB surfactant monolayer, where a larger head group prevents close packing of the aliphatic hy-drocarbon chains [3, 4].

Further information on the surface silanol sites can be obtained by depositing a self-assembled OTES monolayer. This monolayer was deposited on glass sub-strates cleaned by chromic acid or UV/ozone. A monolayer of grafted octadecyl-silane molecules generates high contact angles for water, indicating a hydropho-bic surface of densely-packed aliphatic hydrocarbon chains. Contact angle hysteresis probes the defect density in the self-assembled monolayers. The reced-ing contact angle is primarily influenced by the presence of hydrophilic defects in the hydrophobic monolayer. Measured wetting properties for water drops are given in Table 8.

For the UV/ozone cleaned surfaces, the data show a higher quality hydrophobic monolayer on the silica surface. A densely packed octadecylsilane monolayer gives rise to high contact angles for water. This dense packing is a result of a high degree of ordering of the octadecylsilane molecules and their grafting to each other and to the glass substrate. The aluminoborosilicate glass shows a coating quality intermediate between that on the silica and sodalime glass surface, which shows the lowest coating quality. The chromic acid cleaned glass surfaces show similar coating quality, slightly higher than that seen for the UV/ozone cleaned surfaces. This improved coating quality may be, in part, due to hydration of the chromic acid cleaned glass. The trends in coating quality for all UV/ozone cleaned glass surfaces and all chromic acid cleaned glass surfaces correlate with those seen for contamination of the glass surfaces, as described in Section 3.1.

Table 8.

Wettability towards water of self-assembled monolayers of OTES deposited on the cleaned glass surfaces. The sessile drop, advancing, and receding contact angles (in degrees) are given, with the sessile drop value above the advancing and receding values. The wettability measurements were made in air. For the sodalime glass (SDL), the slides used were float glass. The wettability of the deposited monolayer coatings was measured separately on the air side and the on the float side of this glass

silica ABS SDL (air) SDL (float)

106 101 100 99 UV/O3 cleaned

108 97 109 88 107 90 107 87

108 107 107 108 CHR cleaned

112 99 112 99 112 99 112 99

W. Birch et al. 106

4. CONCLUSION

While cleaned silica-based glass surfaces have similar surface compositions, their susceptibility to strongly adsorbing organic contaminants depends strongly on the glass composition and the cleaning procedure. For the three glass species exam-ined: silica, aluminoborosilicate, and sodalime glass, the glass surfaces behave similarly after chromic acid cleaning. They show significant differences in their properties following a dry cleaning procedure, such as pyrolysis or UV/ozone cleaning. The cleaned silica surfaces show a high susceptibility to adsorbing or-ganic contamination following pyrolysis cleaning, while the pyrolyzed sodalime glass appears to be virtually immune to strongly adsorbing organic molecules. Py-rolyzed aluminoborosilicate glass shows an intermediate susceptibility to adsorb-ing organic contaminants. The chromic acid cleaned glass surfaces all show an in-termediate susceptibility to contamination by adsorbed organic molecules. Thus, it may be an oversimplification to consider a clean glass surface as a high energy substrate that is bound to attract ambient organic contamination.

The wettability behavior of the cleaned glass surfaces showed features associ-ated with their exposed chemical functions. The non-dispersive interaction energy between glass and water as a function of pH showed evidence of charging of the surface silanol groups. The point of zero charge for these surface chemical func-tions was observed at pH 3. An estimate of the non-dispersive interaction energy between glass and water at the point of zero charge enables a reasonable estima-tion of the density of surface silanol groups on the cleaned glass. The trends ob-served for the surface charge as a function of pH correlate with the observed sus-ceptibility for adsorbing organic contamination to the cleaned glass surfaces.

Charge-adsorbed surfactant monolayers indicated a negative surface charge on the cleaned glass, as expected for silica-based glass surfaces at neutral pH. The wettability of grafted self-assembled octadecylsilane monolayers indicated high quality coatings on the cleaned glass surfaces. The coating quality was identical for all three glass species following chromic acid cleaning. The UV/ozone cleaned glass surfaces showed the highest coating quality on the silica surface, followed by the aluminoborosilicate surface and the sodalime glass surface. The trends in coating quality for all chromic acid cleaned surfaces and UV/ozone cleaned surfaces correlate with those seen for susceptibility to organic contamina-tion of the cleaned glass surfaces exposed to unpurified liquid octane.

REFERENCES

1. W.R. Birch, in: Sol Gel Handbook, M. Aegerter (Ed.), Kluwer Academic Publishers, New York (to be published) and references therein.

2. K.L. Mittal (Ed.), Silanes and Other Coupling Agents, Vol. 2, VSP, Utrecht (2000). 3. W.R. Birch, S. Garoff, M. Knewtson, R.M. Suter and S. Satija, Colloids Surfaces 89, 145

(1994). 4. W.R. Birch, S. Garoff, M. Knewtson, R.M. Suter and S. Satija, Langmuir 11, 48 (1995).


5. R.J. Good and M.K. Chaudhury, in: Fundamentals of Adhesion, L.-H. Lee (Ed.), pp. 137-151. Plenum Press, New York (1991).

6. J.J. Jasper, J. Phys. Chem. Ref. Data 1, 841 (1972). 7. F.M. Fowkes, Ind. Eng. Chem. 56 (12), 40 (1964). 8. J.N. Murrell and A.D. Jenkins, Properties of Liquids and Solutions, p. 168, John Wiley and

Sons, Chichester (1994). 9. A. Carré, F. Roger and C. Varinot, J. Colloid Interface Sci. 154, 174 (1992).

10. J.R. Vig, J. Vac. Sci. Technol. A 3, 1027 (1985). 11. J.R. Vig, in: Treatise on Clean Surface Technology, Vol. 1, K.L. Mittal (Ed.), pp. 1-26, Plenum

Press, New York (1987). 12. J. Davidovits, PhD Thesis, Détermination des conditions d’obtention de films monomoléculaires

organisés : application aux silanes auto-assemblés sur silice, Université Paris 6, Chapter 5 (1998).

13. J. Peanasky, H.M. Schneider and S. Granick, Langmuir 11, 953 (1995). 14. R.K. Iler, The Chemistry of Silica, John Wiley & Sons, New York (1979). 15. A. Carré, S. Moll, J. Schultz and M.E.R. Shanahan, in: Adhesion 11, K.W. Allen (Ed.), pp. 86-

87, Elsevier Applied Science, New York (1987).


Ed. K.L. Mittal

© VSP 2003

Decontamination of sensitive equipment

ROBERT KAISER∗ and KYLE HARALDSEN

Entropic Systems, Inc. (ESI), P.O. Box 397, Winchester, MA 01890-0597

Abstract—Most of the electronic and electro-optic equipment fielded by the military is incompati-ble with the standard aqueous based decontamination solutions, such as 5% sodium hypochlorite so-lution, ESI has developed a nondestructive decontamination process for such sensitive equipment. In this process, the components to be decontaminated are immersed in an ultrasonic bath filled with an organic solvent, if they are contaminated with chemical warfare agents (CWA), or in a solution of a surfactant in this solvent, followed by a pure solvent rinse, if they are contaminated with bio-logical agents or radioactive particles.

In both cases, the contaminants are dissolved or suspended in the decontamination liquid in the bath. The contaminants are removed from the decontamination liquid by circulating it through a fil-tration train. In the case of CWA, the filtration train consists of an activated carbon filter, a particu-late pre-filter, and a membrane filter. In the case of biological or nuclear contaminants, the circulat-ing liquid bypasses the activated carbon beds.

A prototype decontamination system has been built and operated to demonstrate the process. In this program, a wide range of sensitive equipment was contaminated with a CWA simulant. The contaminated equipment was immersed and sonicated in a flowing solvent, which recirculated around a purification loop, until the simulant could no longer be detected, and dried. The decon-taminated equipment was then functionally tested. In all cases:

a. no traces of simulant were found on the processed pieces, and

b. the processed items were fully functional.

Keywords: Decontamination; sensitive equipment; chemical warfare agents; biological warfare agents; ultrasonic cleaning; hydrofluoroethers.

1. INTRODUCTION

While much of the military equipment that is susceptible to chemical or biological threat agents can be decontaminated with aqueous decontamination agents, there are broad classes of critical equipment, including optical, electronic and commu-nication devices, that are rendered nonfunctional by such treatment. Historically, such equipment had been decontaminated by spraying and flushing with CFC-113. CFC-113 however, is an Ozone Depleting Compound (ODC) which was

∗To whom all correspondence should be addressed. Phone: 781-938-7588, x 22, Fax: 781-938-

7589, E-mail: [email protected]


eliminated from all DOD activities by the National Defense Authorization Act for Fiscal Year 1993.

Thus, alternate methods and equipment for nondestructively decontaminating water sensitive military equipment, such as avionics, electronic, electrical and en-vironmental systems equipment are needed. These methods need to:

a. Be effective against a wide variety of threat agents,

b. Be nontoxic to personnel,

c. Not degrade the equipment being decontaminated, and

d. Be field deployable.

Decontamination system equipment should be highly mobile and self-sustaining. These methods should also be able to treat equipment that is be-smirched with battlefield soils, including dirt (particulates), dried mud, oil, etc. In a broader context, the methods and equipment should also be capable of perform-ing maintenance cleaning operations in a depot environment. These methods and equipment should also comply with environmental regulations.

In terms of performance, an effective decontamination method has to be able to remove or deactivate the contaminant without affecting the part being cleaned. Because the type of equipment that would likely be decontaminated is both geo-metrically complex in shape and thermally sensitive, the most effective technique will likely use a flowing or agitated liquid, at ambient or modest temperatures, as a means of removing the range of threat agents from the equipment.

Numerous alternate potential decontamination options have been examined, but have been determined to be of limited effectiveness [1]. Heating an article above a modest temperature may not be an option for decontaminating thermally sensitive items, which leads to problems in terms of effectively removing relatively non-volatile contaminants. Decontamination by particle blasting methods, such as car-bon dioxide snow or plastic pellets, are limited to surfaces that are in direct line of sight with the ejection nozzle. Such methods are not effective in terms of cleaning blind holes, crevices, and obstructed surfaces. These types of methods can be abrasive and destructive to the equipment being decontaminated. Capture and processing of nonvolatile contaminant-laden particles may be a problem as well.

2. APPROACH

Until recently, there were no commercially available organic (i.e. nonaqueous) liquids that would be effective cleaning/decontamination media, and that would satisfy current and projected future safety and environmental criteria. Most vola-tile organic liquids that exhibit good solvency for chemical threat agents are flammable, toxic or environmentally unacceptable. The 1990 Clean Air Act, de-signed to eliminate volatile organic compounds (VOCs), ozone depleting com-pounds (ODCs), and other hazardous air pollutants (HAPs), has severely limited the classes of volatile organic solvents that can be considered.

Decontamination of sensitive equipment 111

Table 1.

Properties of fluorinated solvents of interest

Solvent Vertrel-XF [HFC-43-10]

HFE-7100

HFE-7200

Freon TF [CFC-113]

Chemical Formula C5F10H2 C5F9H3O C6F9H5O C2Cl3F3

Supplier DuPont 3M Co. 3M Co. NA

Molecular Weight 252 250 264 163

Boiling Point, °C 54 61 76 48

Freezing Point, °C –80 –135 –138 –3.5

Heat of Vaporization, cal/g @ bp 31 30 30 35

Specific Heat, cal/g @ 25°C 0.27 0.28 0.29 0.21

Specific Gravity (H2O = 1) 1.58 1.52 1.43 1.57

Viscosity, N·s/m2 @ 25°C 0.067 0.061 0.061 0.068

Surface Tension, mN/m @ 25°C 14.1 13.6 13.6 17.3

Vapor Pressure, mm Hg @ 25°C 226 202 109 334

Solubility of

Water in Solvent, ppm 490 95 92 170

Solvent in Water, ppm 140 <12 20 110

Hildebrand Solubility Parameter, MPa 0.5

13.8 13.1 13.5 14.7

VOC, kg/kg 0 0 0 1

Ozone Depletion Potential (CFC-11 = 1) 0 0 0 0.8

Global Warming Potential [100 yr ITH (1)]

1700 320 55 5000

Atmospheric Lifetime, yrs 17.1 4.1 0.8 110

Flashpoint, °C None None None None

Flammability Range in Air, % None None 2.4-12.4% None

Exposure Guidelines, 8 hr TWA, ppm

200 750 200 1000

(1) ITH: Integration time horizon Data Compiled from Published Information

Hydrofluorocarbons (HFCs), including the sub-class of hydrofluoroethers (HFEs), are a new class of organic liquids that have physical properties similar to CFC-113 (Table 1). The principal commercially available products are DuPont’s Vertrel-XF (HFC 43-10mee, 2-3 dihydrodecafluoro-pentane) and 3M’s Novec HFE-7100 (methyl nonafluorobutyl ether). The HFEs contain carbon, hydrogen, and oxygen, but no chlorine; and therefore have zero ozone depletion potential. The presence of a minority of hydrogen atoms gives HFEs many of the character-istics of a perfluoroalkane molecule, but also some characteristics of a hydrocar-bon molecule.


Figure 1. ESI’s process flow chart.

While retaining many of the properties and useful characteristics of CFC-113, such as wide materials compatibility, low toxicity, and lack of flammability, they do not possess the environmental limitations of CFC-113. They are not classified as VOCs, HAPs, or ODCs. HFEs also have a significantly lower global warming potential than CFC-113.

Entropic Systems, Inc. (ESI) developed a process for the decontamination of sensitive equipment that meets current requirements [2]. A conceptual process flow chart for the process is outlined in Figure 1. The contaminated parts are sprayed with a fluorescent marker and immersed in a bath filled with decontami-nation liquid. In this bath, surface contaminants are removed from the surface of the parts and transferred to the decontamination or decon liquid, either by solution or by suspension. Contaminated decon (decontamination) liquid is withdrawn from the bath and sent to a purification module that removes the dissolved or sus-pended contaminants from the liquid. The purified liquid is returned to the bath through spray nozzles to further treat the contaminated parts and decontaminate the cleaning chamber.

The parts remain in the bath until a prescribed cleaning regime is completed or until fluorescence sensors in the fluid circuits can no longer detect the fluorescent marker in the solvent that exits the cleaning chamber. The operator who opens the clean side door can verify that there are no longer any harmful levels of contami-


nants remaining on the treated parts by visually examining the parts for residual fluorescent marker before the parts are removed from the cleaning chamber.

The principal objectives of ESI’s development program were to:

a. Identify and/or develop safe and environmentally compatible cleaning agents and processes that will effectively remove biological and chemical threat agents from sensitive equipment that may be contaminated with other soils, without damaging this equipment.

b. Identify and/or develop means by which the threat agents suspended in the process liquids can be subsequently deactivated, and/or safely removed from these liquids.

c. Identify and/or develop methods of monitoring the presence of threat agents and/or other contaminants in the recovered process liquids.

d. Demonstrate the technology developed in the above objectives on a pilot plant scale.

These objectives were all met, and the major findings of this program are dis-cussed next.

3. MAJOR FINDINGS

3.1. Identify and/or develop safe and environmentally compatible cleaning agents

Ultrasonic solvent cleaning processes can effectively decontaminate sensitive equipment. Methoxyperfluorobutane (3M’s HFE-7100) is the decontamination liquid of choice because it meets the following criteria:

1. It is compatible with a wide range of sensitive equipment – the performance of electronic and optical equipment is not affected by immersion in HFE-7100. In particular, it does not attack components made of poly (methyl methacrylate) or polycarbonate, as does DuPont’s Vertrel-XF.

2. The principal chemical warfare agents (CWAs) of concern are sufficiently soluble in HFE-7100, as indicated in Table 2.

3. HFE-7100 is effective in ultrasonic cleaning baths because it has very low surface tension, which allows it to penetrate small features of the surface, and because it has a low heat of evaporation, which allows the ultrasonic agitation to produce strong shear forces in order to disrupt the boundary layer and en-train the contaminants.

4. The principal CWAs of concern are quantitatively removed from solution in HFE-7100 by activated carbon.

5. When agent-contaminated HFE-7100 is passed through a bed of activated carbon, the agent adsorbs onto the activated carbon, resulting in agent-free HFE-7100 that can be recycled and reused. In comparison, there was poor ad-sorption of chemical agents from Vertrel MCA+ in which mustard agent (HD)


exhibited a higher solubility. Vertrel MCA+ thus was not considered for two reasons – it was too aggressive a solvent to meet material compatibility re-quirements, and could not be reclaimed for recycling by adsorption of chemi-cal agents.

6. It is nonflammable, nontoxic, and environmentally acceptable.

Table 2.

Solubility of chemical agents in solvents of interest [2]

GB GD HD VX

Vertrel MCA+ M (RT) M (RT) 17% (RT) M (RT)

Vertrel XP-10 M (RT) M (RT) 8% (40°C) M (RT)

Vertrel-XF M (RT) M (RT) 8% (40°C) M (RT)

HFE-7100 M (RT) M (RT) 8% (40°C) M (RT)

HFE-7200 M (RT) M (RT) 8% (40°C) M (RT)

CHP M (RT) M (RT) M (RT) M (RT)

M (RT) = Miscible at room temperature

For effective decontamination to occur, sufficient shear has to be provided to result in effective mass and physical transfer of contaminants from the surfaces of the objects being decontaminated to the bulk of the decontamination liquid. In this process:

1. Ultrasonic agitation is a preferred means of providing this shear action.

2. For ultrasonic agitation to be effective, a minimum power density of 60 watts/gallon (15 watts/liter) is required.

3. The ability to generate ultrasonic power over a range of frequencies, from 40 kHz to 170 kHz, is desirable because it rapidly removes a range of particle sizes from the surface of the immersed part.

4. Oils soluble in HFE-7100, but thickened with a nonsoluble additive, are re-moved from exposed surfaces by high intensity ultrasonic agitation.

Biological contaminants are also effectively removed or inactivated by immer-sion and sonication in HFE-7100 or solutions of a fluorinated surfactant, poly-hexafluoropropylene oxide carboxylic acid (DuPont’s Krytox 157FS), in HFE-7100, as shown in Figures 1 and 2. More specifically:

1. Vegetative cells are killed by sonication in HFE-7100.

2. Processing in HFE-7100 with up to 4 to 6% Krytox 157FS can result in the sterilization of slides initially contaminated with approximately 100 spores (i.e. > 105 spores/m2).

3. Processing in these solutions also sterilizes slides that had been initially con-taminated with 104 bacteriophage particles.


4. Immersion in HFE-7100, with or without surfactant, denatures proteins.

5. The physical removal of biological species from a contaminated surface by sonication in HFE-7100 is enhanced by the presence of > 1% Krytox 157FS in the HFE-7100, and by the use of higher frequency ultrasonic (> 100 kHz) agitation.

It should be noted that the mechanism for the removal of radioactive contami-nants is similar to the removal of spores. The decontamination of sensitive equipment contaminated with radioactive contaminants by perfluorinated surfac-tant solutions was initially demonstrated by ESI under the auspices of a program sponsored by the U.S. Nuclear Regulatory Commission [3, 4], and subsequently commercialized as the Sonatol Process [5-7].

3.2. Identify and/or develop means of removing the threat agents from process

liquids

The removal of CWAs and of CWA simulants dissolved in hydrofluorocarbons by adsorption on activated carbon was demonstrated in small-scale batch tests that were performed by ESI for CWA simulants, and under subcontract by a surety laboratory (Battelle Memorial Institute) for CWA.

Figure 2. Circuit boards in culture medium. Circuit boards with environmental contamination were processed in the Cadet and then cultured in trypticase soy broth to assess effectiveness of removal ofmicroorganisms. The circuit board in jar A is an unprocessed control. After incubation, the culture broth is turbid, indicating multiplication of microbial contaminants (bacteria, fungi) present on thecircuit board as a result of exposure to the environment. The circuit board in jar B was decontami-nated using the Cadet as described in the text. The lack of turbidity in the culture indicates that the circuit board was rendered sterile as a result of processing.


Pilot scale continuous flow adsorption tests were performed in order to estab-lish the effect of process parameters on the rate and extent of removal of CWA simulants from a hydrofluorocarbon solution as it flowed through an activated carbon column, and on the adsorption capacity of the activated carbon in the col-umn. Providing sufficient residence time is critical to obtaining a reasonable col-umn loading before breakthrough of the contaminant occurs. A CWA simulant loading of 3 wt-% on granular activated carbon (Norit’s 1240 GAC) was achieved with a residence time of five minutes. The length to diameter ratio of the columns should be larger than 3 so as to minimize liquid bypassing. Activated carbon beds and filters that come into contact with contaminated liquid can be contained in commercially available housings that shield the system operator from any con-tained toxic contents. These sealed containers, and their contents, can be de-stroyed by standard methods, such as chemical deactivation or incineration.

3.3. Identify and/or develop methods of monitoring the presence of threat agents

Spectrographic fluorimetry is an extremely sensitive method of detecting fluores-cent materials. With this method, the detection limit of fluorescent dyes dissolved in HFE-7100 was found to be of the order of 10 parts per trillion (ppt). CWA simulants which contained small amounts (0.05 wt-% to 5 wt-%) of a fluorescent dye (Try-33 made by Day-Glo Corp., Dayton, OH) were used as contaminants in the pilot decontamination studies described in the next section. The presence of the fluorescent dye in the contaminant allowed the decontamination process to be easily monitored, in terms of being able to both:

1. Measure and detect low levels of contaminant in the process liquid leaving the ultrasonic bath and the activated carbon columns.

2. Detect traces of residual contaminant on the parts being processed by illumi-nating these parts with an ultraviolet lamp.

3.4. Demonstrate the technology developed in the above objectives on a pilot

plant scale

One of the major objectives of the program was to design and build a breadboard decontamination system in order to demonstrate:

1. The functionality of representative pieces of sensitive equipment is not af-fected by a process consisting of immersion and sonication in a bath of HFE-7100, followed by drying in super-heated HFE-7100.

2. The chemical agent simulant is effectively removed from such pieces of sen-sitive equipment.

3. The chemical agent simulant dissolved in HFE-7100 is quantitatively re-moved in real time from solution by passing the contaminated solution through a bed of activated carbon.

4. The internal recovery of the contaminated liquid, to allow recycling of the pu-rified process liquid.


5. The capture and removal of waste products (agent simulant, other soluble contaminants, and particulates) from the contaminated solution in fully en-closed, disposable activated carbon and filter cartridges, that are easy to in-stall, remove, and would minimize operator exposure to the contents of the cartridges.

6. The ability to immerse parts in both pure HFE-7100 and in fluorinated surfac-tant or coupling agent solutions in HFE-7100.

Additional objectives were to obtain operating data on specific unit operations, such as the effect of liquid circulation rate on the removal kinetics of simulant from contaminated test parts, the effects of liquid flow rate and contaminant con-centration on the efficiency, and capacity of activated carbon columns.

The breadboard decontamination system consisted of three modules:

1. A Poly-KleenTM Vapor Degreaser

2. An activated carbon column train

3. A circulating water chiller

The Poly-KleenTM Vapor Degreaser is a manually operated vapor degreasing system designed to be used with low boiling fluorinated solvents, such as per-fluorocarbons, hydrofluorocarbons and hydrofluoroethers. Figure 3 is a photo-graph of the inside chambers of the system. Overall dimensions are 1956 mm H x 736 mm D x 904 mm H (77" L x 29" D x 36" H). The height includes the handles on the sliding door which are 76 mm (3") high. The Poly-KleenTM Vapor De-greaser takes full advantage of the ease of fabrication of polypropylene and its compatibility with fluorinated liquids to allow a high performance system to be manufactured at a lower cost than an equivalent stainless steel system.

The basic Poly-KleenTM Vapor Degreaser is a three-sump unit that requires 140 liters (38 gallons) of fluid to operate. The principal components of the system and their effective dimensions are:

An immersion sump 432 mm L x 279 mm W x 254 mm H (17" L x 11" W x 10" H)

A boil sump 483 mm L x 356 mm W x 254 mm H (19" L x 14" W x 10" H)

A drying sump 356 mm L x 279 mm W x 305 mm H (14" L x 11" W x 12" H)

A vapor zone 1854 mm L x 279 mm W x 76 mm H (73" L x 11" W x 3" H)

A chilled condensate zone 1854 mm L x 279 mm W x 203 mm H (73" L x 11" W x 8" H)

A freeboard zone 1854 mm L x 279 mm W x 356 mm H (73" L x 11" W x 14" H)

A water separator 178 mm L x 102 mm W x 229 mm H (7" L x 4" W x 9" H)

A sliding cover 940 mm L x 330 mm W (37" L x 13" W)


Figure 3. Interior of Poly-KleenTM system.

Figure 4. Activated carbon adsorption module.

The activated carbon module is shown in Figure 4. The modules and the chiller are interconnected as shown in the process flow diagram presented as Figure 5. Referring to this process flow diagram, the in-line process components from BV-6 to CV-3 were mounted on an angle iron frame, 1829 mm L x 736 mm W x 762 mm H (72" L x 29" W x 30" H).


Figure 5. Poly-Kleen process flow diagram.

Cleaning trials were performed with the following pieces of sensitive equipment

a. Auto-Ranging LCD Digital Multimeters, Model No. 22-179A, Radio Shack, A Div. of Tandy Corp., Fort Worth, TX

b. Electronic Calculator, Model No. EC-441, Radio Shack, A Div. Of Tandy Corp., Fort Worth, TX

c. Global Positioning System (GPS) receiver, Model No. GlobalNav 212, Serial No.005263360, Lowrance Electronics, Inc., Tulsa, OK.

d. Night Vision Binoculars, Model RO 38, 4 x 48 Nighthawk, Serial No. 982331, with Model RO45, Zoom IR Illuminator, LAN Optics International, Burlington, MA.

e. 7.65 mm semi-automatic pistol, Model PP, Carl Walther GmbH Sportswaf-fen, Ansberg, Germany

f. Inverter Circuit Boards, 38 mm (1.5 in) square, designed by Entropic Sys-tems, Inc.

Numerous tests were performed with digital multimeters, which were consid-ered to be good prototypes for sensitive equipment. These items performed a number of electrical functions, they had a liquid crystal display covered by a clear plastic window, they contained a variety of materials that would be damaged by


many solvents, and were inexpensive enough (about $12.00 each) to be consid-ered disposable test items.

In addition, some tests were performed with other items to test the effects of part geometry. These items included standard 25 mm x 75 mm microscope slides (standard flat surfaces), brass pipe nipples (easily accessible interior surfaces), and magnet assemblies (difficult to access interior surfaces). A magnet assembly consists of a circular piece of stainless screen (typically 100 mesh) that is sand-wiched between two cylindrical Alnico magnets of equal diameter. The magnets were 13 mm (½") in diameter by 6.5 mm (¼" ) in height. The soil is deposited on the screen before forming a magnet assembly. This sandwich was then subjected to a cleaning trial. The changes in weight of the assembly and in the appearance of the screen were measures of the effect of the cleaning trial.

The test pieces were contaminated with a variety of neat and thickened CWA simulants and other soils. CWA simulants used in these tests were diethyl phtha-late (DEP), tributyl citrate (TBC), and Krytox 157 (L) and (H) fluorosurfactants. These materials are all water insoluble oils that have a low vapor pressure at am-bient room temperature. They also all are miscible with HFE-7100. The CWA simulants were all doped with a fluorescent dye, Try-33, that greatly facilitated their detection on the test pieces and in the decontamination liquid.

In some of the tests, a thickener was added to the simulant to mimic the behav-ior of thickened CWA agents. Two different types of thickeners were used: fumed silica (Cabosil LM-130, Cabot Corp.), and an acrylic polymer (Paraloid K-125, Rohm & Haas Corp.). Paraloid K-125 has been used to thicken military CWA. The consistency of the simulant depends on the amount of thickener used. At 1-2 wt-% thickener loading, the simulants flow like honey, while they become semi-solid gels at thickener loading greater than 5 wt-%. One key difference between colloidal silica and an acrylic polymer is that colloidal silica is not soluble in any organic solvent, but the acrylic polymer can dissolve in a polar organic solvent.

In addition to the above simulants, test pieces were also contaminated with rep-resentative soils that could be found on fielded equipment: mineral oil, SAE 30 motor oil (NAPA) thickened with Arizona road dust (Duke Scientific Co, Palo Alto, CA), multi-purpose lithium grease (Lubrimatic), and dried, 50 grit SiC wa-ter based lapping compound (Clover).

The contaminant removal tests were performed in the Poly-KleenTM system ac-cording to the following general procedure:

a. The equipment to be processed was weighed and photographed under visible and UV light.

b. One or more tared pieces of equipment were coated with contaminant(s) or soil(s), photographed under visible and UV light, and re-weighed.

c. The test piece(s) were placed into the transfer basket of the Poly-Kleen sys-tem, which was then covered with a tight fitting screen.

d. The immersion sump of the Poly-Kleen system contained enough HFE-7100 to cover the part in the basket. Sonication for 30 minutes degassed this liquid.


e. The transfer basket containing the items to be cleaned was lowered into the immersion sump, and statically (i.e. no liquid flow) sonicated for a finite pe-riod of time, usually 15 minutes.

f. After static sonication, the rinse pump was turned on and the liquid in the immersion bath was circulated through the activated carbon columns at a rate of 1,700 ml/minute for a finite period of time. The circulation time ranged from 15 minutes to 2 hours, depending on the purpose of the test.

g. The rate of decontamination was monitored by following the concentration of the contaminant in the decontamination liquid (HFE-7100).

h. Steps e and f were repeated until the presence of contaminant in the circulat-ing liquid could no longer be detected.

i. When the immersion sump liquid was free of contaminant, the transfer basket was moved from the immersion sump to the superheat sump and dried for 30 minutes to remove liquid drag out.

j. The transfer basket was removed from the Poly-Kleen system. The test pieces were removed from the basket, visually examined, photographed under visible and UV light, reweighed, and archived.

In order to maximize ultrasonic power density, the minimum amount of liquid needed to cover the parts being cleaned was used. Typically, the sump contained from 130 to 180 mm (5 to 7 inches) of liquid, which corresponds to a liquid vol-ume of approximately 15 liters to 30 liters (4 to 8 gallons) and a corresponding ul-trasonic power density of 26 to 18 watts/liter (100 to 70 watts/gallon). In prelimi-nary tests, it was noted that immersing and sonicating the test samples when the immersion sump was filled to the brim (about 53 liters (14 gallons)) did not result in effective cleaning. At that volume, the ultrasonic power density had dropped to a value of 8 watts/liter (30 watts/gallon). While this value would be considered marginal in a stainless steel ultrasonic bath, where the ultrasonic waves can be re-flected from the walls back into the liquid, in a polypropylene bath in which the walls absorb rather than reflect the ultrasonic waves, this power density level is too low.

If parts were also contaminated with biological agents, after Step h, they would be sonicated in a fluorinated surfactant/HFE-7100 solution that would be circu-lated through microfilters to remove suspended materials. The parts would then be rinsed in fresh HFE-7100 to remove fluorocarbon surfactant residues, and then dried as described above.

Table 3 lists the sensitive equipment decontamination experiments that were

carried out in the Poly-Kleen system during the course of the program. The combination of equipment processed, contaminants used, and monitoring method(s) examined are listed in this table. The results of the various cleaning re-sults are summarized in Table 4. This table records the weights of the items listed in Table 3, before and after contamination, as well as the post-cleaning weight and visual appearance of these items.


Table 3.

List of sensitive equipment decontamination experiments performed

Experiment No.

Sensitive equipment processed

Contaminant(s) Monitoring method

1 Multimeter Diethyl Phthalate Visual

2 2 Microscope Slides & Tributyl Citrate Visual, UV

2 circuit boards

3 2 Multimeters Tributyl Citrate Visual, UV

4 Multimeter Krytox FS

5 Multimeter Tributyl Citrate Visual, UV

6 GPS Receiver & Radio Krytox FS + Mineral Oil Visual, UV

Shack Calculator

7 Multimeter & Night Vision Krytox FS + Mineral Oil Visual

Goggles

8 Multimeter & Circuit Board Tributyl Citrate Visual, UV

9 Walter PP Pistol Krytox FS + Tributyl Citrate Visual

10A-10E Multimeter & Pipe Krytox FS Visual

Nipple (10 D only)

11 Multimeter Motor Oil, Lapping Compound, Krytox FS

Visual

12 2 Magnet Assemblies & Krytox FS Visual

2 Brass nipples

13 Multimeter - Face Down Krytox FS Fluorescence



16 Multimeter - Face Up Krytox FS Fluorescence

17 Multimeter - Face Down Tributyl Citrate Fluorescence

Except for the runs where there was visible attack of the substrate by the simu-

lant (as in run 1 in Table 3), there was an increase of less than a 0.1 gram in the weight of the object after contamination and cleaning and the original (i.e. before contamination) weight of this object. In some cases, there was a weight loss of the order of 0.1 gram (as in the calculator in run 6 and the pistol in run 9). This was attributed to the removal of other soils that were previously present on these test items.

If the ratio of (weight change/contaminant weight) is used as a cleaning crite-rion, this value is less than 10%, except for run 1 (for the reasons cited above), and for runs 13 to 17. For these last five runs, the relatively high values of this ra-tio is attributable to the precision of the weight measurement. The weight meas-urements were performed on a balance that had an accuracy of ± 0.02 gram, which would account for most of the observed weight differences.

Deco

nta

min

atio

n o

f sensitive eq

uip

men

t

12

3

Table 4.

Sensitive equipment cleaning results


Figure 6. Decontamination of night vision goggles (NVG).

While not a quantitative measurement, visual examination under ultraviolet il-lumination was considered to be the most sensitive and accurate means available to ESI of assessing whether traces of fluorescent contamination remained on the processed objects. Fluorescent contamination was observed only for run 1, and runs 3 and 5, where there was no noticeable weight increase.


Figure 7. Decontamination of Walther PP 7.65 mm hand gun.


Photographs of various test objects were taken under both normal and UV il-luminations,

a. Before the application of contaminant(s),

b. After the application of contaminant(s), and

c. After cleaning or decontamination.

The presence or absence of contaminant on the items is clearly evident by sim-ple examination of the photographs taken under UV illumination of the night vi-sion goggles presented in Figure 6 and the pistol in Figure 7.

Process kinetics was examined by monitoring the concentration of contaminant in the liquid in the immersion bath as a function of time. Removing the contami-nant from the surfaces of the object being cleaned and transferring it into the bulk liquid took approximately 15 minutes. Once the circulation pump to the activated carbon columns was turned on, approximately three bath turnovers were required to purge the immersion sump of contaminant. In this test program, turnover time was of the order of 12 to 15 minutes.

4. POTENTIAL BENEFITS OF THE TECHNOLOGY

Potential benefits of the decontamination process discussed in this paper include:

Demonstrated compatibility of sensitive equipment with the hydrofluoroether process liquids under processing conditions.

Demonstrated solubility levels of CWA agents of interest in the process liq-uids.

Demonstrated ability to remove CWA simulants from items of sensitive equipment.

Demonstrated ability of HFE-7100/Krytox 157FS surfactant solutions, in con-junction with multi-frequency ultrasonic agitation to remove/neutralize a range of biological agents from solid substrates, including circuit boards.

Demonstrated ability to quantitatively remove CWA and CWA simulants from solution in HFE-7100 by adsorption on activated carbon.

Simple, safe, automatable decontamination process.

Acknowledgements

The work described in this paper was supported by the U.S. Air Force under the auspices of SBIR PROJECT AF 97-014, “DECONTAMINATION OF AIR-CRAFT ELECTRONIC EQUIPMENT”, Contract Number: F41624-98-C-5061. It was facilitated by the insights and encouragement of Dr. Ngai Wong, AFRL/HEST, the Air Force Technical Monitor.


REFERENCES

1. “Joint Service Sensitive Equipment Decontamination (JSSED), Block I Systems Technology Overview”, Report PAO-01-432, October 2000 (available from DTIC, Fort Belvoir, VA).

2. R. Kaiser, “Decontamination of Aircraft Electronic Equipment”, Final Technical Report, SBIR Project AF97-014, Contract No. F41624-98-C-5061, November 2000.

3. R. Kaiser, C.S. Yam and O.K. Harling, “Enhanced Removal of Radioactive Particles by Fluoro-carbon Surfactant Solutions – Process Development”, Interim Report, Prepared for US Nuclear Regulatory Commission, Washington, DC, Contract No. NRC-04-93-106, March 1995.

4. R. Kaiser, C.S. Yam, S.R. Landahl, P.A. Droof and P.H. Jones, Jr., “Enhanced Removal of Ra-dioactive Particles by Fluorocarbon Surfactant Solutions – Process Demonstration”, Final Re-port, Prepared for US Nuclear Regulatory Commission, Washington, DC, Contract No. NRC-04-93-106, September 1997.

5. R. Kaiser, C.Y. Yam and A.E. Desrosiers, “Decontamination of Electromechanical Parts by the Sonatol Process: II – Results”, Proc. Waste Management (WM) ’98 Conference, Tucson, AZ, March 1998.

6. A.E. Desrosiers and R. Kaiser, “Decontamination of HEPA Filters for Reuse”, Proc. EPRI In-ternational Low-Level Waste Conference, McAfee, NJ, July 1999.

7. A.E. Desrosiers, R. Kaiser and C.B. Voth, “TRU Waste Minimization During Hot Cell Decom-missioning”, Proc. American Nuclear Society Annual Meeting, Boston, MA, June 1999.

Note: Copies of Refs. 2, 3, 4, can be obtained from the sponsoring agencies or the author of this paper.


Ed. K.L. Mittal

© VSP 2003

The fundamentals of no-chemistry process cleaning

JOHN B. DURKEE II∗

Creative Enterprizes, 437 Mack Hollimon Drive, Kerrville, TX 78028

Abstract—This paper describes a new method for cleaning of parts. The method converts soils to water-organic emulsions via pressure waves generated by ultrasonic transducers. Underwater photo-graphs and symbolic representations show that the process consists of: (1) contact with ultrasonic-generated pressure waves produces an emulsion of soil in water, and (2) rinsing the emulsion from the parts.

Keywords: No-chemistry; cleaning; emulsion; ultrasonics.

1. INTRODUCTION

Cleaning without chemistry is not new; however, the information in this paper is

new. This paper is not about flushing with pure water to displace particles in criti-

cal cleaning situations nor is it about slowly dissolving non-volatile residue

(NVR) [1] without rinsing residues in precision cleaning situations.

This paper is about cleaning large or small quantities of oils, greases, or soaps

from metal (chiefly) parts using only water [2]. This paper describes a new

method of solving practical cleaning problems at all levels of cleaning quality [3].

This new method is used most often in continuous processes, though it can be

used easily in batch processes.

In this paper, the new method is described and compared to normal aqueous

cleaning technology. Photographic evidence of its use is shown. Limitations are

defined. Areas for future research are discussed.

2. EXPERIMENTAL

Four photographs, Figures 1a to 1d, were taken at 32°C (90°F) in un-degassed tap

water. They illustrate the experiments carried out here to describe the two-step

process. In the first process step, the parts to be cleaned are bombarded with ultra-

∗Phone: (830)-895-3755, Fax: (612)-677-3170, E-mail: [email protected]

J.B. Durkee 130

sonic waves from a transducer. This produces an emulsion of water and soil on

the parts. In the second step, the emulsion (which contains the soil) is rinsed from

the parts using any combination of water jets.

The circular-shaped ultrasonic transducer produced 2000 watts over about a

30.48 cm (12 inch) length. The part was an easily recognizable form – a bullet

formed of carbon steel. The part was held on a thin copper wire about 1.27 cm

(0.5 inch) away from the circular transducer. The soil was 10W30 motor oil ap-

plied liberally with a paintbrush before the part was immersed in water. The four

photographs in Figure 1 were taken over a period of about five seconds.

The emulsion was analyzed for oil, and part cleanliness was measured.

Figure 1a (Equivalent to Figure 3a) Figure 1b Continued Formation of Emulsion Initial Formation of Emulsion

Figure 1c (Equivalent to Figure 3c) Figure 1d Further Condensation of Gel Material Formation of Gel from Dilute Emulsion

Figure 1. Photographic representation of cleaning with no-chemistry process.

The fundamentals of no-chemistry process cleaning 131

3. RESULTS

Emulsions similar to those seen in Figures 1a to 1d were sampled via a suction

tube from beneath the liquid level. The initial emulsion (Figures 1a and 3a) was

stable over several days and contained around 5% water. The final gel-emulsion

was stable under boiling conditions for at least one week nonstop and contained

approximately 90% water.

After the gel was rinsed off via pressurized water sprays, the parts (bullets)

were free of oil as ascertained by white glove tests and viewing in black light. An

extraction test in hexane showed all parts to have <1.85 mg/m2 (15.6 mg/ft

2) oil

residue after cleaning in this manner. These parts would be very clean for indus-

trial use, after only one step of sonication and rinsing.

4. DISCUSSION

Today’s aqueous and solvent processes both involve the same three factors: sol-

vency, temperature (heat), and mechanical force. This is shown in Figure 2a.

Naturally, solvency dominates solvent cleaning. Mechanical force dominates

aqueous cleaning.

The No-chemistry technology presented in this paper involves a different para-

digm. This is shown simply in Figure 2b. There is no chemistry so there is no di-

rect concern about solvency. The process is practiced from room temperature (or

below) up to about 51.6°C (125°F). Consequently, the addition of heat is not a

significant factor.

4.1. Another kind of mechanical force

Mechanical force is the remaining of the three factors. The force is generated by

ultrasonic pressure transducers which normally produce cavitation bubbles.

Figure 2a. Traditional elements of cleaning. Figure 2b. No-chemistry cleaning.

J.B. D

urkee

13

2

Figure 3a Figure 3b Figure 3c Figure 3d Figure 3e

Pressure Waves Impact Pressure Waves Emulsion Collects Rinse Dislodges Rinsed Cleaned

Contaminated Surface Produce Emulsion into a Gel Gel from Surface Surface

Figure 3. Symbolic representation of cleaning with no-chemistry process.

Small blue arrows in all figures represent pressure waves produced by ultrasonic transducers. Colored shapes in Figure 3a represent individual elements of

soil. Shadowed shapes in Figure 3b represent conversion of soil to water emulsions. The agglomeration of shadowed shapes in Figure 3c represents a wa-

ter-bearing gel with some adhesion to the surface. Large red arrows in Figures 3d and 3e represent jets or sprays of pressurized water as flows of rinse wa-

ter. The part is free of soil in Figure 3e.


Table 1.

Comparison of no-chemistry cleaning to cleaning via cavitation

Item Conventional cavitation No-chemistry cleaning

Water degassing Essential Not done

Optimum temperature ~ 71°C (160°F) Any temperature

Hydraulic conditions Static flow conditions Moving flow conditions

Standing waves Avoided via frequency sweep No frequency sweep

Bubbles Observed Not Observed

Cavitation YES NA

Hence, one might assume that the nature of the force is cavitation-induced implo-

sion of tiny bubbles. This assumption is incorrect. Cavitation plays no role in this

No-chemistry cleaning process. In fact, conditions are deliberately controlled to

be opposite of those which normally produce cavitation bubbles. These conditions

are given in Table 1, and will be later shown in Figure 3a to Figure 3d.

4.2. The no-chemistry cleaning process

The No-chemistry cleaning process is simple. As stated in Section 2, only two

steps are involved. Both must be performed completely, or the cleaning will not

be effective. The steps are to:

(1) Produce an emulsion via ultrasonic pressure waves

(2) Rinse the emulsion off the parts

The pair of steps are shown photographically in Figures 1a to1d, and symboli-

cally as Figures 3a to 3e. They may be repeated as necessary to produce the de-

sired level of cleaning in the time or space allowed.

The separation in time of Figures 1a to1d and Figures 3a to 3d/3e is about five

seconds. In other words, the cleaning process involving these two steps can easily

be completed in twenty seconds, or less.

The process is self-limiting. It is complete when there is no soil left to emulsify.

4.3. Conditions necessary to form and remove emulsion

There are three necessary conditions [2]:

1. The transducers must be located close to the parts. The distance of 1.27 cm to

7.62 cm (0.5 to 3 inches) is recommended.

2. The volumetric power intensity must be high, vs applications involving ultra-

sonic-produced cavitation, (>26.4 watts / liter (100 watts / gal)).

3. There must be motion in the fluid (to complete the rinsing step). Both con-

tinuous flow and batch processes have been built and operated successfully.

A single transducer can be located above or below the parts. A pair of trans-

ducers may be located above and below the parts. Line of sight access is not re-

quired, as with megasonic transducers.

J.B. Durkee 134

If multiple transducers above and below the parts are used, the same frequency

is used for each, vs the referenced patent [4], which requires a difference in fre-

quency. The purpose of the multiple transducers is to expose the surface which

would not ordinarily be exposed to ultrasonic waves. Flat parts, which have two

sides, are a common example.

The rinsing nozzles can be located under water with the transducers or the parts

can be raised above the water and rinsed there. Obviously, a different design of

the rinse nozzle would be used in each case.

The rinsing step is absolutely crucial, as is shown by the following experience.

A scaled-up machine was constructed. It had two steps of emulsion production via

sonication, and no intermediate step of rinsing. It was worthless as a cleaning ma-

chine, despite 2,000 watts of ultrasonic power aimed at a few non-moving parts.

After a great deal of trial and error, an intermediate rinsing zone was added. It

was this inadvertent experiment which showed that No-chemistry cleaning was a

two-step process.

Water quality is not relevant. Excellent cleaning has been obtained in deionized

water, tap water, water produced by reverse osmosis, water containing tramp in-

soluble soil, and water containing ~ 1% emulsion.

There is no practical limit to operating temperature. Issues which determine

temperature [2] are soil and substrate. Lower temperatures make emulsion forma-

tion and removal more difficult because the emulsion is more viscous. Higher

temperatures cause steel surfaces to rust. Typical values of operating temperature

are between 26.6°C (80°F) and 51.6°C (125°F).

Excellent cleaning of molybdenum-based grease from 304 stainless steel has

been done at 90.6°C (195°F) [5]. This temperature was chosen to reduce the vis-

cosity of the grease in order to increase indirectly the rate of formation and re-

moval of emulsion. 10W50 motor oil has been removed from carbon steel at

15.6°C (60°F).

4.4. Process operation

No-chemistry cleaning is not a mere curiosity, like the vortex tube. More than a

dozen machines have been constructed. Each machine transports parts through

zones where the parts can be sonicated to produce an emulsion and then another

zone where the emulsion can be rinsed away. Some machines have operated for

thousands of hours. Some machines have cleaned millions of parts. Nearly all

machines have cleaned hundreds of pounds of parts per hour. No chemicals have

been added to any machine.

These machines differ chiefly in transport of parts, and that is the fundamental

limit on the use of No-chemistry cleaning technology. Large parts are most diffi-

cult to transport and expose to ultrasonic transducers mounted close to the part’s

surface. Tractor axles are the largest part successfully cleaned. Each weighs about

18.8 kilograms (forty pounds) and are about the size of a 13" television set. Two

machines continuously clean oils from these parts.


4.5. Process limitations

No-chemistry cleaning technology has at least three types of limitations.

4.5.1. Technology related

Proximity of the parts to the transducer is the most significant item. Shorter dis-

tances are favored: 1.27 cm (0.5 inches) to 7.62 cm (3 inches). In a recent suc-

cessful application, the transducers were 12.7 cm (5 inches) from the parts. But

the power intensity had to be increased to ~ 2.5 times normal. The practical effect

of this restriction is to limit the dimensional size or weight of parts which can be

cleaned without chemistry.

4.5.2. Inadequately designed process

Design factors must be within limits determined by experiment, such as:

Holdup time under sonication should be at least fifteen seconds. Holdup time

is a tradeoff between cleaning quality and productivity [6].

Too few steps of sonication (formation of emulsion) and rinsing. The number

of steps of sonication is a tradeoff between cleaning quality and machine size

or cost.

Operating temperature can be too low (between 26.6°C [80F] and 51.6°C

[125°F]). This can limit cleaning quality, as a more viscous emulsion is rinsed

less efficiently.

Omission of a rinse step after a sonication (formation of emulsion) step. This

can limit cleaning quality if some emulsion is not removed

Intrusion of free or tramp soil. This situation is worse than having to clean

parts which are more dirty, because the soil content can be unanticipated and

variable. Free soil can be fatal to any proposed application.

4.5.3. Waste management

This is a problem where the soil can be removed from parts, but cannot be re-

moved from the cleaning machine. The problem is unique to No-chemistry clean-

ing technology. This is because there is no chemistry to “surround” the soil, and

“escort” it from the cleaning machine.

Thus soil elements agglomerate or attach themselves to the cleaning machine.

When No-chemistry cleaning removes the soil from the parts and allows it to be

deposited on the machine, the application is a total failure. Fortunately, this hap-

pens only with some soils which do not form a stable emulsion with water. Two

such soils are molybdenum disulfide and lithium-based greases.

5. FUTURE RESEARCH

Outside of developing knowhow and implementing this technology in commercial

cleaning machines, future R&D will focus on achieving a more complete under-

standing in these areas:

J.B. Durkee 136

a. Factors which govern the rate of emulsification.

b. Surface chemistry and physics associated with cleaning ferrous parts which

do not rust after rinsing and drying.

c. Limits of part size on cleaning rate. This will involve the following variables:

weight, area per weight, part geometry, distance from transducers to parts,

and deliberate reflection of ultrasonic pressure waves onto the parts. One

practical example will be an attempt to clean engine blocks.

d. Types of part conveyances which can or cannot be used. Currently, rotating

metal drums and metal belt conveyor are the only part conveyances used.

Planned experiments include hydraulic conveyance of small parts in a duct

and woven spun-bonded fabric materials used as belts.

e. Soil management. Current plans are to develop an efficient evaporation sys-

tem which would concentrate the soil–water emulsion. Water would be re-

covered for reuse in cleaning; soil would be concentrated to minimize dis-

posal costs.

6. SUMMARY

Cleaning of oil and grease soils without chemistry is technically feasible. Photo-

graphs show how these soils are emulsified in water. Cleaning data show how the

emulsions are rinsed to produce clean parts. This two-step cleaning process has

been implemented in continuous or batch machines.

Acknowledgments

I would like to thank Walter Johnson, Flo-Matic Corporation, Rockford, Illinois,

USA, for his sponsorship of this work and for the freedom to experiment without

regard to commercial need. I would also like to thank my colleagues Gary

Kauffman and Vasilly Pekun of Flo-Matic.

REFERENCES

1. “Guide to Inspections Validation of Cleaning Processes”, http;//ww.fda.gov/ora/inspect_ref/igs /valid.html

2. W.J. Johnson, US Patent 6,368,414 (2002). 3. J.Y. Baker and J.B. Durkee, A2C2 Magazine, 4-9 (October 1999), 4-6 (November 1999); and 39-

40 (January 2000). 4. H.B. Swainbank et al., US Patent 4,788,992 (1988). 5. W.J. Johnson, paper presented at the International Fastener Exposition, Chicago, May 27, 1999. 6. J.B. Durkee, W.J. Johnson, G. Kauffman and V. Pekun, US Patent Applications 340769/

900016, 340769/90008, and 340769/900032 (2000).


Ed. K.L. Mittal

© VSP 2003

Development of a technology for generation of ice

particles

D.V. SHISHKIN, E.S. GESKIN∗ and B. GOLDENBERG

Waterjet Laboratory, Department of Mechanical Engineering, New Jersey Institute of Technology,

Newark, NJ 07102-1982

Abstract—The mission of this project was to develop a practical technology for formation of ice particles. Our previous works demonstrated the effectiveness of the use of ice-air mixture as a clean-ing medium. However, a practical technology for fabrication of ice particles of a desired size and at a desired temperature is still under development. The physical properties of ice (tendency to ag-glomerate, melting, etc.) make the formation and transportation of ice particles extremely difficult. We developed a process for controllable generation of ice particles using a rotational crusher em-bedded into a heat exchanger. Water was supplied at the bottom of the heat exchanger and as it moved along the rotating auger ice was formed and crushed. A refrigerant or liquid nitrogen were used as cooling media. At the exit of the heat exchanger the air stream entrained the generated parti-cles. In the course of operation the cooling conditions and the auger rotation were maintained con-stant while the water flow rate varied. The rate of production and the shape and size of the generated particles were monitored. We also investigated the temperature distribution along the heat ex-changer and the corresponding distribution of particle sizes. As the result of this study, process phe-nomenology was developed and a design of the system for formation of ice particles was suggested.

Keywords: Ice; particles; solidification; solid decomposition; brittle fracturing; thermal stresses.

