integration of modeling and simulation into process ... · means to integration of modeling and...

P Ventzek December 2003MOTOROLA and the Stylized M Logo are registered in the US Patent & Trademark Office. All other product or service names are the property of their respective owners. © Motorola, Inc. 2002.

Integration of Modeling and Simulation into Process Development

P.L.G. VentzekS. RaufP.J. Stout

Northern California AVS Section Joint • Plasma Etch Users Group • Thin Films Users Group

Meeting

AGENDA

• Perspective on how modeling and simulation fit as a value propositions in the semiconductor industry.

• Computational process integration– Process integration tool-kit at Motorola– Plasma Chemistry– Surfaces

• FEOL Example• BEOL Example

Simulation and the Manufacturing Life-Cycle

• New Materials– New devices and new materials. Understanding material behavior

dependencies on unit processes. – “Materials science calculations”

• Concept and Feasibility– Tool design and the mating of tools to next technology nodes– “Equipment simulation”

• New Product Development– Unit process development– Process integration – “Linked equipment and feature scale models”

• Manufacturing– Yield enhancement, troubleshooting– “Calibrated models”

• Re-use

Turn-key Models

SpecializedModels

Barriers to Integration of Modeling and Simulation into Development Culture

• Credibility and timeliness issues– Late results, replication vs. truly predictive

• Not quantitative enough• Undersized development groups in industry and university

– Downsize of traditional research sources– Small nature of many end users of M&S

• Communication gap between interested parties: chip makers, vendors

• Communication gap between funding agencies that would fund research , researchers and “customers”.

Means to Integration of Modeling and Simulation into Development Culture

• Focus on predictive work (as much as possible)• Strive to be quantitative – no corner cutting from

fundamental physics data to tool details.• Critical mass group or extended group• As much as possible, engage vendors in joint projects• Take an international view regarding research partners

• Aim where need is greatest: integration.– Point process studies while insight providing have limited value

add potential.– The ensemble of unit processes or process integration is

our view of where impact is greatest.

Process Integration

Unit Process ModelA

Feature Scale ModelA

Unit Process Model B

Feature Scale ModelB

Unit Process ModelC

Feature Scale ModelC

Unit Process Model D

Feature Scale ModelD

Yield Enhancement andSuperior Reliability, Electrical Performance

Materials Science Models

Concept Flow/Device

Sequences of unit process models threaded together can replicate what is observed on Silicon.

Computational Integration Tool Kit

• Models for most commercial tools (AMAT, Lam, TEL) developed in-house.

• Plasma chemistry database assembled through internal computational environment.

• Plasma-surface interaction database developed through internal and external efforts.

• In-house simulations for both chamber and feature scale simulations.

Computational Infrastructure

• Following codes have been utilized for the studies reported here:– IO:1 2-d fluid model for capacitive plasma simulation,– HPEM:2 2-d fluid model for inductive plasma simulation,– Papaya:1 3-d Monte Carlo based feature scale model.

1Motorola; 2University of Illinois

IO HPEM

Papaya

Reactor GeometryOperating Conditions

Plasma Chemistry

Feature Profile

Surface Chemistry

Species FluxesEnergy/Angular Distributions

Reactor GeometryOperating Conditions

Initial Profile

Electron Impact Processes In SF6 Plasmas

SF5 SF5+

SF4+SF4

SF3 SF3+

SF2 SF2+

SF SF+

5

F

SF5 SF4

SF3 SF2 SF S

SF6

F-

SF5-

SF6-

SF5+

SF3+

1

2 3

4

Needed to be calculated!

Processes (4) were calculated by us in 2000 to be bale to optimize critical deep Sietch processes used for e-beam lithography.

C4F6 Plasma Chemistry in Detail

C4F6 + e

+ eNeutral dissociation

+ 2e

[C4F6-] Attachment

Ionization C3F3+, C4F6

+, CF+, C3F4+,

C3F2+, C4F5

+, CF3+, C4F4

+,C2F2

+, C3F+, CF2+

C4F6

C3F4

C4F5

C2F3

CF2

C2F2 C3F3CF

C4F4

C3F2

C4F3

C2F

C, F, C2

Not only electron impact processes for the parent but also the plethora of radicals present in the system must be considered.