1. INTRODUCTION – ICE AS AN ABRASIVE MEDIUM

A number of surface processing technologies based on the use of the air-ice

stream have been previously suggested. A car washing machine constituted the

first attempt towards utilization of the ice particles [1]. Another technology used a

stream of charged ice particles directed toward surfaces [2]. Szijcs [3] proposed

cleaning of sensitive surfaces by the impact of fine grade ice and air. The atomi-

zation of the liquid in the air stream and subsequent freezing of the generated fine

droplets formed the blast material. The freezing was achieved by the addition of a

refrigerant (N2, CO2, Freon) into the stream in the mixing chamber or by the addi-

tion of refrigerant into the jet after the mixing chamber. Another technology in-

∗To whom all correspondence should be addressed. Phone: (973)596-3338, Fax: (973)642-4282,

E-mail: [email protected]

D.V. Shishkin et al. 138

volved the use of ultra-clean ice particles, having a uniform grain size, for clean-

ing the surface and grooves of ferrite block (Tomoji [4]). An ice blasting device

utilizing stored particles was suggested by Harima [5]. Vissisouk [6] proposed to

use ice particles near melting temperature for surface decoating. Mesher [7] de-

veloped a nozzle for enhancement of the surface cleaning by ice blasting. Shinichi

[8] suggested an inexpensive cleaning of various surfaces by mixing ice particles,

cold water and air. Niechial [9] proposed an ice blasting cleaning system contain-

ing an ice crusher, a separator and a blasting gun. Settles [10] suggested produc-

ing ice particles of a size range below 100 µm within the apparatus just prior to

the nozzle.

Although the potential use of ice blasting has been suggested by a number of

inventors, the practical use is much more limited. Herb and Vissisouk [11] re-

ported precision cleaning of zirconium alloys in the course of production of bi-

metallic tubing by ice pellets. It was shown that ice blasting improved the quality

of bimetal. The use of air-ice blasting for steel derusting was reported by Liu [12].

The following operational conditions were maintained: air pressure: 0.2-0.76

MPa; grain diameter: below 2.5 mm; ice temperature –50°C; traverse rate 90

mm/min; and standoff distance 50 mm. At these conditions the rate of derusting

ranged from 290 mm2/min at the air pressure of 0.2 MPa to 1110 mm

2/min at the

air pressure of 0.76 MPa. The quality of the treated surfaces complied with ISO

8501-1 Sa 2.

In the final analysis, the adoption of the ice-jet technology is determined by the

effectiveness of the generation and handling of ice particles. Regular abrasives are

stable at practically feasible operational conditions, while ice particles can exist

only at subzero temperature. Maintaining such a temperature prior, within and

outside of the nozzle is an extremely difficult task. The adhesion between the par-

ticles increases dramatically as the temperature approaches 0°C. At this condition

ice tends to pack and clog the supply lines. Thus these lines must be maintained at

a low temperature. This and other similar problems prevent adoption of ice-

waterjet (IWJ) by the industry. In order to assure the acceptance of IWJ, it is nec-

essary to develop a practical technology for formation, transportation and accel-

eration of ice particles.

2. PROPERTIES OF WATER ICE

The practical application of water ice as a machining medium is determined by

the physical properties of the solid water. The properties of Ice I existing at the

modest pressure (below 200 MPa) have practical importance. An important fea-

ture of Ice I is the reduction of the melting temperature with an increase of the

pressure. The minimum temperature of the liquid water is attained at a pressure

about 200 MPa and is equal to –20°C. Another important feature determining par-

ticle behavior in the course of impact is ice elasticity. In the temperature range of

–3°C to –40°C ice is an almost perfect elastic body. Hooke’s law is obeyed if the

Development of a technology for generation of ice particles 139

stresses in the ice are below a certain level and are applied for a short period of

time. The dynamic elastic properties of ice at –5°C are: Young’s modulus (E) =

8.9-9.9 GPa; Rigidity modulus (G) = 3.4-3.8 GPa; Bulk modulus (K) = 8.3-11.3

GPa; Poisson’s ratio (χ) = 0.31-0.36 according to Hobbs [13]. For comparison,

for an aluminum alloy 1100-H14, E = 70 GPa and G = 26 GPa. For silica glass E

= 70 GPa. If the columnar ice is stressed perpendicular to the long direction of the

column, the static Young’s modulus in bars is determined by the following equa-

tion:

E = (5.69-0.64T)*104 (1)

where, temperature T is given in °C. The dynamic Young’s modulus of ice in-

creases almost linearly from 7.2 GPa at –10°C to 8.5 GPa at –180°C, and is inde-

pendent of the direction of loading. The data above show that the ice powder can

be considered as a soft blasting material and used accordingly.

One of the main issues in the use of the ice powder is sintering of the particles

and their adhesion to the surface of the enclosure. The strength of adhesion de-

pends on ice temperature. This dependence is shown in Figure 1 (a). It follows

from this figure that it is necessary to maintain ice temperature below –30°C to

prevent sintering of the particles. Sintering also depends on the duration of parti-

cles contact. The radius of the neck, which forms between two ice spheres,

Figure 1. (a) Strength of adhesion of ice particles, and (b) schematic of the sintering of ice particles [13].


brought into contact for time t at temperature T (Figure 1 (b)) is determined by the

equation

( )n

m

x A Tt

r r

=

(2)

where x is the radius of the neck, r is the radius of sphere, A(T) is a function of

the temperature, n and m are constants. A(T), m and n depend on the mechanism

of sintering. It follows from equation (2) that it is necessary to prevent the contact

between particles in order to avoid particles sintering.

The moisture contained in the atmosphere in the course of ice transportation

enhances the adhesion of the ice to the walls as well as sintering of ice particles.

Both phenomena result in the plug formation and clogging of the conduits.

3. EXPERIMENTAL SET UP FOR FABRICATION OF ICE PARTICLES

In our previous experiments several systems for ice powder formation were

tested. One of such systems as depicted in Figure 2 was selected for further ex-

periments. The system consists of the following functionally separated blocks:

– ice making block which includes the evaporator, auger, auger driver, sealing

and liquid nitrogen cooling apparatus;

– ice unloading mechanism

– nozzle block which includes parallel nozzles and focusing device.

In our experiments water entered the heat exchanger via a special port. As it

moved along the rotating nozzle it solidified and an ice plug was formed. Decom-

position of this plug led to formation of ice particles. At the outlet of the heat ex-

changer the powder entered into the nozzle block 5 and was driven to the air gun 6.

The heat exchanger and the auger of the icemaker constituted a modified com-

mercial icemaker of the Hoshizaki Co. of America, Peachtree, GA. The design of

these parts will be changed in the next generation of the device. The cooling was

carried out by the refrigerant Galden HT-55 supplied by the cooling TurboJet ap-

paratus or by liquid nitrogen stored in a tank. In both cases the supply of the cool-

ing medium was determined by the characteristics of the source, the refrigeration

system or the nitrogen tank. We replaced Hoshizaki auger driver by a more pow-

erful device in order to prevent jamming of the ice. The rotation momentum of the

auger 4 was provided via a gearbox with gear-ratio 1:100. However, the selected

driver operated at a constant angular velocity of 100 rpm. At these conditions, the

size and temperature of particles were determined by the rate of the water supply

to the port 8, which changed gradually from 0 to 200 ml/min. Water flow rate was

precisely controlled by a special valve (Figure 2 (a)). In the final analysis, we were

able to generate the desired kind of particles at given cooling conditions by the

proper selection of the water flow rate. The attempts to improve ice production by


an increase of the water pressure or a control of the supply of the cooling medium

were unsuccessful. Solidified ice plug moved forward along the auger helical

ways. It was determined, however, that it was absolutely necessary to eliminate

any obstructions to ice flow in order to prevent jamming of the heat exchanger. It

is quite obvious that the conditions of the ice production (specific cost and energy

consumption, process stability, uniformity of the generated particles, output per

cm2 of the outlet cross-sectional area, etc.) will be dramatically improved by the

process optimization.

An unloading mechanism delivered ice particles to the abrasive port of the air

gun. The nozzle block consisted of two nozzles and a special focusing device.

Three different sizes of the nozzles were used, however, in all cases the nozzle to

focusing tube ratio was 1:2. Ice was delivered to the nozzles abrasive port through

insulated flexible plastic tubes.

The ice crystallization was monitored through a set of thermocouples and resis-

tance gauges (10) imbedded into the evaporator (Figure 2(b)). Data acquisition

card processed the signals generated by the gauges.

The large number of control variables (rate of supply of water and coolant, rate

of auger rotation, diameter and length of the heat exchanger, distribution of tem-

perature and heat flow along the refrigerator, geometry of the auger) practically

excluded empirical process optimization. At the same time, the complexity of the

Figure 2. (a) Schematic of auger type IJ system where: 1 – evaporator, 2 – refrigerant coils, 3 – in-sulation, 4 – auger, 5 – ice reloading device, 6 – air gun, 7 – air supply port, 8 – water supply port,9 – cooling medium port, 10 – gauges, A – air flow rate valve, B – water flow rate valve, C – cool-ing medium valve, D – data acquisition card, and (b) schematic of gauges placement inside theevaporator with their distances measured from water inlet port (water zero level).


phenomena that occur in the course of the powder formation (water freezing and

simultaneous decomposition in the turbulent layer) made it impossible to con-

struct a physical model of the process. Thus the only practical approach to the

process design entailed experimental study of the distribution of water properties

within the icemaker, construction of the process phenomenology using the ac-

quired information and then the use of commercial packages (Pro/ENGINEER,

FIDAP, etc) to evaluate the design parameters of the icemaker as well as the cor-

relation between input variables and the ice properties.

Special experiments were carried out in order to examine the state of the solid

ice in the course of solidification and particles formation. In order to attain this

goal the auger was stopped during a normal operation and then removed from the

heat exchanger. The solid phase was visually examined. The state of the ice is de-

picted in Figure 3. The anatomy of the ice in the course of freezing and particles

formation was investigated using the set of pictures developed. This information

was supplemented by the ice temperature and electrical resistance monitored prior

to the auger removal. Extensive database acquired in the course of these experi-

ments was used to develop a hypothetical mechanism of ice particles formation

The suggested hypothetical process phenomenology is described below.

4. PHENOMENOLOGY OF ICE BLOCK DECOMPOSITION

Powder formation is a complex process involving solidification, stress develop-

ment in the solid phase, supercooling of the solid with cracks propagation inside

the block, concentration of the structural and thermal stresses within the solid and

finally decomposition of ice blocks. According to our hypothesis ice block solidi-

fication occurs in the bottom evaporator zone. The formation of a solid block and

its initial decomposition are depicted in Figure 3. At this stage the formation of

Figure 3. Solid ice block decomposition at the two distinct stages of ice powder formation: a) initialstage of block formation, and b) the intermediate stage of block decomposition.


the ice structure is complete and further cooling brings about only reduction of ice

temperature. Thus particles size can be controlled only at the initial stages of the

process.

The following phenomenology of particles formation has been suggested. Ice

nucleation inside the evaporator obeys the rules of granular ice formation (Ice

type I). It originates as water flow reaches the evaporator wall with a temperature

of –196°C. Ice nucleation and crystal growth depend on the conditions of cooling

and the auger spiral movement. The ice plug forms and propagates along the au-

ger stem. The generated ice acquires a multilayer pattern. A supercooled ice layer

directly adjacent to the evaporator and a moderately cooled layer in the vicinity of

auger can be identified. An explosive character of water freezing near the evapo-

rator wall leads to thermal ice expansion inside the volume constrained by the

evaporator and the auger and formation of the ice plug. This results in the com-

pression of the solid and nucleation of cracks. The nucleation of cracks under a

compressive stress is generally due to dislocation pile-up at the grain boundaries

and relief of stress concentration along the grain boundaries. The phenomenon of

crack nucleation has been discussed for low to moderate loading conditions by

Sanderson [14]. His studies indicate that crack nucleation is determined by the de-

layed elastic strain. The generated cracks as well as the cracks initially developed

in the solid ice propagate within the ice body and eventually fracture the solid

block. Block decomposition occurs almost instantaneously and can be considered

as an explosive destruction. The ice temperature in this region ranges from –15°C

to –150°C (distance from water inlet level D = 10-24 mm).

The following hypothetical mechanism of ice formation and decomposition

(Figure 4) was suggested. Liquid water freezes and forms ice plug in the bottom

zone of the evaporator. Thermal expansion essentially stops ice advance, and de-

velops intensive stresses within the plug. This brings about formation and devel-

opment of cracks and eventually fracturing the solid ice plug into particles. The

density of cracks developed in the course of freezing determines the size of parti-

cles. This density, in turn, depends on the rate of water cooling.

This explains the effect of the rate of water supply on the size of particles. At a

low rate of water supply, i.e. at a high rate of freezing the fine particles are gener-

ated. The periodic mode of formation and decomposition of ice blocks determines

the periodical character of driver current oscillations. Indeed, the ice powder exits

the icemaker periodically and the periods of ice exit exactly coincide with those

of the driver operation shown in Figure 5. Before plug formation water and ice

flow freely along the auger and thus loading of the driver is minimal. As long as

the plug stops the ice motion, the torque of the driver increases dramatically, but

drops almost instantaneously when the plug is decomposed (Figure 5). After the

plug decomposition, the geometry of particles does not change and the heat ex-

change results in ice cooling only.


Figure 4. Hypothetical schematic of ice plug formation and decomposition in an auger spiral move-ment: (a) bottom or solidification zone of the evaporator, (b) middle or supercooling zone of the evaporator, (c) ice plug decomposition zone, and (d) ice particles cooling zone.

Figure 5. The correlation between the driver current with time for different water flow rates. Notice distinct frequencies of maximum current value for different water flow rates.


5. CHARACTERIZATION OF GENERATED ICE PARTICLES

Our previous work showed the possibility of using ice particles for a wide variety of

surface processing operations. However, it was found that each specific technology

required a narrow range of particle sizes and temperature. Derusting of steel, for ex-

ample, requires particles ranging from 3 to 7 mm, while the biomedical applications

require highly homogeneous ice powder in the range 0.25 mm-0.3 mm.

Figure 6. Size distribution of ice particles for two different cooling media: (a) the average diameter of ice particles as a function of a temperature at the evaporator outlet, and (b) the average diameter of ice granules as a function of water flow rate.


The proposed mechanism of particles formation enables us to design a technol-

ogy providing a desired kind of particles. In our experiments the control strategy

was developed by properly selecting the cooling medium and water flow rate. The

effects of the cooling medium (refrigerant Galden HT-55 vs. liquid nitrogen) on

the particles size are shown in Figure 6 (a), while Figure 6 (b) shows the effect of

water flow rate on this parameter. It must be pointed out that the exit particle

temperature is also an important operational parameter, because it determines the

stability and hardness of particles.

The actual entrainment of stable low temperature particles is shown in Figure

7. The streams containing low and high concentrations of ice particles are de-

picted in this figure.

6. SELECTED APPLICATIONS OF ICE PARTICLES

A series of the experiments were carried out in order to demonstrate the potential

application of the ice-air jet for various surface-processing operations.

6.1. Biomedical applications of ice-airjet (IAJ) technology

The experiments were conducted on two distinct types of skin, the chicken skin

and the pigskin. The paint was deposited on the skin (Figs. 8 a, b left) in question

by a waterproof marker. Then the IAJ was used to remove this paint. The feasibil-

ity of the paint removal without damaging the underlying layers as well as a

selective removal of the epidermis layer of the skin without damaging the under-

neath layers was demonstrated (Figs. 8 a, b right). The removal was performed

without disturbing the skin structure as well as without creating a temperature

gradient in the impingement zone.

Figure 7. Pictures show the high and low density ice-airjet streams (air pressure is 0.544 MPa) and ice particles flow rate: (a) 20 g/min, and (b) 60 g/min.


Figure 8. (a) Waterproof marker was partially removed from the epidermis layer of pork skin. Then the epidermis layer was removed too. No damage to the underneath layers was detected, and (b) wa-terproof marker was removed from the highly sensitive surface of chicken skin. No damage to the epidermis was observed in course of the cleaning procedure. The pictures were taken with a Sony MVC-FD71 digital camera. Note: the marker (a) was removed without damaging the skin epidermis layer (b).

6.2. Decontamination of heavily contaminated machine parts

The deposit consisted of a mixture of dry grease and dust and moderately adhered

to the substrate (Fig. 9 a left). A selected area of the part surface was decontami-

nated (Fig. 9 a right). The visual inspection confirmed the cleanness of the gener-

ated surface.


Figure 9. (a) Decontamination of a heavily contaminated machine part, and (b) removal of highly adhesive glue layer from a plastic surface. The pictures were taken with a Sony MVC-FD71 digital camera. Note: the highly adherent layer of grease (Fig. 9 (a)) was removed without damaging the painted surface of the machine part (right hand side picture). The residue of glue (Fig. 9 (b)) was removed in course of IAJ cleaning. No damage to the plastic surface was observed (right hand side picture).

6.3. Removal of an highly adherent layer

Two plastic discs were glued together with a highly adherent glue (Fig. 9 b left).

Then the glue remaining on the plastic disc surface was removed by the IAJ (Fig.

9b right). No surface damage was found. Notice that it was not possible to remove

this deposit using mechanical means. Another example involved removal of a thin

layer of fresh rust formed from the steel surface. (Figs. 10 a and b).


Figure 10. (a) Rusted carbon steel surface. Notice that newly formed rust layer is highly adherent, and (b) carbon steel plate was partially derusted using IAJ (middle part of the plate). The pictures were taken with a Sony MVC-FD71 digital camera.


7. CONCLUSION

The suggested mechanism for the formation of ice powder enables us to design an

effective industrial scale device for ice jet formation. The procedure for ice parti-

cles formation was developed and a device readily available for industrial de-

ployment was constructed. An extensive use of this device is envisioned.

Acknowledgement

The study was supported by NSF grant number DDM9312980.

REFERENCES

1. C. Schlosser, L. Muelle and G. McDougal, US Patent 5,752,39 (1950). 2. S. Hitoshi, Japanese Patent 10137707 A (1996). 3. J. Szijcs, European Patent 0509132B1 (1991). 4. M. Tomoji, Japanese Patent 04078477 (1990). 5. I. Harima, Japanese Patent 04360766 A (1992). 6. S. Vissisouk, US Patent 5,367,838 (1994). 7. T. Mesher, US Patent 5,607,478 (1997). 8. H. Shinichi, Japanese Patent 09225830 A (1997). 9. R. Niechial, US Patent 5,820,447 (1998).

10. G. Settles, US Patent 5,785,581 (1990). 11. B. Herb and S. Vissisouk, Proc. Precision Cleaning 1996 held in Anaheim CA, pp. 172-179

(1996). 12. B. Liu, in Jetting Technology, H. Louis (Ed.), pp. 203-211, Professional Engineering Publishing

Ltd., London, UK (1998). 13. P. Hobbs, Ice Physics, Clarendon Press, Oxford (1974). 14. B. Sanderson, Ice Mechanics: Risks to Offshore Structures, Graham & Trotman, London, UK

(1988).


Ed. K.L. Mittal

© VSP 2003

Cleaning with solid carbon dioxide pellet blasting

FRED C. YOUNG∗

Cold Jet, Inc., 455 Wards Corner Road, Loveland, Ohio 45140, USA

Abstract—Blasting with solid carbon dioxide (dry ice) pellets is a technology that is gaining wide acceptance in industry and the military for removing coatings and contaminants from surfaces. Dry ice pellet blasting is also being used to prepare surfaces prior to applying coatings. This paper ex-plains the principles of dry ice pellet blasting, and includes a brief history of the development of the technology. A discussion follows to explain how dry ice pellet blasting works by combining kinetic energy with thermal shock. The two kinds of dry ice pellet blast systems that are commercially available, the induction (venturi) and direct acceleration types, are also discussed. The operating principles and performance characteristics of both types are explained and compared in detail. Blast-ing system control parameters, such as dry ice pellet flow rate, compressed air propellant flow rate, and dry ice pellet density, are defined and their effects on cleaning performance are presented. To conclude, the author provides a review of actual applications for dry ice pellet blasting technology as it is currently used in the semiconductor manufacturing industry.

Keywords: Dry ice; carbon dioxide pellets; blasting; cleaning.

1. INTRODUCTION TO DRY ICE PARTICLE BLAST CLEANING

Dry ice is made from liquid carbon dioxide, a recycled byproduct of several

manufacturing processes. During the blasting process the dry ice sublimates to

carbon dioxide gas, just like that exhaled by humans and found naturally in our

atmosphere. Using dry ice is safe for employees and the environment. Dry Ice

blasting uses extruded dry ice pellets, roughly the size of a grain of rice. The pel-

lets are made on a dry ice extrusion machine called a “pelletizer” or “nuggetizer”.

The pellets can be produced, stored in sealed insulated containers, shipped, and

used for blasting several days after they are produced.

Dry ice blasting accelerates solid carbon dioxide pellets with compressed air in a

subsonic or supersonic blast stream to remove unwanted surface contaminants.

Upon impact with the surface the dry ice sublimates (turns from a solid to gas

without passing through a liquid phase) into carbon dioxide. The process is dry and

non-conductive, non-abrasive and non-toxic, leaving no residue on the part or

equipment being cleaned. All that remains to be collected (by vacuuming and/or

∗Phone: 513-831-3211 ext. 367, Fax: 513-831-1209, E-mail: [email protected]

F.C. Young 152

filtration) is the surface contaminant being removed. Dry ice blasting leaves no

secondary waste as from sand, bead or water blasting. This allows equipment sur-

faces to be cleaned in-place during the manufacturing process. The absence of a

secondary waste stream makes dry ice blasting a perfect non-polluting technology.

2. THE PRINCIPLES BEHIND DRY ICE PARTICLE BLAST SURFACE

CLEANING

With a low temperature of –79°C, dry ice (solid carbon dioxiode) has an inherent

thermal energy ready to be tapped. In addition to the kinetic energy associated

with any accelerated medium blasting, dry ice blasting uses the inherent low tem-

perature to increase shear stress in the surface coating or contaminant, enabling

the particle impact to break-up the coating. Further, the thermal gradient between

two dissimilar materials (the contaminant and the substrate) with different thermal

expansion coefficients can serve to break the bond between the two materials. The

ability of these surface mechanisms to remove the coating or contaminant varies

depending on coating or contaminant. Thermal shock is most evident when blast-

ing a thin, non-metallic coating or contaminant bonded to a metallic substrate.

Thermal shock, a key element that makes dry ice blasting an effective cleaning

method, does not cause thermal stress in the substrate being cleaned. The tem-

perature decrease caused by dry ice blasting is localized at the surface where the

contaminant is bonded to the substrate. (See references [1, 2].)

3. A BRIEF HISTORY OF DRY ICE PARTICLE BLASTING TECHNOLOGY

DEVELPOPMENT

In the early 1930’s, the manufacture of solid phase carbon dioxide (dry ice) be-

came possible. During this time, the creation of dry ice was nothing more than a

laboratory experiment. As the procedure for making dry ice became readily avail-

able, applications for this innovative substance grew. Obviously, the first use was

in refrigeration. Today, dry ice is widely used in the food industry for packaging

and protecting perishable foods.

In 1945, stories exist of the U.S. Navy experimenting with dry ice as a blast

medium for various degreasing applications. In May 1963, Reginald Lindall re-

ceived a patent for a “method of removing meat from bone” using “jetted” carbon

dioxide particles. In November 1972, Edwin Rice received a patent for a “method

for the removal of unwanted portions of an article by spraying with high velocity

dry ice particles”. Similarly, in August 1977, Calvin Fong (then working for the

Lockheed Corp.) received a patent for “sandblasting with pellets of material ca-

pable of sublimation”. The work and success of these early pioneers led to the

formation of several companies in the early 1980’s that pursued the development

of dry ice blasting technology. (See reference [3].)

Cleaning with solid carbon dioxide pellet blasting 153

In recent years, dry ice pellet blasting has found its most commercially success-

ful “market niche” for in-line (in-process) cleaning of molds (rubber, plastics,

aluminum foundry, food baking). Other emerging areas of successful commercial

application are with specialized contract cleaning services, the wood and paper

industries, the semiconductor manufacturing industry, the printing industry, and

the aerospace industry.

4. HOW CARBON DIOXIDE PARTICLE BLASTING WORKS

(SEE REFERENCES [4–7])

4.1. Overview

Carbon dioxide pellet blasting uses compressed air to accelerate frozen carbon di-

oxide “dry ice” pellets to a high velocity. A compressed air supply between 415

kPa and 620 kPa pressure is required. Dry ice pellets can be made on-site or sup-

plied. The pellets are made from liquid carbon dioxide, which is a naturally occur-

ring compound that is non-toxic, non-flammable and chemically inert. Carbon di-

oxide is inexpensive and easily stored at work sites.

4.2. The principal factors contributing to cleaning performance

Carbon dioxide pellet blasting works because of two factors: pellet kinetic energy

(velocity) and thermal shock (temperature). The performance of solid carbon di-

oxide blasting for surface cleaning is optimized by combining these factors and

tuning the parameters of the system specifically for the application. These pa-

rameters are compressed air pressure, type of blast nozzle, pellet size and density,

and the pellet flow rate.

4.2.1. Kinetic energy

High pellet kinetic energy is achieved by using high velocity supersonic nozzles

that are shaped properly to aim directly at the surface of the mold or other article

being cleaned. The “single-hose – direct acceleration” type of carbon dioxide pel-

let- blasting system provides the very high kinetic energy required to remove the

most tenacious contaminants from most surfaces.

Solid carbon dioxide (dry ice) possesses virtually no “hardness” when compared

to sand, glass beads, or even plastic beads. It is estimated that dry ice possesses a

hardness between 1.0 and 1.5 on the Mohs scale. Lack of true hardness deprives

dry ice of the “chiseling” effect which is the prevailing mechanism in all other

forms of particle blasting. This also explains why dry ice blasting is considered

NON-ABRASIVE to most substrates. Since dry ice cannot chisel and erode away

the surface contaminant or coating, it must rely on extremely high initial kinetic

impact energy to create very high instantaneous shear stresses in the coating layer.

Dry ice particles are like tiny “snowballs” traveling at extremely high velocity, yet

possessing no coefficient of restitution, so that ALL of each individual particle’s

impact energy is completely absorbed by the coating layer. Excessive shear in the

F.C. Young 154

coating layer itself, and at the interface between the coating and the substrate,

causes instant fracturing and separation of coating material from the surface.

In general, if a coating or contaminant is CHEMICALLY bonded to a surface,

dry ice particle blasting will NOT effectively remove the coating. If the bond is

PHYSICAL or MECHANICAL in nature, such as a coating of rubber residue

which is “anchored” into the porous surface of an aluminum casting, then there is a

good chance that dry ice blasting will work. Contaminants which are etched, or

stained into the surfaces of metals, ceramics, plastics, or other materials typically

cannot be removed with dry ice blasting. If the surface of the substrate is extremely

porous or rough, providing strong mechanical “anchoring” for the contaminant or

coating, dry ice blasting may not be able to remove all of the coating, or the rate of

removal may be too slow to allow dry ice blasting to be cost effective.

The classic example of a contaminant that does NOT respond to dry ice blast-

ing is RUST. Rust is both chemically and strongly mechanically bonded to steel

substrate. Advanced stages of rust must be “chiseled” away with abrasive sand

blasting. Only the thin film of powderized “flash” rust on a fresh steel surface can

be effectively removed with dry ice blasting.

4.2.1.1. Induction (venturi) and direct acceleration blast systems – the effect of

the type of system on available kinetic energy

In a two-hose induction (venturi) carbon dioxide blasting system, the medium

particles are moved from the hopper to the “gun” chamber by suction, where they

drop to a very low velocity before being induced into the outflow of the nozzle by

a large flow volume of compressed air. Some more advanced two-hose systems

employ a small positive pressure to the pellet delivery hose. In any type of two-

hose system, since the blast medium particles have only a short distance in which

to gain momentum and accelerate to the nozzle exit (usually only 200 to 300

mm), the final particle average velocity is limited to between 60 and 120 meters

per second. So, in general, two-hose systems, although not so costly, are limited

in their ability to deliver contaminant removal kinetic energy to the surface to be

cleaned. When more blasting energy is required, these systems must be “boosted”

at the expense of much more air volume required, and higher blast pressure is re-

quired as well, with much more nozzle back thrust, and very much more blast

noise generated at the nozzle exit plane.

The other type of solid carbon dioxide medium blasting system is like the

“pressurized pot” abrasive blasting system common in the sand blasting and Plas-

tic Media Blasting industries. These systems use a single delivery hose from the

hopper to the “nozzle” applicator in which both the medium particles and the

compressed air travel. These systems are more complex and a little more costly

than the inductive two-hose systems, but the advantages gained greatly outweigh

the extra initial expense. In a single-hose solid carbon dioxide particle blasting

system, sometimes referred to as a “direct acceleration” system, the medium is

introduced from the hopper into a single, pre-pressurized blast hose through a

sealed airlock feeder. The particles begin their acceleration and velocity increase


immediately, and continue to gain momentum as they travel the length of the

hose. At the end of the hose, the spray nozzle “gun” actually consists of a conver-

gent-divergent nozzle, which exchanges pressure differential across the nozzle for

a huge increase in air and particle velocity. Carbon dioxide particle velocities

have been measured and substantiated in excess of 215 meters per second, and up

to as high as 270 meters per second at the nozzle exit plane. This is accomplished

at less than one third of the flow volume (only 3000 L/min compared to 10,000 or

more L/min) required by the most aggressive two-hose systems. In addition to the

lighter weight and less cumbersome hand held applicator and hose of a two-hose

system, the contaminant removal energy delivered to the surface is considerably

higher than that provided by a two-hose inductive system. Even with solid carbon

dioxide particle blasting, a significant component of the contaminant removal en-

ergy is the kinetic energy per unit of area delivered to the surface. Since kinetic

energy is a function of mass and velocity of the particles, i.e., Ke=1/2 mv2, it can

be seen that a two-fold increase in particle velocity, with equal particle mass and

equal nozzle spray area, effectively increases the impact energy delivered to the

surface by a factor of four. A three-fold particle velocity increase, from 90 to 270

meters per second, increases the blast impact energy nine times that of a two-hose

system.

4.2.2. Thermal shock

Unlike other blast media, the carbon dioxide particles have a very low tempera-

ture of –78°C. This inherent low temperature imparts the dry ice blasting process

with unique thermodynamically induced surface mechanisms that affect the coat-

ing or contaminant to a greater or lesser degree, depending on coating type. Be-

cause of the temperature differential between the dry ice particles and the surface

being cleaned, a phenomenon known as thermal stress fracturing (“fracking”) or

THERMAL SHOCK can occur. As the temperature differential between the coat-

ing and the substrate increases, the thermal shear stresses in the coating increase

and couple with the impact induced stresses to increase the coating removal rate.

A good example is the fouling which occurs on rubber, plastic, and tire curing

molds. This contaminant is a chemical compound created by the interaction of

mold release products and the base polymer under high pressure and temperature.

The contaminant or “fouling” resembles a very thin glass-like material which re-

sponds very readily to the thermal shock effect of dry ice pellet blasting. In fact,

hot molds, at or near the cure temperature of 160°C, can be cleaned three to four

times faster than the same dirty molds at room temperature.

4.3. Cleaning performance control parameters

Similar to abrasive grit blasting technology, dry ice pellet blasting performance, or

“cleaning power”, can be adjusted to meet the needs of the application. The change-

able performance control parameters allow the user to increase cleaning aggression,

as for rubber mold cleaning, or decrease the level of aggression for more delicate

applications, such as cleaning soldering flux from printed circuit boards.

F.C. Young 156

4.3.1. Pellet velocity (v)

It can be increased or decreased by changing the blast air pressure or the type of

nozzle selected. There are as many nozzles as there are applications, and they can

be designed for sub-sonic, sonic, or super-sonic air flow and corresponding lower

or higher pellet velocity.

4.3.2. Pellet size (m)

It can be varied by using 1 mm, 2 mm, or 3 mm diameter extruded pellets. The

pellets always break up into smaller particles as they travel through the blast hose

and nozzle. The larger the pellets that you start with, the larger will be the parti-

cles which exit the nozzle and impact with the surface.

Pellet size can also be varied by selecting a smooth bore or rough (convoluted)

bore blast hose. The rough inside surface of convoluted hose can break up the lar-

ger 3 mm diameter pellets into very fine particle sizes.

4.3.3. Thermal shock

It can always be enhanced by heating the substrate surface or the entire mass of

the substrate. Rubber and tire molds, and baking oven molds are good examples

of starting with a hot substrate.

4.3.4. Thermal shock and kinetic energy

These can be varied also by adjusting the flow rate of the pellets in the blast

stream. In the single-hose system, the radial airlock feeder speed can be precisely

controlled to meter out just the right amount of pellets. Sometimes too much pel-

let flow can cool the coating and substrate too quickly, resulting in a performance

drop. Sometimes a higher pellet flow is needed if the application requires more

kinetic energy than the thermal effect, like removing heavily built-up oil, grease,

and grime from machinery.

5. DRY ICE PARTICLE BLAST CLEANING APPLICATIONS IN THE

SEMICONDUCTOR MANUFACTURING INDUSTRY

Dry Ice particle blasting is emerging as a method of choice for critical cleaning

requirements in the semiconductor industry. The areas in the semiconductor

manufacturing process where dry ice particle blasting is currently being applied

are:

(1) Silicon wafers are polished to a high degree of surface flatness in the early

stages of the diffusion process. The polishing compound is deposited randomly on

the internal surfaces of the polishing machines, then dries as a hard abrasive coat-

ing. Flecks of this abrasive contaminant can fall back onto the surface of newly

inserted wafers, causing deep scratches that cannot be polished out, resulting in

costly scrapping of silicon wafers. Dry ice particle blast has been found to be the

best method to remove the abrasive, dried-on polishing slurry from the polishing

equipment, without damaging the expensive polishing equipment components.


(2) Integrated circuits (ICs, or Chips) are produced in molds where the finished

silicon based integrated circuit is encased in a material called EMC (Epoxy Mold

Compound). The EMC eventually builds up a light deposit on the mold surface

which can cause sticking of the finished ICs in the molds, and in other forms of

IC surface defects. The traditional method used to remove EMC deposition is to

coat the molds with melamine, let it cure, then pull the melamine off the surface.

The melamine adhesively bonds to the EMC deposit. The EMC deposit is pulled

off the mold surface together with the melamine as the melamine is removed. The

major drawback is that the melamine tends to also pull the chromium plating off

the mold surfaces, rendering the molds useless. Carbon dioxide particle blast has

been demonstrated to be very effective in cleaning the IC mold surfaces without

removing the chromium plating or otherwise damaging the mold surfaces.

(3) Silicon wafers are photo-etched as part of the process to cut the wafers into

the individual rectangular or square ICs, and as part of the process to produce the

final surface transistor circuits. A compound called photoresist is deposited on

silicon wafers to mask them in areas where the photo-etching action is not de-

sired. After the photo-etching processes, the photoresist must be removed from

the wafer surface. Carbon dioxide particle blast has shown very promising results

in this area, and a great deal of developmental effort is now underway to bring

this process into widespread use in the semiconductor industry.

(4) A contamination problem arising from the etching process is the outgassing

of compounds that redeposit on the surfaces of the etching equipment (e.g., fix-

tures, insides of chambers). This polymer-like deposit can re-contaminate subse-

quent wafers being etched in this high temperature process. Also, the deposit can

build up so thick that the wafer holding fixtures become unusable. Some of the

etching process fixtures are made of very expensive and delicate materials, like

quartz. Traditional deposit removal methods include soaking in chemical baths

with toxic solutions. Carbon dioxide particle blasting has been found to be very

effective for removing the etching contaminants on quartz and some other sub-

strates.

6. CONCLUSION

Dry ice pellet blasting is a surface cleaning and preparation technology that is

gaining increasing popularity and acceptance each year, primarily as an industrial

mold cleaning process. Industry’s acceptance of dry ice blasting is based on ran-

dom trial-and-error testing by a few companies, for the purpose of verifying ac-

ceptability for their own cleaning applications, and the great willingness of the

general industry to “assume” that the mechanics and physics behind dry ice blast-

ing were well understood, documented, and as easy to apply to any given applica-

tion (like a simple cookbook recipe). In fact, there is very little scientific basis

(through investigation, testing, evaluation, and reporting) to support the assump-

tions behind the theories of how and why dry ice blasting works. Industry has ac-

F.C. Young 158

cepted dry ice blasting for mold cleaning because it has been proven to increase

profitability and improve product quality. In the harsh environment of a typical

foundry or rubber parts molding shop, dry ice blasting is a relatively benign mold

cleaning method compared to sandblasting and other crude forms of abrasive grit

blasting.

More recently, other industries, such as semiconductor manufacturing, radioac-

tive waste decontamination, and aerospace, are beginning to view dry ice blasting

as an improvement over current methods of cleaning and surface preparation, or

even as a potential “breakthrough” technology for use in developing completely

new manufacturing or processing methods. Many of these new potential applica-

tions require the cleaning or preparation of very delicate surfaces, such as thin

metal alloys, silicon wafers, composite materials, and even populated printed cir-

cuit boards. It is for these types of applications that a much better understanding

of the dry ice blast contaminant removal phenomenon is required, so that much

more precise control over the process, and predictability of the outcome of using

the process, can be achieved.

REFERENCES

1. L.C. Archibald, “Cold Jet Thermal and Surface Cleaning Characteristics” (June 1988, The Pro-duction Engineering Research Association of Great Britain, business name PERA). PERA is lo-cated in Melton Mobray, Leicestershire, UK. Telephone 44-664-501501. The information in this report is restricted to Cold Jet, Inc., 455 Wards Corner Road, Loveland 45140, USA. Please contact Cold Jet, Inc. at 513-831-3211 to obtain a copy of this report.

2. K. Lay, “An Analysis of Mold Integrity After Carbon Dioxide Blast Cleaning”, published in the proceedings of the International Tire Exhibition and Conference (ITEC), 1996, Akron, Ohio, USA. (Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio 44313-5251. Telephone 330-836-9180. Fax 330-836-1005)

3. J.A. Snide, “Carbon Dioxide Pellet Cleaning - A Preliminary Evaluation”, Materials & Process Associates, Inc., October 12, 1992.

4. D.R. Linger, “Fundamentals of Dry Ice Blast Cleaning Technology”, published in the proceed-ings of the International Tire Exhibition and Conference (ITEC), 1996, Akron, Ohio, USA. (Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio 44313-5251. Telephone 330-836-9180. Fax 330-836-1005)

5. C. Cundiff, “Evaluation of the Cold Jet, Inc. Carbon dioxide Blast System for Paint Stripping”, Battelle, 505 King Avenue, Columbus, Ohio 43201, USA, October 18, 1989.

6. F. Young, “Blast Off” article published in Tire Technology International Magazine, December, 2000. Pages 54–58. (Available by contacting UK & International Press, Abinger House, Church Street, Dorking, Surrey RH4 1DF, UK. Telephone +44 (0) 1306 743744. Fax +44 (0) 1306 742525. E-mail [email protected])

7. F. Young, “Tire Mold Maintenance with Solid Carbon Dioxide Pellet Blasting”, published in the proceedings of the International Tire Exhibition and Conference (ITEC), 1998, Akron, Ohio, USA. (Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio 44313-5251. Telephone 330-836-9180. Fax 330-836-1005)


Ed. K.L. Mittal

© VSP 2003

Development of a generic procedure for modeling of

waterjet cleaning

K. BABETS and E.S. GESKIN∗

New Jersey Institute of Technology, Mechanical Engineering Department, Waterjet Laboratory,

Newark, NJ 07102-1982

Abstract—A practical procedure for utilization of available information, both numerical and lin-guistic and identification of the operational conditions of the waterjet cleaning is presented. Neural Networks based prediction models were constructed using previously available information. The constructed models constituted knowledge base for the procedure. Then a single parameter, the ero-sion strength for cleaning, was determined experimentally. The fuzzy logic technique enabled us to determine a weighted contribution of each preliminary constructed model for the process in ques-tion. Thus, the first approximations of the operational conditions are determined. In the course of the further operation the developed model is improved. The developed procedure will assist a practitio-ner in the selection of a decontamination technology for an unknown surface.

Keywords: Waterjet; cleaning; soft computing; process prediction; process modeling.

1. INTRODUCTION

An effective material decontamination is one of the major industrial concerns to-

day. It is difficult to imagine a single manufacturing process where material de-

contamination is not involved at some level. The field of material decontamina-

tion includes such vital applications as disinfecting and wound cleaning in

hospitals and extends to road deicing, maintenance of building and bridges, paint

stripping from aircrafts, and so on.

Currently, the most usable approach for material decontamination involves

chemical cleansers. Chemical cleansers are comparatively inexpensive, and in

many cases they are readily available and are extremely effective.

The problem with chemical cleansers is that they are potentially hazardous to

worker’s health and are environmentally unfriendly. These and other problems

with chemical cleansers (such as disposal of used agents, separation of debris

∗To whom all correspondence should be addressed. Phone: (973) 596 3338,

Fax: (973) 642-42882, E-mail: [email protected]


from cleaning agents, etc.) require alternative methods for effective material de-

contamination based on physical coating removal techniques.

Physical coating removal techniques take advantage of differences in physical

properties between the coating and the substrate to destroy the bonding and/or

abrade the coating from the underlying substrate. Physical coating removal tech-

niques use one or more of four general types of physical mechanisms [1].

Abrasive techniques wear the coating off with scouring action.

Impact techniques rely on particle impact to crack the coating to remove it.

Cryogenic techniques use extreme cold conditions to make the coating more

friable and induce differential contraction to debond the coating.

Thermal techniques use heat input to oxidize, pyrolyze, and/or vaporize the

coating.

These techniques include but are not limited to: plastic media blasting, wheat

starch blasting, sodium bicarbonate wet blasting, high pressure water blasting and

cryogenic blasting. It is clear that the water blasting constitutes the most effective

technique. Water is readily available, comparatively inexpensive, and induces no

damage to the environment. A complete separation of water and debris facilitates

material recovery. Therefore, complete pollution prevention is feasible.

Although numerous extensive studies of waterjet-based material cleaning have

been implemented [2-4], and this topic is currently of interest to many research-

ers, there is no one universal technique that will allow practitioners to bridge the

gap between the available information about the process and the current need of a

practitioner to remove a specific contaminant. This difficulty renders the process

unusable for most practical applications. Therefore, a goal of this research was to

develop a modeling tool that would assist in practical implementation of such a

technology.

2. DEVELOPMENT OF A GENERIC MODELING TOOL

The experimental studies of material decontamination enabled us to identify the

range of the application of waterjet technology for surface cleaning as well as to

acquire a database for development of empirical modeling and optimization tech-

niques. The theoretical study resulted in the development of corresponding algo-

rithms and computer codes. However, the ultimate goal of the process investiga-

tion is to provide practitioners with an effective and practical approach for

processing all information available to the practitioner, regardless of the form and

accuracy.

2.1. Determination of the erosion strength

In our work the following approach was used to obtain a generic coefficient to

characterize any combination of substrata and deposits. Following the results

Development of a generic procedure for modeling of waterjet cleaning 161

available in the literature we define an area cleaning efficiency, Ea, as the ratio of

area cleaned per unit time and power delivered by the nozzle:

(1)

where area cleaning rate A is in m2/hour, and

P p Q= ∆ ⋅ (2)

where: ∆p – is the pressure drop across the nozzle,

Q – is the flow rate;

Thus Ea has the units of [m2/kW-h], that is a unit area cleaned per unit of en-

ergy expended by the nozzle.

The idea of characterizing a material’s ability to resist erosion is far from new.

Thiruvengadam [5] in his studies of cavitation erosion has suggested the notion of

erosion strength that was based on a strain-energy absorption concept. Heymann

[6] has suggested the concept of relative erosion strength. Thus utilizing ideas of

Thiruvengadam [5] and Conn [7] relates the area cleaning rate, A , the erosion

strength for cleaning Sc and the erosive intensity I, for a given waterjet nozzle and

fixed set of waterjet parameters (water pressure, traverse rate, angle of impinge-

ment, standoff distance, etc.,) as:

(3)

or

(4)

Combining expressions (1) and (4) results in

1

a cE S

−∝ (5)

The relation (5) is the basic relation, used to derive the curves, representing the

dependence of the area cleaning efficiency Ea and erosion strength for cleaning Sc

(Figure 1).

2.2. Determination of the erosion strength based on the available cleaning examples

The available experimental database reflecting material decontamination with

pure waterjet was compiled, and the area cleaning efficiencies were calculated for

deposit types given in Table 1. The data in Table 1 were used to derive the rela-

tionships between area cleaning efficiency Ea and erosion strength for cleaning Sc

Area cleaned per unit time

Power delivered by nozzlea

AE

P= =

c

IA

S=

c

a

SP

IE

⋅=


(Figure 1). The point of departure for the line 276-310 MPa in Figure 1 was the

data for Item 1 from Table 1. The calculated cleaning efficiency for removal of

Epoxy #1 deposit was assigned an Erosion Strength Sc=103, relative units. This

line, per relation (5) was plotted on a log-log chart. The location of the rest of the

data points was based on calculated area cleaning efficiencies and working water

pressure.

To derive line for 138 MPa (Figure 1) items 1 and 2 were compared. Since the

Erosion Strength for the deposit type Epoxy #1 is known (Sc=1000 relative units)

and is constant, the first point for line 138 MPa thus could be located. Similarly to

derive the line for 70-100 mPa, item #6 was located at the 138 mPa line and com-

pared with item #7.

The main result that could be inferred from the graphical relationship between

the area cleaning efficiency Ea and erosion strength Sc shows that there is a defi-

nite relationship between Ea and Sc and that Erosion Strength for cleaning of simi-

lar materials is closely spaced together, and consequently the Sc parameter can

used to characterize an unknown deposit-substrate combination. On the other

Figure 1. Graphical relationship between Ea and Sc.


hand, it should be emphasized that the relations in Figure 1, as presented by Conn

[7] and verified by our experimental studies do not constitute exact relations, at

best they represent an order of magnitude comparison only.

3. DEVELOPMENT OF GENERIC PREDICTION TECHNIQUE

The problem that most of the waterjet practitioners face when dealing with an un-

known surface is the lack of information about the process or, in other words, the

unavailability of a generic technique that could be used as a first approximation of

the process. This section is concerned with the development of such an approach.

The idea behind such an approach is to combine the previous knowledge about

the process in question and based on that make an informed decision as to which

waterjet parameters to apply, as a first approximation.

3.1. Modeling approach

In data / information processing the objective is to gain the understanding of a

complex phenomenon through modeling of the system either experimentally or

analytically. Then after a model of the system has been obtained, various proce-

dures (e.g. sensitivity analysis, statistical regression, etc.) can be used to gain a

better understanding of the system.

Table 1.

Cleaning examples

Item #

Deposit type

Water pressure (MPa)

Nozzle diameter(mm)

Flow rate m3/s

Area cleaning rate (m2/hour)

Power delivered by nozzle kW-h

Area cleaning efficiencing m2/kW-h

1 Hard Epoxy

276 0.305 4.610E-05 0.52 12.71 0.0408

2 Hard Epoxy

138 0.305 3.260E-05 0.04 4.50 0.0085

3 Hard Epoxy

103 0.3556 3.837E-05 0.02 3.97 0.0048

4 Rust 310 0.254 3.391E-05 0.99 10.52 0.0938

5 Weaker1 Rust

310 0.1778 1.662E-05 0.52 5.16 0.1040

6 Oil Based 138 0.254 2.262E-05 0.69 3.12 0.2201

7 Oil Based 69 0.254 1.599E-05 0.17 1.10 0.1541

8 Weaker Epoxy1

276 0.3554 6.259E-05 1.04 17.26 0.0603

9 Auto Paint1 138 0.254 2.262E-05 0.14 3.12 0.0434

1 Deposits were not used in model development.


There are, however, situations in which the phenomena involved are very com-

plex and not well understood and for which the first principle models are not ef-

fective. Even quite often, experimental measurements are difficult and/or expen-

sive. These difficulties led us to explore the application of Soft Computing

(Artificial Intelligence) techniques as a way of obtaining models based on ex-

perimental measurements. The field of Soft Computing is comparatively new, and

it includes fuzzy logic, neural networks, expert system, cellular automata, chaotic

systems, wavelets, complexity theory, anticipatory systems, among others. But

only fuzzy logic, neural networks and genetic algorithms have reached the stage

of development where they are used for real world problems.

Fuzzy logic systems address the imprecision of the input and output variables

directly by defining them with fuzzy sets (fuzzy numbers) that are generally ex-

pressed in linguistic terms. Moreover, they allow for very complex and nonlinear

systems to be described in very simple terms, thus making them easier to under-

stand. Another important feature of fuzzy systems is their ability to accommodate

the existing expert knowledge of a process into a model by expressing it in terms

of fuzzy rules.

Neural Networks, on the other hand, model a system by using sets of input-

output data to train some generic model of the system. Neural Networks are very

good at modeling very complex nonlinear relationships with large numbers of in-

put and output variables. Models based on neural networks are also easy to opti-

mize, since although the model itself is not given in terms of an explicitly defined

function, the gradient of this function can be found numerically.

The combination of the above two techniques often results in greater flexibility

and/or clearer representation of the model than when they are used separately.

This combination is often referred to as neuro-fuzzy model of the system. Neuro-

Fuzzy Reasoning approach also allows overcoming some traditional problems in

using fuzzy logic or neural networks, such as the problem of defining a member-

ship function, extracting fuzzy rules, etc.

We are using the notion of Erosion Strength (Sc) developed in the previous sec-

tion to classify an unknown surface together with Neural Networks Fuzzy Rea-

soning technique, suggested by Takagi and Hayashi [8], for information process-

ing.

The prediction technique construction begins with the development of fuzzy

universe for erosion strength, Sc. The experimental database allows us to construct

three fuzzy sets, based on the number of experimental situations available. The

hard epoxy deposit with erosion strength of 1000 represents fuzzy Class I in Fig-

ure 2. Using the notation of the Fuzzy Logic theory we state that the hard epoxy

deposit has the degree of membership of one in the fuzzy set Class I, or, in simple

terms, this deposit is the most representative of all deposits that might be classi-

fied as belonging to Class I. Similarly, the rust deposit, with erosion strength of

400 is assigned the degree of membership of one in the fuzzy set Class II, and, fi-

nally, the oil based paint deposit is assigned the degree of membership of one in

the fuzzy set Class III. To explain the idea of “degree of membership” we refer


the reader to Figure 2. Carefully inspecting this figure we notice that the classes

III and II, and II and I overlap to some degree. This means that some random de-

posit, with erosion strength of say 100, will belong to both class III and class II.

The extent to which this random deposit type is represented by either of the

classes is expressed in terms of the “degree of membership”. Thus, from Figure 2

we notice that a deposit with erosion strength of 100 belongs to Class III with a

degree of membership of 0.8 and at the same time belongs to Class II with a de-

gree of membership 0.2. The higher the degree of membership in a class, the more

representative this class is for a given deposit type. Clearly, if a deposit has a de-

gree of membership of 1.0 in some class, it belongs only to that particular class,

and to no other class.

Thus each of the classes in Figure 2 is represented (with a degree of member-

ship 1) by a specific deposit type available in our experimental database. In other

words, these three fuzzy sets cover the ranges of all possible values for the mate-

rials with erosion strength for cleaning from 1 to 10000 relative units. Similarly, if

we identify a deposit-substrate combination with some value of erosion strength,

Sc, and this value happens to be inside the range [1,10000], then we can identify

the degree of membership of such a deposit-substrate combination in the three

fuzzy sets (Figure 2). This procedure alone can be very useful when trying to

classify some unknown deposit-substrate combination.

Figure 2. The fuzzy universe for erosion strength, Sc.


It should be emphasized that for each of the three basic deposits (hard epoxy

paint, rust, and oil based paint) an extensive experimental study was performed.

From this point on we will refer to these deposits as “base deposits” and equiva-

lently we will refer to the three fuzzy sets (classes) they represent as “base

classes”. Since we possess a required empirical knowledge for these processes,

the appropriate numerical representation of each process can now be made using

artificial neural networks. The procedure for the application of a neural network

for process modeling and optimization was described in [9]. Thus, after each net-

work has been created, properly trained and tested we have obtained reliable nu-

merical models for the three base deposits, or equivalently, three fuzzy base

classes of erosion strength (Sc). Therefore, it can now be stated that the process of

cleaning a material with erosion strength in the range [1,10000] can now be ap-

proximated as some combination of the numerical models of the three base

classes. The computational procedure is as follows. For an unknown deposit, the

practitioner makes a simple experiment that allows him/her to calculate area

cleaning efficiency. Then using the computed Ea we can determine the corre-

sponding erosion strength (Sc) for this surface from Figure 1, for the given water

pressure. Once a corresponding Sc coefficient has been found, the fuzzy member-

ship in the three classes in Figure 2 can be determined. Separate experiments were

conducted for the auto paint deposit removal with plain waterjet. The computed

area cleaning efficiency was calculated as Ea auto paint = 0.04 m2/kW-h. From Figure

1 the corresponding coefficient Sc was found to be Sc=180 relative units. And

from Figure 2, the degree of membership (µ) in the three basic classes can be cal-

culated as µ (class I) =0, µ (class II) =0.4, µ (class III) =0.59. These degrees of membership

can be interpreted as follows. The erosion strength of the auto paint deposit is ap-

proximately midway between that of the hard epoxy paint and rust deposit. Now

that the surface was identified we could use the numerical models available for

the three base deposits (hard epoxy, rust, and oil based paint) to obtain a first ap-

proximation to the process. We supply a set of input parameters (Water Pressure,

Nozzle Traverse Rate, Nozzle Diameter, and Standoff Distance) as an input into

the numerical models represented by the neural networks. The corresponding out-

put in terms of the single strip width is obtained by each of the network. The final

result is obtained by defuzzifing the output according to Equation (6).

(6)

In equation (6), µ is the membership value of the deposit with erosion strength

Sc in the three base classes, Us is the output of the sth neural network, and y* is the

final defuzzified output.

n ... 1,2i ,

)x(

)x(u)x(

yr

1s

iA

r

1s

iSiA

*

i

S

S

=µ

⋅µ=

=

=


The output of the model is the single clean width of the strip produced on the

surface by the combination of the input parameters, which then can easily be con-

verted into area cleaning rate. This procedure is sketched in Figure 3.

4. EXPERIMENTAL VERIFICATION OF PERFORMANCE

In order to experimentally verify the suggested modeling approach an additional

experimental database was acquired. The experimental samples consisted of three

types of deposits – auto paint, weaker rust, and weaker epoxy paint, (items 5, 8, 9,

Table 1). Waterjet parameters varied in these additional experiments were limited

to the water pressure, nozzle traverse rate, standoff distance, and nozzle diameter.

The experimental setup and procedures were similar to those described in the pre-

vious paper [10]. Area cleaning efficiencies for removal of these deposits were

used to identify the corresponding erosion strength coefficient from Figure 1, and

the degrees of membership of these deposits in the three basic classes were identi-

fied using Figure 2. Table 2 shows the results for the test deposits along with the

deposits representing the basic classes.

Figure 3. Generic modeling approach. NN1, NN2, NN3 – Artificial Neural Network models for the three basic classes. X1, X2, …, Xn – Process input variables.


Table 2.

Cleaning samples

Degree of membership

Sample number

Deposit Area cleaning efficiency (m2/kW-h)

Erosion strength (relative units)

Class I Class II Class III

1 Hard Epoxy Paint 0.04 1000 1 0 0

2 Weaker Epoxy Paint

0.06 665 0.41 0.58 0

3 Rust from Steel 0.0975 400 0 1 0

4 Weaker Rust from Steel

0.105 360 0 0.89 0.1

5 Auto Paint 0.04 180 0 0.4 0.59

6 Oil Based Paint 0.21 30 0 0 1

The model performance was tested on each of the test deposits by providing the

model with a set of waterjet input parameters within the working space, obtaining

the corresponding output in terms of the width of a clean strip and comparing the

results with experiments.

5. DISCUSSION

The model for prediction of the results of waterjet cleaning described in the pre-

vious sections was tested on several additional test deposits. Figures 4-9 present

the results of prediction. Analyzing the results, it is clear that the model prediction

results are acceptable at both relative error of prediction (~ 20%), and at following

the trend of the process, which is also important for any cleaning study. Neverthe-

less, as a first estimation of a cleaning efficiency for a given type of deposit, these

results constitute a reasonable approximation.

However, it should be noted that at the current stage the prediction technique

was tested only in the middle of the problem space. At the outskirts of the prob-

lem space, veritable results could not be obtained. The reason for this lies in the

limitations in the development of the three numerical models that represent the

cleaning of the base deposits (hard epoxy, rust, and oil based paint). Since the

ranges of experimental parameters used for the construction of models were dif-

ferent in each case, and there was no coordinated experimental setup, but rather

the data were compiled at later stages, there are inconsistencies in choosing the

levels of process parameters in case of a test cleaning space. For example, if for a

base model development the nozzle traverse rate was in the range from 1000

mm/min to 2500 mm/min, and a test case was run at 1500~4000 mm/min, the re-

liable model performance will be at the intersection of these ranges. The way to


cure the above problem is to cover all the parameter space (limited by equipment

capabilities) for each process variable in all the base models. Of course, this re-

sults in quite extensive experimentation, but on the good side it needs to be done

only once, when developing the base models. Also the current model does not

cover the full range of all possible erosion strengths of different materials, but by

extending the procedure with additional base models for lower or higher degrees

of erosion strength for cleaning (Sc), this limitation can be reduced or eliminated.

Figure 4. Auto paint removal, standoff distance 13.9 cm. WP – water pressure, ND – nozzle diame-ter, standoff – stand off distance.

Figure 5. Auto paint removal, width of strip vs. traverse rate. Experimental vs. predicted.


Figure 6. Removal of weaker rust. Experimental vs. predicted width of clean strip. Water pressure 241 MPa, nozzle diameter 0.1778 mm.

Figure 7. Experimental vs. predicted width of clean strip. Water pressure 172 MPa, nozzle diameter 0.1778 mm.

6. CONCLUDING REMARKS

The approach presented here for modeling of waterjet cleaning process allows a

user to obtain a reliable process approximation given no or limited information

about process condition. For an unknown surface a practitioner needs to deter-

mine a single coefficient, the erosion strength for cleaning (Sc), based on a simple

experiment(s). The proposed approach utilizes this coefficient and approximates


the cleaning results in terms of area cleaning rate. It is believed that the current

work will assist in practical implementation of waterjet cleaning technology,

where the information deficiency on process conditions is the main reason for in-

effective application of this technology.

Figure 8. Removal of weaker epoxy paint. Nozzle diameter 0.356 mm, water pressure 207 MPa.

Figure 9. Removal of weaker epoxy paint. Nozzle diameter 0.3556 mm, water pressure 276 MPa.


REFERENCES

1. United States Environmental Protection Agency, Office of Research and Development. Guide to

Cleaner Technologies, EPA/625/R-93/015,Washington D.C. (1994). 2. A.F. Conn and G. Chahine, Proc. Third American Waterjet Conference, Pittsburgh, PA, Water-

jet Technology Association, St. Louis, MO (1985). 3. F. Erdman-Jesnitzer, A.M. Hassan and H. Louis, Proc. Third International Symposium on Jet

Cutting Technology, Chicago, IL, British Hydraulic Research Association, Cranfield, UK (1976).

4. S.T. Johnson, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Tech-nology Association, St. Louis, MO (1993).

5. A. Thiruvengadam, in Erosion by Cavitation or Impingement, STP No. 408, 22-36, ASTM, Philadelphia, PA (1966).

6. F. Heymann, in Characterization and Determination of Erosion Resistance, STP No. 474, 212-244, ASTM, Philadelphia, PA (1969).

7. A. Conn, Proc. Fourth American Waterjet Conference, Berkeley, CA, Waterjet Technology As-sociation, St. Louis, MO (1987).

8. H. Takagi and I. Hayashi, Intl. J. Approximate Reasoning, 5, No. 3, 191-212 (1991). 9. K. Babets, E.S. Geskin and B. Chaudhuri, Proc. 10th American Waterjet Conference, Houston,

TX, Waterjet Technology Association, St. Louis, MO (1999). 10. K. Babets, E.S. Geskin, Intl. J. Machining Sci. Technol., 4, No. 1, 81-101 (2000).


Ed. K.L. Mittal

© VSP 2003

Experimental and numerical investigation of waterjet

derusting technology

K. BABETS, E.S. GESKIN∗ and B. GOLDENBERG

New Jersey Institute of Technology, Mechanical Engineering Department, Waterjet Laboratory,

Newark, NJ 07102-1982

Abstract—The study is concerned with the development of effective technology for derusting of a steel surface. We have investigated the surface derusting by high-speed waterjet and determined the optimal operational conditions. This investigation involved topographical and metallographical studies of the substrate surfaces and subsequent classification of the substrates with respect to the degree of rust development. Then the rust was removed by a moving waterjet at various impact con-ditions and the generated surfaces were examined. Soft computing techniques were used to select the optimal conditions for rust removal. Due to the extremely chaotic and fuzzy nature of input in-formation the advanced numerical procedure based on the Neural Network Driven Fuzzy Reasoning was employed. As the result, the realistic procedure for steel derusting was found and a practical technique for process design was suggested.

Keywords: Derusting; fuzzy reasoning; neural network; soft computing; waterjet.

1. INTRODUCTION

The corrosion of metal structures poses a serious technological and economical

problem. It shortens the life span of the steel parts and deteriorates dramatically

their performance.

Corrosion is a chemical or electrochemical process in which surface atoms of a

solid metal either react with or dissolve in a substance that contacts the exposed

surface. Corroding media are generally classified as aqueous or non-aqueous. The

rate of steel corrosion in the atmosphere depends on geographical location, and

can reach 1070 µm/yr. When rust depth reaches 1% of the thickness of the steel,

the strength of the steel reduces by 5-10%. Throughout the world steel corrosion

annually equals to 20-40% of its annual production [1]. The corrosion of all car-

bon steels is most devastating when the metal is subjected to an alternately wet

and dry atmosphere in the presence of chloride salts. Typically this environment

can be encountered on the underbodies of automobiles and trucks. The most con-

∗To whom all correspondence should be addressed. Phone: (973) 596 3338, Fax: (973) 642 4282,


K. Babets et al. 174

ventional way of rust prevention is to coat the metal surface. But prior to the coat-

ing, a complete cleanliness of the surface must be assured to ensure that a surface

is free of any rust.

The conventional methods of rust removal involve acid cleaning or sandblast-

ing. These techniques, though proven to be effective, can be environmentally haz-

ardous. Some new derusting techniques, such as use of rust neutralizers converts

rust into a chemically neutral surface, leaving the surface ready for coating appli-

cation. However, the use of these neutralizers is limited to oil-based types only,

and the chemicals contained in these products can be detrimental to the worker’s

health.

Waterjet surface derusting constitutes rather an efficient way to clean steel sur-

faces. The following experimental study was concerned with optimization, or at

least improvement, of jet based derusting technology.

2. EXPERIMENTAL SETUP

2.1. Experimental procedure

The derusting experiments were carried out at the Ingersoll-Rand waterjet system

(Fig. 1). The nozzle head was mounted on a 3-axes gantry robot whose move-

ments were guided by an Allen Bradley 8200 series CNC controller.

The major obstacle in the experimental study of a derusting technology is the

extreme diversity of the rusted surfaces. It is difficult to find several samples with

Figure 1. Waterjet setup.

Experimental and numerical investigation 175

a similar degree of rusting. As a result, a rather low reproducibility of the experi-

ment is obtained. In order to at least partially overcome this problem we used the

ISO developed standard for classification of the rusted steel surfaces. This stan-

dard specifies four rust grades, designated A, B, C and D [2]. The rust grades are

defined by written descriptions together with representative photographic exam-

ples. The selected steel samples were sorted according to visual similarity and

compared to the representative photographs. Those identified as the class C were

used as experimental samples. In the above-mentioned ISO publication, C is de-

fined as the “steel surface on which the mill scale has rusted away or from which

it can be scraped, but with slight pitting visible under normal vision”. Still it

should be stressed that the existent rust grades classification based on visual com-

parison does not provide a reliable procedure for rust identification and thus

makes it quite difficult to collect uniform experimental samples.

In our experiments the effects of water pressure, traverse rate and nozzle di-

ameter on cleaning effectiveness and surface quality were investigated. The tests

were run at water pressures of 310, 241, 172, 69 MPa (i.e. 45,000, 35,000, 25,000

and 10,000 psi). The water nozzles with diameters 0.127, 0.1778, 0.254, 0.3556

mm were used. In these experiments the effect of the standoff distance (the dis-

tance between the nozzle exit and the sample) as an independent process parame-

ter was not investigated. Instead, the ratio of nozzle diameter to the nozzle stand-

off distance was kept constant. Thus a number of standoff distances were tested to

find a near optimum value for a selected nozzle diameter. Then the obtained ratio

was kept constant for the other nozzle diameters. The study was carried out at the

traverse rates of 635, 2540, 7620 and 12700 mm/min. The upper bound of the

nozzle traverse rate (12700 mm/min) was imposed by the equipment limitations.

In order to study the effect of the waterjet parameters on the surface, a full fac-

torial experimental design was employed. In such a design one process variable is

tested at its different levels, while the other variables are held fixed at some level.

The experimental procedure involved the following steps. For each cleaning situa-

tion (i.e. the combination of water pressure, nozzle traverse rate and nozzle di-

ameter / standoff) the width of clean strip was measured with Mitutoyo Toolmak-

ers Microscope and recorded. These values of strip width (STW) were then used

to calculate process effectiveness (Rate of Area Cleaned) and specific water con-

sumption according to the following expressions:

Rate of Area Cleaned (m2 min

–1) = Traverse Rate * Width of Cleaned Strip (1)

( ) ( )122

3 2

2

Water Consumption / 4 Rate of Area Cleaned

Dw

PC D

m m

π ρ⋅⋅ ⋅ ⋅

=⋅

(2)

where ρw is the water density, CD is the discharge coefficient of the waterjet ori-

fice, whose diameter is D. In present work CD is taken to be 0.7 [3], P – waterjet

pressure.


Then several nozzle passes with 25% overlapping were made at the same op-

erational conditions. The resulting derusted area was evaluated visually and pho-

tographed with Olympus photomicroscope with 12 x magnification in order to de-

termine the degree of cleaning.

2.2. Surface examination

In current study, the X-ray diffraction was used to evaluate the presence of oxides

(rust) on the metal surface after rust removal with waterjet. The Siemens D5000

diffractometer with θ-2θ diffractometer geometry at the Stevense Institute of

Technology was used for this investigation. The experimental samples consisted

of rusted-cleaned pairs. The following procedure was used for sample prepara-

tion. First the metal samples were machined to a block 10.16 x 10.16 x 6.350 mm.

Then the samples with similar rust were grouped in pairs. From each pair one

sample was left as it was, and the other one was cleaned of rust using waterjet.

The following waterjet parameters were employed: Water Pressure 200 MPa,

Nozzle Diameter 0.254 mm. Two cleaning runs were made. At first the rust was

removed from the metal surface at a low nozzle traverse rate, and in the second

run the flash rust was removed at the high traverse rate of 3175 mm/min. Then the

sample was dried in hot air. Each pair was evaluated for the presence of oxides by

the diffractometer. The resulted diffraction patterns enabled us to compare the ox-

ides content on the samples before and after the waterjet treatment.

3. EXPERIMENTAL RESULTS

3.1. Surface classification

The quality of derusting by waterjet was evaluated in accordance with ISO stan-

dards (ISO 8501-1:1988). This standard defines four grades of cleanliness of the

surfaces generated by jet derusting. These surfaces are termed Sa 1, Sa 2, Sa 2 ½

and Sa 3. The qualitative description of each grade along with the representative

photographs of the surfaces are presented. During this experimental study it was

found difficult to follow the ISO classification. Instead, the following “fuzzy”

classification was suggested. According to the developed procedure we divided

the derusted surfaces into two classes: “well cleaned”, and “poorly cleaned”. A

surface is allowed to have a partial degree of membership in both classes. The

class “well cleaned” would roughly correspond to ISO grades Sa 3 and Sa 2.5,

while the class “poorly cleaned” would correspond to surface grades Sa 2.5, Sa 2

and Sa 1. Figure 2 depicts a typical well-cleaned surface. Here two shades of

green can be distinguished. Light green corresponds to the derusted surface, while

dark green corresponds to flash rust, which appears immediately after waterjet

pass. It was found that flash rust could be removed easily by an additional appli-

cation of the waterjet at a high traverse rate, or prevented by immediate drying of

the surface in hot air.


Figure 3 shows a typical “poorly cleaned” surface. Such a surface is free from

lightly adherent mill scale, rust and other contamination, but some firmly adherent

rust remains on the surface. Thus the most representative surface samples were

classified as either belonging to one of these classes, or, not. The “fuzzy” mem-

berships in the two fuzzy classes for the remaining surfaces were determined us-

ing the artificial neural network assisted fuzzy classification method described in

the following section.

Figure 2. Optical photograph of “well cleaned” metal surface (12 x magnification).

Figure 3. Optical photograph of “poorly cleaned” surface (12 x magnification).


3.2. Surface cleanliness

In order to evaluate qualitatively and quantitatively the waterjet-based derusting

technology, several studies were carried out. These examinations included taking

scanning electron micrographs of the surface, performing chemical analysis of the

metal surface, and performing the x-ray diffraction analysis. Figure 4 shows a

rusted surface at 500 x magnification. The main features of the surface are the

oxidized metal grains of different sizes. Figure 5 represents a surface derusted

with waterjet. No oxidized metal grains are observed; the surface is smooth and

visually free of rust.

Figure 4. SEM micrograph of rust covered metal surface.

Figure 5. SEM micrograph of waterjet-derusted metal surface.


In order to estimate qualitatively the effectiveness of derusting, chemical

analysis of the surface prior to and after treatment was carried out. The scanning

electron microscopy was used for this study. The typical results of the analysis are

presented in Figs. 6-7. These figures show that the oxygen content of the surface

Figure 6. Chemical composition of metal surface prior to waterjet rust removal. Oxygen content isat 900 count.

Figure 7. Chemical composition of metal surface after waterjet rust removal. Oxygen content is at 320 count.


was significantly reduced after the water jet treatment. In order to evaluate the de-

gree of derusting, the chemical analysis was supplemented by the x-ray diffrac-

tion. The corresponding diffraction patterns of the metal surfaces before and after

waterjet cleaning are presented in Figs. 8 and 9, respectively. Roughly speaking

each peak in these figures corresponds to a chemical compound. The intensity of

Figure 8. X-ray diffraction analysis. Diffraction pattern of rusted metal surface prior to waterjet treatment.

Figure 9. X-ray diffraction analysis. Diffraction pattern of metal surface after waterjet rust removal.


Table 1.

Results of the metal surface chemical analysis

Chemical compound

Experimental atomic planes spacing values

0

( )d A

Standard tables of atomic planes spacing values

0

( )d A Fe Fe2O3 Fe3O4

3.281 3.24 YES

3.009 2.967 YES

2.795 2.728 YES

2.546 2.532 YES

2.174 2.176 YES

2.106 2.099 YES

2.027 2.03 YES

1.723 1.715 YES

1.624 1.616 YES

1.49 1.485 YES

1.437 1.43 YES

1.289 1.281 YES

1.211 1.212 YES

1.168 1.17 YES

a peak represents the relative amount of this chemical compound. The atomic

spacing values shown just above these peaks allow us to determine the type of the

chemical compound present on the surface. From Fig. 8 and Table 1 it is clear that

a rusted surface in addition to Fe contains significant amounts of Fe2O3, and

Fe3O4. After waterjet cleaning (Fig. 9) the three still remaining peaks represent

Fe, with significantly increased intensity levels. Most of the oxides are no longer

present in the figure, and intensity level of those still present is much lower than

that of Fe content. Moreover, due to low intensity levels these peaks most proba-

bly should be attributed to the noise. The wide base peak at angles 10-25 degrees

in Figs 8 and 9 is due to the presence of the holder clay used to attach the sample

in the holder. Thus, Fig. 9 constitutes a compelling proof of the efficiency of the

waterjet rust removal.

3.3. Effect of water pressure

For each set of operational conditions there is a minimal threshold pressure below

which decoating does not occur. This pressure level depends on the adhesion

strength between the coating and the substrate [8]. The maximum working water

pressure is defined from damage-free cleaning considerations, i.e., where the

cleaning does not result in the damage to the material. In our experiments these

upper and lower pressure bounds were dictated primarily by the equipment capa-


bilities, since at the chosen levels of the nozzle traverse rate and nozzle diameter

the removal of rust could still be performed. Between these two threshold values

of water pressure the clean strip width obtained in a single nozzle pass does not

vary linearly with increasing water pressure (Figure 10). The effectiveness of the

process increases with increasing water pressure (Figure 11).

Figure 10. Experimental clean strip width vs. water pressure for nozzle diameter 0.3556 mm at dif-ferent nozzle traverse rates.

Figure 11. Effectiveness (rate of area cleaned) vs. water pressure for nozzle diameter 0.3556 mm at different nozzle traverse rates.


Figure 12. Water consumption vs. water pressure for nozzle diameter 0.3556 mm at different nozzle traverse rates.

Still the relationship between the effectiveness and water pressure can be ap-

proximated as linear at lower values of the traverse rate. At high nozzle traverse

rates and high nozzle diameters the relationship between the process effectiveness

and water pressure can no longer be considered as linear, but rather as a polyno-

mial. Figure 12 shows that for a large nozzle diameter there exists an extremum of

the water consumption as the pressure increases. This can be attributed to the fact

that the process effectiveness is not a linear function of water pressure (at least in

the considered range of process variables). At small nozzle diameters and for the

pressure range used in these experiments no extremum is seen, although it is rea-

sonable to expect that the extremum will appear at higher values of water pressure.

3.4. Effect of traverse rate

The effect of the traverse rate on rust removal appears to be the most significant.

Figure 13 shows that, as expected, the process effectiveness increases with in-

creasing nozzle traverse rate, while the strip width decreases (Fig. 14). The spe-

cific water consumption (Fig. 15) can be approximated as a power function of

traverse rate. This actually means that there is a range of traverse rates when the

increase in traverse rate results in significant drop in the specific water consump-

tion, while the larger increase in the traverse rate insignificantly reduces water

consumption.

3.5. Effect of nozzle diameter

As expected the increase in nozzle diameter resulted in a higher process effective-

ness (Fig. 16), but also in a higher specific water consumption (Fig. 17). It is in-

teresting to follow the relationship between these important quantities. Let us


Figure 13. Rate of area cleaned vs. traverse rate for nozzle diameter 0.3556 mm and for different water pressures.

Figure 14. Clean strip width vs. traverse rate for nozzle diameter 0.3556 mm and different water pressures (69E – 69 MPa, 172E – 172 MPa, 241E – 241 MPa, 310E – 310 MPa).

consider the experimental results for water pressure 69 MPa, traverse rate 12700

mm/min and nozzle diameters 0.1778 mm (0.007 in) and 0.254 mm (0.01 in).

Here we notice that the area of the second nozzle is almost twice the area of the

first nozzle. The calculated process effectiveness is 0.39 m2/hour and 0.52

m2/hour, respectively, with specific water consumption 0.064 m

3/m

2 and 0.094

m3/m

2. If we now take two small nozzles then the total effectiveness will be 0.78

m2/hour while the water consumption will stay at 0.064 m

3/m

2. Thus it appears

that it would be beneficial to use several small nozzles rather than a big one. This

result is important from practical point of view.


Figure 15. Water consumption vs. traverse rate for nozzle diameter 0.3556 mm and for different water pressures.

Figure 16. Rate of area cleaned vs. nozzle diameter for water pressure 241 MPa at different nozzle traverse rates.

Figure 17. Water consumption vs. nozzle diameter for water pressure 69 MPa and at different noz-zle traverse rates.


4. MODEL OF THE PROCESS

4.1. Choice of modeling technique

In data/information processing the objective is to gain an understanding of a com-

plex phenomenon through “modeling” of the system either experimentally or ana-

lytically. Then after a model of the system has been obtained, various procedures

(e.g. sensitivity analysis, statistical regression, etc.) can be used to gain a better

understanding of the system.

There are, however, situations in which the phenomena involved are very com-

plex and not well understood and for which the first principle models are not ef-

fective. Even more often, experimental measurements are difficult and/or expen-

sive. These difficulties led us to explore the application of Soft Computing

(Artificial Intelligence) techniques as a way of developing models based on ex-

perimental measurements. The field of Soft Computing is comparatively new, and

it includes fuzzy logic, neural networks, expert system, cellular automata, chaotic

systems, wavelets, complexity theory, anticipatory systems, among others. But

only fuzzy logic, neural networks and genetic algorithms have reached the stage

of development where they are used for real world problems [4].

Fuzzy logic systems address the imprecision of the input and output variables

directly by defining them with fuzzy sets (fuzzy numbers), which generally are

expressed in linguistic terms. Moreover, they allow for very complex and nonlin-

ear systems to be described in very simple terms, thus making them easier to un-

derstand. Another important feature of fuzzy systems is their ability to accommo-

date the existing expert knowledge of a process into a model by expressing it in

terms of fuzzy rules.

Neural Networks, on the other hand, model a system by using sets of input-

output data to train some generic model of a system. Neural Networks are very

good at modeling very complex nonlinear relationships with large numbers of in-

put and output variables, and in classification problems. Models based on neural

networks are also easy to optimize, since although the model itself is not given in

terms of on explicitly defined function, the gradient of this function can be found

numerically.

The combination of the above two techniques often results in greater flexibility

and/or clearer representation of a model than when they are used separately. This

combination is often referred to as neuro-fuzzy model of a system. Neuro-fuzzy

approach also allows overcoming some traditional problems in using fuzzy logic

or neural networks, such as the problem of defining a membership function, ex-

tracting fuzzy rules, etc.

Our problem at hand is a good example of a system with highly nonlinear rela-

tionship between process inputs and outputs. The problem of defining the degree

of cleaning is one of the classification types. Therefore, it was found reasonable to

apply an advance artificial intelligence modeling technique based on the combina-

tion of fuzzy logic and artificial neural networks. The method used is known as

NN-Driven Fuzzy Reasoning [5], and was used with only slight modifications.


4.2. Model of the process

The fuzzy classification of derusted surfaces contained two classes: “well

cleaned”, and “poorly cleaned”. Some of the cleaned samples were assigned to

“well cleaned” with degree of membership 1; correspondingly, they had the de-

gree of membership 0 in “poorly cleaned”. The rest of the samples had a non-zero

degree of membership in both classes. We used a special neural network to de-

termine these degrees. The procedure involved training the neural network

(NNmem), using only well-defined samples, i.e., samples having the degree of

membership 1 in either class. Then, after being properly trained, such a network

will not only be able to predict the binary degree of membership (either 0 or 1) for

some input data, but also the dual degree of membership (fuzzy membership) for

the input data points in that neighborhood. As the result of training, we obtain

neural network which is able to determine the degree of membership in each of

the two classes using input conditions, such as water pressure, traverse rate, etc.

This procedure is described in details in Takagi and Hayashi [5] and by Ross [6].

Our actual goal, however, was to determine the process effectiveness and the re-

sultant degree of cleanliness. In order to reach this goal we divided the available

database into two data sets. The first data set contained only the data identified

earlier to clearly (i.e. with degree of membership 1) belonging to the class “well

cleaned”, and, similarly the second data set contained data belonging only to the

class “poorly cleaned”. We then trained two separate networks, on these two data

sets, and as a the result each network was able to determine effectiveness for the

class it was responsible for.

From this, we obtained the model of the process in terms of the three trained

neural networks that were connected according to Fig. 18. According to this fig-

ure, given some input data set, the network NNmem identifies the degree of mem-

bership of a sample in each of the two classes. Network NN1 predicts productivity

for class “well cleaned”, and NN2 for class “poorly cleaned”. Input information is

fed to all three networks. The outputs of all three networks are then fed into spe-

cial elements which process the networks outputs to determine the weighted sum

and as a result predict final process effectiveness and the degree of cleanliness.

Thus, we were able to obtain an accurate prediction of the process effectiveness

(the average error in prediction was within 8%). Also we were able to estimate the

quality of derusted surface, based on the fuzzy degrees of membership in “well

cleaned’ and “poorly cleaned” classes inferred by the NNmem neural network. The

results of the prediction are presented in Figs. 19-22. These figures show the

process effectiveness as a function of different process parameters, without regard

to the quality of resultant surface. The quality (degree of membership in two

classes) for any data point in these figures is obtained using the neural network

NNmem.


Figure 18. Computational approach. (I1, I2, and I3 are the input parameters (water pressure, traverse rate, and nozzle diameter). NNmem is the neural network that decides the membership values (w1, w2) in each class, of the above input parameters. NN1 and NN2 determine the outputs (rate of area cleaned) y for each class.)

Figure 19. Model prediction results for water pressure 310 MPa at different nozzle traverse rates.


Figure 20. Model prediction results for water pressure 241 MPa at different nozzle diameters.

Figure 21. Model prediction results for nozzle diameter 0.3556 mm at different nozzle traverse rates.


Figure 22. Model prediction results for nozzle diameter 0.127 mm at different nozzle traverse rates.

5. CONCLUDINGS REMARKS

We have demonstrated the feasibility and, in fact, effectiveness of steel derusting

by waterjet. The effect of the operational conditions on the process was evaluated

and a set of operational conditions was suggested. This set can be used as an ini-

tial state by a practitioner to search for optimum operational conditions. Because

of wide variations in the states of the rusted surfaces and insufficiency of the

available identification technique, an advanced soft computing procedure (neural

network driven fuzzy identification) has been suggested for surface identification.

The analysis of the results of derusting conditions demonstrates the effective-

ness of the use of several nozzles rather than a single nozzle of the same surface

area.

Acknowledgements

The surface examination was carried out at the Stevens Institute of Technology.

The valuable advice of Josef Karagotskiy of the Electronic Microscopy Center at

the Stevens Institute of Technology is gladly acknowledged. This work was par-

tially supported by NSF grant # DDM-9312980.

REFERENCES

1. B. Liu, B. Jia, D. Zhang, C. Wang, H. Li and H. Yao, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology Association, St. Louis, MO (1993).

2. International Standard ISO 8501-1:1998, Third Edition (1999).


3. M. Hashish, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technol-ogy Association, St. Louis, MO (1993).

4. Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, Third Edition, Springer (1996).

5. H. Takagi and I. Hayashi, Intl. J. Approximate Reasoning, 5, No. 3, 191–212 (1991). 6. T. Ross, Fuzzy Logic With Engineering Applications, McGraw-Hill (1995). 7. B.D. Cullity, Elements of X-ray Diffraction, Addison-Wesley Publishing Company (1978). 8. H. Jun, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology

Association, St. Louis, MO (1993). 9. P. Singh, J. Munoz and W. Chen, Proc. of 11th International Symposium on Jet Cutting Tech-

nology, British Hydraulic Research Group, Dordrecht, The Netherlands (1992). 10. X. Shegxiong, H. Wangping and Z. Sheng, Proc. 7th American Water Jet Conference, Seattle,

Washington, Waterjet Technology Association, St. Louis, MO (1993). 11. C. Suryanarayana and M. Norton: X-ray Diffraction, a Practical Approach, Plenum Press, New

York (1998).


Ed. K.L. Mittal

© VSP 2003

Practical applications of icejet technology in surface

processing

D.V. SHISHKIN, E.S. GESKIN∗ and B. GOLDENBERG

Waterjet Technology Laboratory, Department of Mechanical Engineering, New Jersey Institute of

Technology, Newark, NJ 07102-1982

Abstract—The objective of this work was to acquire knowledge needed for the development and deployment of manufacturing processes utilizing the enormous technological potential of water ice. Material removal by blasting with ice media such as particles, pellets and slugs was investigated. The ice media were accelerated by entertainment in an air stream. The ice-airjet (IAJ) can replace sand blasting and the ice-waterjet (IWJ) can replace the abrasive waterjet (AWJ). The obvious ad-vantage of the ice media is complete pollution prevention in course of materials treatment. With this technique it is possible to eliminate both contamination of the substrate as well as generation of con-taminated waste streams. In addition to the obvious environmental benefits, the use of ice media will improve a number of key operational techniques, such as cleaning, decoating, polishing, deburring, drilling, cutting, etc. The “just-in-time” production of ice media at minimal environmental cost con-stitutes another advantage of ice-based technologies. Our previous studies have shown that the po-tential applications of ice abrasives range from cutting of metals to etching of photo films and preci-sion cleaning of electronic parts. However, the rate of the cleaning and machining operations performed was insufficient. A key objective of this research was to improve ice blasting so that it was not only feasible, but also technologically and economically efficient.

Keywords: Surface processing; cleaning; precision; abrasive; particle; ice.

1. INTRODUCTION

There are a number of suggested air-ice based technologies. One of the firsts of

such technologies was a car washing machine, utilizing ice particles [1]. The

stream of the charged frozen particles controlled by a set of coils was directed at

surfaces to be cleaned [2]. Szijcs [3] proposed cleaning of sensitive surfaces by

the impact of a fine grade blast material and air. The atomization of the liquid in

the air stream and subsequent freezing of the generated fine droplets form the

blast material. The freezing is achieved by the addition of a refrigerant (N2, CO2,

Freon) into the stream in the mixing chamber or by the addition of the refrigerant

∗To whom all correspondence should be addressed. Phone: (973)596-3338, Fax: (973)642-4282,



into the jet after the mixing chamber. The use of ice particles, which have a uni-

form grain size, for cleaning the surface and grooves of ferrite block, was reported

by Tomoji [4]. An ice blasting device using stored particles was suggested by Ha-

rima [5]. Vissisouk [6] proposed to use ice particles near melting temperature in

order to effectively remove the coating from the substrate. Mesher [7] suggested a

nozzle for enhancement of surface cleaning by ice blasting. Shinichi [8] suggested

cleaning inexpensively various surfaces by mixing ice particles, cold water and

air. Niechial [9] proposed an ice blasting cleaning system containing an ice

crusher, a separator and a blasting gun. Settles [10] suggested producing ice parti-

cles of a size range below 100 µm within the apparatus just prior to the nozzle.

Although the use of ice blasting is suggested by a number of inventors, the

practical application is much more limited. Herb and Vissisouk [11] report the use

of ice pellets for precision cleaning of zirconium alloys in the course of produc-

tion of bimetallic tubings. It was reported that ice blasting improved the quality of

the bimetal. The use of air-ice blasting for steel derusting was reported by Liu

[12]. The following operational conditions were maintained during blasting: air

pressure: 02-0.76 MPa, grain diameter: below 2.5 mm, ice temperature –50°C,

traverse rate 90 mm/min, and standoff distance 50 mm. Under these conditions

the rate of derusting ranged from 290 mm2/min at the air pressure of 0.2 MPa to

1110 mm2/min at the air pressure of 0.76 MPa. The quality of the cleaned surface

complied with ISO 8501-1 Sa 2.

The most important problem which actually impedes adoption of the ice-jet (IJ)

technology arises from the difficulties in the generation and handling of ice abra-

sives. Regular abrasives are stable at all practical ranges of operational conditions,

while ice particles can exist only at subzero temperature. Maintaining such a tem-

perature both within the nozzle and the jet is an extremely difficult task. Ice parti-

cles tend to pack and clog the supply lines. The adhesion between the particles in-

creases dramatically as the temperature approaches 0°C. Thus prior to entrance in

the nozzle, ice particles should be maintained at a low temperature. These and

some other problems prevent adoption of IWJ. In order to assure the acceptance

of IWJ by the industry, it is necessary to develop a practical technology for for-

mation of ice-water slurry.

2. SET UP FOR ICE-AIRJET EXPERIMENTAL PROTOTYPE

The experimental prototype depicted in Figure 1 was selected for further experi-

ments. The system consisted of the following functionally separated blocks:

– ice making block which includes the evaporator, auger, auger driver, sealing

and liquid nitrogen cooling apparatus;

– ice unloading mechanism

– nozzle block which includes parallel nozzles and focusing device.

Practical applications of icejet technology in surface processing 195

Figure 1. (a) Schematic of auger type IJ system where: 1 – evaporator, 2 – refrigerant coils, 3 – in-sulation, 4 – auger, 5 – ice reloading device, 6 – air gun, 7 – air supply port, 8 – water supply port, 9 – cooling medium port, 10 – gauges, A – air flow rate valve, B – water flow rate valve, C – cooling medium valve, D – data acquisition card, and (b) picture of the ice reloading device with nozzle block.

In our experiments, water entered the heat exchanger via a special port. As it

moved along the rotating auger water solidified and an ice plug was formed. So-

lidified ice plug moved forward along the auger helical ways. Decomposition of

this plug formed ice powder. The heat exchanger and the auger of the icemaker

constituted a modified commercial icemaker of Hoshizaki America Inc., Peach-

tree City, GA. The design of these parts will be changed in the next generation of

the device. The cooling was carried out by the refrigerant Galden HT-55 supplied

by the TurboJet refrigeration apparatus or by liquid nitrogen stored in a tank. We

replaced Hoshizaki auger driver by a more powerful device in order to prevent

jamming of the ice. The rotation momentum of the auger 4 was provided via a

gearbox with gear-ratio 1:100. However, the selected driver operated at a constant

speed of 100 rpm. Water flow rate was precisely controlled by a special valve

(Figure 1 (a)).

At the outlet of the heat exchanger the powder was entrained by the unloading

mechanism which directed it to the nozzle block (5). The nozzle block consisted

of two air guns (6) and a special focusing device. Three different sizes of the noz-

zles were used; however, in all cases the nozzle-to-focusing tube ratio was 1:2.

An unloading mechanism delivered ice particles via flexible plastic tubes to the

abrasive port of the air gun. In the gun the air supplied into the insulated nozzle

block at the room temperature accelerated the particles. The IAJ was formed and

directed to the substrate surface 3.


3. EXPERIMENTAL PROCEDURE

In the course of IJ cleaning, the air pressure was maintained at 0.544 MPa (80

psi), the nozzle diameter was 2.5 mm, and the nozzle focusing tube diameter was

5 mm. The properties of ice abrasive medium were the following: ice temperature

was in the range from –20°C to –70°C, granulometric composition of ice powder

ranged from 0.3 mm to 7 mm, and ice flow rate was 20 g/min to 150 g/min. Ice-

airjet (IAJ) was used for cleaning various sensitive surfaces covered by moder-

ately adhering deposits. The sensitive elements of the electronic boards were cov-

ered by a conductive copper paste and cleaned by the IAJ. When assembled, these

components performed normally and the normal operational modes of the devices

were demonstrated. The feasibility of using IAJ technology as a blasting medium

for cleaning highly sensitive surfaces was shown. Another experiment involved

depainting of various substrates, including mirror-like surfaces and the surfaces of

soft substrates. A complete removal of the paint and the absence of surface dam-

age were demonstrated. The generated surfaces were inspected visually.

A number of experiments involved the use of the ice abrasive in waterjet (WJ)

cutting applications. The experimental procedure was carried out with the follow-

ing parameters: the water pressure was 306.1 MPa, the diameter of the sapphire

nozzle was 0.178 mm, average standoff distance was maintained in range 7 mm–

10 mm and the traverse rate was 1.06 mm/s. Various metals and composite mate-

rials were cleaned by IWJ. The depth and cutting rate were substantially lower

than in the case with conventional abrasive media. However, the IWJ produced a

very narrow cutting kerf compared with AWJ and had a superior cutting ability

over pure WJ. The main obstacle during ice particles entrainment in the nozzle

abrasive port was their agglomeration at the port entrance and their disintegration

in the mixing chamber due to intensive melting. This technology is still under de-

velopment and requires further investigation.

4. EXPERIMENTAL RESULTS

A series of experiments were carried out in order to evaluate the potential of the

application of IAJ for surface processing. The description of these experiments is

given below.

4.1. Cleaning of electronic boards

A disabled TV set was disassembled (Fig. 2a). The electronic boards were cov-

ered by a heavy dust. Then the boards were decontaminated by IAJ and reassem-

bled. The TV set performed normally (Fig. 2b). The architecture of the boards in

question was extremely complex and contained a number of very sensitive sites,

like electrical contacts and conduits. Any damage to the board components would

result in the TV set malfunction. It is obvious that the ice-air stream induced no

damage. More difficult task, however, was a complete grease removal. Even


Figure 2. (a) photograph of the electronics board of a TV set. Notice the heavy layers of dust and dirt on the electric and electronic components of board, (b) photograph of an assembled TV set. The contaminated board of TV set is shown (a). After cleaning TV set worked normally.


small amount of the grease remaining at hidden pockets will disrupt the TV set

performance. It is clear that the jet was able to remove soil from all the difficult-

to-reach pockets.

Another experiment involved decontamination of computer boards. Various

devices (PC, electronic watches, computer games, etc) were disassembled. The

boards were covered by a mixture of lithium grease and then decontaminated by

IAJ. Clean boards were reassembled and tested. All devices worked perfectly.

Some of the devices above were used for several tests. No deviation in the com-

puter operation was noticed. The boards above were populated by a large number

of rather fragile components such as chips, connectors, etc. Any damage to any of

these components, as well as any presence of grease on the board will disable the

device. In all performed experiments the deposit was removed completely and no

damage was induced to the board components. The examples of the boards decon-

taminated in the course of these experiments are shown in Figs. 3 (a) and 3 (b).

4.2. Decoating of sensitive surfaces

The experiments involved depainting of a compact disc (CD). This involved re-

moval of the paint as well as two layers of the coating originally deposited on the

disk (Figs. 4 (a) and 4 (b)). The paint and then the emulsion layers were removed

separately with no damage to the underlining surface. Another experiment in-

volved painting and subsequent depainting of the mirror-like surface of stainless

steel (Fig. 5 (a)). No change in the surface topography was noticed. Further ex-

periments involved depainting of china (Fig. 5 (b)), egg (Fig. 6 (a)), and glass lin-

ing of a pharmaceutical reactor (Fig. 6 (b)). The most representative experiments,

however, involved depainting of a LC display (Fig. 7 (a)) and degreasing of an

optical glass (Fig. 7 (b)).

4.3. Decoating of soft substrates

These experiments involved depainting of a soft plastic (Fig. 8 (a)) and fabric

(Fig. 8 (b)). Decoating of a substrate having mechanical strength lower than that

of the coating constitutes a challenging task, but IAJ was able to perform this

task.

4.4. Restoration of electromechanical devices

A solenoid valve (Fig. 9 (a)) and a DC motor (Fig. 9 (b)) were completely dis-

abled by painting of all contacts. After IAJ cleaning the devices performed nor-

mally.

4.5. Removal of highly adherent surface layers

An aluminum plate was covered by a thick layer of tar. Then the tar was removed

mechanically from a part of the plate. However, a highly adherent thin tar layer


Figure 3. (a) photograph of the board of an electronic game containing electric conduits, microchip and electronic matrix. The board was covered by a mixture of lithium grease and copper powder. Notice the cross contamination of electric conduits of the board, (b) photograph of the assembled electronic game after IAJ cleaning. The electronic game performed normally.


Figure 4. (a) photograph of the CD-ROM covered by Rust-Oleum gloss protective enamel. The paint was partially removed from the CD ROM surface. No surface damage was observed in the course of IAJ cleaning, and (b) photograph of the CD-ROM partially cleaned using IAJ technique. Notice that layers of both paint and emulsion were removed. No surface damage was observed in the course of IAJ processing.


Figure 5. (a) photograph of the polished steel surface. The polished steel surface was contaminated by Rust-Oleum gloss protective enamel. The paint was partially removed from the polished surface. No surface damage was observed in the course of IAJ cleaning, so the feasibility of the precision cleaning of polished surfaces was demonstrated. (b) photograph of the hand-painted china plate. The plate was covered by Rust-Oleum gloss protective enamel. Part of the deposited paint was removed by ice etching. No modification of the original surface was noticed, and thus the feasibility of IAJ etching of sensitive surfaces was demonstrated.


Figure 6. (a) photograph of an egg. The egg surface was painted by Rust-Oleum gloss protective enamel. After this the egg was partially decontaminated by IAJ technique. No damage to the egg surface or penetration of the ice particles through the eggshell was noticed, so the feasibility of de-contamination of highly unstable and brittle surfaces was demonstrated. (b) photograph of the cover of a pharmaceutical reactor contaminated by the lithium grease. Then the grease was partially re-moved from the surface of the cover by IAJ technique. No damage to the glass in the course of IAJ cleaning was noticed.


Figure 7. (a) photograph of the LC display of a calculator containing electronic matrix and LCD conduits. The display was contaminated by Rust-Oleum gloss protective enamel. Then all elements of the LC display were decontaminated by IAJ technique. On assembly of the calculator the LC dis-play performed normally. (b) photograph of a magnification lens. The lens was contaminated by lithium grease. The grease was partially removed from the lens surface, and no damage to the lens surface was observed.


Figure 8. (a) photograph of a PVC tube contaminated by Rust-Oleum gloss protective enamel. The tube was partially decontaminated by IAJ technique. No damage to the tube surface in the course of IAJ cleaning was noticed, and (b) photograph of a cotton fabric. The fabric was contaminated by Rust-Oleum gloss protective enamel. Then the paint was partially removed from fabric surface, and thus the feasibility of the use of ice particles for decontamination of fabrics was demonstrated.


Figure 9. (a) photograph of an electrical solenoid valve with connectors contaminated by Rust-Oleum gloss protective enamel. The contacts of solenoid valve were cleaned by IAJ technique. Af-ter cleaning the solenoid valve was connected to an electrical supply source and performed nor-mally. This experiment demonstrated the feasibility of using IAJ technique for decontamination and restoration of contacts of different electronic devices. (b) photograph of a DC motor. DC motor was disassembled and all elements were covered by a mixture of lithium grease and copper powder. DC motor was cleaned using IJ technique and the assembled DC motor performed normally.


remained on the surface. It was not possible to remove it using mechanical means.

The layer was removed completely by the IAJ (Fig. 10 (a)). A metal wall was

covered by an oil paint and then was subjected to abrasive-airjet (AAJ) (the car-

rier medium was sodium bicarbonate). Then the same procedure was carried out

using IAJ. The initial state of the graffiti covered metal surface is shown in Fig.

10 (b). The graffiti removal by conventional and ice based technologies are shown

in Figs. 11 (a) and 11 (b). Another experiment involved removal of the residual

highly adhesive Weldbond glue from plastic and rubber jointed surfaces (Fig. 12

(a) and 12 (b)). Average process duration in all these experiments was around two

minutes. The heavily contaminated machine part with grease and dust was decon-

taminated by IAJ too (Fig. 13 (a) and 13 (b)). No damage to the underlying

painted surface was noticed.

4.6. Etching applications

The emulsion of a photo film was removed with no damage to the substrate (Fig.

14 (a)). This demonstrates the feasibility of the use of IAJ as an etching agent.

4.7. Ice-waterjet (IWJ) applications

Various metals and plastic materials were subjected to IWJ cutting. The superior

cutting ability of IWJ over pure WJ was seen. The cutting ability of IWJ was lim-

ited by ice abrasive disintegration in the nozzle mixing chamber. This task re-

quired further investigation. However, it was shown that the IWJ cutting kerf was

thinner (Figure 14 (b) and Figure 15) and showed the potential of IWJ as an alter-

native cutting medium for waterjet industry.

5. CONCLUDING REMARKS

Although the ice-water jet constitutes an effective material removal tool, it is nec-

essary to improve conditions of the jet formation in order to assure its adoption in

practice. However, the ice-air jet is suitable for immediate application. It can be

used for decontamination of very demanding and complex surfaces as well for

such manufacturing applications as etching. Simplicity and complete absence of

environmental damage constitute the main advantages of this process. A further

development of IAJ surface cleaning technology will involve improvement of the

control of ice particle properties and enhancement of the methods for the delivery

of ice particles to the substrate. This enhancement will enable us to modify mate-

rial polishing, surface cleaning, and, perhaps, grinding.


Figure 10. (a) photograph of the aluminum surface contaminated by a thick layer of tar. The bulk of the tar was removed by WJ and knife scrubbing. The highly adherent thin layer was removed by ice etching. No damage to the metal surface was noticed and (b) graffiti covered painted metal surface. The oil paint is highly adherent.


Figure 11. (a) Graffiti was removed with conventional AAJ. Notice discoloration occurred in the treated region and (b) surface was decontaminated by the IAJ. No damage to the underlying paint layer occurred.


Figure 12. (a) The Weldbond glue was used to create a joint between plastic and rubber surfaces. Notice the highly adhesive character of the glue, and (b) the glue residue was removed by IAJ clean-ing. No surface damage was noticed.


Figure 13. (a) picture of the highly contaminated machine part with grease and dust, and (b) part was decontaminated by IAJ cleaning. No damage was seen on the underlying painted surface.


Figure 14. (a) photograph of a strip of a photo film. The photo emulsion was partially removed from the film surface. No surface damage was observed in the course of IAJ cleaning and thus the feasibility of complete and selective emulsion removal from thin photo film was demonstrated, and (b) photographs of cutting of aluminum strip of thickness 3.1 mm (X65). Notice the reduced width of the kerf generated by IWJ cutting. Also note substrate surface erosion in the vicinity of IWJ gen-erated kerf.


Figure 15. Photographs of cutting of titanium sample of thickness 0.7 mm (X65). Notice the re-duced width of kerf in the course of IWJ cutting. Also note the intensive erosion of the substrate sur-face in the vicinity of IWJ generated kerf.

REFERENCES

1. C. Schlosser, L. Mueller and G. McDougal, US Patent 5,752,39 (1950). 2. G. Kanno, US Patent 5,074,083 (1991). 3. J. Szijcs, European Patent 0509132B1 (1991). 4. M. Tomoji, Japanese Patent 04078477 (1990). 5. I. Harima, Japanese Patent 04360766 A (1992). 6. S. Vissisouk, European Patent 05076607 (1995). 7. T. Mesher, US Patent 5,607,478 (1997). 8. H. Shinichi, Japanese Patent 09225830 A (1997). 9. R. Niechial, US Patent 5,820,447 (1998).

10. G. Settles, US Patent 5,785,581 (1998). 11. B. Herb and S. Vissisouk, Proc. Precision Cleaning 1996 held in Anaheim, CA, pp. 172-179

(1996). 12. B. Liu, in Jetting Technology, H. Louis (Ed.), pp. 203-211, Professional Engineering Publishing

Ltd., London, UK (1998).


Ed. K.L. Mittal

© VSP 2003

Correlating cleanliness to electrical performance

TERRY MUNSON∗

Contamination Studies Laboratory (CSL), 201 East Defenbaugh, Kokomo, Indiana 46902

Abstract—This paper explores the correlation between the cleanliness levels on electronic assem-blies and their electrical performance. It documents an experiment conducted to explore this correla-tion. Cleanliness was measured using Ion Chromatography (IC), and electrical performance was measured using Surface Insulation Resistance (SIR) testing under elevated humidity and tempera-ture conditions. Furthermore, this paper discusses electronic assembly cleanliness issues, and a new cleanliness assessment approach for determining cleanliness levels required for the typical flux technology of today. We conclude – from the samples examined, and based on our past 10-years of experience analyzing similar experiments – that circuit board field performance (good or poor) is strongly correlated to the specific amount and type of invisible and visible residues between pads and holes in all areas of active circuitry.

Keywords: Ionic contaminants; residues; electrochemical metal migration; ion chromatography; sur-face insulation resistance; electrical performance testing; cleanliness levels.

1. INTRODUCTION

Since the 1987 Clean Air Act, when Government legislation forced the electron-

ics industry to stop using the ozone depleting chemical Freon® as a cleaner, the

industry has been required to find new chemicals and processes. The industry be-

lieved at the time that Freon® solvent cleaning techniques effectively removed all

surface contaminants. Subsequent research has shown that the old rosin based

fluxes used in the manufacturing processes actually sealed in contaminants,

whereas the new fluxes left contaminants exposed to react with humidity in the

end-user environments. These inherent process residues must be removed to

achieve product reliability.

The change from Freon® cleaning initiated many changes in the manufacturing

processes. There was no direct substitute cleaning chemistry. The new solvents

could not be used for cleaning using many of the old processing chemistries. The

change was an opportunity for many manufacturers to change their overall proc-

esses. As the changes were made to new processing materials and cleaning sol-

∗Phone: 765-457-8095, Fax: 765-457-9033, E-mail: [email protected]

T. Munson 214

vents, product performance data indicated increasing field performance problems

related to surface ionic residues. Understanding surface residues and the effects of

each type became critical for product quality.

Since the fluxes by themselves did not have corrosive activators and passed the

coupon evaluations of the IPC (IPC TP-1042 “Phase 2 No-Clean Flux Study”)

and the water-soluble process passed all the test coupon work published by the

IPC (IPC TP-1043 and TP-1044 “Phase 3 Water Soluble Fluxes Study” Septem-

ber 30, 1992), then the focus shifted to meeting solderability performance with

real product. Process cleanliness had not been an issue when the rosin flux was

used. Manufactures that switched to No Clean (low solids) processes in October

of 1992 would, in some cases, generate a product recall in May 1993. It has be-

come clear that the bare board, component, temporary mask and materials and

rework contaminants are critical to today’s electronic hardware performance.

Before examining the correlation between cleanliness levels and electrical per-

formance on electronic equipment, let us define cleanliness and the factors that af-

fect it.

1.1. Definition of cleanliness

In the case of electronic assemblies, finished cleanliness levels are a measure of

the amount of detrimental residues remaining on completed assemblies. Elec-

tronic assembly cleanliness is a result (signature) of the assembly process, materi-

als and secondary processes required creating the finished assembly.

1.2. Acceptable cleanliness levels

The acceptable levels of cleanliness depend on how the various residue types that

remain on the assembly react under electrical power in poorly controlled envi-

ronmental conditions. The various residue types react differently from pad-to-pad

and hole-to-hole. How they react determines the quality of electrical performance.

Two of the most valuable tools for determining board and assembly cleanliness

are Ion Chromatography (IC) and Surface Insulation Resistance (SIR) testing.

1.3. Why is cleanliness an important issue?

Electronic assembly cleanliness is an important issue for at least four reasons:

1. The trend toward higher operating frequencies and lower operating voltages is

causing circuits to be less tolerant of stray current leakage.

2. Spacing geometries have pushed traces and leads closer together and have in-

creased the probability of power-to-ground pathways due to smaller amounts

of fluids required to bridge the smaller spacing.

3. Component packages have become shorter, spaced tighter to the board sur-

face, and smaller in relationship to the board area.

Correlating cleanliness to electrical performance 215

4. New flux chemistry residues on the newest metal composition result in new

interaction issues that have effects that are inadequately understood.

1.4. Why is determining cleanliness a problem?

Determining if an electronic assembly is adequately clean is difficult for the fol-

lowing reasons:

1. Cleanliness is not easily assessed with today’s production floor tools. Current

industry-standard process cleanliness tools and test methods are not adequate

gauges of product cleanliness when testing today’s low solids and water-

soluble fluxes. Due to poor extraction conditions, these tools give false low

levels. Also, they do not identify residues as corrosive or insulative.

2. Cleanliness is not visually assessable.

3. Adequate cleanliness depends on the circuit design, the processing materials,

the process, and other factors.

4. Cleanliness is not uniform across the assembly surface, but has concentrations

of residues in the critical areas such as between the component leads and the

board and component interface.

5. Cleanliness is not defined in Industry Specifications as it relates to today’s

field performance.

6. Cleanliness is not necessarily the same in a process from day-to-day, due to

different suppliers of bare boards, operator experience, and vendor variations.

1.5. Residues sources

Circuit board cleanliness is a measure of the cumulative process residues. These

residues are the result of the manufacturing process steps and materials used in

each step. Everything used in the manufacturing process has an impact on the

types of residues that will be created. To complicate things further, boards and

components that look visibly clean and dry can actually be absorbing moisture

and reacting with the bare board HASL flux, creating a leakage path. These resi-

dues can come from:

1. Materials

Bare Board Fabrication (Etching chemicals, HASL fluxes, Rinse water (tap))

Component Packaging (Fluxes, Mold releases)

Component Plating or Tinning (Plating, Bath, Rinse water, Fluxes)

Flux, Solder Paste, Cored Solder, Water Quality, Epoxies and Soaps

2. Processes

Fluxing and Soldering (Flux amounts, Thermal effects)

Paste and Reflow (Flux spread, and outgassing effects)

Water Soluble Cleaning or No-Clean Processing (pressure, water, and

saponifier)

T. Munson 216

Production Handling (Glove and hand residues)

Temporary Masking and Removal (outgassing)

Hand Soldering and Rework (cleaning materials and procedures)

1.6. Residue effects

Process residues are the corrosive contaminant source for corrosion cells to de-

velop, which form when residues, a fluid medium, and a voltage differential are

combined. These corrosion cells cause electrochemical migration (metal migra-

tion). Figure 1 shows dendritic growth in a corrosion cell created using a 21-

second Water Drop Evaluation with a high chloride level (1.29, µg/cm2 of Cl

–)

coupon, a 10-volt bias, and DI water, between 1-mm spacing. Corrosion cells like

this cause electrical leakage and shorts, resulting in failures (or No Trouble Found

(NTF) returns) under typical operating conditions (humidity levels of 50% and

higher).

Figure 1. Dendrite Growth at (a) 5 seconds, (b) 10 seconds, (c) 15 seconds, (d) 20 seconds.


2. EXPERIMENTAL DETAILS

CSL designed and conducted an experiment to explore the correlation between

circuit board cleanliness levels and long-term reliability simulated through accel-

erated life testing under powered, high temperature and humidity conditions.

2.1. Materials

To perform the experiment, 30 test coupon printed circuit boards were fabricated.

The Umpire test board (shown in Figure 2) was selected because it uses a mixed

technology assembly process – surface mount technology (SMT) and plated-

through-hole technology (PTH). It can also be used to assess different component

areas, such as the pad-to-pad or lead-to-lead, and assess entrapment effects on a

variety of components (BGA, LCC, QFP, DIP, PGA, 1206 chip, 0805 chip), and

comb patterns (B24, Bellcore).

2.2. Processing conditions (three levels of bare board cleanliness with a

no-clean assembly process)

All 30 bare boards were fabricated using a Hot Air Solder Leveled (HASL) with

three levels of cleanliness. The assembly process used a low residue solder-paste,

a low residue liquid flux (alcohol-based), and a low-solids core solder (with addi-

tional liquid flux followed by alcohol and brush localized cleaning). This process

also used a peelable temporary soldermask. The samples were processed as one

Figure 2. Umpire SIR Test Board.

T. Munson 218

batch until they reached the final cleaning process. Cleanliness Level 1 (L1) used

a saponifier in the cleaning solution, known to be the cleanest method. Level 2

(L2) used deionized water cleaning, known to be a marginal cleaner. Level 3 (L3)

used tap water cleaning, known to be the dirtiest cleaning method (standard rinse

water process at most fabricators).

These two tests were performed on each sample: 1) Ion Chromatography and 2)

Surface Insulation Resistance with visual inspection performed per IPC 610 pro-

tocol.

2.3. Cleanliness testing using ion chromatography

Ion Chromatography is used to assess cleanliness levels of electronic equipment.

Ion Chromatography is a process of separating ionic and organic residues sus-

pended in a liquid. This separation is achieved through a finely balanced system

of liquid phase eluent and resin columns. The resin has a charge opposite to the

ions, causing different ions to travel through the column at different rates. As each

species leaves the column (illustrated in Figure 3), a conductivity cell measures its

concentration in microSiemens (µS). The IC system records this information on a

chart for the duration of the analysis, and quantifies the area under the curve of

each species detected. Typical species detected include: Fluoride, Chloride, Bro-

mide, Phosphate, Sulfate, Formate, Acetate, Methane Sulfonic Acid, Weak Or-

Figure 3. Ion Chromatography System Illustration.


ganic Acid, Sodium, Calcium, Potassium, and NH4

–. Before testing, the IC system

is calibrated to NIST (National Institute of Standards and Technology) traceable

standards. The level of sensitivity is 0.01 part per million.

CSL is able to test specific areas of boards by combining IC with an effective

extraction procedure that conforms to IPC protocol TM 650 2.3.28. Each sample

was IC tested in four location areas: 1) a top-side surface mount technology

(SMT) area, 2) a bottom-side wave solder area, 3) a rework area, and 4) a peelable

solder mask area.

2.4. Surface insulation resistance (SIR) testing

SIR testing with the Umpire board allowed CSL to subject the processed samples

to accelerated environmental conditions under applied power, which allowed

evaluation of the electrical effects of trapped process residues. Additionally, the

Umpire board allowed CSL to analyze each of the four location areas separately.

Figure 4 diagrams the SIR system used.

According to the IPC J-Std SIR pass-fail criterion, the patterns must maintain

resistance values above 1.0e8 ohms measured at 96 and 168 hours.

3. RESULTS

The experimental results are grouped into the four test area groups: 1) top-side SMT

area, 2) bottom side wave solder area, 3) rework area, and 4) solder-mask area.

3.1. Top-side SMT area results

The data in Table 1 show the ionic and electrical performance mean values from

each group (5 boards per condition) of samples relative to the effects of the fac-

tors on the SMT top-side.

Figure 4. Surface Insulation Resistance Testing System Illustration.

T. Munson 220

Based on the data in Table 1, we concluded that;

1. L1 boards performed well ionically and electrically.

2. L2 boards failed electrically due to the high amount of bare board contamina-

tion (chloride HASL flux). The BGA, LCC and QFP failed due to the chlo-

ride / flux residues from the bare board fabrication process, and not due to the

assembly process since L1 boards passed.


tion (chloride HASL flux).

3.2. Bottom side wave solder area results

The data in Table 2 show the ionic and electrical performance of the samples rela-

tive to the effects of the factors on the bottom side wave solder areas.


1. L1 boards performed well ionically and electrically with the exception of the

bottom side B-24 comb pattern with a layer of WOA residue over the entire

comb thick enough to be visibly obvious. This comb failure occurred because

too much flux was applied and not all of it was complexed (all the flux carrier

driven off and the crystals melted forming an insulative residue), leaving a

partially dried but conductive moisture-absorbing residue between the leads.

Table 1.

Top-side SMT area results for all three levels of cleanliness (L1-L3)

Ionic Test using Ion Chromatography (all values are in µg/cm2)

Sample # Description Chloride Bromide Sulfate WOA Sodium Potassium

L1 SMT paste areas only 0.18 0.34 0.00 7.03 0.33 0.03



Electrical Performance using Surface Insulation Resistance Testing

Sample # Initial measure-ment (in Ohms)

Measurement (in Ohms) at 168 hours

(85°C/85%RH)

BGA LCC LCC

(comb)

QFP QFP (comb)

Head1 (control)

L1 1.3e11 4.3e10 Passed Passed Passed Passed Passed Passed

L2 1.4e10 6.7e7 Failed Failed Failed Failed Failed Failed

L3 2.2e10 1.0e6 Failed Failed Failed Failed Failed Failed

1 = Header pattern topside WOA = Weak Organic Acids BGA = Ball Grid Array LCC = Leadless chip carrier QFP = Quad Flat Pack RH = Relative Humidity



tion (chloride HASL flux) with the exception of the PGA and DIP area. The

failure was due to the insulative flux effects creating a barrier between the

moisture and the leads.


tion (chloride HASL flux).

3.3. Rework area results


tive to the effects of factors on the rework areas.


L1, L2 and L3 boards failed due to the un-reacted flux and the distribution of

this flux residue during the cleaning process. This failure occurred because ex-

tra flux was applied by an operator during a reworking process and not all of it

was complexed (heated correctly). It then spread around leaving a conductive

moisture-absorbing path, resulting in high levels of current leakage, and test

failure.

Table 2.

Bottom side wave solder area results for all three levels of cleanliness (L1-L3)



L1 Wave soldered areas only

0.66 0.36 0.00 50.57 0.26 0.08


1.28 0.43 0.00 65.60 0.17 0.05


2.61 0.36 0.00 60.08 0.20 0.09




(85°C/85%RH)

B-241 PGA DIP1 DIP2 Head1 (control)

L1 1.2e11 3.3e8 Failed Passed Passed Passed Passed

L2 1.3e10 6.7e9 Failed Passed Passed Failed Failed

L3 2.6e10 1.0e6 Failed Failed Failed Failed Failed

1 = Dip pattern 2 = Dip pattern

T. Munson 222

Table 3.

Rework area results for all three levels of cleanliness (L1-L3)



L1 Reworked areas only 0.50 0.44 0.00 59.15 0.20 0.06






(85°C/85%RH)

B-243 Head2 Head1 (control)

L1 1.2e11 3.3e8 Failed Failed Passed

L2 1.3e10 6.7e9 Failed Failed Failed

L3 2.6e10 1.0e6 Failed Failed Failed

1 = Header pattern topside 2 = Header pattern bottomside 3 = B24 board bottomside stripped soldermask comb

Table 4.

Temporary Soldermask area results for all three levels of cleanliness (L1-L3)



L1 Temporary Masked areas only

0.64 0.42 0.00 0.75 0.50 0.17


1.60 0.43 0.00 0.82 0.53 0.20


3.35 0.42 0.00 0.80 0.47 0.17


Sample # Initial measurement (in Ohms)


(85°C/85%RH)

Head3 Head1 (control)

L1 8.e10 1.7e7 Failed Passed

L2 4.2e10 6.9e6 Failed Failed

L3 3.9e10 1.0e6 Failed Failed


3.4. Temporary soldermask area results


tive to the effects of factors on the temporary solder-masked areas.


L1, L2 and L3 boards failed electrically due to the high amount of bare board

contamination (chloride HASL flux) and contact residues from the peelable

soldermask. This failure occurred because no flux residues were present to cre-

ate an insulative barrier.

4. DISCUSSION

This case study shows that the low-residue assembly process used works well in

the areas of SMT and Wave Solder with clean (level 1) bare boards. However, the

secondary processing (rework and temporary soldermask) areas, even with clean

bare boards, showed high levels of electrical failure due to excess partially or un-

reacted flux.

The data presented here are only a small snapshot of the information gathered

through this testing. This assessment is not intended to replace actual product

validation or environmental testing. This case study is intended to assess the proc-

ess effects as a baseline and to determine if process changes are good or bad in

regards to the electrical effect. As an assessment tool, this will help establish the

actual level of cleanliness required in building reliable hardware.

A cleanliness assessment approach such as this will allow electronic assemblers

the opportunity to document the effects of the process and materials as a baseline

from which to make improvements. This case study and others we have per-

formed over the last ten years have shown us that the residues from fabrication do

have a large effect on electrical performance. In addition, the residues from sec-

ondary assembly processes have just as much effect on the field performance of

the product.

Although ionic and organic IC analyses of component areas on electronic as-

semblies detect a specific amount of flux and processing residues, determining

whether the levels are good or bad is based on results from an electrical SIR

evaluation of the same areas. Over the last 10 years, CSL has developed a large

cleanliness level database, and has used this data to determine general levels of

cleanliness, but we are constantly adjusting acceptability levels because of chang-

ing component packages, increasing operating frequency designs, lower voltage

designs, and changing processing materials. The result of all the changes in the

industry is electronic equipment that is more sensitive to process residues, and has

less long-term reliability, especially in harsh environments, unless detrimental

process residues are identified, measured, and reduced.

T. Munson 224

5. CONCLUSION

Knowing cleanliness levels is a vital part of understanding product quality. The

reliability of a product can be improved with adjustments in processing that im-

prove cleanliness. Testing at various steps of the process or locations on the prod-

uct can show what processing steps are affecting the final cleanliness and reliabil-

ity. It is important to be particularly cautious with incoming product cleanliness

and secondary assembly processes.

Acknowledgments

We would like to acknowledge Dr. K.L. Mittal for constructive feedback and

support. We would also like to acknowledge Diversified Systems Inc. of Indian-

apolis, Indiana for their manufacturing support during the experiment.

REFERENCES

1. M.G. Fontana, Corrosion Engineering, 3rd edition, McGraw-Hill Book company, New York (1986).

2. H. Small, Ion Chromatography, 1st edition, Plenum Press, New York (1989). 3. IPC Committee, Post Solder No-Clean Handbook (IPC-SC-62A), IPC, Chicago, IL (1999). 4. IPC TM 650, Test Methods Handbook, “Ion Chromatography,” IPC, Chicago, IL (1995). 5. IPC TM 650, Test Methods Handbook, “Surface Insulation Resistance”, IPC, Chicago, IL

(1995).


Ed. K.L. Mittal

© VSP 2003

Qualifying a cleaning system for space flight printed

wiring assemblies

J.K. “KIRK” BONNER∗ and ATUL MEHTA

Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena,

CA 91109, USA

Abstract—Cleaning critical and high reliability printed wiring assemblies (PWAs) continues to be important to ensure high reliability performance and to prevent premature failure. The necessary steps to qualifying both a cleaning system and an appropriate chemistry for cleaning such PWAs are set forth. This paper addresses a centrifugal cleaning system used in conjunction with a water-based cleaning medium to achieve optimally low levels of contaminants on PWAs. Ionograph data, ion chromatography profiling, residual rosin determination, and outgassing data are presented demon-strating the effectiveness of the centrifugal cleaning system and the aqueous cleaning agent for space flight printed wiring assemblies. It is concluded that a centrifugal cleaning system coupled with a suitable aqueous chemistry can be successfully employed to clean high reliability PWAs.

Keywords: Aqueous cleaning (AC); conformal coating; ionic contamination testing (ICT); green-house warming potential (GWP); multilayer board (MLB); ozone depletion potential (ODP); printed wiring assembly (PWA); printed wiring board (PWB); rosin mildly activated (RMA); semi-aqueous cleaning (SAC); surface mount technology (SMT); volatile organic compound (VOC).

1. INTRODUCTION

During the last decade, the challenges of cleaning printed wiring assemblies

(PWAs) have grown. Today printed wiring boards have grown more complex to

meet the continuing challenges posed by the increasing uses of microdevices, fine

pitch packages, and array devices, such as ball grid arrays, microball grid arrays,

and flip chips. Multilayer boards with a large layer count and narrow trace widths

and spaces are commonplace. The ball grid arrays, microball grid arrays, and

other small devices generally have a large number of inputs/outputs (I/Os), small

standoffs, and small pitches. The small standoff and small pitch, coupled with the

complex circuitry needed to route such components, makes cleaning an ever more

critical operation. High reliability PWAs cannot tolerate contaminants since their

∗To whom all correspondence should be addressed. Phone: (818) 354-1320, Fax: (818) 393-

5456, E-mail: [email protected]

J.K. Bonner and A. Mehta 226

presence can potentially degrade the board, thus compromising the intended mis-

sion.

Cleaning for high performance PWAs is normally performed as a minimum at

the following stages:

(1) At the bare printed wiring board (PWB) stage prior to the application of sol-

der mask;

(2) Immediately after the PWB + components are soldered to form the PWA;

(3) Immediately prior to the application of conformal coating.

If the PWAs are properly stored, the second and third operations are sometimes

combined. A number of contaminants are potentially introduced on the PWB sur-

face. These contaminants can be classified into three broad categories:

(1) Particulates;

(2) Ionic residues;

(3) Non-ionic residues, chiefly organic in nature [1-4].

To ensure the reliability of a PWA, cleaning is mandatory to remove these con-

taminants after the soldering operation and also directly prior to the application of

a conformal coating. In addition to cleaning, some sort of cleanliness verification

method, such as ionic contamination testing (ICT), is normally employed. ICT

can be used to ascertain that a certain level of cleanliness has been achieved. In-

dustry-recognized devices, such as an Ionograph® or Omega-Meter

®, have com-

monly been used for this purpose. In addition, determining the amount of residual

rosin (assuming that a rosin-based flux or paste was used) is often done. Another

useful technique is to remove some of the components and examine for flux resi-

dues both visually and by use of a microscope.

The last decade has also seen the dramatic decrease and continuing disuse of

ozone-depleting solvents. The common chlorofluorocarbon solvents, such as

Freon® TMS, have been discontinued, and many PWA assemblers have switched

to more environmentally-friendly cleaning agents, such as a wide variety of semi-

aqueous and aqueous-based materials. To enhance the performance of such mate-

rials, the proper equipment selection plays a critical role.

2. BACKGROUND

Ten years ago the Electronic Packaging and Fabrication section at the Jet Propul-

sion Laboratory (JPL) established a dedicated facility for producing very low vol-

ume but high performance surface mount technology (SMT) assemblies known as

the SMT Laboratory. This laboratory has successfully assembled SMT PWAs for

such important JPL programs as these:

Cassini;

ChuG Microgyro;

Caltech-ACE;

Qualifying a cleaning system for space flight printed wiring assemblies 227

MISER;

SEAWINDS;

Pathfinder.

and many others. Since the assemblies produced in this laboratory always fall in

the high performance, high reliability category, cleaning is mandatory, not op-

tional. With the demise of the ozone-depleting solvents that were the mainstay of

the electronics industry for twenty years, it was necessary to turn to alternative

chemistries and cleaning systems to ensure cleanliness and high reliability of the

surface mount assemblies (SMAs).

The initial cleaning system chosen for the SMT Laboratory was a two-stage

batch semi-aqueous (SA) cleaning system. Although this system worked satisfac-

torily for a number of years, the decision was reached recently to replace it. Part

of the reason was the increasing complexity of the SMT PWAs. Equipment to en-

sure that the cleaning solution would successfully penetrate under the small

standoffs and tight spacings found under the newer components now being in-

creasing employed was considered mandatory. Another factor in the decision was

that the initial equipment manufacturer sold off this portion of the business and no

longer supported the equipment. It proved increasingly more difficult to maintain

it in good working condition. In addition, isopropyl alcohol (IPA), used in the

original equipment, came under increasing scrutiny by the South Coast Air Qual-

ity Management District (SCAQMD). Because IPA is a volatile organic com-

pound (VOC), its emission into the atmosphere is tightly controlled. The decision

was made to investigate a new cleaning system and a chemistry that would sup-

port JPL’s need for clean PWAs to meet the newer challenges.

3. PERTINENT PROCESS INFORMATION

The following JPL process information is pertinent to the discussion:

Rosin-based fluxes and pastes are used to produce all electronic hardware.

Using the terminology of Mil-F-14256, the classification of these products is

rosin mildly activated (RMA).

The solder paste is applied using a semi-automated screen printer ensuring that

the paste is deposited in a uniform and consistent manner. Only stainless steel

stencils are used in conjunction with a stainless steel squeegee. All boards are

visually inspected for proper paste deposition after the stencil operation.

A laser-based solder paste height and width measurement system is used with

a resolution of 0.0001 inch (2.5 µm). This system provides real time informa-

tion on the uniformity of solder paste deposition. All boards are subjected to

this measurement prior to the reflow operation.

A batch reflow operation is used to create the solder joints of the SMT PWAs.

The SMT PWAs are thermally profiled using a M.O.L.E.® – a thermocouple


is attached to the PWB and to the M.O.L.E. The latter is a microprocessor-

based data logger attached to a computer. Thermal profiling is done to elimi-

nate thermal shock during preheat and reflow. This operation consists of a va-

por phase reflow machine using a constant boiling perfluorocarbon material

(3M Perfluorocompound FC-5312®) (b.p. 216°C) for soldering the SMT

PWAs. The SMT PWAs are preheated to remove paste volatiles and to initi-

ate the activation stage of the paste. The reflow liquid, since it boils at a con-

stant temperature, minimizes the possibility of overheating the SMT PWAs

during reflow and ensures that the vapor blanket performs a uniform and con-

sistent soldering operation.

4. CRITERIA FOR CHOOSING A NEW CLEANING SYSTEM

The key criteria in choosing a new cleaning system were:

Safety and ease of handling;

Performance;

Cost.

Since JPL’s need is low throughput, a batch cleaning system was acceptable.

After various preliminary trials, a centrifugal cleaning system was chosen based

both on performance and versatility. In addition, several new aqueous cleaning

chemistries seemed very promising.

One of these is based on an aqueous chemistry containing a mixture of some

alkoxypropanols with one to three alkoxy units (ether linkages). The molecules

are not particularly large (C2 to C4), so the hydrophobic portion is not too large.

The hydrophilic part of the molecule is due to one alcohol group (-OH) and sev-

eral ether groups (-O-). Overall the organic molecules exhibit excellent solubility

in water. Thus, the cleaning agent in water is herein referred to as an aqueous

cleaning solution. The material itself is easily biodegradable. It has zero ozone

depletion potential (ODP), virtually no greenhouse warming potential (GWP), and

is classified as non-flammable. The following information is supplied by the

manufacturer of the aqueous solution. Although the concentrate is 91% by weight

volatile organic compound (VOC), the material as used in the cleaning system is

only 13.6% by weight VOC.

A broader description of aqueous cleaning systems is provided in the refer-

ences [5-7].

5. NEW CLEANING SYSTEM

The new cleaning system consists of the following equipment and materials. A

brief description of its operation is also given below.


5.1. Equipment

The following equipment is required:

Centrifugal cleaning system;

Vacuum oven;

Refractometer.

5.1.1. Equipment description

The equipment consists of an enclosed stainless steel cylindrical process chamber

with a series of spray nozzles located vertically. A robotic arm containing a fix-

ture holds the PWA and moves it in and out of the chamber vertically. During the

cleaning cycle, the PWA is lowered into the process chamber until it is com-

pletely sealed from ambient. (See Figure 1).

5.2. Materials

The following materials are used in the centrifugal cleaning system:

Aqueous system containing the mixture of some alkoxypropanols with one to

three alkoxy units (ether linkages). – 20% by volume (see Section 4 above for

further details);

Corrosion inhibitor – 1% by volume;

Defoamer – 0.1% by volume;

Figure 1. View of the centrifugal cleaning system.


Deionized (D.I.) water;

High purity nitrogen gas (N2).

Note: Hereafter, the term “aqueous cleaning solution” shall refer to the entire

aqueous system consisting of water, the mixture of alkoxypropanols with one to

three alkoxy units (ether linkages), corrosion inhibitor, and defoamer. The pH of

the working solution is 10.5 (per the vendor).

5.3. Principle of operation

The centrifugal cleaning machine uses centrifugal energy to clean PWAs. Energy

is produced when PWAs to be cleaned are rotated inside an enclosed process

chamber filled with the aqueous cleaning solution (see Section 5.2). This energy

causes penetration of the solution under the components, including low profile

components such as ball grid arrays (BGAs), dissolving the contaminants. The

contaminants are subsequently removed during the rinse operation.

5.4. Overall process description

The cleaning process consists of four-stage operation. The first stage is a nitrogen

purge of the process chamber. The second stage is a wash cycle with aqueous

cleaning solution. The process chamber is filled with appropriate amount of aque-

ous cleaning solution. The PWA, while immersed in the solution, is rotated in the

chamber for a predetermined duration. At the end of the cycle, the solution is cy-

cled back to the storage tank. During the third stage, the deionized (D.I.) water

rinse sprays are activated while the PWA is rotating in the chamber. During this

cycle any remaining material is removed, and final cleaning is achieved. In the

fourth stage, filtered hot air is pumped in the chamber as the PWA rotates and

dries. During these cycles, the PWA rotates alternately, clockwise and counter

clockwise, to achieve optimum cleaning and drying.

6. TESTING OF THE NEW CLEANING SYSTEM

In order to investigate the new cleaning system, a comparison was made between

it and the initial cleaning system. The following objectivcs were pertinent to this

investigation.

6.1. Objectives

The two chief objectives were:

Investigate the new centrifugal cleaner using the aqueous cleaning solution

for flight PWAs;

Establish the optimal cleaning cycle for the new equipment.


To be able to recommend the new centrifugal cleaner using the aqueous clean-

ing solution, the procedure used was to compare the cleaning data of the older

cleaning process using the semi-aqueous (SA) solution with the centrifugal

cleaner using the aqueous cleaning solution.

6.2. Test procedure

The test procedure consisted of assemblying a test PWA that would prove chal-

lenging to clean. Several alternative cleaning runs using the new centrifugal

cleaning equipment were made. The data so obtained were compared with (1) the

test PWA used in the semi-aqueous (SA) cleaning system using the standard SA

cleaning cycle, and (2) a control PWA not cleaned at all.

6.2.1. Test PWA

The test PWA was populated with ball grid arrays (BGAs), a chip scale package,

quad flat packs (QFPs) (20-mil pitch and 25 mil pitch – the nearest metric sizes

are 0.5 mm and 0.625 mm), a plastic leaded chip carrier (PLCC), a flat pack, a

small outline integrated circuit (SOIC) and several discrete chip capacitors and re-

sistors. Both sides of the PWA were populated. The test PWA was assembled us-

ing Sn 63 paste with rosin mildly activated (RMA) flux and soldered in a vapor

phase reflow system operating at constant temperature of 216°C. (See Figures 2

and 3).

6.2.2. Test parameters

The following test parameters were employed: The basic equipment parameters of the centrifugal cleaning machine such as

the temperature of the solution, the rotational speed of PWA and the drying

temperature of the air were kept constant for all the tests.

The only parameters that were varied were the cycle times:

1. Wash cycle time;

2. Rinse cycle time.

6.2.3. Cleanliness determination methods

The following methods were used to assess the achieved cleanliness levels:

Ionic contamination levels were determined using an Ionograph® 500 ionic

contamination tester. In addition, testing was performed using ion chromatog-

raphy (IC) to profile the various ionic species.

Total low volatile residue (LVR) determination consisted of an extraction

with Freon® TF and isopropyl alcohol (IPA) followed by a gravimetric deter-

mination. The total LVR was considered to be equal to organic rosin residue

since rosin residue predominates in flux residue.

Residual chloride analysis (Cl–) using ion chromatography (IC) was em-

ployed. For one run, residual fluoride (F–) and bromide analyses (Br

–) were

also performed.


Figure 2. Top view of the Test PWA.

Figure 3. Bottom view of the Test PWA.


Outgassing per ASTM E595, “Standard Test Method for Total Mass Loss and

Collected Volatile Condensable Materials from Outgassing in a Vacuum En-

vironment.”

Either ionic contamination testing was performed using the Ionograph® 500 or

total LVR was determined for a given sample, but not both, i.e., the tests are mu-

tually exclusive of each other. This is because in the process of conducting the

ionic contamination test, the PWA is cleaned, thus rendering it unfit for further

cleanliness testing. This is indicated in the results (Tables 1-3) using the symbol

N/A (not applicable) in one or the other column.

However, the total low volatile residue (LVR) analysis and the residual chlo-

ride analysis (Cl–) are not mutually exclusive, and both examinations can be per-

formed on the same sample. They are not mutually exclusive because first an ex-

tract is made using D.I. water to remove the very soluble anions present (Cl–, F

–,

Br–), and then an extract is made using the Freon TF/IPA to remove the rosin

residue which is insoluble in water.

The results for the new centrifugal cleaning system using the aqueous chemis-

try are reported in Tables 1-3. Table 4 gives ionic contamination levels using the

older SA cleaning system.

The outgassing test method per ASTM E595 determines the volatile content of

materials when exposed to a vacuum environment. Two parameters must be

measured: Total mass loss (TML) and collected volatile condensable material

(CVCM). In addition, since polyimide printed wiring board material can absorb

moisture, an additional parameter was determined, namely, the amount of water

vapor regained (WVR). The results for the TML, CVCM and TML-WVR values

are reported in Table 5.

6.2.4. Acceptable cleanliness levels

Per JPL D-8208, “Spacecraft Design and Fabrication Requirements for Electronic

Packaging and Cabling”, the ionic contamination level as determined by the Io-

nograph® must not exceed 10 micrograms per square inch (10 µg/in

2 = 1.5

µg/cm2). If it does, the entire lot of PWAs must be recleaned and one PWA per lot

retested until this ionic cleanliness level is achieved. No acceptable standard has

been agreed upon for the amount of residual rosin; however, a limit of no more

than 150 micrograms per square inch (150 µg/in2

= 22.5 µg/cm2) seems appropri-

ate. In the case of ionic profiling using ion chromatography (IC), no acceptable

standards have been agreed upon for the amount of individual ionic species, but

one would expect that the sum of the various ionic species would be less than the

limit obtained from ionic contamination testing. Based on the outgassing determi-

nation per ASTM E595, the acceptable level for the TML must be no more than

1.00%, and the CVCM must be no more than 0.10%. If the WVR is determined,

then TML-WRV is also reported.


6.2.5. Test runs

Three test runs made with the new centrifugal cleaning machine using the aque-

ous cleaning solution are presented in Tables 1-3. As a comparison, a test run us-

ing the older SA cleaning system is presented in Table 4.

6.2.5.1. Test run #1 – new cleaning system

Centrifugal cleaning system with the aqueous cleaning solution was used. The

aqueous cleaning system consisted of 20% by volume of the long chain alcohol

solution, 1% by volume of the corrosion inhibitor, and 0.1% by volume of the de-

foamer. The wash solution temperature was 50°C; the rinse solution temperature

was 50°C; the dry air temperature was 200°C; the wash cycle rotational speed =

150 RPM. The centrifugal cleaning system parameters were: Wash time = 5.0

min.; rinse time = 2.5 min.; dry time = 2.5 min.

Note: The four different batches signify that the run was repeated at four different

times. Ionic contamination testing using the Ionograph® was done as a cleanliness

check on some of the test PWAs. In addition, total low volatile residue (LVR)

analysis and the residual chloride analysis (Cl–) were performed on other PWAs.

The results are presented in Table 1.

Table 1.

Cleanliness data from test run #1 (new cleaning system)

Test PWA Serial No.

Batch No. Ionograph®

results µg/in2*

Low volatile residue µg/in2*

Remarks

16 1 0.40 N/A

14 1 0.23 N/A Batch 1 mean ionic contamination level = 0.32 (µg/in2)

18 2 0.00 N/A


8 3 0.60 N/A

24 3 0.40 N/A

9 3 0.04 N/A


26 4 N/A 3.2 Cl– residue < 0.001 (µg/in2)

10 4 N/A 6.5 Cl– residue < 0.001 (µg/in2)

Components removed

4 N/A 0.5 Cl– residue < 0.005 (µg/in2)

7 Uncleaned PWA

152.2 N/A

* In the U.S., process engineering results are typically given in µg/in2. 1 µg/in2 = 0.155 µg/cm2.









min.; rinse time = 2.0 min.; dry time = 2.0 min.

Note: The two different batches signify that the run was repeated at two different

times. Ionic contamination testing using the Ionograph® was done as a cleanliness

check on some of the test PWAs. In addition, total low volatile residue (LVR)

analysis was performed on several PWAs. The results are presented in Table 2.

Table 2.


Test PWA Serial No.

Batch No.

Ionograph® results µg/in2

Low volatile residue µg/in2

Remarks

15 1 1.35 N/A


27 1 N/A 6.5


29 2 N/A 0.5 Parts were removed from PWB first








min.; rinse time = 6.0 min.; dry time = 3.0 min. Ionic contamination testing using

the Ionograph® was done as a cleanliness check on some of the test PWAs. Also,

total low volatile residue (LVR) analysis, residual chloride analysis (Cl–), and in

addition, residual fluoride (F–) and bromide analyses (Br

–) were performed on

other PWAs. In Table 3 the two aluminum plates were cleaned along with the

PWAs, but they were not exposed to the solder paste. The results are presented in

Table 3.


Table 3.


Test PWA Serial No.

Batch No.



Remarks (Residue results are given in µg/in2)

4 1 N/A 8.4 Rework was simulated and some flux applied to this PWA

5 1 N/A 1.9

11 1 N/A 1.6

12 1 N/A 2.3

102 2 N/A 5.8 Cl– residue < 0.000; F– residue < 0.000; Br– residue < 0.000 (µg/in2)





Al plate #1

2 N/A 0.7 Cl– residue < 0.003; F– residue < 0.002; Br– residue < 0.062 (µg/in2)

Al plate #2


Solvent (Control)


105 3 3.19 N/A

106 3 1.12 N/A


122 Uncleaned PWB

N/A 2462 Bare PWB with solder paste printed on it (bare PWB means no components).

123 Uncleaned PWB

N/A 33 Bare PWB with solder paste printed on it and then reflowed (bare PWB means no components).

6.2.5.4. Test run #4 – old cleaning system

The two-stage batch semi-aqueous (SA) cleaning system was used. The first stage

consisted in placing the PWAs vertically in a suitable conventional rack followed

by cleaning using a terpene-based SA material and water. The PWAs were then

transferred to the second machine and rinsed using a suitable saponifier, isopropyl

alcohol (IPA), and D.I. water. The conventional wash/rinse/dry cycle was used.


The SA cleaning system (old cleaning system) parameters were: Wash time = 5.0

min. (with saponifier); rinse time = 10.0 min. (D.I. H2O); 5.0 min. (D.I. H2O/IPA

mixture); dry time = 5.0 min. The results are presented in Table 4.

Table 4.

Cleanliness data from test run #4 (old cleaning system)

Test PWA Serial No.

Batch No.



Remarks

21 1 8.05 Not performed

22 1 3.32 Not performed

23 1 2.76 Not performed Batch 1 mean ionic contamination level = 4.71 (µg/in2)

6.2.5.5. Outgassing data

The two samples on which the ASTM E595 outgassing test was performed were

part of Test Run #3.

Table 5.

Outgassing data from test run #3

Test PWA Serial No.

Batch No.

TML %

CVCM %

TML- WVR %

Remarks

121 2 0.260 0.002 0.187 This sample was a printed wiring board only.

101 2 0.253 0.000 0.184 This sample was a printed wiring board assembly.

7. SUMMARY OF RESULTS

The results are summarized as follows:

PWAs cleaned with 5.0 minutes wash and 2.5 minutes rinse had average ionic

cleanliness level of 0.27 micrograms per square inch, far below the JPL

maximum acceptable ionic cleanliness level of 10 micrograms per square

inch. This result is much lower than that obtained by the older cleaner. (See

Tables 1 and 4).

PWAs cleaned with 5.0 minutes wash and 2.5 minutes rinse had average low

volatile residue (LVR) cleanliness level of 4.84 micrograms per square inch.

Although no standard exists for LVR, it is less than 6.45 micrograms per


square inch, which is the lowest level of the flight hardware determination

standard MIL-STD-1246C Level A.


cleanliness level of 0.93 micrograms per square inch, far below the JPL


inch. This result is much lower than that obtained by the older cleaning sys-

tem. (See Tables 2 and 4). The results, however, are not optimal.


cleanliness level of 1.97 micrograms per square inch, still far below the JPL


inch. These somewhat higher results may be due to the inadvertant contami-

nation by handling of some of the boards. Since the ionic cleanliness level is

still significantly lower than 10 micrograms per square inch, this result does

not vitiate the overall performance of the new cleaning system.

The anion profile analysis performed with ion chromatography showed ex-

ceedingly low levels of anion species, thus indicating very low levels of re-

maining contamination.

One PWA after cleaning had its components removed to examine for flux

residues. Both visual and 10x magnification were used to detect residues.

Nothing was noted.

The outgassing data for the boards cleaned using the new centrifugal cleaning

system/aqueous chemistry indicates that the total mass loss (TML) is much

less than 1.00% and the collected volatile condensable material (CVCM) is

much less than 0.10%.

• The optimal cleaning cycle suggested by the data is:

• Wash solution temperature 50°C

• Rinse solution temperature 50°C

• Dry air temperature 200°C

• Wash cycle rotational speed 150 RPM

• Wash time 5 min.

• Rinse time 2.5 min.

• Dry time 2.5 min.

Note on ESD

There was some concern that during the hot air drying stage there might be an

ESD (electrostatic discharge) problem. A medical device manufacturer that pur-

chased a Speedline Technologies ACCEL MicrocelTM Centrifugal Cleaning Sys-

tem was concerned about this and performed a thorough investigation. They

found no ESD problem. In addition, at JPL a normal cleaning cycle was run and

an ESD meter was used to see if there was any ESD build-up. No ESD was de-


tected either on the boards or on the equipment. Hence, it is concluded that no

ESD problem exists.

8. CONCLUSION

The centrifugal cleaner using the new aqueous cleaning solution based on long-

chain alcohols shows a marked improvement in cleanliness of PWAs over the

previous two-stage batch semi-aqueous (SA) cleaning system using a terpene-

based SA material and water in machine #1 for cleaning and saponifier, isopropyl

alcohol (IPA), and D.I. water in machine #2 for rinsing. The centrifugal cleaner

using the new aqueous cleaning solution not only cleans at a higher degree of

cleanliness level compared to the older SA cleaning system, but also it is cost ef-

fective to use. The total cycle time is about 50% less than the older SA cleaning

system. Also, it uses single chemical (the long-chain alcohol/aqueous solution)

with very small amount of additives compared to three chemicals used by the

older SA cleaning system. The use of hazardous isopropyl alcohol is also elimi-

nated.

Acknowledgements

The research to qualify this new cleaning system was performed at the Surface

Mount Technology Laboratory at the Jet Propulsion Laboratory, California Insti-

tute of Technology, under a contract with the National Aeronautics and Space

Administration. The authors wish especially to thank Mr. Charles J. Bodie and

Mr. Amin Mottiwala for their support and encouragement.

REFERENCES

1. J.K. Bonner, in Cleaning Printed Wiring Assemblies in Today’s Environment, L. Hymes (Ed.), pp. 65-119, Van Nostrand Reinhold, New York (1991).

2. L. Hymes (Ed.), Cleaning Printed Wiring Assemblies in Today’s Environment. Van Nostrand Reinhold, New York (1991).

3. C.J. Tautscher, Contamination Effects on Electronic Products. Marcel Dekker, New York (1991).

4. C.J. Tautscher, The Contamination of Printed Wiring Boards and Assemblies. Omega Scientific Services, Bothell, WA (1976).

5. F. Cala and A.E. Winston, Handbook of Aqueous Cleaning Technology for Electronic Assem-

blies. Electrochemical Publications, Isle of Man (British Isles) (1996). 6. J.B. Durkee, The Parts Cleaning Handbook without CFCs: How to Manage The Change. Han-

ser-Gardner, Cincinnati, OH (1994). 7. M.C. McLaughlin and A.S. Zisman, The Aqueous Cleaning Handbook: A Guide to Critical-

Cleaning Procedures, Techniques and Validation. The Morris-Lee Publishing Group, Rose-mont, NJ (1998).


Ed. K.L. Mittal

© VSP 2003

Investigation of modified SC-1 solutions for silicon

wafer cleaning

CHRISTOPHER BEAUDRY∗ and STEVEN VERHAVERBEKE

Applied Materials, 974 E. Arques Ave, M/S 81307, Sunnyvale, CA 94086, USA

Abstract—The RCA clean is widely used in the semiconductor industry for many wet-chemical cleaning processes. The RCA clean consists of a particle removal step, the Standard Clean 1 or SC-1 and metallic impurity removal step, the Standard Clean 2 or SC-2 step. In this work we have inves-tigated the addition of chelating agents in SC-1 solutions to prevent metallic deposition during the SC-1 step as well as to remove metallic contamination. We also have studied the effect of surfac-tants in such solutions on sub-micrometer particle removal. This leads to the development of a very fast and efficient single step RCA replacement clean. The use of a single step cleaning strategy in a single wafer mode dramatically reduces the cycle time of cleaning.

Keywords: RCA clean; silicon wafer cleaning; chelating agent; modified SC-1.

1. INTRODUCTION

SC-1 cleaning is widely used in the semiconductor industry during various wet-

chemical cleaning processes due to its outstanding particle removal efficiency.

Although SC-1, a mixture of NH4OH/H2O2/H2O, is an efficient particle removal

solution, it inherently allows some metallic impurities in solution to deposit on the

wafer surface [1]. For this reason a conventional SC-1 is typically followed by

SC-2, a mixture of HCl/H2O2/H2O, which exhibits excellent metallic impurity re-

moval efficiency [2]. This sequence of SC-1 and SC-2 is known as the RCA clean

and has been in use for over 30 years.

The most obvious advantage of adding an appropriate chelating agent to SC-1 is

to prevent the deposition of metallic impurities during the particle removal step and

thus to eliminate the need for a follow-up metallic impurity removal step. Not only

does this reduce the number of chemical cleaning steps required, saving money and

time, it also avoids the adverse effect of particle re-deposition during typical metal-

lic impurity removal steps, such as SC-2 or an HF dip. Furthermore, an appropri-

ately chelate enhanced SC-1 solution can potentially remove metallic contamination

∗To whom all correspondence should be addressed. Phone: (408) 584-0957, Fax: (408) 584-1132,



even more efficiently than SC-2, and its ability to bind free metal ions in solution

will potentially isolate process excursions from affecting process yield.

To understand the effect of adding a chelating agent to an SC-1 solution, it is

important to study the interaction of metallic impurities in solution and the sub-

strate in that solution. In this case, the substrate of interest is silicon. In aqueous

solutions, such as SC-1, a silicon wafer surface is hydroxyl terminated (silanol

groups: -Si-OH). The interaction between the metal ions in solution and the si-

lanol surface groups can be described by the following equation:

-Si-O-H + Mx+

-Si-O-M(x-1)+

+ H+ (1)

where Mx+

is the metallic ion. From equation (1), one can see that there are two

ways to reduce metallic ions from depositing on the wafer surface. The first way

is to increase the concentration of H+, shifting the reaction to the left. Unfortu-

nately, acidifying SC-1 will degrade particle removal effectiveness of the solution

(the high pH provides electrostatic repulsive forces while lowering the pH may

result in attractive forces between particles and the substrate). The second way to

prevent or reduce metallic ion deposition is to decrease the free metal ion concen-

tration in solution. For many years, suppliers have supported such an approach by

the development and use of ultrapure materials, chemicals, and de-ionized water.

Due to the increasingly stringent requirements of wafer surface cleanliness, this

approach alone cannot reach today’s required level of surface metals. In order to

reduce metal deposition in SC-1 solutions to meet and even exceed the current

surface metal specifications, it is necessary to not only use ultrapure components,

but to also add chelating agents to bind the free metal ions present forming com-

plexes which will remain soluble in solution. Typical chelating agents can reduce

the free metal ions in solution by 6 orders of magnitude [3].

In addition to enhancing the metallic cleaning ability of SC-1 solutions, we

have also investigated the use of a surfactant in our modified SC-1 solution. Al-

though SC-1 inherently removes particles quite effectively, megasonic energy is

often applied which dramatically increases particle removal efficiency. This is in-

creasingly important as the dimensional size of semiconductor devices continues

to decrease to even smaller sizes. With this mind, the addition of surfactants to

SC-1 will become an important component to prevent particles removed from the

wafer surface from re-deposition, thus increasing the particle removal efficiency

for small particles.

In liquids, the attraction or repulsion of particles to the wafer surface is de-

pendent on the van der Waals interaction (always attractive) and the electrostatic

double layer forces (usually repulsive). The combination of these interactions will

determine the potential energy of interaction and thus the barrier to adhesion [4,

5]. The barrier to adhesion is related to the particle size, pH of the solution, and

the respective charges on the wafer surface and particle. Cleaning down to sub-

micrometer and smaller sizes becomes increasingly difficult as the barrier to ad-

hesion decreases with decreasing particle size. Thus the tendency to re-deposit on

the wafer surface increases as the particle size decreases. Surfactants may prevent

Investigation of modified SC-1 solutions for silicon wafer cleaning 243

deposition in two ways (i) electrostatically by increasing surface potentials and

(ii) physically by steric hindrance not allowing particles to get close enough to the

surface for van der Waals interaction to dominate.

The focus of this work was to study the effectiveness of chelate and surfactant

modified SC-1 solutions for reducing metallic ion deposition, removing metallic

ions, and particle removal. In addition, we studied the potential for organic con-

tamination residue from both the chelating agent and surfactant.

2. EXPERIMENTAL

We carried out experiments using a modified SC-1 solution with a composition of

1:2:40 (NH4OH:H2O2:H2O) to 1:2:80 (by volume). The concentration of chelating

agent (carboxylic acid based) was varied, but was less than 1wt% of the solution.

The concentration of the surfactant (Valtron SP2200 manufactured by Valtech

Corporation, USA) was also varied, but was less than 1wt%. The measured pH

value was approximately 9.6. Megasonic energy was applied during the modified

SC-1 step (power density 1.13 W/cm2). The process time was 30 to 60 seconds at

a temperature of 50°C or 80°C followed by a rinse at the same temperature and a

spin dry. All wafers were cleaned in a single wafer mode.

Sample wafers for particle removal studies were prepared with an automated

aerosol particle deposition tool made by MSP Corporation, USA (Model 2300D).

The particle deposition pattern was a combination of full random coverage and a

spot (see Figure 1 for an example). In total, approximately 2300 Si3N4 particles

Figure 1. Particle Removal - Example of Si3N4 particle wafer maps before (left) and after (right)

modified SC-1 clean (particles ≥ 0.12 µm, measured on Tencor SP-1).


were deposited on prime 300 mm wafers. The particle measurements were per-

formed on a Tencor SP-1 instrument. Surface metal measurements were obtained

with the vapor phase decomposition-ion coupled plasma mass spectroscopy

(VPD-ICPMS) technique. Time-of-flight secondary ion mass spectrometry (TOF-

SIMS) was used to assess if any residual chelating agent or surfactant remained

on the wafer surface (after the rinsing and drying).

3. RESULTS

Figure 2 shows the particle removal efficiency for an optimized modified SC-1

solution. For one lot, consisting of 13 wafers, the average particle removal effi-

ciency was 99.5% (1σ = 0.28; measured at ≥ 0.12 µm). Typical wafer maps illus-

trating the combination of full and spot Si3N4 particle deposition pattern before

and after processing are shown in Figure 1. In order to determine the effectiveness

of the surfactant studied, Valtron SP2200, we compared final particle counts with

and without surfactant present (Figure 3). In this example we see an average of 50

less particles per 300 mm wafer. The particle cleaning performance for the modi-

fied SC-1 solution was excellent and the selected surfactant reduced the average

final particle count after SC-1 cleaning.

Figure 4 illustrates the effectiveness of the studied chelating agent for reducing

metallic deposition. In particular, it is interesting to look at the level for Al, Fe,

and Zn. These are some of the metals that readily deposit from conventional SC-1

Figure 2. Particle Removal - Average Si3N4 particle removal after modified SC-1 clean (particles

≥ 0.12 µm, measured on Tencor SP-1).


Figure 3. Particle Removal - Final particle counts after modified SC-1 clean with and without sur-

factant (particles = 0.1-0.14 µm, measured on Tencor SP-1).

Figure 4. Metal Deposition - Surface trace metals levels after modified SC-1 clean as determined by VPD-ICPMS (1 sigma error bars are generally within data points).


solutions [1]. The average surface metals levels after the modified SC-1 clean was

equal to or below today’s VPD-ICPMS detection limits. For reference, typical

levels for a conventional SC-1 last clean are: Al ~ 1x1011

, Fe ~ 2x1010

, and Zn ~

1x1011

(atoms/cm2). The chelating agent under investigation is efficiently binding

the free metal ions in solution reducing their deposition onto the wafer surface

and thus can eliminate the need for an additional metal removal step. Although

the modified SC-1 solution did not deposit metals from the solution we also char-

acterized the metal removal efficiency of this solution (Figures 5 and 6). The con-

centration of chelating agent and exposure time were varied while the

NH4OH:H2O2:H2O volume ratio and temperature were held constant at 1:2:80 and

80°C. Figure 5 shows the results for a 30 s processing time. The final concentra-

tion of all metals investigated was typically greater than 1E+10 atoms/cm2. Fe

removal was found to be a function of chelating agent concentration while other

Figure 5. Metal Removal - Surface trace metals levels after a 30 s modified SC-1 clean using differ-ent chelating agent concentrations as determined by VPD-ICPMS.

Figure 6. Metal Removal - Surface trace metals levels after a 10 minute modified SC-1 clean using different chelating agent concentrations as determined by VPD-ICPMS.


metals did not exhibit any dependence. Figure 6 shows the final concentrations of

metal after 10 minutes of exposure. All metals were reduced by 2-3 orders of

magnitude to close to or below 1E+10 atoms/cm2. Exposure time is obviously an

important consideration for metal removal. Methods to increase the metal removal

rate are now under investigation.

One of the concerns with the use of chelating agents and/or surfactants in SC-1

last clean is the potential for organic contamination remaining on the surface of

the wafer. However, in a typical spin cleaning equipment, the rinse process can be

optimized to eliminate such concerns. The use of heated DI water rinse and high

spin rates during rinsing can effectively remove all traces of the chelating agent

and surfactant. TOF-SIMS measurements were carried out to confirm the absence

of both organic additives. No trace of additive-specific residues was observed on

the processed wafers. Thus, organic contamination through use of appropriately

selected chelating agents and/or surfactants can be eliminated through process op-

timization. The absence of any heavy metal signature in the TOF-SIMS data also

confirms our VPD-ICPMS results.

4. SUMMARY

In this paper we have shown that the addition of an appropriately selected chelat-

ing agent to SC-1 solutions can eliminate the need for an additional metal removal

step, potentially saving time and money. In addition, the use of a surfactant can

enhance particle removal efficiencies for very small particle sizes (<0.14 µm).

This modified SC-1 solution, containing both additives, was shown to have excel-

lent particle removal efficiency, to reduce metal deposition on the wafer surface

to below VPD-ICPMS detection limits, and to remove surface metal contamina-

tion. Methods to improve metal removal efficiency are currently underway. Fur-

thermore, rinsing can be optimized to eliminate all traces of the chelating agent

and surfactant residues.

REFERENCES

1. H. Hiratsuka, M. Tanaka, T. Tada, R. Yohsimura and Y. Matsushita, Ultra Clean Technol., 3, No. 3, 18-27 (1991).

2. W. Kern, in: Proceedings of the First International Symposium on Cleaning Technology in

Semiconductor Device Manufacturing, J. Ruzyllo and R.E. Novak (Eds.), Vol. 90-9, pp. 3-19, Electrochemical Society, Pennington, New Jersey (1990).

3. A. Ringborn, Complexation in Analytical Chemistry, John Wiley & Sons, New York (1963). 4. R. Donovan and V. Menon, in: Handbook of Semiconductor Wafer Cleaning Technology: Sci-

ence, Technology, and Applications, W. Kern (Ed.), pp. 152-197, Noyes Publications, West-wood, New Jersey (1993).

5. M. Itano and T. Kezuka, in: Ultraclean Surface Processing of Silicon Wafers: Secrets of VLSI

Manufacturing, T. Hattori (Ed.), pp. 115-136, Springer-Verlag, Berlin (1995).


Ed. K.L. Mittal

© VSP 2003

Performance qualification of post-CMP cleaning

equipment in a semiconductor fabrication environment

MICHAEL T. ANDREAS∗

Micron Technology, Inc., Mail Stop 306, 8000 S. Federal Way, Boise, ID 83707-0006, USA

Abstract—An inexpensive qualification technique is described for wafer cleaning tools used after chemical-mechanical polishing (CMP). Pipette deposition of slurry onto a monitor wafer can pro-vide the particle challenge needed to qualify the performance of a post-CMP wafer cleaning tool. In addition to gauging the performance of these tools, this pipette method is faster and less expensive than many common particle deposition techniques, including immersion, polishing and aerosol deposition.

Keywords: Brush cleaning; chemical-mechanical polishing; CMP; particle deposition; PVA; slurry; wafer cleaning.

1. INTRODUCTION

Surface preparation and cleaning is one of the most critical steps in semiconduc-

tor manufacturing [1]. For all wafer cleaning tools, routine qualification is neces-

sary to ensure that no particle contamination is introduced by the wafer cleaning

equipment [2]. The broad category of wafer cleaning tools includes the poly(vinyl

alcohol) (PVA) brush scrubbing tool [3]. The brush scrubber has been increas-

ingly utilized [4] in semiconductor fabrication as a preferred technique for particle

removal after CMP. Because of the high particle removal performance required of

post-CMP cleaning tools, it is critical to monitor and maintain the performance of

such tools [5]. The most direct measure of tool performance is inline inspection of

actual product wafers [6]. While inline defect analysis is invaluable, it may re-

quire a time lag of hours or even days between wafer cleaning and discovery of

high wafer defectivity. In a high-volume manufacturing environment, this delay

can lead to hundreds of product wafers with possible contamination. For this rea-

son, inline inspection of product wafers is supplemented by regular tool qualifica-

tion using less expensive particle monitor (PMON) wafers. This PMON qualifica-

tion should provide an accurate measure of the tool performance with the quickest

possible turnaround time. For post-CMP cleaning tool qualification, it is neces-


M.T. Andreas 250

sary to use prepared PMON wafers with contamination analogous to that found

on polished product wafers. There are several methods for preparing these con-

taminated monitor wafers. One way is to use polished monitor wafers [7]. In this

method, test wafers can be selected which represent the surface chemistry of

product wafers without the expense of underlying circuitry. The test wafer surface

can be homogeneous or heterogeneous, depending on the process being qualified.

These test wafers can be polished under the same CMP conditions as product wa-

fers. This method requires the same resources as inline product inspection, there-

fore providing an accurate measure of post-CMP tool performance but not neces-

sarily decreasing the turnaround time. Another common method is the slurry dip

[8, 9], where monitor wafers are dipped in a wet process tank of diluted slurry.

Again, monitor wafers can be selected to represent the surface chemistry of prod-

uct wafers. The diluted slurry can be selected to simulate the CMP chemistry.

This method has the advantage of providing contaminated monitor wafers without

the time or expense of using a CMP tool. Also, it is possible to deposit dry parti-

cles using an aerosol deposition technique [10]. While it is claimed that this

method is more controllable and repeatable than aqueous slurry immersion, these

dry particles may not represent polishing residue as accurately as a CMP slurry.

Here we describe an extremely simple contamination technique – direct pipette

deposition of a small volume of undiluted slurry onto a monitor wafer.

2. EXPERIMENTAL

Bare silicon and blanket oxide wafers were used for all tests. Blanket oxide wa-

fers were prepared by plasma enhanced chemical vapor deposition (PECVD) us-

ing tetraethoxysilane (TEOS). These TEOS derived films were deposited to 350

nm thickness on 200 mm diameter silicon substrates. All wafers were cleaned us-

ing OnTrak DSS-200 Series II brush cleaning tools. These tools were run using a

dilute (<1%) basic cleaning solution. All wafers were inspected with a Tencor

SurfScan 6420 laser scattering wafer inspection tool [11, 12]. Bare silicon moni-

tor wafers were inspected for all light-scattering point defects (LPDs) >0.16 µm.

Blanket oxide wafers were inspected for LPDs > 0.18 µm.

3. RESULTS

3.1. Deposition of the slurry drop

Slurry drop testing was first used to investigate the scrubber response to different

slurry types. The slurries investigated contained abrasive materials of alumina,

ceria, fumed (furnace-grown) silica, and colloidal (solution-grown) silica. In this

experiment, a large (~ 0.2 ml) drop of each slurry was deposited directly onto the

center of each blanket oxide wafer. These contaminated wafers were processed

through the wafer scrubber. The total brush cleaning time was varied as an ex-

Performance qualification of post-CMP cleaning equipment 251

perimental factor. Uncontaminated silicon wafers were processed immediately be-

fore and after the oxide wafers to determine any slurry particle carryover. The re-

sults from this test are summarized in Table 1. A plot of post-scrub LPD count

versus brush cleaning time is shown in Figure 1. Here we determined that among

all the slurry types, alumina slurry provided the highest level of wafer contamina-

Table 1.

Initial slurry drop test results

Experimental conditions SurfScan total LPD counts > 0.16 µm

Run order

Wafer surface

Contamination Brush time, sec

Before slurry deposition

After slurry deposition and PVA scrub process

Difference

1 silicon none 80 4 256 252

2 CVD oxide alumina 2 34 18596 18562





7 silicon none 80 12 596 584

8-24 silicon none 80 – – –

25 silicon none 80 6 113 107


27 CVD oxide fumed silica 2 32 786 754

28 CVD oxide fumed silica 20 32 54 22

29 CVD oxide fumed silica 40 65 32 -33

30 CVD oxide fumed silica 80 120 32 -88


32-48 silicon none 80 – – –

49 silicon none 80 2 296 294


51 CVD oxide ceria 2 62 2394 2332

52 CVD oxide ceria 20 57 40 -17





57 CVD oxide colloidal silica 2 29 2680 2651

58 CVD oxide colloidal silica 20 35 77 42

59 CVD oxide colloidal silica 40 40 39 -1




M.T. Andreas 252

tion (as measured by laser scatterometry) for a given drop size. Maps showing the

LPD distribution after 2 second scrubbing are shown in Figures 2 and 3 for wafers

contaminated with alumina and colloidal silica, respectively. In all cases, the 80

second brush cleaning time was sufficient to attain particle levels below the pre-

defined production limit of 100 LPDs. Silicon monitor wafers run before and after

each group of slurry drop test wafers did not show significant slurry carryover.

Due to the simplicity of this procedure, slurry drop deposition was investigated

further as a method for routine tool qualification.

3.2. Development of the slurry drop qualification method

To evaluate the resolution of scrubber qualification methods (SQMs), several ex-

perimental scrubber recipes were created which simulated sub-optimal tool per-

formance [13]. These scrubber recipes are described in Table 2. Several different

SQMs were evaluated using these sub-optimal scrubber recipes. These methods,

including slurry drop, slurry immersion and CMP polishing, are described in Ta-

ble 3. The oxide polish method (SQM index 6) had been in use in our production

line for some time prior to this experiment. Silica-containing slurries were used

for all contamination methods because these slurries were the most readily avail-

able at the time. For the slurry immersion methods, 10 ml of slurry was diluted

with 18 L of deionized water, and the wafers were immersed for 10 sec immedi-

ately before cleaning. For the polished oxide wafers, a 60 second pre-clean using

~ 0.5% HF was utilized between polishing and scrubbing. Many trials were re-

peated using new (freshly installed) and old (near end of service) PVA brushes.

For each combination of qualification method and experimental scrubber recipe,

Figure 1. Plot of post-scrub LPD total vs. brush time.


Figure 2. Wafer map showing LPD distribution after alumina slurry drop and 2 second scrub.

Figure 3. Wafer map showing LPD distribution after colloidal silica slurry drop and 2 second scrub.

M.T. Andreas 254

Table 2.

Experimental scrubber recipes

Recipe Brush height, mm

Chemical flow, L/min

Brush rotation, 1/min

Brush time, sec/brush

Rinse time, sec

C 3.5 0.5 139 40 9

E1 1.0 0.5 139 40 9

E2 3.5 0.5 40 40 9

E3 3.5 0.5 139 10 9

E4 3.5 0.5 139 20 9

E5 3.5 0.0 139 40 9

E6 3.5 0.5 139 40 5

E7 3.5 0.5 139 40 13

Table 3.

Experimental scrubber qualification method details

SQM index Wafer surface Contamination

1 silicon 2 drops (~ 0.1 ml) slurry

2 silicon dilute slurry immersion

3 silicon none

4 CVD oxide 2 drops (~ 0.1 ml) slurry

5 CVD oxide dilute slurry immersion

6 CVD oxide 30 sec polish and 60 sec dilute HF clean

the difference in LPD counts (dLPD) was determined whereby dLPD =

LPD(post cleaning) – LPD(pre-contamination). Comparing dLPDs for all ex-

perimental scrubber recipes for each series of SQM, slurry type and brush con-

dition, a method dynamic range (MDR) was determined as the range between

the highest and lowest dLPD results for that series. This dynamic range gives an

indication of the utility for a given procedure to “catch” sub-optimal tool per-

formance. The results from all trials are presented in Table 4. Considering

dLPDs for new brush vs. old brush conditions, the most sensitive method for

monitoring brush wear is clearly the bare silicon SQM (index 3). Considering

the method dynamic range across scrubber recipes, the silicon wafer methods

(indices 1-3) in general are more sensitive than the oxide wafer methods (indi-

ces 4-6). This may be due, in part, to the higher sensitivity of the SurfScan in-

spection used for bare silicon wafers. Among the silicon wafer methods, the

silicon drop SQM (index 1) provides the most accurate measure of scrubber per-

formance independent of brush age.


Table 4.

Experimental details, dLPD results and method dynamic range (MDR) for each series

Series conditions dLPD for each scrubber recipe MDR

SQM index

Silica Brush age

C E1 E2 E3 E4 E5 E6 E7

1 colloidal old 105 242 69 365 103 66 48 1516 1468

1 fumed new 5 545 10 8 5 30 20 17 540

2 colloidal old 114 324 95 817 84 84 121 1275 1191

2 fumed new 2 26 40 32 8 12 7 8 38

3 – old 3857 1107 1547 – 7495 >30000 8421 3446 >30000

3 – old 357 565 149 1020 403 228 204 3485 3336

3 – new 68 126 – – 33 72 – – 96

3 – new 10 25 15 10 4 11 24 19 21

4 colloidal old 5 21 -2 8 9 0 41 1 43

4 fumed old 39 68 83 – 92 3 – 30 89

4 fumed new 3 55 120 38 -26 26 26 99 146

4 fumed new 5 7 11 – 1 10 2 0 10

5 colloidal old 243 13 22 34 43 133 50 2 241

5 fumed new -75 -46 5 1 14 -40 -71 -58 89

6 colloidal old 4 51 -1 43 8 66 82 56 82

6 colloidal new 3 48 -18 -17 -50 6 -32 -3 98

3.3. Improvement of the slurry drop scrubber qualification method

After implementation of SQM index 1 for a period of time, a new defect pattern

was discovered on product wafers that was related to brush-induced residuals at

tungsten CMP scrub. This defect pattern was characterized by a radial pattern of

slurry residuals. Although this defect pattern was detectable on product wafers, it

did not appear on scrubber qualification wafers. This detection gap led to further

optimization of the scrubber qualification method. First, the full experimental

space of wafer type, slurry type, and slurry amount versus scrubber performance

was revisited. To simulate sub-optimal scrubber performance, two new scrubber

recipes were created: A1 and A2. Both recipes feature a reduced brush pressure

(adjusted by way of brush height) with shorter brush process times. Also, recipe

A2 uses a slower brush rotation. The experimental conditions and inspection re-

sults are given in Table 5. Several wafer maps from this group are shown in Fig-

ures 4 and 5. The reduced efficiency of recipe A1 provided good discrimination

between qualification parameters (e.g. slurry type or wafer type). In general, alu-

mina slurry was more sensitive than silica slurry to radial defect pattern forma-

tion. As for slurry quantity, three drops of alumina appears optimal. Silicon wa-

fers worked better than oxide wafers because the dLPD and MDR results

corresponded more accurately to the expected particle removal performance of

M.T. Andreas 256

Table 5.

The full factorial of scrubber qualification tests using a known good brush installation


Wafer surface Slurry Quantity C A1 A2

silicon silica 1 drop 25 125 >29997 >29972

silicon silica 3 drops 38 125 >28245 >28207

silicon alumina 1 drop 69 508 >29997 >29928

silicon alumina 3 drops 108 2574 >29997 >29889

CVD oxide silica 1 drop -14 2850 701 2864

CVD oxide silica 3 drops -4 376 1535 1539

CVD oxide alumina 1 drop -2 >20188 >29932 >29934

CVD oxide alumina 3 drops 247 >29959 >29933 >29712

Figure 4. Several wafer maps from the SQM optimization tests. These silicon wafers were scrubbed using recipe A1. The lower left note in each frame indicates drop size (1d = 1 drop, 3d = 3 drops) and slurry composition (S = silica, A = alumina).


Figure 5. Wafer map for a CVD oxide wafer contaminated with one drop of alumina slurry and scrubbed using recipe A1.

Table 6.

Recipe details for more sub-optimal scrubber recipes used to test the sensitivity of optimized SQMs. Recipe C is the control scrubber recipe.

Recipe Brush Module 1 Brush Module 2

Rotation, 1/min

Height, mm Time, sec Rotation, 1/min

Height, mm Time, sec

C 139 3.5 40 139 3.5 40

A1 139 1.5 30 139 1.5 30

A2

38 1.5 30

38 1.5 30

T1 139 1.5 40 139 1.5 40

T2 139 1.5 40 139 3.5 40

T3 139 3.5 40 139 1.5 40

experimental scrubber recipes A1 and A2. The improved SQM using 3 drops of

alumina slurry was selected for comparison to the previous SQM using 1 drop of

colloidal silica slurry. Based on results with experimental recipes A1 and A2,

three more sub-optimal scrubber recipes were created. These recipes (T1, T2 and

T3) were designed to perform somewhere between recipes C (control) and A1.

These recipes are described in Table 6. Bare silicon wafers were used for all fur-

ther tests. Wafer run order was randomized to average out any brush carryover ef-

fects. Process details and results are shown in Table 7. Wafer maps showing sen-

sitivity to radial defect pattern formation are shown in Figure 6. The dLPD results

comparing optimized alumina and silica slurry drop methods are plotted in Figure

7. The alumina slurry drop SQM shows better sensitivity to inefficient scrubber

M.T. Andreas 258

operation, especially using recipe T1. All wafers using silica and alumina SQMs

met contol levels for dLPDs when using the control scrubber recipe. One of the

silica drop trials showed anomalously high residue. This may have been carryover

from the previously scrubbed wafer, which brought alumina slurry contamination.

Further tests confirmed that carryover from 3 drops of alumina appeared when us-

ing recipe T1. All experiments confirmed the improved alumina drop SQM as

more sensitive to conditions which may cause radial defect patterns. After imple-

menting this improved SQM, no further radial defect patterns were discovered on

product wafers.

Table 7.

Process details and inspection results for improved slurry drop SQM comparison. All tests were run using bare silicon wafers. The high dLPD result for 1dS wafer 19 (5312 adders) may be due to carryover from 3dA wafer 18


Slurry Quantity Run order C T1 T2 T3

silica 1 drop 1, 4, 9, 11 30 34 67 236 206

2, 5, 13, 15 30 23 69 308 271

3, 6, 22, 19 38 34 83 5312 5278

alumina 3 drops 7, 12, 14, 8 19 3711 80 202 3692

17, 18, 16, 10 100 30553 169 580 30453

20, 24, 21, 25 92 31077 128 468 30985

Figure 6. Representative wafer maps from the 3 drops alumina (3dA) qualification method on sili-con wafers, showing sensitivity to radial defect patterns when using recipes T1 and T3.


Figure 7. Plot of dLPD versus scrubber recipe for optimized alumina and silica slurry drop SQM trials.

4. CONCLUSION

Manual pipette deposition of CMP slurry onto a monitor wafer is a quick and ef-

fective way to provide qualitative particle challenges to wafer cleaning equip-

ment. This method is much faster and cheaper than other common particle deposi-

tion techniques, including polishing, aerosol deposition and immersion in dilute

slurry. We have shown that this method provides enough particle loading to de-

termine whether or not a post-CMP cleaning tool will perform within acceptable

particle removal limits. This, in turn, translates to lower defects on product wafers

and improved yields at a lower quality control cost.

REFERENCES

1. W. Kern (Ed.), Handbook of Semiconductor Wafer Cleaning Technology, pp. 416-419, Noyes Publications, Park Ridge, NJ (1993).

2. F.W. Kern, Jr. and G.W. Gale, in: Handbook of Semiconductor Manufacturing Technology, Y. Nishi and R. Doering (Eds.), pp. 87-104, Marcel Dekker, New York (2000).

3. W. Krusell, J.M. de Larios and J. Zhang, Solid State Technol., 38, No. 6, 109-114 (1995). 4. R. DeJule, Semiconductor Intl., 56-64 (Nov. 1998). 5. J.M. de Larios, J. Zhang, E. Zhao, T. Gockel and M. Ravkin, MICRO, 15, No. 5, 61-73 (1997). 6. C. Dennison, MICRO 16, No. 2, 31-42 (1998). 7. D.W. Cooper, R.C. Linke and M.T. Andreas, MICRO 17, No. 7, 55-64 (1999). 8. A.A. Busnaina, N. Moumen, M. Guarrera and J. Piboontum, in: Semiconductor Fabtech – 9th

Edition, M.J. Osborne (Ed.), pp. 279-282, ICG Publishing, London (1999). 9. S. Ramachandran, A.A. Busnaina, R. Small, C. Shang and Z. Chen, in: Semiconductor Fabtech

– 13th Edition, G. Oliver (Ed.), pp. 271-277, ICG Publishing, London (2001).

M.T. Andreas 260

10. Y.H. Liu, S.H. Yoo, S.K. Chae, J.J. Sun, K. Christenson, J. Butterbaugh, J.F. Weygand and N. Narayanswami, Semiconductor Intl., 145-152 (June 2000).

11. R.S. Howland, Semiconductor Intl., 164-170 (Aug. 1994). 12. J.J. Shen and L.M. Cook, MICRO 15, No. 3, 53-66 (1997). 13. N. Moumen, M. Guarrera, J. Piboontum and A.A. Busnaina, in: Proceedings, 10th Annual

IEEE/SEMI Advanced Semiconductor Manufacturing Conference, pp. 250-254 (1999).


Ed. K.L. Mittal

© VSP 2003

Spatial and temporal scales in wet processing of deep

submicrometer features

MOSHE OLIM∗

Seagate Technology, 7801 Computer Avenue, South Bloomington, Minnesota 55435, USA

Abstract—Liquid-phase processing is commonplace in manufacturing of thin films. Typically, the surface processed has distinct topological features such as trenches and vias. A typical liquid phase process cycle starts with a dry surface and consists of the following steps: (1) wetting of the surface, (2) dispensing a mix of chemical reagents, (3) rinsing the surface, and (4) drying the surface. Step 2 may consist of a sequence of chemical reagents either with or without a rinse in between. Each of the steps is governed by different physical processes which may have distinctly different spatial and temporal scales. These scales are addressed in the paper. A trench is used as a representative feature.

Keywords: Microscale transport; hydrophilic surface; thin films; wet processing.

1. INTRODUCTION

The typical processed surface is dry as the process starts. If a chemical reagent is

dispensed onto a dry surface it is likely that different parts of the surface will be

subjected to the reagent for different time intervals thus resulting in a nonuniform

processing result. Therefore, prior to dispensing chemical reagents on the surface,

it is imperative that the surface be covered with an inert liquid. This liquid is typi-

cally deionized water which covers the whole surface and fills the trenches. The

time required to fill a trench depends strongly on the characteristics of the surface

processed, the surface tension of the water, and the width and depth of the trench.

If the trench is hydrophilic, the capillary action of the water/air interface will en-

sure that the trench fills with water. The geometry of the process is shown in Fig.

1. The pressure of the gas trapped in the feature is increased due to the capillary

force, and this increase in pressure enhances the diffusion of the gas into the liq-

uid. The process continues until the gas trapped in the cavity is completely con-

sumed by this diffusion process. The upper limit on the time required to fill the

trench can be estimated (for details see [1]) as follows:

∗Phone: 952-402-5888, E-mail: [email protected]

M. Olim 262

Figure 1. Geometry of the trench filling process. PL

is the pressure in the liquid, ∆P is the capillary pressure difference across the interface, w and h are the trench width and depth, respectively, and θ is the contact angle.

Figure 2. Time required to complete the trench filling process as a function of trench width for con-tact angle values of 30 and 60 degrees. The surface tension of the liquid is 70 mN/m, and trench as-pect ratio h/w = 10.

21 hfill

k Pt h

DRT P

+ = ∆ (1)

where k, D, R, T, and h are, respectively, a proportionality constant, vapor/air dif-

fusion coefficient, universal gas constant, temperature, and trench depth, and Ph

and ∆P are the atmospheric pressure and capillary pressure difference, respec-

Spatial and temporal scales in wet processing of deep submicrometer features 263

tively, across the interface. The results of the calculation are shown in Fig. 2. It

can be seen that the time required to fill a 0.25 µm wide trench is well below one

second even for a trench whose aspect ratio is 10.

2. TRANSFER OF REAGENTS INTO AND OUT OF THE TRENCH

With the trench full of DI water, a chemical reagent is dispensed onto the sub-

strate. In the interest of uniformity of processing along the full depth of the

trench, it is important that the spatial concentration of the reagent along the trench

depth be kept as uniform as possible. For analysis purposes, one may assume that

the trench is full of water and the top of the trench is covered with liquid reagent.

The reagent may penetrate the trench through either (a) convective or (b) diffusive

mixing. In order for convective mixing to take place, the flow characteristics must

allow for vortices to exist. The possibility of vortex existence may be ruled out by

comparing the relevant geometric parameters to the smallest vortex diameter pre-

dicted by Kholmogorov scales (see [2]) using the following equation:

3/4ul

l v

− ≈

η (2)

where η is the smallest length scale that can sustain turbulence (i.e. the smallest

vortex diameter possible), l is a characteristic length of the system (in this case it

is the width of a trench), and u and v are the flow velocity and kinematic viscos-

ity, respectively. With relatively large trench width and velocity values of l = 0.5

µm and u = 1 m/s one obtains η/l ≈ 1, and with more realistic values of l = 0.25

µm and u = 0.01 m/s one obtains η/l ≈ 100. Since the smallest possible vortex di-

ameter is noticeably larger than the trench width, it is clear that turbulent mixing

cannot take place in the trench. This implies that the reagent is transferred into the

trench by diffusion only. The same argument applies to transfer of reagents out of

the trench when DI water is dispensed onto the surface in order to stop the reac-

tion.

Since it is desired that the results of the chemical process be uniform along the

depth of the trench, it is clear that the exposure time of any point on the trench

wall to the reagent should be as close as possible to that at any other point along

the trench wall. The uniformity of the process may be estimated by comparing the

time it takes for the reagent concentration at the bottom of the trench to equalize

with that at the top of the trench. For practical purposes, let us consider the con-

centrations equalized when the concentration at the bottom reaches 95% of the

concentration at the top. The time required for the reagent concentration at the

trench bottom to reach a given concentration level can be estimated (see [1]) us-

ing the following one-dimensional diffusion equation:

M. Olim 264

Figure 3. Normalized chemical concentration at the trench bottom vs time. u is the velocity of the liquid at the top of the trench, and D is the diffusion coefficient of the chemical in the liquid.

2

20

C C

t

z

∂ ∂+ =∂ ∂

(3)

where D = 1.e-9 m2/s is the diffusion coefficient and C is the reagent concentra-

tion subject to C(t,z=h) = 1, Cz(t,z=0) = 0, C(t=0,z<h) = 0. The results shown in Fig. 3

clearly show that the concentration at the trench bottom reaches 95% of the

concentration at the trench top in approximately 25 ms.

The transport of the reagent out of the trench is described by the same equation

with different initial and boundary conditions C(t,z=h) = 0, Cz(t,z=0) = 0, C(t=0,z<h) = 1

and calculations show that within 100 ms the maximum concentration is reduced

by seven orders of magnitude.

3. DRAINAGE

Typically, the de-ionized water used to rinse the substrate must be removed (in

the liquid state) from the substrate and out of the features fabricated in the sub-

strate. Water that is not removed in the liquid state would evaporate leaving origi-

nally dissolved contaminants to coagulate on the substrate and in the features thus

adversely affecting the yield of the manufacturing process.

Spatial and temporal scales in wet processing of deep submicrometer features 265

Substrate rotation, since it increases the body force on the water in the features,

is often utilized as a means of water removal enhancement. However, as the typi-

cal size of the features decreases, the importance of body forces compared to sur-

face tension (manifested in the Bond number) also decreases thus reducing the ef-

ficiency of rotation as a mechanism for water removal. It has been shown [4] that

the amount of water removed from a trench for a given contact angle does not

change significantly below Bond number of 0.1 which may, therefore, be defined

as the critical Bond number. The parameter that determines the amount of water

removed when the Bond number is below the critical value is the contact angle θ,

and the maximum amount of water removed from the trench of width L does not

exceed πL2/4 per unit length of a long trench. Since the depth of the trench is sig-

nificantly larger than its width, very little water is removed from the trench in the

liquid phase and most of the water must evaporate.

4. EVAPORATION

This step is typically achieved by purging the process chamber using dry nitrogen

at ambient temperature and pressure. The equations describing the time interval

required to dry by evaporation a feature of a given depth are developed in [3] and

they allow for any temporal variation of vapor concentration in the ambient.

The rate of evaporation through a stagnant gas in one dimension is expressed in

Equation (17.2-15a) in [5]. A slightly modified version of this equation is pre-

sented below:

,

,

lnv top

v int

P PPDN

RTz P P

−= −

(4)

where P and T, respectively, are the pressure and temperature in the system, D is

the diffusivity of the vapor in the ambient gas, R is the universal gas constant, z is

the distance between the top of the feature and the liquid/gas interface, and Pv is

the partial pressure of the vapor either at the top of the feature (top) or at the liq-

uid/gas interface (int). This may be converted into an equation showing the rate of

recession of the air/liquid interface which, in turn, may be integrated (for details

see [3]) to yield

,2

0

1ln

1

s v top

sat

Pz ds

P

∆ −= −

(5)

where

1/2

,

,; ; ,2

v topL satv top sat

L

PRT Pz z P P

M PD P P

= = =

ρ

τ (6)

M. Olim 266

and s = t/τ is nondimensional time with τ being a characteristic time scale. To

solve Equation (5) the temporal variation of the vapor pressure at the top of the

feature must be known. Two potentially realistic situations in which the vapor

concentration as a function of time is known are (a) Constant vapor pressure Pv,top

= αPsat where 0 ≤ α ≤ 1, and (b) Exponentially decreasing vapor pressure Pv,top =

Psate–Qt/V

where Q and V are, respectively, the flow rate of the dry nitrogen and the

volume of the chamber. Situation (b) is the slower of the two, and, for a realistic

set of values Q = 1200SLPM and V = 100L, a 1 µm deep trench would require ≈

300 ms to evaporate.

5. SUMMARY

The four main steps in wet processing of submicrometer features on hydrophilic

surfaces are: (1) wetting of the surface, (2) dispensing a mix of chemical reagents,

(3) rinsing the surface, and (4) drying the surface. A trench was used as a repre-

sentative feature. The mechanisms driving each of these steps have been analyzed

analytically in this paper. Introduction of typical dimensions and physical values

into the results of the analysis yielded the time scales relevant to each of the proc-

ess steps. The mechanisms and their time scales are summarized in the table be-

low:

step driving mechanism time scale [ms]

wetting capillary action 100

chemical in diffusion 10

chemical out diffusion 100

drying evaporation 100

REFERENCES

1. M. Olim, J. Electrochemical Soc., 144, 4331-4335 (1997) . 2. H. Tennekes and J.L. Lumley, A First Course in Turbulence, The MIT Press, Cambridge, MA

(1990). 3. M. Olim, J. Microscale Thermophys. Eng., 3, 183-188 (1999). 4. M. Olim, J. Microscale Thermophys. Eng., 4, 223-230 (2000). 5. R.B. Bird, W.E. Stewart and E.N. Lightfoot, Transport Phenomena, John Wiley, New York

(1960).


Ed. K.L. Mittal

© VSP 2003

Microdenier fabrics for cleanroom wipers

JOHN SKOUFIS∗ and DOUGLAS W. COOPER

†

ITW Texwipe, 650 E. Crescent Ave., Upper Saddle River, NJ 07458

Abstract—As the state of technology advances in data storage and integrated circuits, the need to remove smaller particles becomes more critical in order to maintain economical yields and avoid product failures due to contamination. Present textile materials are approaching the limits of their ability to achieve particle removal. New materials are being developed and investigated for provid-ing the high level of cleaning efficiency required. These new materials fall into the general class known as microdeniers. They are being shown to have the properties required to overcome the shortcomings of traditional textiles.

Keywords: Contamination control; wipers; microdenier; cleanroom wipers; particle removal; fiber construction.

1. INTRODUCTION

The critical dimensions in data storage and integrated circuit technologies con-

tinue to get smaller and smaller, putting these high-tech products at risk from sub-

micrometer particles. As flying heights approach 25 nm, contamination of disk

media and read-write heads during their manufacture becomes even more of a

concern. A half-micrometer (500 nm) particle is more than ten times the read-

write gap, perhaps leading to a read-write error or a crash. Similarly, integrated

circuits continue to have ever-decreasing “line width” ground rules, requiring the

control of particles of ever decreasing size limits. Particles hundredths of a mi-

crometer in size can be “killers” [1].

The relatively large fiber diameters of standard wipers make them less efficient

in picking up these small particles. Microfibers have been shown to be more effi-

cient in this application based on their geometry and the physics of particle re-

moval [2].

∗To whom all correspondence should be addressed. Phone: 201-327-9100 X330, Fax: 201-327-

5945, E-mail: [email protected] †Now retired.


2. WIPERS AND SWABS

During plant and cleanroom construction, and then throughout operations, rigor-

ous cleaning is needed, using such consumable materials as wipers, swabs, clean-

ing compounds, and cleaning tapes. Such materials need to be selected carefully

both with respect to contaminating potential and ability to clean. Cleanliness is

measured by examining materials, extracting particulate matter in agitated liquid

or in air, and by extracting chemical constituents such as ionics, hydrocarbons,

non-volatile residue (NVR or residue after evaporation) etc., with appropriate sol-

vents [3].

Cleanroom wipers (and swab heads) selected for cleanrooms of different levels

of cleanliness generally follow these guidelines:

A. Cleanest, Fed-Std-209E Class 1 (= Class M1.5) to Class 10 (= Class M2.5)

rooms:

laundered sealed-edge polyester knit (lowest contaminants) or Nylon knit

(somewhat higher non-volatile residues). A portion of such a wiper is shown

in Figure 1.

B. Class 10 to Class 100 (= Class M3.5) rooms:

laundered polyester knit or hydroentangled polyester (somewhat higher parti-

cles and fibers due to unsealed edges).

C. Class 100 to Class 10,000 (= Class M5.5) rooms:

hydroentangled polyester/cellulose blends.

Figure 1. Laundered sealed edge wiper representing the highest level of cleanliness for criticalcleanroom applications.

Microdenier fabrics for cleanroom wipers 269

D. Classes >10,000:

These consist of various composites, including polypropylene and natural fi-

bers such as cotton and polyurethane foam.

The progression of wiper offerings is logical:

1. sealed-edge products are cleaner than unsealed;

2. laundered products are cleaner than unlaundered;

3. continuous filaments give less fiber contamination than do cut (“staple”) fi-

bers;

4. natural fibers contain contaminants harder to control than those in synthetic

fibers, and natural fibers are only available as staple fibers.

The efficacy of a wiper includes absorbency and ability to pick up and retain

particulate matter. This will depend on the fiber or foam base material and on

construction details. Inevitably, some contaminants will reach the work surface

where non-contaminating materials are needed to remove them.

3. MICRODENIER FABRICS

The denier of a yarn or fiber is the linear density expressed as the number of

grams in 9000 m of the yarn or fiber. Microdenier fibers (filaments) are defined as

1 denier or less, where 1 denier for polyester corresponds to a circular cylinder

with a diameter D of about 11 µm [4].

For fabrics of the same fiber composition of denier d:

– the total length of fibers will be proportional to 1/d

– the radius of a fiber will be proportional to d

– the pores created by bundles of such fibers will have dimensions proportional

to .d

– the total surface area of fibers will be proportional to 1/ d

– the cross-sectional area of a fiber will be proportional to d.

The techniques for creating such fine fibers include:

1. extruding a two-component mixture, then dissolving one component

2. extruding a two-component mixture, then fracturing the fiber with high-

pressure water, mechanical action, or chemical stress

3. extruding a single component, then elongating and thinning the fibers with

high-temperature gas jets.

The first two types, known as “islands in the sea” and “pie,” respectively, are

the ones normally used for microdenier and ultra-microdenier (below 0.1 denier)

materials used for cleanroom applications. They are generally made into woven or

knitted goods. The third type is used for “nonwovens” only, because of the diffi-

culties associated with weaving and knitting these fibers.


Mochizuki et al. [5] of Unitika Ltd. of Japan have described the formation of

ultra-fine fibrous materials by splitting bi-component fibers after they have been

formatted as a spunbonded fabric (molten continuous fibers are laid down and ad-

here where the fibers cross). They noted that three methods were conventionally

used (Figure 2) to form microdenier fibers from bi-component fibers: dissolution

of one of the components, separation of the components by swelling or shrinking,

and separation by mechanical distortion, the last being the approach they used. A

“sunflower” pattern (Figure 3) of six polyester fibers surrounding a polyethylene

core was split apart by flexing, taking a 3-denier fiber and making six 0.25-denier

fibers and a 1.5-denier core segment. The thin fibers contributed softness and

flexibility; the core allowed convenient thermal bonding. The density of the fabric

decreased, giving greater porosity. Triboelectrification of the fibers from dissimi-

lar materials can produce charging that enhances the effectiveness of dry wiping

in picking up dust particles [6]. Of course, this is not recommended for contact

with semiconductors.

Teijin Ltd. of Japan markets its Micro-StarTM

material, made of bi-component

nylon/polyester fibers, that are arranged into 16 filaments that have wedge-shaped

cross sections (Figure 4). Such filaments are typically 0.16 denier, but can be made

smaller. The materials are useful for absorbing both oil and water. Tests done on

cleaning tapes made from these materials generally showed lower levels of ionic

Figure 2. Various manufacturing techniques for forming microdenier fibers. Other technologies can be used but generally will fall into one of the four shown. The first is a single component while the others are multicomponent fibers.


contaminants than from tapes made with conventional denier fibers (1.5 denier).

This is not an inherent quality but the result of a more rigorous cleaning step.

Kuraray Ltd. of Japan produces its SOLIV (R)TM

fibers by splitting polyester

fibers longitudinally, forming fibers that have roughly rectangular cross sections,

claimed to facilitate wiping through a scraping mechanism (Figure 5) [7].

Toray Industries, Inc. of Japan markets its microdenier cleaning cloth as Toray-

seeTM

and LuminexTM

. The polyester fibers are produced by splitting thicker fibers,

creating material with a 2-micrometer diameter (0.06 denier) and having sharp

edges rather than being round. Figure 6 shows a comparison against materials

made with larger, traditional microdenier fibers (ca. 5 micrometers in diameter).

The current technology allows the formation of woven and knit goods from fi-

bers having deniers as low as 0.06. It also allows the formation of nonwoven

items, typically synthetic leathers, having deniers as low as 0.0001. While these

“nanofiber” materials have not yet been utilized into cleanroom wipers, the tech-

nology is nearly available to allow weaving and knitting these fibers into cleaning

materials. These fine fibers will probably be much more fragile in an unbonded

state and their usefulness as wiping cloths need to be determined.

Figure 3. Sunflower pattern showing six polyester fibers surrounding a polyethylene core. Splittingresults in making 3 denier fiber into six 0.25 denier fibers with a 1.5 denier core. The left figureshows an oblique view and the right figure shows a cross section of the bicomponent filaments. Courtesy of Unitika Ltd.

Figure 4. Formation of microdenier fibers by splitting bicomponent polyester/nylon fiber into 16wedge shaped filaments of 0.16 denier. Courtesy of Teijin Ltd.

J. Sko

ufis a

nd

D.W

. Coo

per

27

2

Figure 5. Longitudinally split fibers showing how the rectangular shapes facilitate wiping through a scraping mechanism. The upper left shows the

construction of a single filament while the upper right shows the same filaments split and knit into a wiper.

Courtesy of Kuraray Ltd.


Figure 6. Ultra microdenier polyester cleaning cloth showing split wedge shaped fibers approxi-mately 2 micrometers in diameter (0.06 denier) is shown on the left. Comparison is made with two traditional larger microfiber cloths on the right. Courtesy of Toray Ltd.

4. FABRIC STRENGTH AND RIGIDITY

The force needed to break a fiber (in tension) is generally proportional to its

cross-sectional area, a, so that the tensile strength of a textile fiber is often given

as gpd, grams per denier, which should be roughly independent of fiber diameter.

For polyester this is about 3-10 gpd and is commonly called the fiber “tenacity”.

The yarns of interest to us are in the range of 50 to 150 denier (composed of many

filaments), meaning a single yarn could suspend 150 to 1500 g without breaking,

depending on the tenacity of the particular polyester material [8].

The strength of an individual filament determines whether it breaks when

snagged; the strength is proportional to (denier)/(# filaments) = d/n. Bending

strength is usually also proportional to cross-sectional area, so the same propor-

tionality can be expected. To prevent pilling (formation of lint balls), most

polyester in commercial use is low tenacity. Higher tenacity fibers are needed for

wipers required to have abrasion resistance.

The geometry of a twisted yarn is complicated. The densest plausible packing

would be that of a unit cell that is a hexagon of contiguous cylinders surrounding

a central cylinder, which would give less than 10% open cross-sectional area. A

more plausible approximation is a square unit cell, with sides 4r in length, cir-

cumscribing four cylinders, of radius r, shown in Figure 7. This gives a porosity


e = 1-(π/4) = 0.215 = 21.5%. The spaces around these cylinders are complicated,

but they are roughly cylindrical pores with radii that are about half that of the

filament cylinders, r/2. r/2 will be taken here to be the characteristic dimension of

these inter-filament pores in the yarn. Note that this model is quite approximate,

as this geometry would in fact let liquid enter only from the ends of the yarn, and

not from the periphery, which clearly is not the case in practice.

The packing density is a determining factor in many of the properties of the

wiper: the absorbency, ability to pick up particles, and feel are all affected. Once

the filament size is determined, the science of fabric construction comes into play

in order to provide optimum properties.

5. LIQUID REMOVAL EFFECTIVENESS

The liquid to be wiped up usually contains particulate and molecular contami-

nants that will be removed roughly in proportion to how much of the liquid is re-

moved [9].

There is reason to believe that the residual liquid (“boundary-layer”) left by

wiping with a fabric made from a yarn would be roughly the size of the inter-

filament spacing, r/2. The filament radius, thus the inter-filament spacing, and, by

Figure 7. Calculation of cross-sectional porosity, e, or open space, of a twisted yarn from the dens-est possible packing of filaments.


inference, the boundary layer will be proportional to the square root of the fila-

ment cross section and therefore to (d/n) . Some boundary-layer reduction ad-

vantage is expected for microdenier fabrics based on this analysis.

The height to which a non-volatile liquid can rise in a fabric by absorption is

inversely proportional to the pore size. Microdenier fabrics, because of their

smaller pore sizes compared to standard denier fabrics, should raise liquids

higher. The speed the liquid will travel horizontally through the fabric is propor-

tional to the pore size. Microdenier fabrics provide less speed of absorption. This

may not be critical for many applications but in cases where it is, fabric construc-

tion can overcome this disadvantage.

The amount of liquid a wiper can hold is roughly proportional to its thickness.

For the same knit, the thickness can be expected to be roughly proportional to the

square root of the yarn denier, so wipers with larger deniers should offer greater

absorption capacity. However, the weight of the wiper will be the product of the

denier and the length of yarn used to knit it. To keep the basis weight the same,

the length of yarn will have to be inversely proportional to the denier. This would

mean fewer or shorter loops. In the limit of a very large denier, one would have

almost a very loose weave, rather than a knit, which will adversely affect absorp-

tion capacity. Larger denier yarns would also feel stiffer and more difficult to get

into tight spaces.

It is not completely clear what the implications of denier are in absorption ca-

pacity as so much is dependent on basis weight and construction. Experience has

shown that the ability to hold liquid is of less interest in critical cleaning than the

ability to pick up small particles and reduce boundary layers.

For different fabrics with the same basis weight (g/m2), there will be the same

volume of polyester contained in a total length of filaments L'. The volume of the filaments would be πr2

L'. The surface area would be 2πrL'. The surface area per unit volume (thus per unit weight) would be 2/πrL', inversely proportional to the filament radius, thus the surface area would be proportional to (n/d) . The greater the surface area, the higher the liquid absorption, but cleaning the fabric

may be more difficult.

6. PARTICLE REMOVAL EFFECTIVENESS

A simple geometrical model can be used to suggest the importance of smaller

wiping element size in lifting particles from a surface being wiped. Figure 8

shows two circles. One of the circles represents a particle. The other represents a

fiber. The larger object (particle or fiber) has a radius = R. The smaller object (fi-

ber or particle) has a radius = r. Arguably, to remove the particle away from the

surface we need a vertical component of force. [One could also cause the particle

to roll, not considered here.]


Figure 8. If one circle represents a particle and the other a fiber, the force, Fv, required to lift the particle can be calculated. The smaller the fiber, the smaller the particle it can push up from the sur-face.

The contact (idealized) between the cylindrical fiber and the spherical particle

occurs at their surfaces, along a line from one center to the other. The vertical

component of the force is then

Fv = F sinθ,

where F is the force at the contact and θ is the angle with respect to the surface or

Fv = F (r - R) / (r + R),

with the smaller object being pushed toward the surface and the larger object be-

ing pushed away. The smaller the denier, the smaller the particle it can push up

from the surface.

Besides the size advantage microdeniers offer in picking up very small parti-

cles, microdeniers can be made in a variety of shapes, even within the same

unsplit bicomponent filament. Common shapes include: shovel, wedge, star,

multi-lobe, and interior hollows. The wide variety of shapes provides opportuni-

ties for investigation into their suitability for particle removal. Specific shapes

may be optimal only for specific particle types.


7. FUTURE ACTIVITIES

While microdenier materials have been in wide use throughout Asia in clean-

rooms, they have been commercialized slowly in the United States. One reason is

the lack of US producers; another is their higher cost compared to standard denier

products.

Interest in microdenier materials is gradually increasing in the U.S. More so-

phisticated microelectronic products demand better contamination control. The

price of microdeniers is decreasing due to competition as more microdenier mate-

rial manufacturers are entering the market.

Several manufacturers who intend to revolutionize fiber production and signifi-

cantly reduce manufacturing costs have designed new equipment in the U.S. Sev-

eral companies and universities have purchased pilot-scale and production equip-

ment. Initial trials look promising. The equipment is able to change polymer types

quickly, and configurations not previously possible are now feasible. These ma-

chines can produce nanofibers of various shapes and various polymers by simple

economical screen changes, rather than the time-consuming changes in expensive

spinnerets previously needed.

8. CONCLUSIONS

The new microdenier fibers being used for textiles should be able to produce

highly absorbent and very clean and effective wiping materials for cleanroom use.

The technical developments of the industry are driving the need for fibers and

fabrics which are more effective in removing ever smaller particles. Asia has led

in development and use of microdenier fibers, but in the U.S. manufacturers are

catching up, using economical and flexible designs and methods.

REFERENCES

1. D.W. Cooper, Microcontamination, 3(8), 48–54, 73 (1985). 2. Kuraray Ltd. Publication 1241–58, “Wiping Cloth for High Class Cleanrooms” (1998). 3. “Evaluating Wiping Materials Used in Cleanrooms and Other Environments,” IES Publication

RP–CE–004–2 (1992). 4. J. Skoufis, “Fabrics for Disk Media,” Internal Training Document PMG–1A (1998). Available

for the author. 5. M. Mochizuki, K. Nagaoka and M. Hirai, “A Sunflower Comes into Blossom,” The Technical

Progress, Unitika Publication, undated. 6. D.W. Cooper, A2C2 (October 1998). 7. Teijin Ltd. Publication 93.6.2000, “Microstar Wiping Cloth.” 8. B.P. Saville, Physical Testing of Textiles, Woodhead Publishing, Abington, Cambridge, England

(1999). 9. R. Wang, Microcontamination, 14(2), 39–47 (1996).


Ed. K.L. Mittal

© VSP 2003

Fine particle detachment studied by reflectometry and

atomic force microscopy

ADAM FEILER

1,∗ and JOHN RALSTON

2

1KTH, Royal Institute of Technology, Department of Chemistry, Surface Chemistry, Drottning

Kristinas Väg 51, SE-100 44 Stockholm, Sweden 2Ian Wark Research Institute, The ARC Special Research Centre for Particle and Material

Interfaces, University of South Australia, Mawson Lakes, Adelaide, SA 5095, Australia

Abstract—Optical reflectometry was used to study the attachment and subsequent detachment of silica particles (diameter 25 nm) from the surfaces of titanium dioxide wafers under well-defined hydrodynamic conditions. The rate of detachment and maximum detached amount was studied as a function of both pH and added linear polyphosphate solutions. The latter have the general formula [PnO3n+1]

(n+2)- where n is the number of phosphorous atoms in the molecule. The maximum detached amount increased with increasing pH. The maximum detached amount also increased with n.

Atomic force microscopy was used to measure the interaction between silica spheres (diameter

7 µm) and titanium dioxide wafers under the same solution conditions. The detachment force needed to separate the surfaces decreased with increasing pH as well as with n in direct agreement with the reflectometry data. It was shown that, in addition to repulsive electrical double layer forces, adsorbed polyphosphates provided a short-ranged steric layer that reduced the lateral interaction be-tween the surfaces. The use of these two complementary techniques has given valuable insight into the processes responsible for fine particle detachment and has particular application to surface cleaning.

Keywords: Fine particle detachment; reflectometry; atomic force microscopy; particle adhesion.

1. INTRODUCTION

Surface contamination due to submicrometer particulate matter is of concern in

many areas including silicon wafer fabrication, mineral processing, water purifi-

cation and detergency. The permanent removal of these particles from surfaces is

a critical factor in these processes [1]. In solution, the combined effects of van der

Waals attractive forces and electrical double layer forces govern the interaction

between particles and a surface. By varying the solution conditions it is possible

to alter the surface chemistry of interacting materials and change their interaction

from attractive to repulsive. In this work, the attachment and detachment of

∗To whom all correspondence should be addressed. Phone: +46 8 790 9971, Fax: +46 8 20 89 98, E-mail: [email protected]


nanosized silica particles onto titanium dioxide surfaces has been studied as a

function of pH and addition of solutions of linear polyphosphates. Previous in-situ

infra-red studies [2] have shown that linear polyphosphates selectively adsorb

onto titanium dioxide forming strong chemical bonds. Streaming potential meas-

urements

[3] have shown that for a fixed polyphosphate concentration, the tita-

nium dioxide becomes more negative with increasing n, which was attributed to

an increased charge density with n. In addition, direct force measurements

[3-5]

have shown that adsorbed polyphosphate introduces a steric layer which leads to a

short-ranged repulsive interaction.

Reflectometry combined with a stagnation point flow cell was used to measure

silica particle attachment and detachment. The stagnation point flow is ideally

suited for the study of colloidal particle attachment and detachment processes in

which their transport is governed only by diffusion [6].

2. EXPERIMENTAL

2.1. Materials

2.1.1. Titanium dioxide wafers

A titanium dioxide layer was deposited on the surface of silicon wafers by sput-

tering (prepared at Philips Research, The Netherlands) [3]. Ellipsometry meas-

urements showed the titanium dioxide layer to be 40 nm thick and XPS analysis

showed the composition of the deposited layer to be pure titanium dioxide. X-ray

diffraction showed the deposited TiO2 to be amorphous. Imaging by AFM showed

an rms roughness of 0.3 nm over an area of 1 µm

2 with a maximum peak height

of 2 nm. The wafers were cleaned by detergent washing followed by rinsing with

high-purity water, ethanol, heptane and copious amounts of more high-purity wa-

ter. Finally, the wafers were blown dry in a stream of nitrogen and plasma cleaned

(Harrick Plasma Cleaner/Steriliser PDC-32) for 1 minute immediately prior to

use. The isoelectric point of the titanium dioxide covered wafers was determined

to be pH 4.2 using streaming potential measurements [3].

2.1.2. Silica

Suspensions of silica particles used for reflectometry experiments were prepared

from Ludox

TM AS40 (DuPont). The particles were dialyzed for 2 days in Milli Q

water and then suspended in solutions of KNO3 (10

-3 M) with a particle concentra-

tion of 100 mg/l. The mean particle radius was R = 12 ± 2 nm, determined by

transmission electron microscopy (CESMA, Adelaide University). The silica

spheres used for AFM measurements were obtained from Allied-Signal, (Chi-

cago, Illinois). XPS analysis (CSIRO Division of Molecular Science) showed the

composition of the sample to be pure silica. The typical diameter was found to be

7 µm. AFM imaging of the spheres over an area of 500 nm

2 showed the rms

roughness to be 0.8 nm.

Fine particle detachment studied by reflectometry and atomic force microscopy 281

2.1.3. Linear polyphosphates

The linear polyphosphates were provided by Albright and Wilson (Australia) as

dry sodium salts and were made up into solutions with a polyphosphate concen-

tration of 10

-5 M with background electrolyte concentration of KNO3 (10

-3 M).

The linear polyphosphates referred to here as P1, P2, P3 denote the monodisperse

species (Na3PO4), (Na4P2O7) and (Na5P3O10) respectively, P<10> refers to a

polydisperse sample with n ranging from 1-19 with an average of 10 P atoms.

2.1.4. Other reagents

Analytical grade KNO3, HNO3, and KOH were obtained from BDH Chemicals

(Australia) and were used as supplied. High purity water (surface tension 72.8

mN/m and resistivity 18 MΩ at 20°C) was from obtained from an Elga UHQ sys-

tem. The solution pH was adjusted with drop wise addition of 0.01 M HNO3, or

KOH via a micropipette.

2.2. Methods

2.2.1. Reflectometry

The experimental setup for the combined reflectometry and stagnation point fluid

cell is shown schematically in Figure 1. A detailed description of the experimental

procedure and the theory behind the technique is given elsewhere [7, 8]. Here a

Figure 1. Schematic diagram of the combined reflectometry and stagnation point flow cell.


brief description of the essential features is given. The reflected light from a plane

polarized He/Ne laser off an adsorbing surface is measured. The adsorbing sur-

face in this work was titanium dioxide covered silicon wafers. The reflected light

is split into its parallel (p) and perpendicular (s) components and the measured

signal, S, is the ratio of these intensities:

S

P

I

IS = (1)

This ratio depends on the refractive index profile close to the surface of the

substrate. Material adsorbed at the interface, in this case silica nanoparticles, will

change the refractive index profile and hence result in a change in S. Quantitative

measurements of the attached particle amount (Γ) can be obtained from the

change in signal via:

0s

1

S

S

A

∆=Γ (2)

where S0 is the intensity ratio prior to adsorption and ∆S is the change in intensity

ratio. The sensitivity factor, As, takes into account the explicit refractive index

contributions from the surface, the adsorbed material and the aqueous medium.

2.2.1.1. Stagnant point flow

In the stagnant point fluid cell, the collector surface is positioned at a critical dis-

tance from the inlet tube such that a stagnant point flow is generated at the point

where the fluid impinges the surface. Under these conditions the hydrodynamics

can be very well defined [9]. The particles arrive at the surface under the influ-

ence of surface forces and Brownian diffusion only. Solutions were gravity fed

into the reflectometry cell from high-density polyethylene (HDPE) containers

(250 ml) mounted on adjustable laboratory jacks. The height of the liquid above

the inlet port to the fluid cell determined the flow rate. The flow rate was main-

tained at 1.5 cm

3/min with a height of the liquid 13 cm above the inlet port. A

valve was used to switch between solutions entering the cell. The cell volume was

30 cm

3. A vacuum pump was used to suck excess solution from the cell.

2.2.2. AFM

A sphere attached to the end of an AFM cantilever comprises a colloid probe. A

silica sphere was attached to a cantilever using a heat softening resin (Epikote

1004, Shell) using a micromanipulation arm attached to a metallurgical micro-

scope (Olympus BH2). The cantilevers were silicon nitride, tipless, 200 µm long,

wide legged from Nanoprobe (Park Scientific, USA). Spring constants were de-

termined to be 0.1 ± 0.05 N/m by measuring the resonance frequency of the canti-

levers with added known masses [10]. Prior to measurement the colloid probe was

rinsed with ethanol, dried in a stream of nitrogen and plasma cleaned (Harrick

Plasma Cleaner/Steriliser PDC-32) for 1 minute. A Nanoscope III controller and


Atomic Force Microscope (Digital Instruments, Santa Barbara, CA, USA)

equipped with a fluid cell was used to measure the forces of interaction. Electro-

lyte solutions at the required pH and containing the desired polyphosphate solu-

tions were introduced into the fluid cell via Teflon tubing. The solution was al-

lowed to equilibrate for at least 15 minutes prior to measurement. The

experiments were conducted employing standard measurement procedures com-

prehensively described by other authors [11-13]. Measurements of the cantilever

deflection against scanner (piezoelement) displacement were taken. The piezo-

element was calibrated via an optical interference technique [14]. The cantilever

deflection data were subsequently converted to force (F) as a function of apparent

surface-surface separation (h), simply called separation hereafter. The force of in-

teraction was normalised by the radius of the sphere, i.e. F/R, employing the Der-

jaguin approximation for sphere-flat interactions [15].


An example of typical reflectometry data is shown in Figure 2. Electrolyte solu-

tion at pH 4 was flowed into the reflectometry cell for 20 minutes before the start

of the measurement. This ensured that the solution in the cell had reached thermal

Figure 2. Typical reflectometry raw data showing change in signal with time upon introduction of aparticle suspension at arrow (a) followed by the introduction of a displacing solution at arrow (b).


equilibrium with the solution in the reservoirs. The baseline signal, S0 was moni-

tored during this time to ensure that the signal was stable prior to measurement.

The background electrolyte solution was permitted to flow into the cell for a fur-

ther 100 seconds after the measurement began. After this time (point (a) in Figure

2), the valve was switched to allow silica particles into the cell (the electrolyte

concentration remained unchanged). The reflectometer signal increases due to the

attachment of the silica particles. Initially the signal increases linearly with time.

The rate of attachment decreases markedly close to saturation coverage. At point

(b), the valve was switched to introduce a new solution, containing either a parti-

cle-free electrolyte solution at high pH or a solution of linear polyphosphate. The

decrease in the signal at point (b) is due to the detachment of the silica particles.

Initially the decrease in signal is very rapid but the rate of detachment slows down

as the maximum detached amount is reached. The signal reaches a “plateau de-

tached amount” before complete detachment of the particles has been obtained.

The attachment of silica particles to a titanium dioxide wafer at pH 4, followed

by their detachment upon introduction of electrolyte solutions at higher pH’s, is

shown in Figure 3. The rate of attachment and the maximum attached amount,

Γmax, was the same for each experiment. The linear attachment regime is indica-

tive of a rate limited only by the mass transport of the particles from solution to

Figure 3. Amount of silica particles attached at the titanium dioxide surface as a function of time af-ter the introduction of a particle suspension in 10-3 M KNO3 at pH 4 followed (at t = 1000 s) by in-troduction of particle-free solutions of 10-3 M KNO3 at higher pH’s.


the surface [16, 17]. A similar observation was made in other reflectometry stud-

ies of nanometre-size particles [8, 18, 19]. No detachment was measured when

electrolyte solutions below pH 6 were introduced into the cell. Upon switching to

electrolyte solutions at pH 6 and higher, detachment of silica particles was de-

tected. Both the rate of detachment and the maximum detached amount increased

with increasing pH. At pH 6, only a small quantity (10%) of silica particles were

detached and the detachment rate was slow compared to initial rate of attachment

at pH 4. At pH 9, half of the pre-attached particles were detached and the initial

rate of detachment was faster than the initial rate of attachment.

In Figure 4 the detachment force measured by AFM between a silica sphere

and titanium dioxide substrate as a function of pH under the same solution condi-

tions as in Figure 3 is presented. A pH dependent adhesion force is evident. Pre-

vious studies have shown [3, 20] that the interaction force between silica and tita-

nium dioxide surfaces is well described by the DLVO theory of colloidal stability.

The adhesion results presented here may be rationalised in terms of the combined

effects of an attractive van der Waals force and pH-dependent electrical double

layer interactions. The isoelectric point (iep) of the silica particles and the tita-

nium dioxide wafers have been measured to be at pH ~ 2 and pH ~ 4.5 respec-

tively [3, 20]. At pH values below the iep of the titanium dioxide, in addition to

Figure 4. Force-distance curves for the retraction of a titanium dioxide wafer from a silica sphere (R

= 3.5 µm) as a function of pH in 10-3 M KNO3. The curves correspond to, from top to bottom, data at pH 9, 8, 7, 6, and 5.6.


the attractive van der Waals force, there will be an attractive electrical double

layer interaction between the oppositely charged surfaces. At pH values above the

iep of titanium dioxide both surfaces will be negatively charged and the electrical

double layer interaction will be repulsive. A tensile force is needed to separate the

surfaces from intimate contact at pH 5.6. This is indicated by the negative value

of the normalised force at the point at which the surfaces jump out of contact. At

this pH, the electrical double layer interaction will be weakly repulsive and the

adhesion is due to the attractive van der Waals forces. At higher pH values the

electrical double layer interactions become increasingly repulsive and the surfaces

are seen to separate from contact even in the presence of a positive applied force.

The separation force curves correlate well with the detachment data seen in Figure

3 and explain why no detachment of silica particles was detected below pH 6 and

also why the detached amount increased with pH.

The attachment of silica particles at pH 4 onto titanium dioxide substrates fol-

lowed by their detachment upon switching to solutions of linear polyphosphates

(10

-5 M) is presented in Figure 5. As discussed in the Introduction, adsorption of

polyphosphate onto titanium dioxide modifies the surface rendering it negatively

charged. Note that whereas changes in pH affected both the silica and titanium

Figure 5. Amount of silica particles attached at the titanium dioxide surface as a function of time af-ter the introduction of a particle suspension in 10-3 M KNO3 at pH 4 followed (at t = 1000 s) by in-troduction of solutions of linear polyphosphates (10-5 M) at pH 4 in 10-3 M KNO3. The subscript in P

n

refers to the number of phosphorous atoms in the molecule.


dioxide surface potentials, the specific adsorption of polyphosphate onto titanium

dioxide leaves the silica unmodified. Thus, any detachment of silica particles can

be attributed solely to chemical changes at the titanium dioxide surface. For n = 2,

3 and <10> the rate of detachment and the detached amounts of particles are simi-

lar to each other and larger than P1 on the time scale of the experiments. The dis-

crepancy seen in the detachment profile in the presence of P1 can be understood in

terms of the adhesion measurements, see below.

Figure 6 shows the interaction force curves during separation between a silica

colloid probe and titanium dioxide at pH 4 in the presence of solutions of poly-

phosphate (10

-5 M) of varying n. For comparison, the interaction force between

silica and titanium dioxide at pH 4 in the absence of polyphosphate is also shown.

At pH 4 in the absence of polyphosphate there is a large adhesion due to com-

bined attractive electrical double layer and van der Waals forces. The presence of

polyphosphate clearly modifies the interaction force, dramatically reducing the

adhesion. In the presence of P1 the detachment force is negative indicating a sig-

nificant adhesion force. For n > 1 the detachment force is positive and the force

curves show that the separation is dominated by a repulsive interaction. The mag-

nitude of the repulsive interaction increases with n. The trends seen in the interac-

tion force curves are in accord with previous AFM studies and streaming potential

Figure 6. Force-distance curves for the retraction of a titanium dioxide wafer from a silica sphere (R

= 3.5 µm) at pH 4 in 10-3 M KNO3 in the presence of linear polyphosphate solutions (10-5 M) from top downwards P<10>, P3, P2 and P1.


measurements [3] and can be directly related to chemical modification of the tita-

nium dioxide surface due to adsorbed polyphosphate. The increase in the repul-

sive interaction with n is due to an increased negative surface potential on the ti-

tanium dioxide and also the presence of a steric layer, whose thickness, δ,

increases with n (δ ≈ 0.4-0.6 nm for n = 1-3 and δ ≈ 1.6 nm for n = <10>) [3].

A good correlation between the detached amounts of silica particles measured

by reflectometry and the normalised detachment force measured by AFM is seen

in Figure 7. The detached amount of particles is plotted as a percentage of the to-

tal attached amount of particles prior to switching to the displacing solution. The

detached amount of particles increases as the normalised detachment force be-

comes more positive (more repulsive). The detachment of the silica particles is

clearly sensitive to the variation in the electrical double layer interactions brought

about by changes in pH as well as due to adsorbed polyphosphate. The fact that

detachment of silica particles is detected at all in the presence of P1 despite the

force curves showing an adhesional interaction is evidence that the steric layer

due to adsorbed polyphosphate is important in the detachment process. The slight

discrepancy from a linear trend between the detached amount and detachment

force seen for the P<10> data point can be understood in terms of the polydisperse

Figure 7. Amount of detached particles measured by reflectometry against the normalised detach-ment force measured by AFM under similar solution conditions. The symbols refer to the data

measured as a function of pH (♦) and in the presence of polyphosphate (•).


nature of the P<10> sample which contains a range of linear polyphosphates from n

= 1-19. In the reflectometry measurements, the diffusion rate of the polyphos-

phate species becomes a critical factor. The diffusion rate of the polyphosphate

species will decrease with n. Thus, although the larger n species will impart a

more repulsive force on the silica-titanium dioxide system, the smaller n species

will diffuse to the surface more quickly. The consequence is a competition be-

tween the rate of polyphosphate adsorption and the modification of the resultant

polyphosphate adsorption.

Finally, it is of interest from an application viewpoint as to the most efficient

use of a dispersing agent such as polyphosphate in preventing particle attachment.

In Figure 8 the order of addition of silica particles and P3 solution (10

-5 M) to the

cell is investigated. Curve I shows the attachment of particles in the absence of

polyphosphate followed by their detachment upon introduction of a P3 solution

(10

-5 M). This sequence of addition is identical to that shown in previous figures.

Curve II shows the attachment of silica particles on a titanium dioxide wafer that

Figure 8. The effect of the sequence of addition of silica particles and P3 solution. Curve I is for the introduction of silica particles followed by the introduction of P3 (10-5 M) at t = 1500 s. Curve II is for the introduction of silica particles after pretreatment by flowing P3 solution into the cell for 1000 seconds and rinsing with electrolyte solution for 100 seconds. Curve III is for the introduction of asuspension containing a mixture of silica particles and P3 (10-5 M). All experiments were conducted in 10-3 M KNO3 at pH 4.


was conditioned with polyphosphate. For this experiment, a solution of P3 (10

-5

M) was introduced into the cell for 1000 seconds followed by rinsing with elec-

trolyte solution for 100 seconds immediately prior to introducing the silica parti-

cles. Particle attachment was detected; however, the initial rate of attachment was

less than that seen in Curve I onto the bare surface. The maximum attached

amount is also reduced. The most striking effect is seen in Curve III, in which sil-

ica particles and P3 solution were introduced simultaneously into the cell. No par-

ticle attachment was detected, even after 50 minutes. The results indicate that

equilibrium considerations are relevant in determining the amount of attached

particles and adsorbed polyphosphate that takes place at the surface. Precondition-

ing the titanium dioxide surface (Curve II) reduces particle attachment by reduc-

ing the available surface sites at which the particles can attach. In Curve III the

bulk solution contains both excess silica particles and polyphosphate. As a much

smaller species, the polyphosphate possesses a much higher diffusion coefficient

than the particles. Thus it is expected that the polyphosphate would arrive and ad-

sorb at the titanium dioxide surface before the particles. The excess concentration

of polyphosphate ensures that even if polyphosphate desorbs from a surface site,

another polyphosphate molecule will quickly adsorb in its place and so prevent

the attachment of particles over long periods of time.

4. CONCLUSIONS

The attachment of nanosized silica particles onto titanium dioxide surfaces and

their subsequent detachment due to changes in solution pH or in the presence of

linear polyphosphates was studied using reflectometry. The rate of detachment

and maximum detached amount increased with both pH and n. It was seen that in-

creasing the pH led to repulsive electrical double layer interactions, which were

responsible for the detachment of the particles. In addition to electrical double

layer interactions, adsorption of linear polyphosphate onto titanium dioxide pro-

vided a steric component, which facilitated the particle detachment. A good corre-

lation was seen between the amount of detached particles and the AFM measured

detachment force. Furthermore it was shown that a solution containing excess lin-

ear polyphosphate could prevent particle attachment all together.

REFERENCES

1. K. L. Mittal (Ed.), Particles on Surfaces 5 & 6: Detection, Adhesion and Removal, VSP, Utrecht (1999).

2. A. P. Michelmore, W. Gong, P. Jenkins and J. Ralston, Phys. Chem. Chem. Phys., 2, 2985 (2000).

3. A. Feiler, P. Jenkins and J. Ralston, Phys. Chem. Chem. Phys., 2, 5678 (2000). 4. A. Feiler, I. Larson, P. Jenkins and P. Attard, Langmuir, 16, 10269 (2000). 5. Y. K. Leong, P. J. Scales, T. W. Healy, D. Boger and R. J. Buscall, Chem. Soc. Faraday Trans.,

89, 2473 (1993).


6. Z. Adamczyk, B. Siwek, M. Zembala and P. Belouschek, Adv. Colloid Interface Sci., 48, 151 (1994).

7. J. C. Dijt, PhD thesis, Wageningen University, The Netherlands (1993). 8. M. R. Bohmer, J. Colloid Interface Sci., 197, 251 (1998). 9. T. Dabros and T. G. M. van de Ven, Colloid Polym. Sci., 261, 694 (1983).

10. J. P. Cleveland, S. Manne, D. Bocek and P. K. Hansma, Rev. Sci. Instrum., 64, 403 (1993). 11. W. A. Ducker, T. J. Senden and R. M. Pashley, Nature, 353, 239 (1991). 12. I. Larson, C. J. Drummond, D. Y. C. Chan and F. Grieser, J. Am. Chem. Soc., 115, 11885

(1993). 13. P. G. Hartley, I. Larson and P. J. Scales, Langmuir, 13, 2207 (1997). 14. M. Jaschke and H.-J. Butt, Rev. Sci. Instrum., 66, 1258 (1995). 15. J. N. Israelachvili, Intermolecular & Surface Forces, 2nd ed., Academic Press, London (1992). 16. Z. Adamczyk, L. Szyk and P. Warszynski, J. Colloid Interface Sci., 209, 350 (1999). 17. N. Kallay, M. Tomic, B. Biskup, I. Kunjasic and E. Matijevic, Colloids Surfaces, 28, 185

(1987). 18. M. R. Bohmer, E. A. van der Zeeuw and G. J. M. Koper, J. Colloid Interface Sci., 197, 242

(1998). 19. R. A. Hayes, M. R. Bohmer and L. G. Fokkink, J. Langmuir, 15, 2865 (1999). 20. I. Larson, C. J. Drummond, D. Y. C. Chan and F. Grieser, J. Phys. Chem., 99, 2114 (1995).


Ed. K.L. Mittal

© VSP 2003

Dust removal from solar panels and spacecraft on Mars

S. TRIGWELL, M.K. MAZUMDER,∗ A.S. BIRIS, S. ANDERSON and

C.U. YURTERI

Department of Applied Science, Donaghey College of Information Science and Systems

Engineering, University of Arkansas at Little Rock, 2801 South University Avenue,

Little Rock, AR 72204-1099

Abstract—In Lunar or Martian habitat systems it is impossible to avoid contact with dust. Martian

dust storms, containing submicrometer to 50 µm particles, are an environmental threat to solar cells, spacecraft, and spacesuits. Because of the high electrostatic charge of the dust and its strong adhe-sion properties, its deposition onto life support equipment could damage or degrade equipment, re-ducing the mission duration and endangering personnel. The inhalation of electrostatically charged airborne dust is also a health hazard to astronauts inside the habitat. Ways to minimize or eliminate the potential hazards caused by charged particles on space life support equipment are therefore needed. Specifically, the following topics are discussed in this paper: (1) tribocharging of insulating materials, (2) the design of a sensor to measure particle size and electrostatic charge distributions of Mars dust on a single particle basis and in real-time, (3) an experimental plan to minimize deposi-tion of charged particles on solar cells and life support equipment, and (4) a novel method for re-moving deposited dust particles.

Keywords: Mars dust; solar panels; electrostatic; charged particles.

1. INTRODUCTION

The atmosphere of Mars contains significant amounts of suspended dust, and in

any mission to Mars it will be impossible to avoid contact with this dust. Martian

dust storms containing fine particles (submicrometer to 50 µm in diameter) are a

serious problem to solar cells, spacecraft, and spacesuits [1, 2]. The dust may also

possess a high electrostatic charge due to tribocharging by contact with other par-

ticles or materials, or photoionization by the intense UV radiation. Because of the

possibility of high charge on dust particles and resulting strong adhesion forces,

deposition of dust onto support equipment could damage or hinder correct func-

tionality of the equipment, reducing the mission lifetime.

∗To whom all correspondence should be addressed. Phone: 501-569-8007, Fax: 501-569-8020,


S. Trigwell et al. 294

The settling of this dust, especially during a Martian dust storm, can have a

significant effect on the efficiency of solar panels, due to the settled dust imped-

ing the sunlight from the cells. Results from the Materials Adherence Experiment

(MAE) on the Mars Pathfinder mission measured an obscuration of the solar ar-

rays due to dust deposition at a rate of about 0.28% per day [3] with an estimate

that settling dust may cause degradation in performance of a solar panel of be-

tween 22% and 89% over the course of two years [4].

Particles may also settle on the solar arrays by a process known as saltation, in

which particles are lifted from the surface by the wind. These have a size range of

1–200 µm in diameter and an average trajectory of 10 to 20 cm off the surface [5].

Due to the low barometric pressure in the atmosphere, of about 10 mbar, saltation

occurs at wind velocities greater than 15 m/s, which has been recorded at Viking

lander sites [5]. Dust accumulation can also occur due to settling from the atmos-

phere. However, the real deposition rate will also depend on the geographical lo-

cation and from season to season.

The removal of dust settled on an array by natural wind forces on Mars has been

ruled unlikely due to the low atmospheric pressure, which will necessitate high wind

velocities of the order of 35 m/s [6]. The measurements of wind velocities at Viking

sites showed that maximum peak wind velocity was only 25 m/s, with winds over 15

m/s occurring only 1% of the time [6]. Therefore, it was concluded that for long-

duration missions, prevention of deposition or periodic removal of accumulated dust

must be performed to maintain the efficiency of the solar power arrays.

The problem with the design of any mechanism that has to work on a Mars

spacecraft is the hostile environment in which it is expected to perform. The at-

mosphere of Mars is quite different from that of Earth in that it is composed pri-

marily of carbon dioxide (95.3%) with minor amounts of other gases (nitrogen –

2.7%, argon – 1.6%, oxygen – 0.13%, and trace amounts of water and neon) [2].

Although the water content of the atmosphere is about 1/1000 that of Earth, it can

condense out forming clouds and even ground frost in the winter. The most sig-

nificant factor is the temperature on the surface. The average recorded tempera-

ture on Mars is –63°C, with a maximum and minimum of approximately 20°C

and –140°C, respectively. However, the temperature variation depends on the lo-

cation. Temperatures of –133°C are observed at the winter poles, while tempera-

tures as high as 27°C are observed on the dayside during the summer [2].

In this study, the goal was to develop an understanding of the principles of par-

ticle charging and to perform theoretical and experimental studies on the adhesion

and removal of charged particles. Specifically, the following are presented: (1) a

study of the effects of tribocharging of insulating materials and how it can play a

role in dust accumulation on solar panels, (2) design of a sensor to measure parti-

cle size and charge distributions of Mars dust, (3) development a self-cleaning

panel with electrodynamic screens to repel charged dust from settling on solar

panels, and (4) development of an electrostatic wiper type brush for removing de-

posited particles utilizing minimum mechanical parts.

Dust removal from solar panels and spacecraft on Mars 295

2. TRIBOCHARGING OF PARTICLES

There exist two primary mechanisms of charge transfer for contact or tribocharg-

ing between two dissimilar materials. The first is electron transfer in that a linear

relationship is observed between charge transferred during contact of two dissimi-

lar materials. The second is ion transfer in that real surfaces of metals or insula-

tors are covered by adsorbed layers, which are frequently ionic in nature, and that

charge transfer is by positive or negative ion transfer between the materials. A

third theory is postulated that involves material transfer that carries an associated

charge.

2.1. Electron transfer

In a metal at absolute zero, all the states below the Fermi level in the metal are

filled, and all those above are empty. When two dissimilar metals with different

work functions, A and B, are brought into contact, then electrons will flow from

metal A into metal B decreasing the potential difference until equilibrium of the

Fermi levels is reached. Metal B will now have a net negative charge, and metal

A will have a net positive charge of equal magnitude, where the contact potential

difference, Vc, is given by:

VC = (φB - φA) /e (1)

where φA and φB are the work functions of metals A and B, respectively. How-

ever, in metal/metal tribocharging, a back tunneling current exists when the two

materials are separated, resulting in a net zero charge on the two metal surfaces.

In a metal-insulator contact electrification, which is likely to be found on Mars as

the dust comes into contact with spacecraft parts and instrumentation, electrons

may pass from the metal into the empty states in the insulator, or from occupied

insulator states into the metal. Insulators, specifically polymers, have been con-

sidered to have a wide forbidden band gap where very few extrinsic states exist.

However, there are likely to be localized surface states, surface impurity states,

bulk defect states, or bulk impurity states [7-9]. These states may emit or accept

electrons in contact electrification. Bulk defect levels and surface states give rise

to an “effective” work function for an insulator φI. Before contact, the surface

states are filled to the equivalent Fermi level, EFP. A simple surface states theory

of contact electrification of insulators is shown in Figure 1.

Surface states on an insulator can be intrinsic or extrinsic. In either case, contact

with a metal will cause empty states below the metal Fermi Level EF to be filled,

and full states above it will be emptied. The number of electrons that transfer to the

insulator will be equal to the number of surface states with energies between φI and

EF. However, there is still a considerable uncertainty in this description of insulator

charging [7-9], but most theories assume that the amount of back tunneling of

charge when the materials are separated is negligible, and that the final charge upon

separation is approximately the same as when the surfaces were in contact.


In covalently bonded solids, impurity atoms provide most of the additional

available energy levels [1], but in molecular solids, electron traps may be associ-

ated with the ends of molecular chains or cross-links. Duke and Fabish proposed a

model [10] to interpret contact electrification of pendant group polymers. This

model suggests that side groups on a polymer chain can form intrinsic charge car-

rying sites, which may be electron donors or acceptors. The model states that the

electronic states are localized and represented by double Gaussian distributions

representing electron acceptor and donor states. The distribution of the states is

suggested to be due to differences in the local environment for each molecule.

A number of factors are involved in contact electrification under different con-

ditions. When contacting a metal surface, it has to be considered that a metal ox-

ide layer is always present. Similarly, the insulator surface may also be covered

with an oxide layer or at least other contaminants. For this, the contact charge ex-

change density, σ, on the insulator is given by [11, 12];

σ = -feNS[φI - φ][1 + (fe2aNS/ε)] (2)

where f is the fraction of area that makes intimate contact, e is the electronic

charge, NS is the surface state density per unit area per unit energy (eV), φI is the

insulator surface work function, φ is the metal surface work function, a is the

thickness of the oxide layer, and ε is the permittivity of the oxide layer.

In the case of a low surface state density, NS << ε/fe2a,

σ = -feNS[φI - φ] (3)

and in the case of a high surface state density, NS >> ε/fe2a,

Figure 1. (a) The insulator is uncharged with states filled below the neutral level. (b) On contact,

empty states below the Fermi level are filled and the insulator charge is now proportional to φ - φI.


σ = -fεNS[φI - φ]/ea (4)

It is the concentration of the surface states, NS that determines whether the elec-

trons occupy bulk states in addition to surface states during the time of contact

[11]. If the charge transfer is completed in a very short time, then only surface

states are involved. If charging is notably dependent upon time, then charge trans-

fer into the bulk is more probable. However, the above equations (3) and (4),

show that contact charging depends upon both surface oxidation and density of

surface states. The physical meaning of the surface work function of an insulator,

φI, and the distribution of the surface states within the forbidden gap are still not

clearly established.

This uncertainty in describing insulator charging as it applies to contact with

metals, therefore, leads to uncertainty in the understanding of insulator contact

charging with other insulators. The accepted theory that even insulators higher up

in the triboelectric series will charge positive when contacted with insulators lower

down has led to several published triboelectric series [7, 13, 14]. However, no two

series agree absolutely, with positions on the list of some materials varying widely

between different series, and only a rough agreement as to the relative positions of

several polymers. There is some uniformity for certain selected polymers such as

Nylon and poly(tetrafluoroethylene) (PTFE), which are consistently found on the

opposing ends of the series. Similarly, polyethylene and polystyrene are usually

found in consistent positions among the different series. This uniformity between

different triboelectric series suggests that for insulator/insulator charging, a similar

mechanism as for metal/insulator exists. Therefore, an insulator/insulator contact

charging theory may be constructed. Charge exchange between insulators can be

predicted from the knowledge of the charge acquired by contact with metals, and

so the general conclusion is that insulator-insulator charging is caused by the same

basic mechanism as metal-insulator charging.

2.2. Ion transfer

Real surfaces are always covered with an adsorbed layer. This layer is frequently

ionic in nature or contains a charged double layer. This covering layer can act as a

significant potential barrier through which the electrons must travel. However, ion

exchange between two covered contact layers can take place. In this case, the pos-

sible mechanisms for ion transfer include the difference in the affinities of the two

contacting surfaces for specific ions and the abundance of a particular ion on one

surface. In addition to the above, material transfer may also be considered. Frag-

ments of one material may break off one surface and be deposited on the other.

The break point is a few molecules beneath the surface, and mass transfer has

been detected between combinations of certain polymers [1]. However, in this

case, the amount of material transferred exceeded that considered necessary for a

typical measured charge transfer. At this point it is emphasized that all theories of

charge transfer for both metal-insulator and insulator-insulator contacts are still

poorly understood, and a much better understanding of the nature of the surfaces


of polymers and insulators regarding the electron energy levels and the role of

impurities is needed.

2.3. Experimental

In order to better understand the factors involved in how the surface properties of

materials influence the charge that can be transferred to a material, the surface

work function and the surface chemical composition of various metallic, ceramic,

polymeric, and mineral materials were analyzed. The metal and polymer materials

used were typical of those that are or may be used in a Mars mission. Pulverized

quartz (SiO2), alumina (Al2O3), and pyrite (FeS2) were used as examples of min-

erals as spectroscopic analysis of Mars dust had shown it to be composed of sili-

cates and iron and magnesium rich sulfates [15]. The samples were analyzed in

the as-received condition with no prior cleaning except for the polymeric materi-

als that were scraped with a clean scalpel blade to expose a fresh surface. The

samples were then analyzed by X-ray photoelectron spectroscopy (XPS) to de-

termine the surface chemistry, and by ultra-violet photoelectron spectroscopy

(UPS) in air to measure the surface work function.

The XPS data were obtained on a PHI Quantum 2000 ESCA Spectrometer us-

ing a focussed monochromatic Al Kα (hν = 1486.7 eV) x-ray source. The x-ray

beam used was a 100 W, 100 µm diameter beam and was rastered over a 1.5 mm

by 0.2 mm area. The survey scans were collected using a pass energy of 117.4 eV

producing a Full Width at Half-Maximum (FWHM) of less than 1.6 eV for the

Ag 3d 5/2 peak. The high energy resolution data were collected using a pass en-

ergy of 23.5 eV, producing a FWHM of less than 0.75 eV for the Ag 3d 5/2 peak.

The collected data were referenced to an energy scale with binding energies for

Cu 2p 3/2 at 932.67 +/– 0.05 eV, and Au at 84.0 +/– 0.05 eV. On some insulating

samples, positive charging of the surface was observed due to the loss of elec-

trons, causing the peaks to shift during data acquisition For these cases, low en-

ergy electrons were used to flood the specimen to neutralize the surface.

The UPS data were obtained on a Riken Keiki AC-2 UV photoelectron spec-

trometer. The samples and detector were placed in open air. The UV source was a

deuterium (D2) lamp with a spot diameter of 2 mm by 2 mm. For the samples with

a high efficiency of photoemission (the metals and graphite) the light source

power was 49.9 nW, and for those samples that have a low efficiency of photo-

emission (the polymers, coal, pyrite, and vitrinite) the light source power was in-

creased to 600.2 nW. The resolution of the instrument for precision measurements

is given as 0.02 eV. The samples were analyzed at a temperature of 22°C, a rela-

tive humidity of 40%, and a pressure of 1 MPa.

2.4. Results

The XPS data are presented in Table 1. The relative atomic concentrations of the

observed elements as reported in Table 1 were obtained by integrating the area

under each peak of interest and normalizing with sensitivity factors supplied by


the instrument manufacturer. The data showed that the metal specimens had par-

ticularly high levels of surface contamination (carbon and oxygen) compared to

the ceramic and polymeric specimens. For example, copper showed 70 atomic %

of carbon on the as-received surface, compared to the quartz (SiO2) specimen

with only 9.9 atomic % of carbon. For PTFE, the XPS data showed the surface

composition to be 30 atomic % carbon and 70 atomic % fluorine, which is very

close to the CF2 stoichiometric composition of PTFE. This supports the fact that

polymers do not pick up surface contamination in air as readily as metals. The

graphite specimen shows only a minimal oxygen concentration (2.3 atomic %).

The measured UPS data are shown in Table 2. The value of the work function

for each material is compared with the value reported in the literature.

It was observed in Table 2 that the measured work function for each material

was higher than the reported values. A closer examination showed that the ratio of

the measured-to-reported work function value for copper (1.17:1) was higher than

for the other metals; silver (1.08:1), and aluminum (1.05:1); and these metals

showed less carbon surface contamination by XPS than the copper. In contrast,

the measured-to-reported work function value for PTFE was approximately unity,

and the PTFE showed no discernible carbon contamination by XPS. Clearly, a

correlation between the amount of surface contamination and increase in work

function can be observed.

Table 1.

Surface element concentrations (atomic %) of selected materials as measured by XPS

C O N Na Si Cu Fe Cr Ag Al Mg F Ca Zn Cl

S

Copper 70 21 – – 1.4 7.4 – – – – – – – – – –

316L steel 53 35 0.8 – 0.9 – 8.7 1.7 – – – – – – – –

Electro. 316L 69 24 2.1 1.8 1.3 – 0.8 0.7 – 1.1 – 0.2 – 0.3 0.3 –

Silver 40 16 – – – – – – 40 – – – – – 2.6 –

Aluminum 23 53 – – – – – – – 24 1.2 – – – – –

PTFE 30 0.2 – – – – – – – – – 70 – – – –

Nylon 66 79 13 7.8 – 0.2 – – – – – – – – – – –

Polystyrene 75 21 2.4 1.2 – – – – – – – – – – – 0.7

Glass 15 55 – 9.5 17 – – – – – 2.9 – 0.3 – – –

SiO2 9.9 64 – 0.6 26 – – – – – – – 0.7 – – –

Pyrite 51 27 0.4 0.6 – – 5.4 – – – – – – – 1.4 14

Graphite 98 2.3 – – – – – – – – – – – – – –

Polycarbonate 84 16 – 0.1 0.2 – – – – – – – – 0.1 0.1 –

Acrylic 73 27 – – – – – – – – – – – – – –


Table 2.

Measured work functions by UPS in air compared to work function values reported in the literature. Measurement errors were +/– 0.002 eV

Material Measured work function [eV] Reported work function [eV]

Aluminum 4.53 4.30

Silver 4.66 4.30

Graphite 5.09 4.50

Copper 5.11 4.38

Stainless steel 5.37 Unknown

Polystyrene 5.48 4.90

Pyrite 5.50 5.40

Acrylic 5.52 4.72

Polycarbonate 5.57 4.80

Nylon 66 5.61 4.30

PTFE 5.80 5.75

Figure 2. XPS survey scan of pyrite (FeS2).

As previously mentioned, pyrite was chosen as an example of a mineral found

on Mars. For this sample, high resolution scans of the carbon, sulfur, and iron

peaks were examined, and the chemical composition rather than just the elemental

composition was determined. The XPS survey scan for pyrite is shown in Figure

2, and high resolution scans of the carbon, sulfur, and iron peaks are shown in

Figure 3.


Figure 3. High resolution XPS scans of peaks for pyrite (FeS2). a) Carbon C1s peak, b) Sulfur S2p peak, and c) Iron Fe 2p peak.

(a)

(b)

(c)


The XPS survey spectrum shows the peaks associated with the surface compo-

sition. In addition to the peaks for carbon and oxygen, minor amounts of sodium,

nitrogen, and chlorine were also detected. In the high resolution scans, the peak

for each element was curve fitted with the component peaks as reported in nu-

merous reference literature and data bases [16-18]. What can be observed from

the high resolution peaks is that the carbon present on the mineral is predomi-

nantly C-C or C-H bonding indicating carbonaceous contamination, with minor

amounts of carbon-oxygen species. A small amount of organic carbon-sulfur was

also detected. However, the sulfur peak shows the sulfur to be mainly in the sul-

fate form with very little, if any, iron sulfide present. This is confirmed in the iron

peaks where the iron is present predominantly as iron sulfate, and in this case no

iron sulfide is detected. The data show that for the case of pyrite what is actually

chemically present on the surface is very different from the bulk composition,

which can affect the value of the work function. This is important in understand-

ing how materials charge relative to each other in the triboelectric series. How-

ever, these data were taken in an Earth environment, and so it is of great interest

to determine what the surface composition of minerals would be in a Martian en-

vironment.

3. DESIGN OF SENSOR

There are a number of instruments that can be used to characterize the aerody-

namic size distribution of particles. Instruments such as a Faraday cup are avail-

able to estimate the net average electrostatic charge on particles samples. How-

ever, the choice of instruments for real-time simultaneous measurements of both

aerodynamic diameter and electrostatic charge distributions of particles on a sin-

gle particle basis is limited. The Electrical Single Particle Aerodynamic Relaxa-

tion Time (E-SPART) analyzer is used extensively for simultaneous characteriza-

tion of particle size distribution (PSD) and electrostatic charge distribution [19].

The analyzer can be used in the diameter range from 0.5 to 50 µm and charges in

the range from 0 to their saturation charge limit.

The E-SPART analyzer, as shown in Figure 4, uses an AC electric drive to os-

cillate the particles in air. The resultant oscillatory motion of the particle lags be-

hind the external AC field. The phase lag (φ) relates to the aerodynamic diameter

of the particle, and the amplitude of the particle trajectory determines the particle

charge and polarity, as shown in Figure 5(a). Figure 5(b) shows a still from a

video image taken of charged particle tracks in the chamber. The particle tracks

are analyzed by Laser Doppler Velocimetry. The details of the operation of the

instrument are available [20-23].

Figures 6(a) and (b) show typical particle size and charge distribution, respec-

tively, for a positive copier toner with a mean particle size of 8 µm. Although the

data show the overall particle count, individual particle data can be extracted.


Figure 4. Schematic of the E-SPART analyzer.

(a)

(b)

Figure 5. (a) Principle of operation of E-SPART, and (b) video image of particle tracks. The E-SPART analyzer uses an AC electric drive to oscillate the charged particles.


(a)

(b)

Figure 6. (a) Particle size distribution, and (b) charge distribution for a positive toner.


4. DEVELOPMENT OF A SELF-CLEANING PANEL USING

ELECTRODYNAMIC SCREENS

Dust settling out of the atmosphere onto any horizontal surface is a potential prob-

lem in the obscuration of solar arrays. A method is required to periodically re-

move the dust, or prevent the dust from settling in the first place. Ideally, a

method that requires no moving parts and is robust in operation is most desirable,

as shown in Figure 7.

The static charges on the particles provide an opportunity to prevent dust deposi-

tion by using an AC voltage driven electrode screen. This type of screen creates a

repelling force to the charged particles regardless of their polarity. The device con-

sists of an electrode screen that contains a number of parallel electrodes placed

equidistant from each other, embedded in a insulating coating as shown in Figure 7.

Figure 7. Solar panel with embedded electrodynamic screen.


5. DEVELOPMENT OF A METHOD FOR REMOVING DEPOSITED DUST

At present, there are four categories of dust-removal methods, namely, natural,

mechanical, electromechanical, and electrostatic [1]. It has been observed that

wind velocities on Mars are insufficient to remove settled dust on its own. A pos-

sible aid to natural dust removal would be to devise a movable array so it can be

turned vertical such that gravity in addition to wind may remove the dust. By de-

signing an array with a vibration frequency that would correspond to the wind

would assist in the dust removal. Electromechanical methods are similar to natu-

ral methods, and include vibrating, shocking, or using ultrasound, in combination

with tilting the array to remove the dust. These methods would require sophisti-

cated mechanisms.

The problem with any mechanical device is the risk of mechanical failure of

one of the components that will be catastrophic in an alien environment where it

cannot be accessed to be repaired. In this section, a method is described to me-

chanically clean the solar panels by a mechanism using a NiTi (Nickel-Titanium,

also know as nitinol) shape memory alloy. This method is unique in that it in-

volves only a thermoelastic process (no motors or electrical components) and thus

considerably reduces the probability of failure.

5.1. Proposed method for the removal of dust particles

A proposal for the development of an electrostatic brush for removing particles

from different surfaces was based on empirical studies using materials such as

polystyrene (PS), poly(tetrafluoroethylene) (PTFE), and polyamides (Nylon). A

brush made of Nylon and PTFE fibers may be effective in getting tribo-charged

and in removing particles with both negative and positive polarities. The objective

is, therefore, to build a device that would utilize such a brush but would have

minimal moving and mechanical parts that would reduce the susceptibility to fail-

ure. The proposed device would be built using a smart material, namely NiTi al-

loy, which will move the wiper/brush to clean the surface of the solar array when

subjected to heating by the sun.

Shape memory alloys, such as nickel-titanium, are a class of unique alloys that

can be deformed, but then recover their original shape when heated. This is due to

the occurrence of a martensitic phase transformation and its subsequent reversal.

Figure 8 [24] shows a typical plot of property changes versus temperature for a

shape memory alloy. Basically, the parent phase is an austenite, and the alloy is de-

formed into the martensite phase. Upon heating through its transformation tempera-

ture, it reverts back to austenite and recovers its previous shape with great force.

This process can be repeated millions of times. The shape recovery process oc-

curs over a range of a few degrees. The temperature at which the alloy “remem-

bers” its higher temperature form when heated can be adjusted by slight changes

in its composition. The lowest active transformation temperature for commer-

cially available NiTi alloy is at present 0–10°C, containing 55.8 wt% Ni (Alloy C

– Shape Memory Applications, Inc., Santa Clara, CA); however, alloys with


lower transformation temperatures (< –50°C) are available. Data from the Path-

finder mission showed a high level of consistency in the surface temperature

range at the Ares Vallis landing site [25], as shown in Figure 9. From these data, a

NiTi alloy with a transformation temperature of approximately –50°C (223°K)

would suffice.

Figure 8. Property change versus temperature for martensitic transformation in NiTi alloy.

Figure 9. Daily temperature variation at Ares Vallis landing site on Mars.


(a)

(b)

(c)

Figure 10. (a) NiTi spring in tension. (b) Upon heating through the transformation temperature, and (c) after heating.


NiTi is also superelastic, in that it possesses incredible amounts of flexibility

and kink resistance. It has a strain recovery of about 8%, which makes it more re-

silient than stainless steel. Work on the corrosion resistance of NiTi in the case of

biomedical implants [26] has shown it to be very resistant to corrosion in the

harshest environments due to the formation of a passive TiO2 layer on the surface.

The design of the cleaning device involves attaching the brush between two sets

of springs, one set normal stainless steel, and the opposing set a NiTi alloy. The ten-

sion would be set so that the steel springs would be contracted, holding the brush to

one side of the solar array, and the opposing NiTi spring set would, therefore, be ex-

tended against the strain. Upon heating of the device by the sun, the NiTi alloy

would pass through its transformation temperature and contract, the recovery strain

now being greater than that of the steel springs, and would effectively pull the brush

across the array. A prototype device is shown in Figure 10, where a heat gun was

used to heat the alloy spring. When the temperature drops back past the transforma-

tion temperature, the NiTi alloy will relax, allowing the steel spring to now contract

again, pulling the brush back. The cleaning would be repeated several times a day

as the temperature fluctuates depending upon the conditions.

6. CONCLUSIONS

The tribocharging of Mars dust can contribute to strong adhesion of dust particles to

solar panels, spacecraft, and spacesuits. Both the polarity of the charged dust parti-

cles as well as the amount of charge depend upon the surface composition of the

particles and the contacting materials. It has been shown that in an ambient Earth

environment, surface contamination and oxidation produce significant changes to

the surface composition and hence the work function of a variety of materials.

An instrument has been developed that can simultaneously measure both size

and charge distributions on an individual particle basis. A smaller, more robust

version of which is proposed to measure size and charge distributions of Mars

dust in situ.

An electrodynamic screen shows promise for preventing deposition of charged

dust particles, and an electrostatic brush for dust removal has been developed us-

ing NiTi shape memory alloy. The electrodynamic screen and cleaning device

proposed have no mechanical parts so the probability of failure is minimized.

7. CONTINUING AND FUTURE WORK

In order to understand the tribocharging properties of Mars dust, an environ-

mental chamber has been constructed that can effectively control the relative hu-

midity from 0 to 98% with a +/– 1% stability. The atmospheric composition

within the chamber can simulate that on Mars (predominantly CO2). The chamber

can be mounted on top of an E-SPART analyzer. The charging properties of a

Mars dust simulant, obtained from NASA Johnson Space Center (JSC Mars-1),


will be investigated against stainless steel and PTFE to determine the charging

characteristics in a Martian type environment. Simultaneously, an Ultra-violet

Photoelectron Spectrometer is being developed to allow work function measure-

ments to be taken as a function of relative humidity within the chamber.

A compact E-SPART analyzer that is robust enough for space flight is also be-

ing developed, as well as the development of the electrodynamic screen is being

continued.

REFERENCES

1. G.A. Landis, Paper presented at the Intersociety Energy Conversion Engineering Conference, Honolulu, HI, July 27–August 1, 1997.

2. Mars News, www.marsnews.com/planetology, November 1999. 3. G.A. Landis and P.P. Jenkins, Proc. 26th IEEE Photovoltaic Specialists Conference, 865-869

(1997). 4. G.A. Landis, Acta Astronautica, 38, 885-891 (1996). 5. R. Greeley, N. Lancaster, S. Lee and P. Thomas, Mars, pp. 835-933, University of Arizona

Press, Tuscon, AZ (1992). 6. J.R. Gaier, M.E. Perez-Davis and M. Marabito, Paper presented at the 16th AIAA/NASA/

ASTM/IES Space Simulation Conference, Albuquerque, NM, November 5–8, 1990. 7. J. Cross, Electrostatics: Principles, Problems and Applications, Adam Hilger, Bristol, England

(1987). 8. Y. Murata, Jap. J. App. Phys., 18, 1-8 (1979). 9. J. Lowell and A.C. Rose-Innes, Adv. Phys., 29, 947-1023 (1980).

10. C.B. Duke and T.J. Fabish, Phys. Rev. Lett., 37, 1075-1078 (1976). 11. D.A. Hays, Proc. International Conf. on Modern Electrostatics, Ruinian Li (Ed.), Beijing, China,

pp. 327-330, International Academic Publishers, New York (1988). 12. H. Bauser, Dechema Monographs, 72, 11-28 (1974). 13. W.R. Harper, Contact and Frictional Electrification, Laplacian Press, Morgan Hill, CA (1998). 14. D.M. Taylor and P.E. Secker, Industrial Electrostatics: Fundamentals and Measurements, John

Wiley & Sons, New York (1994). 15. C.D. Cooper and J.F. Mustard, Paper #6164 presented at The Fifth International Mars Science

Conference, Pasadena, CA (1999). 16. Ph. De Donato, C. Mustin, R. Benoit and R. Erre, Appl. Surface Sci., 68, 81-93 (1993). 17. C.D. Wagner, W.M. Riggs, L.E. Davis and J.F. Moulder, Handbook of X-Ray Photoelectron

Spectroscopy, Perkin-Elmer Corp., Eden Prairie, MN (1983). 18. NIST XPS Database, http://srdata.nist.gov/xps/Bind_e_spec_query.asp 19. M.K. Mazumder and R.E. Ware, US Patent #4633714 (1987). 20. P.A. Baron, M.K. Mazumder and Y.S. Cheng, in: Aerosol Measurements: Principles, P. Baron

and K. Willeke (Eds.), Chap. 17, Van Nostrand Reinhold, New York (1992). 21. M.K. Mazumder, S. Banerjee, R.E. Ware, C. Mu, N. Kay and C.C. Huang, IEEE Trans. Ind.

Applications, 30, 365-369 (1994). 22. M.K. Mazumder, S. Banerjee and C. Mu, in: Dispersion and Aggregation, B.M. Moudgil and P.

Somasundaran (Eds.), Engineering Foundation, New York (1994). 23. M.K. Mazumder, R.E. Ware, J.D. Wilson, R.G. Renniniger, F.C. Hiller, P.C. McLeod, R.W.

Raible and M.K. Testerman, J. Aerosol Sci., 10, 561-569 (1979). 24. C.R. Wayman, MRS Bull., 49-56, April 1993. 25. Mars Pathfinder website: http://www.mars.jpl.nasa.gov (1997). 26. S. Trigwell, R.D. Hayden, K.F. Nelson and G. Selvaduray, Surface Interface Anal., 26, 483-489

(1998).

Surface Contamination and Cleaning, Vol. 1, pp. 311–334Ed. K.L. Mittal VSP 2003

Laser cleaning of silicon wafers: Prospects and problems

M. MOSBACHER,1 V. DOBLER, M. BERTSCH, H.-J. MÜNZER,J. BONEBERG and P. LEIDERERUniversity of Konstanz, Department of Physics, Fach M676, D-78457 Konstanz, Germany

Abstract—We report on experiments on the underlying physical mechanisms in the Dry- (DLC)and Steam Laser Cleaning (SLC) processes. Using a frequency doubled, Q-switched Nd:YAG laser(FWHM=8 ns) we removed polystyrene (PS) particles with diameters in the range of 110 nm to2000 nm from industrial silicon wafers by the DLC process. The experiments have been carried outboth in ambient conditions as well as in high vacuum (10−6 mbar) and the cleaned areas have beencharacterized by atomic force microscopy for damage inspection. In DLC we have determined thecleaning laser fluence thresholds for a large interval of particle sizes. Additionally we could show thatparticle removal was due to a combination of at least three effects: substrate thermal expansion, localsubstrate ablation as a consequence of field enhancement at the particle, and explosive evaporationof moisture adsorbed from the air. Which effect dominates the process depends on the boundaryconditions. For our laser parameters no damage-free DLC was possible, i.e. whenever a particle wasremoved by DLC we damaged the substrate by local field enhancement. In our SLC experimentswe determined the amount of superheating of a liquid layer adjacent to surfaces with controlledroughness. On silicon wafers the water layer could be superheated to 250 C prior to the onset of laserinduced bubble nucleation. The heat transfer from the silicon substrate into the liquid was found to belimited by a thermal boundary resistance which can be characterized by a heat transfer coefficient of3 · 107 W/(m2K). Based on the knowledge about the particle removal mechanisms and the determinedcleaning efficiency we discuss the advantages and disadvantages of DLC and SLC as possible futureindustrial surface cleaning procedures.

Keywords: Particle removal; laser cleaning; field enhancement; cleaning mechanisms.

1. INTRODUCTION

The removal of particle contamination from surfaces is one of the crucial prerequi-sites for a further increase in the integration density of ICs and for the progress innanotechnology. At all stages of the production of ICs, e.g., from the bare Si waferto the patterned chip, particles even smaller than 100 nm in size can cause a damageto the produced structure and hence be responsible for the failure of the final device.

In the late 1980s, the experts in the field of cleaning technology predicted thattraditional cleaning methods such as ultrasonics and wet techniques would reach

1To whom all correspondence should be addressed. Phone: +49-7531-882627, Fax: +49-7531-883127, E-mail: [email protected]

312 M. Mosbacher et al.

their limit of capability [1, 2]. In addition, these traditional techniques were andstill are harmful to the environment as they consume large quantities of aggressivechemicals and water. Although the traditional methods have been continuouslyimproved [3], still particle contamination causes considerable production losses[4], and with further shrinking of line widths [5] there is a definite need to replacetraditional methods by new cleaning technologies.

One of these new approaches is called laser cleaning. In Dry Laser Cleaning(DLC) [6–8] the surface to be cleaned is irradiated by a short laser pulse. In SteamLaser Cleaning (SLC) [6, 7, 9, 10] prior to the application of the laser pulse a liquid,e.g. a water-alcohol mixture, is condensed onto the surface.

After the first attempts of implementing laser cleaning in prototype cleaningtools [11], this strategy was not pursued any further as there were too many openquestions related to the underlying physics. In the following years several researchlaboratories around the world [12–20] started to investigate the physical processesinvolved both in SLC and DLC, starting from the simple scenarios suggested bythe authors of the first publications on the subject. These scenarios also formed thebasis for certain models to describe laser cleaning and to interpret the experimentalresults obtained [6, 7, 10, 18, 19, 21–33]. However, recent experiments [31, 33–41] show that both the DLC and SLC scenarios that have been taken as commonsense so far do not incorporate all the important cleaning mechanisms and henceare oversimplified.

In this article we will first summarize our knowledge on the cleaning processesinvolved in laser cleaning and their interplay, and then present the results ofsystematic measurements of cleaning efficiencies in both DLC and SLC for particlesizes from 110 nm up to 2000 nm. The interpretation of these results will clearlypoint out the importance of the cleaning mechanisms neglected in the original SLCand DLC scenarios. Against this background we will discuss briefly the state oftheoretical modeling of laser cleaning. Based on the previous sections we will finishthe article with a statement of the prospects and problems of laser cleaning as anindustrial cleaning process from today’s state of knowledge.

2. EXPERIMENTAL ASPECTS

2.1. Sample preparation

In our quantitative studies on the cleaning efficiency we did not use irregu-larly shaped particle contaminants commonly used in many laser cleaning stud-ies (Al2O3, Si3N4,...), but spherical colloidal polystyrene (PS; Interfacial DynamicsCorporation, Portland, OR, USA) and SiO2 (Bangs Laboratories Inc., Fishers, IL,USA and Duke Scientific Corp., Palo Alto, CA, USA) particles. These particlesare advantageous for investigation of the underlying physical processes involved inlaser cleaning due to their narrow size distribution (standard deviation ± 5% forPS, ± 20% for SiO2) as compared to irregular particles. This enables studies of

Laser cleaning of silicon wafers: Prospects and problems 313

Figure 1. Typical sample as used in the laser cleaning experiments imaged in a scanning electronmicroscope. The displayed area is 4.8 µm × 4.8 µm and the particle size is 110 nm.

removal efficiencies for various, well-defined sizes. Their spherical shape addition-ally facilitates a comparison with theoretical models, as adhesion forces of particlesare mostly calculated for the geometry of a sphere on a flat substrate. Some ex-periments were also performed using irregularly shaped Al2O3 particles (SummitChemicals Europe GmbH, Düsseldorf, Germany) as contaminants.

As substrate we used industrial silicon (100) wafers (Wacker Siltronic, Burghau-sen, Germany) that were cleaned in isopropyl alcohol (IPA) in an ultrasonic bathbefore applying the contaminants.

The particles were deposited on the silicon substrate by a spin coating process,described in detail in [17, 41]. We were able to prepare samples with more than95% of isolated spheres at particle densities above 1000 per cm2. A typical examplecan be seen in Fig. 1 where 110 nm sized PS particles were deposited onto aSi wafer. Care was taken to prevent particle agglomeration, which is importantfor quantitative experiments, as agglomerates exhibit a different cleaning behaviourcompared to single particles [15, 35].

2.2. Laser sources

For all experiments we used a frequency doubled, Q-switched Nd:YAG-laser (λ =532 nm, FWHM = 8 ns).


2.3. Evaluation of the cleaning efficiency

Particle removal in the cleaned area (about 1 mm2) was determined under ambientconditions by a light scattering technique. A 5 mW HeNe laser (λ=633 nm)illuminated a spot with a diameter of 0.5 mm, which corresponds to severalhundred particles monitored. The light scattered by the particles was detected by aphotomultiplier. The monitored area was much smaller than the illuminated area,therefore in this case the laser fluence can be considered as almost homogeneous.The cleaning efficiency, defined as the fraction of particles removed, was determinedby comparing the scattered light intensities before and after the cleaning with a cleanreference sample [17].

In high vacuum (HV) we determined the fluences necessary for particle removalby inspecting the illuminated spot with an optical microscope prior to and after thelaser pulse. In addition, we measured the cleaning laser fluence threshold relative tothe melting threshold of Si by monitoring the reflected light of the HeNe laser withns time resolution. As the laser fluence for the onset of melting of silicon is wellknown, this can be used for conversion of relative fluences into absolute numbers.

2.4. Determination of laser fluence

The determination of laser fluences for the nanosecond pulses is described in detailin [17]. Briefly, the laser fluence was determined relative to the well-known meltingthreshold fluence of Si, making use of the higher reflectivity of the molten withrespect to the solid phase. This was done by time-resolved monitoring of thereflected light of a HeNe laser (λ=633nm). Simultaneously the laser’s scatteredlight was detected in order to probe the cleaning efficiency.

The techniques described above are not suitable for determining laser fluencesin bulk liquids. Hence for the bubble nucleation experiments we chose another,intrinsic laser fluence calibration method. During the experiment we found a distinctchange in the reflectivity of the water/silicon system. As computer simulations ofthe temperature dependency of this reflectivity show, this change is primarily causedby the temperature change in the water. Based on this, it is possible to derive [42]that the integral

W :=∫ 9µs

4µs

(Rp(t)

R0,p

− 1

)dt ∼ F (1)

containing the reflectivity change in p-polarization (R0,p initial reflectivity, Rp(t)

reflectivity after bubble nucleation) is directly proportional to the applied laserfluence F . The upper boundary of the integral is given by the end of the detectiontime, the lower boundary of 4 µs was chosen as the time where no gas bubbleswere present on the surface any more. Hence, the applied local laser fluence rightin the area where bubble nucleation is probed can be determined from the measuredreflectivity changes of the system.


Figure 2. Interferometric measurement of the laser-induced surface displacement. The frequencydoubled Nd-YAG pulse is attenuated by glass plates (A) and guided to the silicon substrate byseveral prisms (P). A heterodyne interferometer (IF, B.M. industries, SH-130, bandwidth 200 kHz-45 MHz) measures the surface displacement of the sample. The temporal pulse shape is captured by aphotodiode (PD). Displacement and pulse shape are recorded on a digital storage oscilloscope. A lens(L) was used to increase the laser fluence at the substrate.

2.5. Surface acceleration measurement setup

A detailed description of the setup shown in Figure 2 which was used for the deter-mination of surface accelerations can be found in [43]. Briefly, we illuminated thesilicon substrate by a frequency doubled Nd:YAG pulse and determined the surfacedisplacement of the sample by a heterodyne interferometer. Both displacement andpulse shape were recorded on a digital storage oscilloscope.

2.6. Bubble nucleation experiments

Figure 3 shows the setup for the bubble nucleation experiments. In order to nucleatethe bubbles a Q-switched Nd:YAG laser (λ=532 nm, FWHM=8 ns) heated thesilicon sample. The pulse energy was split (BS) and measured for each individualpulse by an energy meter (FM: Field Master, Coherent). Sample and liquid wereplaced in a fused silica cuvette that could be heated up to 360 K. Bubble growthwas monitored by a cw Ar-ion laser (λ=488 nm, P=175 mW) which was focusedonto the sample.

Both the specularly reflected beam of the Ar-ion-laser and the light scattered bythe nucleated bubbles were collected in forward direction (as shown in the diagram),as well as the scattered light perpendicular to the incident ray (similar setup, notshown in the diagram). For all the light detection we used fast photodiodes (PD:FND 100, rise time < 1 ns) covered by interference filters (IF). By a polarizingbeam splitter (PBS) the reflected beam was decomposed into its p- and s-polarizedconstituents that were detected individually.


Figure 3. Optical determination of laser induced bubble nucleation (for details see text).

3. PARTICLE REMOVAL MECHANISMS

Already the authors of the first publications on laser cleaning [6, 7, 9, 10] suggestedphysical mechanisms to describe the particle removal process. In DLC the removalprocess was ascribed to the thermal expansion of the substrate due to the heatingwith the laser pulse. SLC was explained as a consequence of the explosiveevaporation of the applied liquid. However, recent research [40, 44, 45] in the fieldof laser cleaning has shown that several other mechanisms are of importance forthe process. In the following sections we will describe the cleaning mechanismsexperimentally verified so far.

3.1. Thermal substrate expansion

Most authors explain the particle removal process in DLC in the following way:during the laser pulse its energy is absorbed in the substrate. Due to the subsequentthermal expansion the surface with the adhering particle is accelerated and theparticle gains kinetic energy. Depending on their elastic properties some energyis also stored in elastic deformation of both particle and surface. At the end of thepulse the expansion of the surface stops and the particle leaves the surface due to itsinertia.


From this it is clear that a measurement of the dynamics of the substrate expansionupon illumination with a laser pulse provides insight into this cleaning mechanism.Detailed discussions of our experiments can be found in [41, 43], therefore we willonly highlight the results here.

Using an interferometric setup as described in Section 2 we measured theexpansion of a silicon substrate after illumination with a Nd:YAG laser pulse(FWHM = 10-20 ns, λ=532 nm) with ns time resolution. Typical values of thisdisplacement were a few nm. By numerical derivation of the averaged displacementas a function of time of 600 individual experiments we obtained the deceleration ofthe substrate surface. This deceleration of the surface, rather than its acceleration asdiscussed by most authors in the description of DLC, is the relevant parameter forthe particle removal [45]. For typical laser cleaning fluences of a few 100 mJ/cm2

the decelerations are in the order of about 107 m/s2, just the order of magnitude thatis thought to be necessary for removal of such small particles [7].

The experimental values are in good agreement with a simple theoretical descrip-tion of the thermal expansion of the substrate. Taking into account only the 1Dheat equation2 and denoting by R the reflectivity, by α the linear thermal expansioncoefficient, by CP the specific heat of the substrate, by ρ its density, by f (t) thetime dependent normalized intensity of the laser, and by F its fluence the surfacedisplacement is given by

d(t) = α

ρ CP

(1 − R)F

∫ t

−∞f (t ′) dt ′ (2)

as the Grüneisen parameter α/(ρ CP ) is almost temperature independent. Expres-sion (2) can be derived for a known pulse shape f (t), and for a gaussian pulse withFWHM τ it yields a maximum deceleration of

amax = −1.71 · 106gns2

mJ/cm2

F

τ 2(3)

with a strong 1/τ 2 dependence on the pulse length.Experimentally we were able to verify the linear dependence of both displacement

and maximum deceleration on the applied laser fluence. However, the 1/τ 2-dependency on the pulse length was not found, which might be explained bydeviations of the actual pulse shape from an ideal Gaussian especially for longpulses. In general, the experimental data were in good agreement with the onescalculated from equation (3).

2Thermoelastic effects [46, 47] are not taken into account, as the area where the displacement isdetected by the interferometer was at least ten times smaller than the illuminated spot and was locatedin its center. Thus lateral thermal gradients are neglected at this spot.


Figure 4. Intensity (E2) of the electromagnetic field around a PS particle (diameter 1700 nm,refractive index 1.59) in vacuum when illuminated by a laser at λ=800 nm. The laser enters fromthe left, and the light is polarized in the image plane. At the light averted side (i.e., not directlyilluminated by the laser) of the particle the intensity is enhanced by a factor of about 30.

3.2. Local substrate ablation

Besides the thermal expansion of the substrate due to heating by the absorbed laserenergy we experimentally identified [35, 38, 40] a second cleaning mechanismwhich is pronounced particularly in DLC.

3.2.1. Optics of particles with sizes comparable to the wavelengthIn 1908, the German physicist Gustav Mie described the scattering of light at adielectric sphere in vacuum [48]. Two predictions of his theory are particularlyimportant for the application of laser light to the removal of particles comparable insize with the wavelength.

First, one would expect a “focusing” effect, i.e. the laser intensity underneath theparticle and hence in the substrate plane should be higher than the incoming laserintensity that hits the bare part of the surface. For particles of sizes much largerthan the wavelength (about one order of magnitude or more) this focusing can beexplained in terms of geometrical optics as focusing by a spherical lens. However,geometrical optics fails to explain the intensity enhancement underneath particlessmaller in size than the wavelength. In this case the near field distribution of the


electromagnetic field at the particle has to be computed numerically. An exampleof such a computation using a program based on [49] is given in Fig. 4.

But not only does the Mie theory describe an enhancement of the laser intensity inthe particles’ near field, it also predicts that for certain values of the size parameterπd/λ (d denoting the particle diameter, λ the laser wavelength) the enhancementshould be particularly efficient, resulting in a resonant intensity enhancement, theso-called “Mie-resonances”.

3.2.2. Near-field induced substrate damageWhen inspecting contaminated samples by scanning electron microscopy (SEM) oratomic force microscopy (AFM) after DLC using ns laser pulses, the consequencesof the field enhancement process became obvious: all over the cleaned areaswe found substrate damages localized exactly at the former particle positions[35, 37–39]. These damages manifested as melting pools or even holes in thesurface, typical examples can be seen in Fig. 5.

The consequences for the laser cleaning process are obvious. The intensityenhancement reduces the maximum laser fluence that can be applied in the process.Usually in laser cleaning studies [19, 31] the laser fluence corresponding to themelting threshold of a bare surface is taken as the damage threshold fluence. Ourexperiments show clearly that this is an inadequate definition. Instead one musttake into account the enhanced laser fluence underneath the particles, as it will bediscussed in Section 4.

From the obtained AFM images we were able to analyse in detail the surfaceprofile at the damaged sites. Here we found that for high field enhancement factorsthe silicon substrate was not only molten, but that some material was even ablated(see Sec. 4). The momentum transfer to the particles during the ablation processsignificantly contributes to the cleaning process and hence local substrate ablation

Figure 5. Surface damage caused by local melting and ablation of the substrate. The melting wasinduced by the enhancement of laser intensity in the near field of the particles at the surface. Shownare AFM images in top view (left) and a 3D illustration (right) of the same site.


must be considered as a particle removal mechanism. But of course, due to theaccompanying substrate damage, this removal mechanism should be suppressed inany practical application of laser cleaning.

3.2.3. Resonances in the laser cleaning threshold fluenceAs a second consequence, the optical resonances in the enhancement should giverise to a resonant, non-monotonous dependency of the laser cleaning thresholdfluence on the size parameter. In case of a resonant enhancement, the laser fluencefor particle removal should be less than in absence of resonance. In Section 4 wewill demonstrate the existence of the resonances from experimental determinationof threshold fluences for a variety of particle sizes. At this point it is worth tomention that the observation of the size dependent Mie-resonances in the thresholdfluence was only possible due to the use of spherical colloidal particles with a smallsize distribution (± 5%).

3.3. Explosive evaporation of a liquid

Although laser cleaning has been known for more than ten years now, still manyquestions related to the underlying physical processes are not answered yet. Theauthors of the first publications on the subject [6–10] suggested simple physicalscenarios that were accepted for the interpretation of experimental data thereafterand were taken as a basis for theoretical modeling: DLC was explained solely bythe thermal expansion of the substrate and the adhering particle, SLC by explosiveevaporation of the liquid condensed onto the surface [6, 7, 10, 18, 19, 21–33].However, it is still not clear whether these scenarios accurately reflect the actualparticle removal process for all laser parameters and cleaning environments studiedso far, even in the experiments that are quoted to justify the models. This isbecause the models suffer from major drawbacks, such as the neglect of the cleaningmechanisms field enhancement induced local substrate ablation.

In addition to the thermal expansion of the substrate the early publications onlaser cleaning [6, 7, 9, 10] suggested a second mechanism for the removal of dirtparticles: the explosive evaporation of a liquid such as water or alcohol. Here theliquid layer is heated directly [10] (laser energy absorbed in the liquid) or indirectly[6, 7] (laser energy absorbed in the substrate and transferred to the liquid via heatconduction) by the laser pulse. Due to the short time scale of this heating the liquidis superheated to a certain extent, becomes intrinsically unstable and subsequentlyevaporates explosively. The momentum transfer of this explosion onto the particleis then thought to lift the particles off the surface. In the following, we will discussthis cleaning mechanism in three variations: the evaporation of liquid adsorbed atthe particle-surface area as it is the case for DLC in ambient air, the evaporation ofa liquid film as it is found in SLC, and the laser induced nucleation of bubbles in abulk liquid that provides considerable information on the underlying physics.


3.3.1. Liquid adsorbed at the particle-surface contact areaSeveral authors report that in their DLC experiments an increase in ambienthumidity led to an increase in cleaning efficiency [31, 33, 35]. This was attributedto the additional cleaning force provided by the explosive evaporation of the wateradsorbed at the interstice between particle and surface.

In a recent publication on the DLC of polyimide [33] the authors systematicallyincreased the humidity from 30% to 90% and found an efficiency increase, a strongsupport for the above interpretation. However these experiments were not carriedout on silicon and the obtained data originated only from two particle sizes.

One straightforward way of minimizing the amount of water present at theinterstice of particle and surface is the heating of the substrate to a temperatureabove the boiling point of water. We used arbitrarily shaped Al2O3-particles 300 nmin diameter as contaminants in order to minimize cleaning by local ablation and toprevent a deformation of the particles due to heating. For comparison we performedtwo different sets of DLC (Nd:YAG laser, λ=532 nm, FWHM=8 ns) experiments.In the first one we carried out DLC at ambient conditions, as commonly done inthe investigations published in literature. During the second set of experiments,we heated the sample to a temperature of 120 C prior to the application of thecontaminant particles and during the whole experiment. As a result, we found anincrease in the cleaning threshold laser fluence for the heated sample of more than20% versus the sample cleaned in ambient conditions. This result can be taken as afurther indication for the involvement of liquid adsorbed from the atmosphere andstimulated a more systematic and detailed investigation as described in Section 4.

3.3.2. Liquid filmsLaser cleaning can not only be promoted by the adsorption of atmospheric humidity,but also by the condensation of a liquid onto the sample on purpose. This process,called Steam Laser Cleaning (SLC), relies very much on the explosive evaporationof the liquid film. So far the reported studies on this subject [15, 17, 36] lacka systematic investigation of the parameters of this film, i.e. its thickness andcomposition. However, this control on the film parameters is not easy to achieve,and thus all research on the dynamics of laser induced bubble nucleation in liquids,and hence the physical processes underlying SLC, published in the literature so farhas been carried out in bulk liquids.

3.3.3. Laser induced bubble nucleation in bulk liquidsThe study of laser induced bubble nucleation in liquids provides the key for theunderstanding of liquid enhanced laser cleaning processes and their increasedefficiency in comparison to dry techniques. Different methods were used for thesestudies: optical methods such as the change of the substrate reflectivity due to thebubble film [50–54] and the detection of light scattered by the bubbles [50, 52, 54],the detection of the pressure wave nucleated during the growth or collapse of thebubbles [52, 54–56] or surface plasmon spectroscopy [54].


Typically, these studies were carried out on rather rough surfaces such as metalfilms. Here quite moderate superheatings of about 20 K were found to besufficient for the nucleation of laser induced bubbles, a value much lower than thetheoretically predicted one of 200 K for water [57]. This has been attributed to thesurface roughness of these substrates. But to our knowledge there are no systematicstudies that confirm this roughness-dependency of the threshold. Yet this is thecrucial point in the transfer of the obtained results from the water/metal film systemto the water/silicon wafer system under consideration in laser cleaning studies. Withrespect to this several open questions arise: 1) Supposing rough metal films providenucleation sites for gas bubbles and cause only moderate superheatings of the liquid,what will in contrast be the degree of superheating on smooth substrates like siliconwafers? 2) Is it justified to determine the temperature of the superheated liquidby measuring [58] or computing [28] the temperature of the silicon substrate andsimply assuming a perfect heat transfer from the substrate into the liquid?

3.3.3.1. Bubble nucleation thresholds. We, therefore, studied [40, 42] the nu-cleation of bubbles at a superheated liquid/solid interface under controlled surfaceroughness. When the incident laser fluence reached a well defined threshold, scat-tered light was observed. This indicates a sharp nucleation threshold. The thresh-old decreases with increasing starting temperature T0 of the water as less energyis needed to reach the nucleation temperature. Figure 6 shows these temperaturedependent thresholds for the examined systems. The lines are linear extrapolationsto vanishing fluence and yield the nucleation temperature by their intersection with

Figure 6. Comparison of the bubble nucleation thresholds and extrapolated superheating limits ofwater on substrates of different roughnesses.


the temperature axis. As a consequence of the interpolation over a large temperatureinterval this procedure yields a rather large error of about 20%.

In a first series of experiments we determined the nucleation thresholds andsuperheating temperatures for water on a 50 nm silver film thermally evaporatedonto glass. This system was chosen for comparison with previous experiments in[54, 56], and indeed we found a nucleation temperature of (130 ± 30) C, onlyslightly above the one reported in the literature. In contrast, water on conventionallypolished silicon wafers (rms roughness 0.2 nm) exhibited a bubble nucleationtemperature of (250 ± 30) C in close agreement with the expected value fromtheory and far above that one measured on the silver film. To verify the influenceof the surface roughness we patterned a silicon wafer with holes (diameter approx.500 nm, depth approx. 40 nm, hole density approx. 0.05/µm2, for the preparationmethod see reference [39]). On this substrate the nucleation temperature decreasedstrongly to (160 ± 13) C, close to that of the rough silver film, a clear evidence forthe influence of surface roughness.

3.3.3.2. Heat transfer coefficient. It is well known that there exists a discontinu-ity in the temperature profile at boundaries between different materials. This is dueto the finite heat conductivity of the boundary region. The so-called heat transfercoefficient

ζ = Q

AT(4)

quantifies this thermal boundary resistance as functions of the heat flow Q, theboundary area A and the temperature jump T . Clearly, in laser cleaning thethermal resistance can limit the heat flow from the substrate into the liquid and thuslower the liquid temperature considerably. None of the published computations oftemperature profiles in laser cleaning incorporates this fact. Probably one reason isthat although the phenomenon is well investigated for low temperatures below 50 K(Kapitza resistance, see [59]) and for technical applications at room temperature,long time scales (several seconds) and macroscopic dimensions (Nusselt-number,see [60]), there are no data at ns time scales and nm length scales, as needed for theinterpretation of laser cleaning data.

The data obtained in the bubble nucleation experiments allow us now for the firsttime to give the heat transfer coefficient for these scales. Figure 7 shows computedmaximum temperatures of the water layer adjacent to the silicon surface. Ourcomputations are based on the 1D heat equation implemented in a finite elementalgorithm taking into account different values for the heat transfer coefficient ζ

between silicon and water. The calculations have been repeated for different startingtemperatures of the water and the corresponding experimentally determined laserthreshold fluence for bubble nucleation was applied.

At this threshold fluence the water layer must reach a specific temperature whichis the same for all starting temperatures. If the assumed value of ζ is too smallor too large, however, the water layer will reach maximum temperatures that are


Figure 7. Computed maximum temperatures in water on a silicon surface as function of the heattransfer coefficient (HTC) ζH2O for different starting temperatures T0. As laser fluence we used theexperimentally determined threshold fluence for bubble nucleation.

different for each starting temperature. It can be seen from these computations thatfor all starting temperatures of the water the graphs of the calculated maximumtemperatures as a function of the assumed heat transfer coefficient intersect atone single point. This point is given by a value of ζH2O = 3 · 107 W/(m2K)and a maximum water temperature of 250 C – just as determined experimentally.Therefore, this ζ -value represents the heat transfer coefficient in the studied system.Interestingly, the computations exclude values of ζ < 3 · 106 W/(m2K), as in thiscase the equilibrium boiling temperature of 100 C is not reached for all startingtemperatures.

3.4. Further mechanisms

The mechanisms described above are those that could be identified by our experi-ments (see Sec. 4) so far. However, this does not mean that they are the only onesinvolved in laser cleaning! There are at least three more mechanisms that may playa role in the process and have been suggested: particle vibrations/elastic deforma-tions, light pressure, and surface acoustic waves. In a very recent publication [45]Arnold et al. suggested that elastic deformation of the particles might lead to parti-cle removal if the particles were stimulated in resonance with their eigenfrequency.Vereecke et al. [61] suggested that at grazing incidence of the laser pulse the lightpressure might roll the particles over the surface. Also surface acoustic waves mayfacilitate DLC [62]. Although in our opinion so far there is no conclusive experi-mental evidence for those mechanisms, they might well be found in laser cleaningas well as others unknown today.


4. EFFICIENCY MEASUREMENTS

4.1. Steam laser cleaning

For the SLC we used [17] the same experimental setup as in the DLC experiments,but supplemented it with a steam providing unit [11]: a controlled flow of filtered,pressurized air was directed through a reservoir of a water/IPA mixture (90% water)heated to 330 K. Then the steam/air mixture was directed to the sample via a nozzleat a distance of 1.5 cm from the area to be cleaned. The IPA’s role was to improvethe wetting of the steam condensed onto the silicon wafer leading to the formationof a liquid film. We estimated the film thickness using ellipsometric measurementsto be about 200-400 nm.

The cleaning process was found [35] to be statistical in a way that the number Nr

of the remaining particles after n cleaning steps is given by Nr/N0 = (1 − p)n asfunction of the single shot cleaning efficiency p and the original particle number N0.Therefore, in Fig. 8 we plotted only the cleaning efficiencies after 20 cleaning steps(steam and laser pulse) for the sake of clarity. This figure shows the dependenceof the cleaning efficiency on the applied laser fluence for particles of different sizes(60 nm-800 nm), materials (PS, SiO2, Al2O3) and geometries (spherical PS, SiO2

and arbitrarily shaped Al2O3).Some important results can be obtained from this graph. A predominant feature

is the existence of a universal cleaning threshold for all particles investigatedat a laser fluence of about 110 mJ/cm2. For slightly higher laser fluences we

Figure 8. Cleaning efficiency for various colloidal particles in SLC with λ=532 nm, FWHM=7 ns.The cleaning threshold is found to be the same for all particles.


observed a very steep increase in the cleaning efficiency, and values above 90%after 20 cleaning steps (laser pulse plus steam jet) are reached well below themelting threshold of a bare silicon substrate, even for particles as small as 60 nm.The threshold cleaning fluence of 110 mJ/cm2 found in SLC is larger than thethreshold fluence of 80 mJ/cm2 determined for bubble nucleation in bulk water(see. Sec. 3.3). This needs further experimental investigation, however there aretwo possible explanations. First, the bubble nucleation process may depend on thefilm thickness, and hence there would be differences in the process dynamics whencomparing bulk liquid to liquid films. Second, it is not clear so far whether the merenucleation of bubbles is sufficient for an efficient cleaning or whether more laserfluence is needed to create larger or faster growing bubbles.

4.2. Dry laser cleaning in ambient conditions

A first step in the investigation of DLC is the study of particle removal in ambientconditions (relative humidity 30-40%) with a ns Nd:YAG-laser. This environmentrepresents the conditions that may be found in a possible future application of theprocess. A flow of pressurized, filtered air was used to blow away the removedparticles and to prevent their redeposition.

In Figure 9 the thresholds in applied laser fluence for the removal of PS particlesare plotted as a function of the particle size.

First we would like to draw the reader’s attention to two very important thresholdsin the laser cleaning process. One of them is the threshold for the onset of meltingof the bare substrate. As DLC is aimed for an industrial application, any change

Figure 9. Thresholds in the applied laser fluence for particle removal in DLC in ambient air. Particlessmaller in diameter than 110 nm could not be removed.


of the structure of the silicon wafer, i.e. the silicon substrate and the native oxidelayer of specified thickness, as induced by melting has to be strictly avoided. Fromexperiments [63] this melting threshold is known to be about 280 mJ/cm2 for ourlaser parameters, which therefore represents the upper limit of applicable laserfluence. The second threshold, also indicated in Figure 9, is the cleaning thresholdof the SLC process as discussed in Section 4.1.

In order to obtain information on the dependence of the cleaning threshold onthe particle size we investigated many different particle diameters in the range of110-2000 nm. It turned out that only by the investigation of this large variety ofparticle sizes some main characteristics of the dependence of the cleaning thresholdon the particle diameter could be revealed. At first sight the shape of the curveroughly follows a 1/rk-trend, r denoting the particle radius and 1< k <2. Thismonotonic behavior was predicted by the first DLC models, and in fact already thefirst publications on DLC reported that smaller particles were harder to remove thanlarger ones due to the nature of the adhesion forces [6–8].

Taking a closer look, however, one discovers an oscillating behavior of the thresh-old fluences which we attribute to optical resonances as discussed in Section 3.2.3.This is illustrated in the graph by the line connecting the data points. However, itshould be pointed out that this line is just a guide to the eye and does not describethe field enhancement efficiency as function of the particle diameter. The numberof particle sizes used in our experiments is not sufficient to resolve this dependency.

In DLC using sub-ns pulses the removal of a particle is always accompanied bythe formation of a hole [38], i.e. the hole formation threshold is identical withcleaning threshold. From this, one can conclude that here the dominant cleaningmechanism is local substrate ablation. Against the background of field enhancementas the origin of surface damage, it is a natural consequence also for nanosecond laserpulses to determine not only the cleaning threshold fluences in DLC, but also thelocal melting/ablation thresholds. The latter ones, instead of the melting thresholdof the bare silicon surface, represent the true upper limit for the applicable laserfluence and are by their nature particle dependent. For its determination we madeuse of the Gaussian spatial beam profile of our cleaning laser. Due to this profile aspatial variation of the position in the cleaned area corresponds to a variation in thelocally applied laser fluence. In a post process analysis we investigated the cleanedareas of our samples with an atomic force microscope (AFM, Digital Instruments).

By imaging damage sites such as displayed in Figure 5 at different locations inthe cleaned areas and especially at their borders, corresponding to the cleaningthreshold fluence, we determined the damage threshold for each particle size. Forall particles investigated the cleaning threshold was identical with the damagethreshold. Damage-free DLC was not possible by applying the laser parameterswe used!

The AFM images of damage sites contain even more information on the particleremoval mechanism as they reveal quantitative topographical information. All theinvestigated damage sites showed the same features: a “trench” surrounding a


Figure 10. Trench depth at the damage sites as a function of the particle size when the particles wereremoved in DLC applying the threshold cleaning fluence. For particles larger than 250 nm in diameterthe depth increases strongly with the particle size, for smaller particles it remains almost constant.

central “hillock”. Generally speaking, at high laser fluences the hillock was lowerand the trench deeper, whereas for low laser fluences a hillock was detectable butthe trench almost disappeared. In Figure 10 we plotted for the investigated particlesthe mean trench depths for damages that occurred at the cleaning/damage thresholdin a double logarithmic graph.

This plot clearly shows two regimes: for particles smaller than about 250 nm indiameter the depth remains almost constant at about 1 nm. For larger particles wefound a strong increase in the trench depth and the volume of the hillock was smallerthan that of the trench, i.e., ablation of substrate material had taken place.

From this observation we conclude that even for DLC using ns pulses localablation of the substrate plays a role as a cleaning mechanism for “large” particleswhere the field enhancement is high and thus provides fluences high enough forablation. For smaller particles field enhancement probably causes local melting, butno ablation at the threshold cleaning fluence.

4.3. Dry laser cleaning in high vacuum

As already discussed in Section 3 there exists experimental evidence that “dry” lasercleaning in ambient air is facilitated by adsorbed humidity from the surroundingatmosphere. In order to minimize this effect and hence to exclude explosiveevaporation of adsorbed moisture as a cleaning mechanism, we repeated theexperiments reported above in high vacuum (HV, 10−6 mbar). The samples wereallowed to degas approximately for 10 hours at HV before cleaning.


Figure 11. Comparison of the cleaning thresholds in ambient air and high vacuum. In vacuum thethresholds are distinctly higher.

The results and the comparison with the ambient values are presented in Figure 11.In vacuum the laser fluences necessary for particle removal are higher for particlessmaller than about 800 nm in diameter. For larger particles no difference in thethreshold was detected between ambient conditions and HV.

We attribute this to the predominance of different cleaning mechanisms for differ-ent particle sizes. Cleaning in ambient conditions is facilitated by explosive evapo-ration of adsorbed moisture from the air. When exposed to vacuum for several hoursthe amount of moisture at the particle-surface contact is significantly decreased, thecontribution of explosive evaporation to the cleaning forces decreases, and conse-quently the cleaning threshold increases. This argument should also be valid forparticles larger than 800 nm in diameter. However, as discussed above, for largeparticles local ablation induced by field enhancement appears to strongly contributeto the cleaning process. And, of course, this is the case as well in ambient air as invacuum and hence no large difference in the cleaning thresholds for the two differ-ent environments is detected.

5. PROSPECTS AND PROBLEMS

Against the backdrop of the results presented above we now discuss the prospectsas well as the problems of using laser cleaning as an industrial cleaning technique.In this context we would like to highlight three main aspects: the systematicdetermination of cleaning thresholds, the role of different cleaning mechanisms,and the consequences for the applicability of theoretical models proposed so far.


5.1. Cleaning thresholds and process efficiency

A systematic determination of cleaning thresholds in both DLC and SLC shouldprovide key information for the application of laser cleaning, as it allows to predictthe minimum particle size that can be removed and to judge which of the twoprocesses DLC or SLC is more efficient. On the basis of our measurements thiscomparison can be done for the first time and for a large size interval of particles.

Perhaps the most striking difference in the two laser cleaning methods is thedependence of the cleaning threshold fluence on particle size. Whereas in SLCthis threshold appeared to be universal, i.e. size- and material-independent forthe investigated particles, in DLC we found in agreement with other authors a sizedependent threshold, with smaller particles being harder to remove than larger ones.From this it is obvious that SLC is a more efficient method for small particles, i.e.for particles smaller than about 400 nm in diameter (for particles larger than 400 nmsee below) which is the most interesting size regarding the cleaning of bare siliconwafers in the semiconductor industry. In addition, SLC is superior to DLC in theminimum particle size that could be cleaned from silicon wafers. Recalling that thecurrent minimum line width in ICs is 130 nm, which means that particles of about60-70 nm in size have to be removed, this is a key information on the quality of acleaning method. The lower size limit of particles that could be removed by DLCwas found to be 110 nm, compared to 60 nm and an efficiency above 90% in SLC.Summarizing the above, SLC is superior to DLC due to three crucial characteristics:its universal cleaning threshold, its lower threshold fluences for the relevant particlesizes, and its capability of removing sub 100 nm-particles.

5.2. Consequences of cleaning mechanisms involved

Although in DLC no particles smaller than 100 nm could be removed, at a firstglance it seems to be the more appropriate method for larger particles as its cleaningthresholds are distinctly lower than the universal SLC threshold.

However, for a judgement of the perspectives of SLC and DLC it is not sufficientto solely determine and compare cleaning efficiency and laser cleaning thresholdfluence. On the contrary, as our studies above show very clearly, this comparisonmust be put into perspective by taking a closer look at the cleaning mechanismsinvolved. The most important physical process not taken into account in traditionalinvestigations and only recently [34, 35, 38–40] studied is the local substrateablation due to the enhancement of the laser intensity in the near field of theparticles.

The first, and most obvious, consequence of field enhancement is a locallyincreased laser fluence underneath the particle, and hence a decrease in the incidentlaser fluence necessary for particle removal. At a first sight this looks like apositive effect, but obviously a locally enhanced laser intensity drastically lowersthe threshold for surface damage, and indeed we did observe surface damagecaused either by melting (small particles) or local substrate ablation (large particles)


whenever a particle was removed in DLC. This means that damage-free DLC wasimpossible with our laser parameters. In fact, the local substrate ablation combinedwith a momentum transfer to the particles was found to be the dominating cleaningmechanism with large size parameters and sub-nanosecond laser cleaning [40].

A second consequence of field enhancement also argues against a technologicalapplication of DLC. The DLC results obtained both in ambient conditions and inHV confirm a general trend already obtained by other authors [6, 7, 19, 31, 33, 64]and predicted by their models: cleaning thresholds for smaller particles tend tobe higher than for larger ones. Yet this is not a strict rule. In contrast to theabove cited experiments we used a large variety of particle diameters ranging from110 nm to 4100 nm and could show that the dependence of the cleaning thresholdas a function of the particle diameter was non-monotonous as a consequence of theoptical resonances in the near field of the particles. This non-monotonous behaviourin DLC makes it difficult to apply the correct cleaning fluence for the removal of aspecific particle size, unlike the universal threshold in SLC.

It should be pointed out that field enhancement is an intrinsic physical effect thatcannot be avoided in laser cleaning. However, it can be minimized by choosingsuitable wavelengths to reduce the size parameter or by carrying it out in anenvironment with a refractive index close to that of the particles.

5.3. Models

Models which describe the laser cleaning process accurately would be an importanttool in the real world application, as they could predict the optimum cleaningconditions for various substrates, particles, lasers and cleaning environments.However, such models must inevitably incorporate all the knowledge gainedregarding the physical processes behind both SLC and DLC.

So far in DLC the particle removal has always been solely ascribed to thethermal expansion of the substrate. Field enhancement was taken into accountonly as increased laser fluence [19, 34], but not as the origin of an additionalcleaning mechanism via local ablation. The third cleaning mechanism, evaporationof adsorbed ambient moisture, is not incorporated into any model published sofar. The latter is even more important, as most of the experiments carried outto compare with laser cleaning models were performed in ambient conditions[22, 24, 26–28, 30, 64]. Only in [19] the experiments were conducted in vacuum,and Vereecke et al. [31] varied the relative humidity and reported a higher cleaningefficiency at increased humidity levels. In addition to this, only recent models[44, 45] consider the elastic properties of the substrate/particle system which mighthave a great influence on the cleaning process as well.

Future models should incorporate at least these three cleaning mechanisms andtreat DLC as an interplay of all of them. Depending on the process parameters(laser wavelength, pulse duration, optical constants of the materials, etc.) theiroverall contribution to the cleaning will vary. It should also be pointed out that


the cleaning mechanisms are not independent of each other, e.g. adsorbed moisturemay influence the field enhancement pattern.

Compared to DLC the state of modeling the SLC process is at a rather initialstage. Two groups [28, 65] have suggested models to describe it. Yet these modelsrely on far-reaching assumptions in the description of the processes of laser inducedbubble nucleation and growth as well as on the assumption of the temperature of thesuperheated water layer as growth medium. As our experiments on the last aspectshow, it is impossible to transfer the results gained on rough metal films [50, 55, 56]to the water film/silicon system. Furthermore, it is not clear neither qualitatively norquantitatively how the explosive evaporation differs between bulk water (as in ourinvestigations) and water films (as in SLC) or even small water menisci as they canbe found in ambient environment DLC. Therefore, a good deal of future researchon the dynamics of laser induced bubble nucleation and the explosive evaporationin all these systems is necessary to accurately describe SLC.

6. SUMMARY

In this paper we have described our state of knowledge on the cleaning mechanismsresponsible for particle removal in laser cleaning. Besides the well-known thermalexpansion of the substrate and the explosive evaporation of a water film weidentified local substrate ablation as another cleaning mechanism. Additionallywe have shown the significant impact of the explosive evaporation of atmosphericmoisture adsorbed at the particles for DLC.

Local substrate ablation caused by field enhancement in the particles’ near fieldnot only causes particle removal in DLC, but inevitably also causes substratedamage. Furthermore a damage-free DLC process was not possible with the laserparameters we used in our experiments.

Steam laser cleaning, on the contrary, proved to be superior to the DLC processdue to its higher efficiency, universal cleaning threshold and its capability to removemuch smaller particles.

These findings argue for the application of SLC in wafer cleaning and underlinethe need for further research on the physics of both DLC and SLC as only thisknowledge will ensure a successful implementation of the technique in futureindustrial applications.

Acknowledgements

We thank Prof. B. Luk’yanchuk (DSI, Singapore) and Dr. Nikita Arnold (Johannes-Kepler-University, Linz, Austria) for useful discussions. The authors would alsolike to thank Dr. Bernd-Uwe Runge, Christof Bartels, Johannes Graf, Florian Lang,and Michael Olapinski (all of University of Konstanz) for constructive discussionsof the findings of our experiments. Financial support by the EU TMR project “LaserCleaning” (No. ERBFMRXCT98 0188) and the Konstanz Center for Modern


Optics is gratefully acknowledged. Wacker Siltronic supplied the industrial siliconwafers.

REFERENCES

1. S.A. Hoenig, in: Particles on Surfaces 1: Detection, Adhesion and Removal, K.L. Mittal (Ed.), p.3, Plenum, New York (1988).

2. T. Hattori, Solid State Technol., 8 (July 1990).3. R. Kohli, in: Particles on Surfaces 7: Detection, Adhesion and Removal, K.L. Mittal (Ed.), VSP,

Utrecht, in press.4. R. DeJule, Semiconductor Intl., 65-68 (August 1998).5. International Technology Roadmap for Semiconductors, Sematech International (2000).6. W. Zapka, W. Ziemlich and A.C. Tam, Appl. Phys. Lett. 58, 2217 (1991).7. A.C. Tam, W.P. Leung, W. Zapka and W. Ziemlich, J. Appl. Phys. 71, 3515 (1992).8. A.C. Engelsberg, Mater. Res. Soc. Symp. Proc. 315, 255 (1993).9. I.V. Beklemyshev, V.V. Makarov, I.I. Makhonin, Y.N. Petrov, A.M. Prokhorov and V.I. Pustovoi,

JETP Lett. 46, 347-350 (1987).10. K. Imen, J. Lee and S.D. Allen, Appl. Phys. Lett. 58, 203-205 (1991).11. H.K. Park, C.P. Grigoropoulos, W.P. Leung and A.C. Tam, IEEE Trans. CPMT A, 17, 631 (1994).12. J.B. Heroux, S. Boughaba, I. Ressejac, E. Sacher and M. Meunier, J. Appl. Phys. 79, 2857 (1996).13. S. Boughaba, X. Wu, E. Sacher and M. Meunier, J. Adhesion, 61, 293-307 (1997).14. A.C. Tam, H.K. Park and C.P. Grigoropoulos, Appl. Surf. Sci. 127-129, 721-725 (1998).15. M. She, D. Kim and C.P. Grigoropoulos, J. Appl. Phys. 86, 6519 (1999).16. M. Mosbacher, N. Chaoui, J. Siegel, V. Dobler, J. Solis, J. Boneberg, C.N. Afonso and P. Leiderer,

Appl. Phys. A 69[Supp.], 331 (1999).17. M. Mosbacher, V. Dobler, J. Boneberg and P. Leiderer, Appl. Phys. A 70, 669 (2000).18. J.M. Lee, K.G. Watkins and W.M. Steen, Appl. Phys. A 71, 671-674 (2000).19. Y.F. Lu, Y.W. Zheng and W.D. Song, J. Appl. Phys. 87, 1584 (2000).20. D. Bäuerle, Laser Processing and Chemistry, 2nd Ed., Springer (1996).21. J.D. Kelley and F.E. Hovis, Microelectronic Engg. 20, 159 (1993).22. Y.F. Lu, W.D. Song, M.H. Hong, B.S. Teo, T.C. Chong and T.S. Low, J. Appl. Phys. 80, 499

(1996).23. Y.F. Lu, W.D. Song, B.W. Ang, M.H. Hong, D.S.H. Chan and T.S. Low, Appl. Phys. A 65, 9

(1997).24. Y.F. Lu, W.D. Song, K.D. Ye, M.H. Hong, D.M. Liu, D.S.H. Chan and T.S. Low, Appl. Surf. Sci.

120, 317 (1997).25. Y.F. Lu, W.D. Song, Y. Zhang and T.S. Low, Proc. SPIE 3550, 7 (1998).26. Y.F. Lu, W.D. Song and T.S. Low, Mater. Chem. Phys. 54, 181 (1998).27. Y.F. Lu, W.D. Song, M.H. Hong, D.S.H. Chan and T.S. Low, Proc. SPIE 3097, 352 (1997).28. Y.F. Lu, Y. Zhang, Y.H. Wan and W.D. Song, Appl. Surf. Sci. 138-139, 140 (1999).29. Y.F. Lu, Y.W. Zheng and W.D. Song, Appl. Phys. A 68, 569 (1999).30. W.D. Song, Y.F. Lu, K.D. Ye, C.K. Tee, M.H. Hong, D.M. Liu and T.S. Low, Proc. SPIE 3184,

158 (1997).31. G. Vereecke, E. Röhr and M.M. Heyns, J. Appl. Phys. 85, 3837 (1999).32. X. Wu, E. Sacher and M. Meunier, J. Appl. Phys. 87, 3618 (2000).33. T. Fourrier, G. Schrems, T. Mühlberger, J. Heitz, N. Arnold, D. Bäuerle, M. Mosbacher,

J. Boneberg and P. Leiderer, Appl. Phys. A 72, 1-6 (2001).34. B.S. Lukyanchuk, Y.W. Zheng and Y.F. Lu, Proc. SPIE 4065, 576-587 (2000).


35. P. Leiderer, J. Boneberg, V. Dobler, M. Mosbacher, H.-J. Münzer, N. Chaoui, J. Siegel, J. Solis,C.N. Afonso, T. Fourrier, G. Schrems and D. Bäuerle, Proc. SPIE 4065, 249-259 (2000).

36. P. Leiderer, J. Boneberg, M. Mosbacher, A. Schilling and O. Yavas, Proc. SPIE 3274, 68 (1998).37. M. Mosbacher, presentation at the 5th International Conference on Laser Ablation COLA 99, July

1999, Göttingen, Germany.38. M. Mosbacher, H.-J. Münzer, J. Zimmermann, J. Solis, J. Boneberg and P. Leiderer, Appl. Phys.

A 72, 41 (2001).39. H.-J. Münzer, M. Mosbacher, M. Bertsch, J. Zimmermann, P. Leiderer and J. Boneberg, J.

Microscopy 202, 129 (2001).40. M. Mosbacher, M. Bertsch, H.-J. Münzer, V. Dobler, B.-U. Runge, D. Bäuerle, J. Boneberg and

P. Leiderer, Proc. SPIE (in print).41. M. Mosbacher, H.-J. Münzer, M. Bertsch, V. Dobler, N. Chaoui, J. Siegel, R. Oltra, D. Bäuerle,

J. Boneberg and P. Leiderer, in Particles on Surfaces 7: Detection, Adhesion and Removal, K.L.Mittal (Ed.), VSP, Utrecht, in press.

42. V. Dobler, PhD thesis, University of Konstanz, Germany (2002).43. V. Dobler, R. Oltra, J.P. Boquillon, M. Mosbacher, J. Boneberg and P. Leiderer, Appl. Phys. A 69,

335 (1999).44. B.S. Lukyanchuk, Y.W. Zheng and Y.F. Lu, Proc. SPIE 4423, 115 (2001).45. N. Arnold, G. Schrems, T. Mühlberger, M. Bertsch, M. Mosbacher, P. Leiderer and D. Bäuerle,

Proc. SPIE (in print).46. S.V. Buntsents, S.G. Dmitriev and O.G. Shagimuratov, Phys. Solid Stat., 38, 552-557 (1996).47. M. Vicanek, A. Rosch, F. Piron and G. Simon, Appl. Phys. A, 59 407-412 (1994).48. G. Mie, Ann. Physik 4, 377-445 (1908).49. P.W. Barber and S.C. Hill, Light Scattering by Particles: Computational Methods, World

Scientific Publishing, Singapore (1989).50. O. Yavas, P. Leiderer, H.K. Park, C.P. Grigoropoulos, C.C. Poon, W.P. Leung, N. Do and A.C.

Tam, Phys. Rev. Lett. 70, 1830 (1993).51. H.K. Park, C.P. Grigoropoulos, C.C. Poon, A.C. Tam, O. Yavas and P. Leiderer, Proc. SPIE 2498,

32-40 (1994).52. O. Yavas, P. Leiderer, H.K. Park, C.P. Grigoropoulos, C.C. Poon, W.P. Leung, N. Do and A.C.

Tam, Appl. Phys. A 58, 407-415 (1994).53. H.K. Park and C.P. Grigoropoulos, C.C. Poon and A.C. Tam, Appl. Phys. Lett. 68, 596-598

(1996).54. O. Yavas, A. Schilling, J. Bischof, J. Boneberg and P. Leiderer, Laser Phys. 7, 343-348 (1997).55. H.K. Park, D. Kim, C.P. Grigoropoulos, J. Appl. Phys. 80, 4072-4081 (1996).56. O. Yavas, A. Schilling, J. Bischof, J. Boneberg and P. Leiderer, Appl. Phys. A 64, 331-339 (1997).57. C.T. Avesidian, J. Phys. Chem. Ref. Data 14, 695-729 (1985).58. A.C. Tam, W.P. Leung and W. Zapka, in: Particles on Surfaces: Detection, Adhesion and

Removal, K.L. Mittal (Ed.), pp. 405-418, Marcel Dekker, New York (1995).59. E.T. Swartz and R.O. Pohl, Rev. Mod. Phys. 61, 605-668 (1989).60. G. Cerbe and H.J. Hoffmann, Einführung in die Thermodynamik, Munich (1999).61. G. Vereecke, E. Röhr and M.M. Heyns, Appl. Surf. Sci. 157, 67 (2000).62. A.A. Kolomenskii, H.A. Schuessler, V.G. Mikhalevich and A.A. Maznev, J. Appl. Phys. 84, 2404-

2410 (1998).63. J. Boneberg, Metallische Dünnfilmschmelzen von Halbleiterschichten nach ns-Laser-Annealing,

PhD thesis, University of Konstanz, Germany (1996).64. D.R. Halfpenny and D.M. Kane, J. Appl. Phys. 86, 6641 (1999).65. X. Wu, E. Sacher and M. Meunier, J. Appl. Phys. 87, 3618 (2000).


Ed. K.L. Mittal

© VSP 2003

Particle removal using resonant laser detachment

KEVIN KEARNEY and PETER HAMMOND

∗

Lightforce Technology, Inc., 125 Tech Park Drive, Rochester, NY 14623

Abstract—A new photonic cleaning process that minimizes exposure of the substrate is introduced. The concepts of the Resonant Laser Detachment (RLD) are described. The RLD process uses a laser light source with intensity modulation to remove sub-micrometer contaminant particles from a sub-strate. Unlike other laser removal methods, which eject the particle from the surface with a single high-intensity laser pulse, RLD uses a continuous series of low-intensity laser pulses. The timing and shape of these laser pulses are tuned to exploit the kinematic properties of the particle-surface system and laser-material interactions. Theoretical analysis suggests a correlation between system resonant frequency and the particle separation mechanism. This technique results in an efficient par-ticle removal mechanism that minimizes stress and heat loading to the underlying substrate.

Keywords: Particle detachment; laser cleaning; resonance.

1. INTRODUCTION

The trends in the manufacturing of Integrated Circuits (IC’s) and Flat Panel Dis-

plays (FPD’s) and in other related manufacturing industries and the requirements

for environmentally safe cleaning technologies are driving the need to investigate

advanced dry photonic based cleaning methods.

Contamination remaining on the substrate after a processing step may create

device defects, rendering final end products useless. It is recognized that sub-

micrometer particles can cause major defects that can result in defective compo-

nents and lower production yield. With the trend towards reduced feature size in

IC’s, contamination will have an even greater impact on manufacturing yield. The

need is to increase yield, thus lowering production cost and increasing manufac-

turing competitiveness.

This investigation focuses on the use of a light based cleaning technology for

the purpose of developing a commercially viable critical cleaning process.

∗To whom all correspondence should be addressed. Phone: 585-292-5610, Fax: 585-427-8422, E-mail: [email protected]


2. PARTICLE ADHESION MECHANISMS

Strong adhesion forces exist between particles and surfaces due to van der Waals,

electrostatic, and capillary attraction mechanisms. Bowling [1] has discussed

these forces in detail, while a concise summary can be found in Tam et al. [2].

van der Waals forces are comprised of London-dispersion force, dipole-dipole

and dipole-induced dipole interactions. Capillary forces arise when atmospheric

moisture condenses in the gap between a particle and a substrate, and are a func-

tion of particle radius and liquid surface tension. Coulomb electrostatic forces

originate from the electrostatic double-layer formed between the particle and the

substrate.

Tam et al. [2] note that as particle size decreases, the relative adhesion force

increases dramatically. Particle mass (m) decreases as the cube of the diameter,

while adhesion forces (F) decrease directly with diameter; hence, the acceleration

(a=F/m) required to detach a particle from a surface scales inversely with the

square of the diameter. As will be shown later, adhesion forces on a micrometer-

size particle greatly exceed gravitational forces.

As noted by Bowling [1], the contact area of a particle in contact with a surface

is quite small, resulting in tremendous pressures at the particle-substrate contact

point. For a typical 1-µm particle, force per unit area is estimated to be 10.9 N/m

2

which is enough to deform the particle and increase the particle to surface contact

area. Since the force of adhesion depends on the contact area, this effect will fur-

ther strengthen the binding of the particle to the surface.

2.1. van der Waals forces

For particles in the semiconductor-processing environment, van der Waals forces

predominate. Visser’s [3] treatment of van der Waals forces between a particle of

radius r and a flat surface at a distance z away from the surface shows that,

26 z

dhFvdW

=π

, (1)

where h is the material dependent Lifshitz-van der Waals constant.

Notice that for a given separation z, the van der Waals forces scale with the

particle diameter d, while the mass of the particle scales as (d/2)

3. Hence, the

force per unit mass (acceleration) scales with particle size as 1/d

2: smaller parti-

cles require greater acceleration than larger particles to remove them from the

substrate. From Eq. (1), we see that the van der Waals forces on a sub-micrometer

particle can greatly exceed the gravitational force on the same particle resting on

the surface of the substrate.

As an example, consider a 1 µm diameter particle of silicon resting on a silicon

substrate. The gravitational force on the particle is approximately 1.5x10

–15 N,

while the van der Waals forces at a distance of 0.4 nm from the substrate (the meas-

ured equilibrium separation distance) are 7x10

–8 N, a nearly 50x10

7 times greater.

Particle removal using resonant laser detachment 337

2.2. Particle deformation

One consequence of the strong adhesion force is particle deformation. The adhe-

sion forces distort or flatten the particle contact thus increasing the particle con-

tact area. Again following Visser [3], the force due to particle deformation is a

function of the particle contact radius r,

3

2

6 z

rhF ndeformatio

=π

(2)

Krishnan et al. [4] have shown that as a particle is allowed to rest on a substrate

the increase in contact area due to deformation begins immediately upon contact

and can increase by a factor of 50% within ten minutes.

As shown in Figure 1, for a substrate being contaminated during the manufac-

turing process it is apparent that if particles are not removed shortly after being

deposited they will become increasingly more difficult to remove and may ulti-

mately be impossible to remove. In production, as the substrate is moved through

the manufacturing process particles not removed shortly after being deposited

may cross-contaminate subsequent processes or ultimately accumulate on the sub-

strate, therefore contributing to potential product defects. Thus, it is important to

remove particles immediately after they land on the substrate.

2.3. Particle-surface potential energy

As a particle comes in close contact with the substrate, van der Waals attractive

forces begin to be offset by the quantum mechanical repulsion force associated

Figure 1. The particle contact area increases with the time the particle remains on the surface due to deformation forces, thus increasing the total particle adhesion force.


with the orbital electrons of the particle and surface molecules. The two forces are

estimated to be equal at a distance of 0.4 nm from the surface [1]. This distance,

indicated in Figure 2, represents the minimum of the potential well in which the

contaminant particle is trapped. Although the exact shape of the potential well is

difficult to calculate precisely, it will have the general functional form shown in

Figure 2. A particle bound to a surface may thus be considered an oscillator. As

discussed below, the RLD technique exploits this kinematic behavior by exciting

the resonant frequencies of the oscillator.

3. EXISTING LASER PARTICLE REMOVAL TECHNIQUES

Laser particle removal methods may be differentiated in a number of ways. The

first distinction is between ablative vs. non-ablative methods. Ablation is an ener-

getic phase transformation from a solid to a gaseous state. At high laser intensities

a thin surface layer of the substrate can be removed – carrying any surface con-

taminants away with it. As a cleaning method, this process may be considered

analogous to chemical etching, in that a thin surface layer of material is removed.

Like chemical etching, laser ablation tends to “micro-roughen” the surface. In

typical precision cleaning applications, micro-roughening is a very serious con-

cern for every cleaning method – wet or dry. In fact, any uncontrolled changes in

surface morphology, chemistry or physical properties are usually undesirable in

these applications.

Non-ablative particle removal methods use laser-generated particle-substrate

interactions to break the physical and chemical bonds holding the particle to the

Figure 2. A particle resting on a substrate surface is trapped in a particle-surface potential well. The potential energy of a particle at a given separation distance from a surface is shown.


substrate. These methods may be further differentiated by the presence or absence

of an energy transfer medium used to mediate the substrate-particle interaction.

In direct exposure methods (no energy transfer medium), a UV laser is used to

directly irradiate both the particle and an area of the surrounding substrate [5, 6].

The method acts to both photochemically break surface bonds [7, 8] and to induce

a rapid thermal expansion of the particle and/or a thin substrate layer, which

forcefully ejects the particle from the surface [5, 9].

The alternative method involving an energy transfer medium uses a several mi-

crometer thick layer of a liquid on the substrate [10-12]. This liquid layer is then

directly or indirectly rapidly heated by a laser, causing explosive evaporation that

removes the transfer medium and trapped surface particles.

4. RADIATION FORCES

4.1. Photophoretic force

The photophoretic force has two aspects: radiation force due to photon momen-

tum coupling to matter, and radiometric force arising from the interaction of a la-

ser-heated particle and an ambient (liquid or gaseous) medium. The radiation

force operates independent of gas pressure, while the radiometric force increases

from zero in vacuum to a maximum when the ambient gas molecules mean free

path (mfp) is approximately equal to the particle diameter while decreasing at

higher gas pressures (shorter mfp) [13].

For spherical, dielectric particles, the radiation force due to photon momentum

has been found to have two orthogonal components [14, 15]. The first component

Figure 3. Laser light sources utilize photothermal energy for a direct energy transfer to the particle and surface. The net effect is a rapid thermal expansion of the particle and substrate surface, dis-lodging the particle from the surface.


(axial) is directed along the laser beam axis. The second component (transverse)

arises due to the transverse light intensity gradient (e.g., a gaussian beam profile).

The combined effects of refraction and the transverse intensity gradient are such

that an unequal angular distribution of light (and hence momentum) exits at the

spherical surface of the particle, producing a net effective force towards the center

of the beam. This effect has been well demonstrated experimentally, and can be

used to form an optical “trap” for small particles [14]. In practice, the presence of

a viscous medium is required to stabilize oscillations about the center of the po-

tential well formed by the intensity profile. For typical beam diameters and light

intensities, the transverse force can be made comparable to the axial force, and

both can exceed the particle’s weight by many orders of magnitude.

The radiometric force has been exhaustively verified, and is commonly used in

aerosol studies [13]. The radiometric force is caused by uneven heating of the par-

ticle surface, and the interaction of the heated particle with the ambient back-

ground gas. Under optimum gas pressure conditions, the radiometric force can be

several orders of magnitude greater than the radiation force.

4.2. Photophoresis

Photophoresis has been investigated by Periasamy [16] as a mechanism to inhibit

particle attachment to a substrate. The motivation in their research was to prevent

particles from initial contamination. Our proposal is to use the mechanism of pho-

tophoresis to transport particles, which have been laser-detached from the sub-

strate, away from the surface and, ultimately, to a particle collection device such

as an electrostatic filter or vacuum line.

5. THE RLD TECHNIQUE

The Resonant Laser Detachment technique utilizes a low-intensity, amplitude-

modulated (or repetitively pulsed) laser light directed onto the particle and sub-

strate. Energy absorption by the particle-surface system is maximized when the

laser light (energy) is applied at a pulse rate equal or near to the natural resonant

frequency of the system. Thus applying the laser light at a pulse rate equivalent to

the resonant frequency of the system allows for lower laser fluences and greater

energy transfer to the particle-surface system.

As noted earlier, laser light interacts with the system in a number of ways.

First, the incident light imparts direct momentum pressure to the particle. Second,

it deposits heat energy into both the particle and the substrate. This heat, in turn,

generates photothermal expansion in the particle and a photothermal surface ex-

pansion in the substrate. Depending on the laser parameters, other interactions

such as bond breaking and non-equilibrium carrier (electron) energy-distributions

may also be present.

The RLD technique uses the electrostatic field from low energy fluence laser

pulses as a driving force to induce a resonant motion of the particle. The resulting


forces and heat loads are kept very small. Figure 4 demonstrates the application of

a single light source at a given incident angle to the substrate surface. Several in-

cident light sources may be required to achieve an effective particle detachment.

Defining the particle movement to be along a path orthogonal to the surface,

we can assume that the particle is bound in a potential well with an equilibrium

distance of approximately 0.4 nm above the surface. By applying a small, pertur-

bative force to the particle it is possible to cause the particle to oscillate about this

equilibrium position. For very small displacements, we can represent the potential

near the minimum as a parabola – i.e., a harmonic oscillator potential. This ap-

proximation will break down as the oscillations increase in amplitude. However,

it is sufficient to obtain an order-of-magnitude estimate of the resonant vibrational

frequency of the particle in the potential well.

For a 1 µm silicon particle resting on a silicon substrate with a contact diameter

of 0.03 µm we can estimate the resonant frequency by calculating the total adhe-

sion force (van der Waals and deformation forces, F) and applying Hooke’s law

to calculate the effective spring constant (k) for the particle-surface system:

kxF −= (3)

Again, assuming the distance x at which the particle is initially separated from the

substrate to be 0.4 nm gives a resonant frequency

MHzm

kf 94

2

1

2

1 ≈==π

ωπ

(4)

The actual frequency scales with the effective particle separation distance as

x/1 . For a separation distance of 5 nm the frequency decreases to approxi-

mately 1.5 MHz. For a 0.1 µm particle the initial resonant frequency can exceed

1 GHz. In any case, these order of magnitude frequencies are within the limits of

Figure 4. A number (n) of light sources (L) of various pulse frequencies and incident angles (Q) canbe applied to achieve the desired resonant response.


available laser modulation techniques. Figure 5 shows calculated resonant fre-

quency of 0.5 µm, 1 µm and 5 µm particles at a given surface separation distance.

The curves in Figure 5 suggest that to employ RLD will require a shift or

“chirping” of applied pulse frequency as the particle-to-surface separation dis-

tance increases. As the particle-to-surface distance is increased the modulation

frequency is reduced to match the particle-surface system resonant frequency. In

practice, the incident laser light sources (suggested in Figure 4) would be spatially

and temporally coordinated with each other and repetitively chirped on a continu-

ous basis.

The RLD technique used in conjunction with a method to prevent particle reat-

tachment (such as photophoresis, as discussed earlier) can provide an efficient

particle removal process. Gettering techniques can be applied to remove the parti-

cles from the surrounding environment.

6. CONCLUSION

A unique and novel photonic cleaning technique is proposed and investigated.

The RLD process induces motion of the particle relative to the surface by apply-

ing a pulsed light beam as a driving force that is tuned to the kinematic character-

istics of the particle-surface system. The resonant mechanism is described using a

simple harmonic oscillator model. The calculated resonant frequency of a particle

at rest on a substrate is dependent on the particle and substrate material, particle

size, and particle-to-surface separation distance. Not accounting for substrate and

Figure 5. The calculated resonant frequencies for 0.5 µm, 1 µm, 5 µm silicon particles on a silicon substrate as a function of particle-to-surface separation distance is shown. The curves show the de-pendence of resonant frequency on initial separation distance, particle size, and strongly suggest that the applied driving force must be adaptive in relation to separation distance.


particle surface roughness the initial resonant frequency can exceed 1 GHz for a

0.01 µm silicon particle resting on a silicon substrate. The resonant frequency of a

particle on a surface decreases as the particle-to-surface distance increases requir-

ing an adaptive, chirped light source(s) for complete particle detachment. Particle

removal is accomplished by collecting particles, after resonant detachment from

the surface, using electrostatic gettering and vacuum suction to prevent reattach-

ment.

REFERENCES

1. R.A. Bowling, J. Electrochem. Soc., 132, 2208 (1985). 2. A.C. Tam, W.P. Leung, W. Zapka and W. Ziemlich, J. Appl. Phys., 71, 3515 (1992). 3. J. Visser, Particulate Science Technol., 13, 172 (1995). 4. S. Krishnan, A.A. Busnaina, D.S. Rimai and D.P. DeMejo, J. Adhesion Sci. Technol, 8, 1357

(1994). 5. T.J. Magee and C.S. Leung, in: Particles on Surfaces 3: Detection, Adhesion, and Removal,

K.L. Mittal (Ed.), pp. 307-316, Plenum, New York (1991). 6. T.J. Magee, J.F. Osborne, P. Gildea and C.S. Leung, U.S. Patent 4758533 (1988). 7. A.C. Engelsberg, Mater Res Soc. Symp. Proc, 315, 255 (1993). 8. A.C. Engelsberg, Proceedings Microcontamination ‘93 Conference (1993). 9. J.D. Kelley and F.E. Hovis, Microelectronic Eng, 20, 159-170 (1993).

10. S.D. Allen, S.J. Lee and K. Imen, Optics and Photonics News, 3, No. 6, 28-30 (1992). 11. S.J. Lee, K. Imen and S.D. Allen, Microelectronic Eng, 20, 145-157 (1993). 12. W. Zapka, W. Ziemlich, W.P. Leung and A.C. Tam, Microelectronic Eng, 20, 171-183 (1993). 13. O. Preining, in: Aerosol Science, C.N. Davies (Ed.), pp. 111-135, Academic Press, London

(1966). 14. A. Ashkin and J.M. Dziedzic, Appl. Phys. Lett., 19, 283 (1971). 15. A. Ashkin, Science, 210, 1081-1087 (1980). 16. R. Periasamy, U.S. Patent 5472550 (1995).


Ed. K.L. Mittal

© VSP 2003

The future of industrial cleaning and related public

policy-making

CAROLE LEBLANC∗

Toxics Use Reduction Institute, University of Massachusetts Lowell, One University Avenue, Lowell,

MA 01854-2866

Abstract—In this paper, the author presents some of her findings in the pursuit of safer and greener chemical solvents for hard-surface cleaning, as well as some of the new directions that the science of cleaning may take in the next five to ten years. Specifically, innovative methods of research and development into cleaning alternatives are explained, including molecular modeling, data mining, and the use of ionic liquids. A discussion on chemical risk assessment ensues, in light of the scien-tific concepts of hormesis and endocrine disruption. This is followed by a comparative analysis of the European approach to policy-making, known as the Precautionary Principle and recent events pertaining to cleaning issues in the U.S. Finally, conclusions are drawn based on a hypothetical case of ‘over-cleaning’.

Keywords: Data mining; designer molecules; endocrine disruption; hormesis; ionic liquids; molecu-lar modeling; precautionary principle.

1. INTRODUCTION

“Every chemical is potentially a pharmaceutical.” A. Warhurst [1] If the physical properties of liquids were any indication, successful solvent re-placement with aqueous cleaners would never be possible. The components and the behavior of the water molecule are nothing like those of a typical organic sol-vent used for cleaning, as evidenced by the high polarity and dipole moment of water. To illustrate, some important physical properties of the chlorinated solvent trichloroethylene (TCE) and water are compared in Table 1.

Chemists often try to come as close as possible to the physical characteristics of the original solvent in formulating new cleaners for the same application. This is entirely understandable. Devising test protocols in which the solvent behaves in a certain known fashion and expecting a water-based cleaner to function in the same manner, however, is neither very realistic nor even a fair analysis of its potential performance.

∗Phone: 978-934-3249, Fax: 978-934-3050, E-mail: [email protected]

C. LeBlanc 346

Table 1.

Physical properties of a chlorinated solvent (TCE) and water

Properties TCE Water

Chemical Formula C2HCl3 H2O

Molecular Weight 131.39 18.02

Boiling Point 87oC 100oC

Density 1.46 g/cm3 0.99 g/cm3

For this reason, the author described the significance of choosing the right

piece of mechanical equipment in process conversions involving aqueous clean-ing in her thesis, “The Search for Safer and Greener Chemical Solvents in Sur-

face Cleaning: A Proposed Tool to Support Environmental Decision-Making”.a

Nevertheless, the author is familiar with at least two situations in which an over-dependence on chemical properties led investigators to a much narrower field of replacement candidates than the computer program, The Aqueous Way to Go,b the tool developed during her doctoral research, would have recommended. In both cases, an organic solvent was replaced with yet another organic solvent.

As a result, only incremental improvements, if any, were made to the health and safety of workers and to the protection of the environment upon implement-ing the alternative cleaners. The replacement cleaners shared a variety of traits with the original solvents and the inherent dangers in using any organic and/or chlorinated compound remained the same.

2. RESEARCH METHODS AND DEVELOPMENTS

2.1. Molecular modeling

The advent of ‘designer molecules’ has led to the development of products with-out the drudgery of comparing and matching chemicals’ physical properties at every step on the bench. The term designer molecule is used by various chemical disciplines in much the same way as the term designer gene is applied in the field of genetic engineering. Designer molecules allow scientists to visualize molecular structures and how they behave under certain conditions as well as in the presence of other molecules before they are actually synthesized. Chemical modeling soft-

aCompleted for Erasmus University’s (Rotterdam, the Netherlands) international program in

cleaner production, cleaner products, industrial ecology and sustainability, in conjunction with the Toxics Use Reduction Institute at the University of Massachusetts Lowell, leading to a Ph.D. in Sus-tainable Development and Management (www.eur.nl/fsw/gsem/phd), the first awarded to an Ameri-can woman.

bA web-based, interactive matrix, i.e., a tool designed to enhance decision making in solvent sub-stitution (www.angelfire.com/band2/greencleaners/doctoralthesis.html).

The future of industrial cleaning and related public policy-making 347

ware reveals important molecular functions, just as engineering modeling pro-grams reveal stress and metal fatigue patterns in, for example, aircraft parts.

RasMol and CHIME are two popular computer programs for the visualization of molecules. RasMol was written by Roger Sayle of Glaxo-Wellcome and CHIME, a web browser plug-in based on RasMol, is a product of MDL Informa-tion Systems (www.mdli.com). Instructions for downloading RasMol can be found at www.umass.edu/microbio/rasmol and instructions for downloading CHIME are located at www.mdli.com. Originally intended for biological systems such as pro-teins and nucleic acids, both programs could be used for less complicated systems such as cleaning chemicals in the development of safer and greener alternatives.

An important aspect of these programs is the researcher’s ability to manipulate structures. Chemical models can be displayed as traditional stick figures, ball-and-stick figures or space-filled structures. They can be controlled by three-dimensional rotation, size alteration and color coding. Different parts of a model (for example, the asymptotic or active site of an enzyme) can be selected and treated separately. Coupled with the information obtained from the Surfactant Virtual Library at www.surfactants.net, this ability could be very useful in creat-ing new surface-active agents or, for that matter, new composite materials. Scien-tists could conceivably formulate chemicals designed to disassociate into benign forms of their components after performing certain tasks, like cleaning. Figure 1 presents a three-dimensional model, capable of rotation, of the simplest chlorin-ated solvent. For comparison, Figures 2 and 3 are traditional one-dimensional, line representations of more complex surfactant formulations and an aqueous metal-cleaner in action, respectively.

Much work has already been done in molecular modeling and ‘virtual com-pounds’ can be ordered from web-based suppliers listed at www.umass.edu

/microbio/rasmol/whereget.htm or ‘synthesized’ via molecular mechanics calcula-tions with a computational chemistry package such as Chem3D. Once a model is displayed in RasMol, it can be saved in other documents as well as printed. CHIME’s program allows for the dissemination of ‘live’ molecular models on the World Wide Web.

Combined with other sources of data, RasMol and CHIME are powerful mechanisms for global communication among scientists and should be helpful for improving the understanding of chemical information among all stakeholders. Computer modeling of chemical structures also advances the cause of nanotech-nology, the study and control of matter at the atomic or molecular level. The ma-nipulation of substances at the nano-level to the precise site and at the exact mo-ment they are needed should decrease the amounts of chemicals required to achieve a certain response, thereby decreasing the generation of wastes and the likelihood of over-exposure of humans or the environment to toxic substances. Nanotechnology may make possible the bio-inspired design of enzymatic or pro-tein-based cleaners more cost effective.

C. LeBlanc 348

Figure 1. Ball-and-stick rendition of carbon tetrachloride (CCl4).

Figure 2. Anionic (A) and nonionic (B) surfactants.

Figure 3. Saponification of a fatty oil with a strong alkali.

2.2. Data mining of cleaning performance criteria

One of the first attempts to generalize chemical behavior for solvency was the Hansen method [2]. In this method, a battery of chemical reactions is conducted and monitored in test tubes. The results are ranked visually and recorded numeri-cally. Based on the Hansen methodology, DuPont scientists developed a proprie-tary computer program for the selection of semi-aqueous cleaners in the 1980s. Its


application was limited to the company’s Axarel® line of products. The Aqueous

Way to Go further merges the function of a computer program with Hansen-like, actual performance criteria.

While application-specific testing is still required, the results of pertinent clean-ing tests from the Toxics Use Reduction Institute’s Surface Solutions Laboratory (SSL) are stored in The Aqueous Way to Go program. Additional performance in-formation from other databases is also inserted into the program and serves to (1) further decrease the time required to identify greener chemical cleaners and (2) further increase the proficiency of the final selection. Like the molecular model-ing used to accelerate chemical formulating described in the previous section, a mechanism is needed, preferably computer-based for speed and accuracy, to (1) sort through a plethora of data that may, or may not, be relevant and (2) determine what chemical interactions, if any, reveal important trends for cleaning. Table 2 contains some of the newer tools available to conduct this kind of research.

Recently, algorithmic programming has been applied to advance the cause of solvent substitution. In March of 2000, three simulation programs with different algorithms were reviewed for designing greener solvents by Cabezas, Harten and Green [3]. The three simulations were: (1) the U.S. EPA’s Program for Assisting the Replacement of Industrial Solvents (Paris II), (2) the Technical University of Denmark’s software, Computer Added Molecular Design (CAMD) and (3) Mo-lecular Knowledge Systems’ chemical design software, Synapse. The Paris II al-gorithm (www.tds.cc) uses chemicals from the Design Institute for Physical Prop-erty Research (DIPPR) database and “looks for potential replacement solvents whose properties are as close to the required parameters as possible”. The CAMD solvent-design algorithm (www.capec.kt.dtu.dk) operates in a five-stage process using valence (i.e., molecular charge) rules. The Synapse algorithm (www.molknow.com) “generates candidate chemical structures, which are then screened as potential solvent replacements in a four-step methodology”.

Unlike these programs that focus on theoretical scenarios with data that are primarily intended for the scientific community, The Aqueous Way to Go concen-trates on actual performance data of existing cleaners for the end-user community, in addition to applications development. Data mining, or knowledge discovery in databases, offers the best approach for manipulating this kind of information to arrive at meaningful insights from observed tendencies (for example, the per-formance of certain surfactants) in would-be relational databases.

Nevertheless, it would still be possible to use any, or a combination of, the re-maining computer tools described in Table 2 for research and development into greener cleaners. The web site http://surfactants.net/huibers/greenchem.html lists a number of computer programs developed for property prediction, solvent re-placement studies, and reaction design as well as additional solvent substitution resources on the World Wide Web, in particular, the U.S. EPA’s Envio$en$e’s (http://es.epa.gov) links to data systems.

C. LeBlanc 350

Table 2.

Examples of math-based/computer-enhanced research tools

Research method or principle

Description and uses

Algorithmic Programming

Algorithmic, or procedural, languages are designed for solving a particular type of problem. They are called high-level languages because they are largely inde-pendent of hardware. Unlike machine or symbolic languages, they vary little be-tween computers. The first such language was FORTRAN (FORmula TRANs-lation), developed for scientific calculation followed by the first commercial language, COBOL (Common Business Oriented Language). ALGOL (ALGO-rithmic Language), is used primarily in mathematics and science. The latest generation of languages is an outgrowth of artificial intelligence.

Chaos Theory

Also known as nonlinear dynamics, chaos theory is an interdisciplinary science that attempts to reveal structure in seemingly unpredictable dynamic systems. In a linear system, a small change produces a small and easily quantifiable system-atic change, but a nonlinear system exhibits a sensitive dependence on initial conditions: small or virtually immeasurable differences in initial conditions can lead to wildly differing results. (This is sometimes called the butterfly effect, in reference to a 1979 address by meteorologist E.M. Lorenz entitled, “Predictabil-ity: Does the Flap of a Butterfly’s Wings in Brazil Set Off a Tornado in Texas?)”. Uses include the study of diverse phenomena, such as dripping fau-cets and population growth.

Fuzzy Logic

Whereas, classical logic holds that everything can be expressed in binary terms: 0 or 1, black or white, yes or no; fuzzy logic allows for values between 0 and 1, shades of gray, and maybe it also allows partial membership in a set. When used with an expert system, logical inferences can be drawn from imprecise relation-ships. Uses include automatic optimization of household appliances by sensors, automobile subsystems and smart weapons.

Visualization Software

Similar to geographic information systems (GIS) or mapping, visualization software displays sets of interconnected data, often in animation-like format. Uses include aerospace obstacle detection and landscape evaluation.

Data Mining

Data mining (or knowledge discovery in databases, KDD), is a new research area developing methods and systems for extracting interesting and useful in-formation from large sets of data. Uses include commercial/financial databases, telecommunication alarm sequences and epidemiological research.

2.3. Ionic liquids as solvents

Recent advances in ionic liquids show promise in improving the environmental soundness of surface cleaning. Ionic liquids are salts that exist in liquid form at ambient temperature. Like all salts, they possess a positive and a negative charge. Ionic liquids do not occur naturally and must be manufactured. While not much information has been published about them yet, Song and Roh reported the use of a room temperature ionic liquid for the immobilization, recovery and recycling of a chiral catalyst [4]. The ionic liquid used was 1-butyl-3-methylimidazolium hexafluorophosphate depicted in Figure 4.


Figure 4. Structure of ionic liquid, 1-butyl-3-methylimidazolium hexafluorophosphate.

Figure 5. The combination of an ionic liquid and supercritical CO2 to separate an organic com-pound from solution.

Unlike water-soluble compounds that can be extracted with water, or the re-moval of chemicals with high vapor pressures by distillation, ionic liquids require very high temperatures to effect separation of compounds. This post-separation of chemical products from ionic liquids may be difficult to achieve since the heat needed may cause the products to degrade. Furthermore, the energy needed to drive these reactions may be too expensive. If these problems can be solved, ionic liquids may become safer, greener solvents since they do not possess any measur-

N

C. LeBlanc 352

able vapor pressure and so, unlike chlorinated/organic solvents, do not evaporate to be inhaled by workers or to be emitted into the atmosphere and cause air pollu-tion. (The dermatological consequences of exposure to ionic liquids as well as their impact on water pollution are currently unknown.)

To address this separation issue, Brennecke and Beckman performed experi-ments using a combination of an ionic liquid and supercritical carbon dioxide at room temperature [5]. Their experiment, first reported in 1999, is diagramed in Figure 5.

Both the carbon dioxide and the ionic liquid are recoverable for reuse. The same system used for the separation of naphthalene could theoretically be used for the removal of organic surface contaminants. While liquid-liquid (as opposed to liquid-CO2) extractions would still be possible, they would invariably return the system to the use of organic solvents, depending on the coefficient of partition, or to the use of water, which would be almost entirely ineffectual for the separation of most hydrocarbons. The use of various polymers, surfactants or solubilizers may enhance the extraction/cleaning process.

To date, no toxicological or environmental fate studies have been published on ionic liquids. This is urgently needed before much more additional application work is done.

3. DISCUSSION

3.1. Risk assessment and policy making

The preceding sections dealt with the future of industrial cleaning, in terms of chemical and scientific innovation. The subsequent sections are devoted to the underpinnings of public policies that either foster or impede these advances. No other topic is as germane to the issue of chemical discovery, manufacture and use as risk assessment. And no other aspect of risk assessment has been as overlooked as hormesis.

3.1.1. The case for hormesis

Hormesis may be defined as the phenomenon observed in science that the effects of chemical exposure produced at high doses are the inverse or apparent inverse of those produced at low doses in a population [6]. The study of hormesis dates back to the German physician Paracelsus (1493-1541) and father of toxicology who coined the phrase “the dose determines the poison” [7].

It is estimated that approximately 350 studies contain evidence of hormesis. These studies involve a number of different species (fungi, protozoa, bacteria, plants and animals), cover a wide range of chemical types (alcohol and its me-tabolites, hydrocarbons, metals and pesticides) and exhibit varying effects (altera-tions in growth rates, reproduction, longevity and cancer). The hormetic effect of hydrocarbons on plant growth, where growth stimulation occurred at low doses and inhibitory effects at high doses is illustrated in Figure 6.


Figure 6. Hormetic effect of organic solvents on oat seedling growth [8].

Figure 7. Identification of hormetic zone of zinc affecting cell reproduction. Source: H. Rubin, Proc. Natl. Acad. Sci. (USA), 72, 1676 (1975).

Currently, chemical risk assessments are primarily conducted by studying high-level exposures and extrapolating to predict safe levels. Inclusion of hormesis in risk assessments would reveal hormetic zones where the chemical/biological re-sponses may be significant. An example is given in Figure 7. Nowhere is this phenomenon more important than in the study of cancer. Approximately twenty toxicological studies have been conducted whereby hormesis occurred, followed

C. LeBlanc 354

by the onset of cancer. All three stages of the disease – initiation, promotion and proliferation – have been linked to hormetic behavior [8].

Whether or not a chemical is said to exhibit a dose-dependent beneficial or deleterious reaction depends upon the conditions defined at the time of the expo-

sure. For example, many chemicals used in the treatment of Acquired Immune Deficiency Syndrome (AIDS) are considered toxic under almost every other non-diseased circumstance; AIDS patients themselves need to be monitored closely for toxicity levels during treatment. Problems arise when conditions are not de-fined prior to a chemical’s release into the general environment, turning the bio-sphere, if not the patient, into a laboratory. This is descriptive of the use of most of mankind’s synthesized chemicals, including the detergents and solvents used for cleaning.

The point of this discussion on hormetic behavior is that exposure may be more harmful at lower, as opposed to higher, concentrations for the same chemical, tox-icity notwithstanding. In fact, hormesis contains the root word hormones, which are very powerful, biologically-active compounds that function effectively at low concentrations. This refutes the principle learned by most chemists trained before 1990 that “dilution is the solution to the problem” and demonstrates the impor-tance of identifying potential chemical hazards before they enter the biosphere, to avoid the difficulty of separating minute amounts of powerful toxins (for exam-ple, dioxin) from various waste streams.

3.1.2. Surfactants and endocrine disruption

Surfactants are surface-active chemicals that are very important to the cleaning process. Their concentrations in aqueous cleaners are deceptively low (< 10%), given that they are the power horses of the cleaner’s formulation. It should, there-fore, come as no surprise that some of these surface-active agents may exhibit the kinds of effects described above at very low concentrations. The proven health hazards associated with organo-chlorinated cleaning solvents were described by the author [9] while the suspected health hazards involving some surfactants in some aqueous/semi-aqueous cleaners, acting as endocrine disruptors, were only briefly mentioned. More investigative work needs to be done.

The endocrine (or hormonal) system is made of glands throughout the body that synthesize and secrete hormones into the bloodstream and various receptor sites in target tissues that recognize and respond to hormones, especially the sex organs. The endocrine system controls a complex interplay between the sex hor-mones of the oestrogens and androgens, and other hormones, such as those of the thyroid system. The immune and nervous systems are also affected by hormonal regulation. In general, hormonal signaling is more long-lived than neural trans-mission. It is precisely because of these systems’ complexities that it is extremely difficult to accurately predict the behavior of a single chemical compound or its metabolites on the body’s organs.


Figure 8. The structure of oestradiol.

Oestrogens such as oestradiol, pictured in Figure 8, influence the development and maintenance of female sex characteristics, and the maturation and function of the sex organs. Chemicals that can imitate an oestrogen are known as oestrogenic chemicals. Androgens such as testosterone serve a similar purpose in males.

Chemicals can disrupt the endocrine system in several ways, with the degree of disruption being influenced by timing, especially with regard to the stages of the recipient organ’s development and the age of the organism. The main mecha-nisms include (1) binding or activating the oestrogen receptor or other receptors, (2) modifying the production or metabolism of natural hormones and (3) modify-ing the number of hormone receptors. Besides endocrine disrupting, other terms used to describe this chemical behavior include: xenoestrogenic, oestrogenic (es-trogenic) and hormone mimicking.

Synthetic substances implicated as oestrogenic include the alklyphenols (and their derivatives) used in industrial detergents for wool washing and metal finish-ing, various laboratory detergents, including Triton X-100 and some liquid laun-dry detergents. The alkylphenols, nonylphenol and octylphenol are mainly used to make alkylphenol ethoxylate (APE) surfactants.

Alkylphenols were first thought to be oestrogenic in the 1930s [10] and more evidence of such effects was published in the 1970s [11, 12]. However, it was not until 1991 that publication of the effects of nonylphenol on cultured human breast cells led to human health concerns [13]. As reported by Warhurst [1], research has shown that alkylphenols increase the growth of these cells 1000 to 10000 times greater than the natural oestradiol levels required to produce the same growth. Oestrogenic effects have also been shown on rainbow trout hepatocytes, chicken embryo fibroblasts and a mouse oestrogen receptor [14, 15]. Oestrogenic effects are present at tissue concentrations of 0.1 µM for octylphenol and 1 µM for non-ylphenol [16]. A screen for recombinant yeast, using the human oestrogen recep-tor, has shown similar results [17]. Recent research has shown oestrogenic effects of nonylphenol at still lower concentrations and levels of 0.05 mg/L were suffi-cient to increase the number of eggs produced by minnows, as well as an increase in vitellogenin levels (this research also suggested that nonylphenol may lead to an increase in natural oestrogen levels) [18].

C. LeBlanc 356

Alkylphenol ethoxylate surfactants are not effectively degraded in sewage treatment plants or in the environment, tending instead to lose some of their eth-oxylate groups and to also bio-accumulate up the food chain, as does dioxin (the resultant alkylphenols, alkylphenols with one or two ethoxylate groups and alkyl-phenoxy carboxylic acids, APEC, persist even longer). Alkylphenols accumulate where there is inadequate oxygen (for example, in sediments) and APEC persist in rivers and effluents (for example, in sewage).

Human exposure to these chemicals can occur by (1) absorption through skin from shampoos, cosmetics, spermicidal lubricants and domestic and industrial de-tergents, (2) contaminated drinking water, (3) inhalation and ingestion from pesti-cide sprays and (4) contamination of food from fields treated with sewage sludge. Nonylphenol has been detected in human umbilical cords at concentrations up to 2 ppt, which may or may not be correlated to the predisposition of the infant’s sex as a consequence of exposure.

3.2. Status of related public policy

The U.S. EPA formed the Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC) to develop recommendations for a screening program, which were finalized in August 1998. As a result, an “Endocrine Disruptor Screening Program”, was designed with a focus “on providing methods and pro-cedures to detect and characterize the endocrine activity of pesticides, commercial chemicals and environmental contaminants”. By the agency’s own admission, however, “there currently is not enough scientific data available on most of the estimated 87,000 chemicals in commerce to allow us to evaluate all potential risks”, with the exception of some pesticides [19]. A number of papers from the research initiative of the National Science and Technology Council’s (NSTC) Committee on the Environment and Natural Resources (CENR) can be found at www.epa.gov/endocrine/pubs.html.

3.2.1. The precautionary principle: the European model The precautionary principle may be defined as the approach whereby lack of full scientific certainly is not used as a reason for postponing pollution prevention measures to prevent environmental degradation. It was first endorsed in 1987 by European environmental leaders concerned with toxic discharges into the North Sea. They reasoned that releases of chemicals should be reduced/eliminated if they were suspected to be harmful, even before there was clear scientific proof, hence the term precautionary. In a 1992 report, the John Snow Institute, Center for Environmental Health Studies, reported that a number of factors contributed to this scientific uncertainty [20]. These factors are listed in Table 3.

That same year, the United Nations Conference on Environment and Develop-ment (UNCED) adopted Principle 15, which states that “where there are threats of serious irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degrada-tion”. A version of this principle was also incorporated into the Cartagena Proto-


col on Biosafety under the Convention of Biological Diversity. At the Interna-tional Conference on Biotechnology in the Global Economy held at Harvard Uni-versity in September 2000, a discussion was facilitated by the University’s Center for International Development (CID) that “supported efforts to better understand the insti-tutions of precaution through which governments move from science to policy… high-lighting the institutional differences among OECD (Organization for Economic Coop-eration and Development) countries, sub-Saharan countries and international institutions”. The precautionary principle is important to industrial cleaning since its implementation in Europe has led to a ban of some surfactants while the U.S. contin-ues to allow these chemicals in cleaners; many scientists believe that the safer, albeit more expensive, alcohol ethoxylate (Figure 2) is as effective and readily available as the suspect alkylphenol-ethoxylated surfactants. The computer program or tool, The Aqueous Way to Go can be used to ‘screen’ nonylphenol ethoxylate from potential solvent substitutes in much the same way. An overview of current policies covering chemical usage throughout the world, in particular suspect endocrine disrupters, is pre-sented in Table 4.

The author suggests that the proactive stance of the precautionary principle, rather than a variety of reactive policies, should form the basis of technical inno-vation paired to chemical regulation/trade. This is especially true in areas such as the production of genetically-modified organisms (GMOs) and the development of solvent alternatives, where the risks are so high for so many. More information on hormone disrupting chemicals and chemicals policy can be found at Tulane University’s web site, www.tmc.tulane.edu/ecme/eehome in the report, “Environ-mental Estrogens and Other Hormones”.

3.2.2. The effects of a recent policy change: the United States

In a different, but related matter, a recent change in U.S. regulations has led to a loosening of the use of an important solvent chemical, the de-listing of acetone as a volatile organic compound (VOC). This may lead to an increase in the use of acetone and other hydrocarbons as cleaning agents, even though the quantity and quality of safer and greener alternatives continues to rise.

Table 3.

Reasons for scientific uncertainty [20]

I. The complexity of dose and exposure relationships

II. The unknown cumulative effects of exposure

III. The unknown effects of combined exposures to multiple chemicals

IV. The vast number of chemicals about which we have little or no health effects information

V. Individual differences among humans in their receptivity and propensity for diseases

VI. Limitations of scientific knowledge

VII. Delays between exposure and occurrence of disease

C. LeBlanc 358

Table 4.

Overview of global policies affecting suspect endocrine disrupting chemicals [1]

European Union (EU)

The European Commission (EC) published its strategy on endocrine disruption in Dec. 1999. Originally expected to include a list of 20-30 sus-pected endocrine disrupters, the list was postponed to April 2000. In March 1999 the EC’s Scientific Committee on Toxicity, Ecotoxicity and the Envi-ronment published the report, “Opinion on Human and Wildlife Health Ef-fects of Endocrine Disrupting Chemicals, with Emphasis on Wildlife and on Ecotoxicology Test Methods”. The EU also published a communication on the precautionary principle in February 2000. Several endocrine disrupt-ers are under review as part of Existing Substances process. The EU’s chemicals policy in under review as well, having been accepted that it is not currently effective enough.

United Kingdom (UK)

The UK Government published its new chemicals strategy in December 1999. The Environment Agency of England and Wales is currently review-ing its policy towards endocrine disrupters.

United Nations (UN) and NGOs

The OECD has a programme on endocrine disrupters, mainly focusing on the development of testing procedures. The UN is currently negotiating a global treaty covering certain persistent organic pollutants (POPs), includ-ing PCBs, dioxin and DDT, with criteria for adding new chemicals. A simi-lar agreement, the POPs Protocol, has already been negotiated among the UN Economic Commission for Europe. The International POPs Elimina-tion Network is a non-governmental organization (NGO) coalition against POPs.

United States (US)

No signs yet of any new controls on existing chemicals, even on the alkyl-phenols, which are already being phased out in Europe. The US National Academy of Sciences published the report, “Hormonally Active Agents in the Environment” in July 1999.

Chemical Industry

Most relevant industry associations have issued statements about hormone disrupting chemicals relaying their concerns, but calling for more research before any action is taken. Industry claims that effects are not likely to be as significant as those of phytoestrogens. Some companies have stopped us-ing suspect chemicals while others will continue to use them unless they are banned, considering endocrine disruption to be a hypothesis, rather than scientific fact.

In fact, consultants to the U.S. space agency have recommended the use of bu-tane (lighter fluid) in some part-cleaning operations since this policy change. It would appear that as older scientists retire and/or are replaced by younger, inex-perienced researchers/contractors, there is a lack of a common understanding of the past lessons learned from the use of these solvents. This may cause American society to repeat some of the same mistakes made earlier. In other words, the U.S. public may be facing a retreat to increased exposure to hydrocarbon products, and their associated health hazards, used for cleaning prior to the discovery of the de-struction of the ozone layer by chlorofluorocarbons (CFCs). These developments are especially troubling in light of the United States’ active opposition to the Kyoto Protocol to decrease global-warming (i.e., carbon-based) emissions. The


re-introduction of brominated cleaners, notably n-propyl bromide (nPB), is like-wise a concern. UNEP’s STOC considers nPB to be ozone depleting and is not recommending it as a solvent substitute since “non-ozone-depleting solutions ex-ist for all cleaning applications for which nPB is being promoted”.

Aberrations in legal structures, especially liability issues, are no doubt at the root cause of how some societies approach environmental decision-making for cleaning applications. Consumerism, (i.e., the educated consumer) and organized labor (i.e., trade unions representing various segments of the workforce perform-ing cleaning duties) also have roles to play. Various chemical formulators have become more adept in addressing worker safety and the environment due to these concerns. Partly because of these advances, the lines separating parts, precision and institutional (i.e., maintenance and janitorial) cleaning have blurred and are il-lustrated in Figure 9. As workplaces approach the safety of households in clean-ing operations, overlaps among cleaning standards and performance guidelines may become more commonplace. Ironically, these same developments may also tend to increase multiple chemical sensitivities to certain, at-risk, individuals within a given population.

4. A CONCLUDING SCENARIO: PROVOCATIVE POSSIBILITIES INVOLVING

SEMICONDUCTORS

In no industry is the efficacy of cleaning/rinsing cycles more essential than in the semiconductor industry. Cleanrooms, maintained at various levels of cleanliness under U.S. Standard 209E and now ISO standards according to the number and size of airborne particulates, generally require cleanliness levels many times greater than surgical fields. This is because a semiconductor, a silicon wafer with diodes and transistors, must act as a circuit at near atomic levels. Contamination is the primary cause of product failure. Moreover, the increased storage capacity of miniaturized computer chips has caused cleanliness requirements to increase ex-

Figure 9. Illustration of the three principal cleaning fields.

C. LeBlanc 360

ponentially. Table 5 contains the water quality guidelines for chip manufacture and blood dialysis as a means to compare each system’s level of desired contami-nation control.

The goal of this treatise is not to expose the semiconductor industry’s over-reliance on energy and water resources; other researchers such as Ted Smith, founder and director of the Silicon Valley Toxics Coalition, are far more familiar with the industry and have documented this dependency. The purpose here is to reflect on the unsustainable nature of the current technology (indeed, most com-puters are considered obsolete within eighteen months of manufacture), more spe-cifically, the unhealthy, unnatural conditions to which workers are exposed in cleanrooms.

Table 5.

Water quality guidelines for semiconductors (A) and hemodialysis (B)

Source: P. Cartwright, Proc. of the Precision Cleaning meeting held in Rosemont, IL, May 1995. *Measurement system chosen by industry.

A

Contaminant Maximum concentration*

Suspended solids: residue / particulates

0.1 ppm / 500 counts per liter

Dissolved solids: organic (TOC)

0.020 ppm

Ionic: resistivity / dissolved silica (SiO2)

18.3 megohm-cm / 3 ppb

Aluminum (Al) cations 0.2 ppb

Ammonium (NH4) cations 0.3 ppb

Chromium (Cr) cations 0.02 ppb

Copper (Cu) cations 0.002 ppb

Iron (Fe) cations 0.02 ppb

Manganese (Mn) cations 0.05 ppb

Potassium (K) cations 0.1 ppb

Sodium (Na) cations 0.05 ppb

Zinc (Zn) cations 0.02 ppb

Bromide (Br) anions 0.1 ppb

Chloride (Cl) anions 0.05 ppb

Nitrite (NO2) anions 0.05 ppb

Nitrate (NO3) anions 0.1 ppb

Phosphate (PO4) anions 0.2 ppb

Sulfate (SO4) anions 0.05 ppb

Bacteria 0 counts per 100 ml

B

Contaminant

mg/l*

Aluminum (Al) 0.01

Arsenic (Ar) 0.005

Barium (Ba) 0.1

Cadmium (Cd) 0.001

Calcium (Ca) 2.0

Choramines 0.1

Chlorine (Cl) 0.5

Chromium (Cr) 0.014

Copper (Cu) 0.1

Fluoride (F) 0.2

Lead (Pb) 0.005

Magnesium (Mg) 4.0

Mercury (Hg) 0.0002

Nitrate (AS N) 2.0

Potassium (K) 8.0

Selenium (Se) 0.09

Silver (Ag) 0.005

Sodium (Na) 70.0

Sulfate (SO4) 100.0

Zinc (Zn) 0.1


4.1. Hospitals, cleanrooms and cleaners: could there be a connection?

“The globalization of infectious diseases is not a new phenomenon. However, in-creased population movements, whether through tourism or migration or as a re-sult of disasters; growth in international trade in food and biological products; so-cial and environmental changes linked with urbanization, deforestation and alterations in climate; and changes in methods of food processing, distribution and consumer habits have reaffirmed that infectious disease events in one country

are potentially a concern for the entire world <italics added>”. So begins the World Health Organization’s (WHO) report by the Secretariat on global health security – epidemic alert and response (November 2000).

Staphylococcus Aureusc has been implicated in hospital-acquired infections

since the 1950s when 50% of the organism’s strains developed resistance to peni-cillin. It has been considered a serious bacterial pathogenic threat since that time. Known as a ‘super bug’, the organism has also become resistant to newer and more powerful antibiotics such as tertracycline and the aminoglycosides. It is common, even in the cleanest healthcare facilities, with the elderly, the seriously ill and those patients with compromised immune systems being at greatest risk [21].

The MRSA (methicillin-resistant Staphylococcus Aureus) super bug responds only to the antibiotic vancomycin, whose use is now restricted due, at least in part, to its apparent role in producing the ‘super bug’ VRE (vancomycin-resistant Enterococci

d), following its application in European cattle, for which there is no

known treatment. ‘Flesh-eating’ disease or Necrotizing Fasciitis is another antibi-otic-resistant bacterial infection of the Streptococcus Type A variety associated with surgical or wound patients. This variant super bug is more powerful than other strains, with stronger m-protein serotypes.

Benign forms of Staphylococcus Aureus are natural habitants of skin and mu-cus membranes of humans and can be found throughout the environment from dust to door knobs. An infectious disease expert at the American Society for Mi-crobiology in Miami, Florida reported that even personnel who do not come into

direct contact with patients can accumulate and spread bacteria, including resis-

tant strains. Thus, the WHO’s concern is justified for the pandemic spread of these infections to the general population, including apparently young and other-

wise healthy subjects, as illustrated in the ‘mad cow’ disease or Bovine Spongi-form Encephalopathy (BSE), now in its fatal human form called new variant Creutzfeldt-Jakob Disease (nvCJD). Inadvertent cannibalism (that is, cow mate-rial being added to cattle feed without the ranchers’ knowledge), not an overuse

cStaphylococci are gram positive bacteria that are typically arranged in clumps or grape-like clus-

ters. They can be distinguished from streptococci in that only the staphylococi are catalase-positive (catalase is an enzyme that liberates oxygen from hydrogen peroxide).

dEnterococci are gram-positive bacteria that are widely distributed in nature and are part of the normal flora of the gastrointestinal and genital tracts.

C. LeBlanc 362

of antibiotics, appears to be the root cause of these infectious mutations. The dis-ease involves prions. Prions are tiny biological bits that may or may not be alive and so are impossible to ‘kill’ (in the conventional sense of the term) to prevent the infection’s spread. Some of these prions can be viewed three-dimensionally with RosMol/Chime at www.mad-cow.org/prion_structure_folder/viewers.html.

4.2. The ‘cleaner’ connection

Recently, Americans have been introduced to a line of household cleaners, adver-tised as antibacterial, for applications where no antibacterial activity is warranted. The more popular the cleaners become, the more product varieties appear on gro-cers’ shelves. Studies at Tufts University’s School of Medicine (Boston, Massa-chusetts) revealed that the antibacterial agent triclosan, used in many of these products, acts like an antibiotic to promote bacterial resistance and, potentially, the spread of untreatable infections. Furthermore, the U.S. Food and Drug Ad-ministration (FDA) reports that antibacterial soaps kill the beneficial bacteria that live on skin. Unlike their pathogenic counterparts, these bacteria apparently strengthen the immune systems of children [22].

These seemingly unconnected events or trends have one or two things in com-mon: they are related to the things we chose to clean well or not to be able to keep clean enough. Meanwhile, microscopic forms of life have been found in Arctic-

like conditions and other species such as the archaea have evolved near volcanic

emissions under the sea, both environs thought to be uninhabitable by the scien-

tific community not too long ago. Somehow, life finds a way. In hospitals, sterility is maintained (most notably for surgery), cleaning is per-

formed and dressings (gowns, masks, gloves) are donned to protect the person from pathogens. In cleanrooms, sterility is maintained, cleaning is performed and dressings donned to protect the product from the person. Cleanrooms, whose ster-ile environment routinely outrivals the surgeon’s needs, may offer the next best habitat for a super bug, perhaps of prion-nature, to establish a foothold. The tech-nical staff of these high-tech establishments may already have damaged immune systems due to the unnatural conditions in which they work on a daily basis (there is no such thing as a ‘good’ bacterium in a cleanroom). Regardless, the spread of a hypothetical cleanroom-acquired infection may not require an at-risk host, as has been previously discussed.

4.3. Closing statement

Several aspects of the search for surface cleanliness are neither simple nor straightforward. The Alliance for the Prudent Use of Antibiotics located in Bos-ton, Massachusetts, an international organization with members from more that ninety countries, has been monitoring the worldwide emergence of treatment-resistant microbial strains since 1981. This group, and others in the scientific community such as the U.S. Center for Disease Control (CDC), should be made


aware of the developments in the cleaning industry so that other studies, like those conducted at Tufts medical school, can be undertaken.

In the meantime, unless and until the chemical industry provides complete chemical disclosure on a global basis, institutions such as the Toxics Use Reduc-tion Institute’s research facility at the University of Massachusetts Lowell should assist in formulating green chemical cleaners, in addition to providing education and training programs and state-of-the-art laboratory testing of existing products. This could be accomplished through partnerships with commercial enterprises and/or other research facilities. Remaining pertinent issues, some having nothing to do with cleaning performance, could then be addressed. These include studies on the chemical additives of fragrances (over 80% of the odorants now used are synthetic in origin) and dyes or colorants (often added for worker safety in prod-uct recognition). Tighter quality control on percentages of ingredients could also be maintained, since currently the concentrations of a cleaner’s components re-ported on its Material Safety Data Sheet can vary by as much as 400%. Most im-portantly, chronic, hormetic and synergistic chemical-exposure tests need to be developed and implemented before cleaners are marketed.

The development of the computer program, The Aqueous Way to Go revealed many of these trends and potential hazards in cleaning processes and chemicals. The successful technical diffusion of this tool will require an educational compo-nent, the topic of an upcoming paper.

REFERENCES

1. A. Warhurst and the Friends of the Earth, London. http://website.lineone.net/~mwarhurst/policy.html (May 19, 2000).

2. C.M. Hansen, Hansen Solubility Parameters: A User’s Handbook, CRC Press, Boca Raton, FL (1999).

3. H. Cabezas, P. Harten and M. Green, Chemical Engineering, 107 (3), 109 (2000). 4. “Epoxidation Catalyst Immobilized In Ionic Liquid”, Chemical & Engineering News, 78, 21

(May 22, 2000). 5. J. Brennecke and E. Beckman, Nature, 399, 28-29 (May 6, 1999). 6. “Chemical Hormesis: Beneficial Effects at Low Exposures/Adverse Effects at High Exposures”,

Brochure, Texas Institute for the Advancement of Chemical Technology, College Station, TX. 7. Paracelsus, “Seven Defensionmes: The Reply to Certain Calumniations of His Enemies,” trans.

by C. Temkin in Four Treatises of Theophrastus von Hohenheim called Paracelsus, John Hop-kin Press, Baltimore, MD (1994).

8. E. Calabrese and L. Baldwin, Chemical Hormesis: Scientific Foundations, Documentation and Implications for Risk Assessment, School of Public Health, University of Massachusetts Am-herst (June 1997).

9. C. LeBlanc, “The Search for Safer and Greener Chemical Solvents in Surface Cleaning: A Pro-posed Tool to Support Environmental Decision-Making”, Ph.D. Thesis, Erasmus University, Rotterdam (2001).

10. E. Dodds and W. Lawson, Proc. Royal Soc. Lon. B. 125, 222-232 (1938). 11. R. Kociba, Toxicol. Appl. Pharmacol., 46, 279 (1978). 12. G. Mueller and U. Kim, Endocrinol., 102, 1429-1435 (1978). 13. A. Soto, H. Justicia, J. Wray and C. Sonnenschein, Environ. Health Persp., 92, 167-173 (1991).

C. LeBlanc 364

14. S. Jobling and J. Sumpter, Aquat. Toxicol., 27, 361-372 (1993). 15. R. White, S. Jobling, S. Hoare, J. Sumpter and M. Parker, Endocrinol., 135, 175-182 (1994). 16. A. Soto and C. Sonnenschein, Environ. Health Persp., 103, 113-122 (1995). 17. E. Routledge and J. Sumpter, Environ. Toxicol. & Chem., 15, 241-248 (1996). 18. “Industry Glimpses New Challenges as Endocrine Science Advances”, ENDS Report 290, 26-

30 (1999). Contact TURI’s library at telephone 978-934-3390 for information to obtain a copy of this report.

19. U.S. EPA Endocrine Home Page, www.epa.gov/scipoly/oscpendo (Jan. 15, 2001). 20. John Snow Institute, “Toxics Use in Massachusetts: Implications for Public Health”, Center for

Environmental Health Studies, Boston, MA (Jan. 1993). 21. “Staphylococcus Aureus: An Emerging Super Bug” (Feb. 12, 2001).

http://srd.yahoo.com/goo/S.+Aureus+AND+super+bug/2/*http://members.tripod.com/LouCaru/ index-9.html

22. Tufts Health Maintenance Organization, “WELL!” Magazine, Waltham, MA (Winter 2000).

surface contamination and cleaning.pdf

Documents

cleaning volume

cleaning system

influence of cleaning

modeling of waterjet

silicon wafer cleaning

future of industrial

chemistry process cleaning

blank surface contamination