Neutral Dissociation Processes in C5F8 plasmas

c-C5F8

C5F7

Direct dissociation

C5F8 Isomerization

C5F8

Ringopening,

FCF2 C2F3

CF C2F2C2F4+ C4F6

Previously calculated data

3

4

5 1

2

Like C4F8 which is cyclic, the cross-sections depend on the bond that opens and where the electron hits.

e.g., Co reduced energy level scheme“Effective”

Relevant Metastable Short-LivedLevels Levels Levels

Ta >300 4 4Fe >480 6 4Co >320 3 4Ni >200 3 4Sn >100 4 3Mg >80 1 3B >40 1 3

Metallization: Complex Excited State Plasma Chemistry

By Product: Data for Plasma Diagnostics

929.06

931.38

942.62

905.48

938.73

0

1

2

3

4

5

6

7

1100 1000 900 800 700

938.03

Wavenumber (cm-1)

Inte

nsity

Synthetic, RuO3

Synthetic, RuO4

Synthetic, RuO

Synthetic, RuO2

Experimental, suspect RuO4

Experimental, unknown feature

FTIR spectra/cross section study for RuOx Etch:

Plasma-Surface Interactions

• We are using molecular dynamics following the approach of Graves and Hamaguchi to develop plasma-surface interaction datasets.

Crystal Passivated with CF2

+ IonsCrystal with Polymer

Layer

Red – Si, blue – oxygen, light gray – F, dark gray - C

Plasma-Surface Interactions: Thermal Processes• Very complex problem!• We seek relative sticking coefficients and reactions and try to cover a large

parameter space using quantum chemistry calculations and small clusters. The final arbiter on these parameters is ultimately comparison of feature scale predictions with experiment.

SiO2

CFx

Direct barrierlessreactions

SiO2

CFx

Transition state

( )θγ −= 1s

Plasma-Surface Interactions: Yields

• An example of macroscopic parameters that are inserted into our models. This constitutes a part of an ever-evolving array of species-surface parameters.

Si atoms removed per incident CF2+ ion

0 10 20 30 40 50 60 70 80 90θ , degrees

Si/C

F 2

E = 50 eVE = 150 eVE = 300 eV

Plasma-Surface Interactions: Kinetics• Example angular and energy distribution functions for SiF associated with

polymer bombardment. Both are needed for feature scale simulations. Understanding the energetics of ejected species (even apart from their nature) is important for the thermal and chemical properties of the plasma.

0

0.01

0.02

0.03

0 10 20 30 40 50 60 70 80 90θ, degrees

0

0.05

0.1

0.15

0.2

0.25

0.3

0 2 4 6 8 10 12 14 16 18 2Ek, eV

Dry Etch Achievements- Transistors

• At the heart of a CMOS transistor’s performance is the gate patterning process. Product requirements demand innovative patterning techniques which produce gate lengths far smaller than the theoretical limits of lithography wavelengths.

Supernominal, nominal, and subnominal Poly-Si lines patterned with conventional 193nm lithography, and etched down to <20nm Lpoly. Gate dielectric is <13Å oxynitride.

Supernominal, nominal, and subnominal Poly-Si lines patterned with conventional 193nm lithography, and etched down to <20nm Lpoly. Gate dielectric is <13Å oxynitride.

Supernominal

Nominal

Subnominal

Top-down

Advent of novel structures and new materials brings huge challenges

FEOL Computational Integration

• Calculated (w/ Papaya) evolution of an SRAM gate after 7 etch steps.• Shown here is an over aggressive trim blows out the gates. These would

be dealt with through lithographic adjustments or etch engineering.

trim oeARCinitial

PRARC poly Si oxide

Applications to Process Development for Novel Devices

An example of fin process step gone awry!

Computational Process Integration (FEOL)MASK

LITHOGRAPHY-ST-LITH

TRIM ETCH-PAPAYA

Ma s

k-to

-Etc

h ed

F eat

ure

P re d

i ct io

n s

Ma s

k M

anu f

a ct u

re a

nd D

e si g

n0 nm

500 nm

Initial Height500 nm 200 nm

time

Example of Trim Etch for Different Initial Resist Heights

Dry Etch Achievements- Interconnect

• Advanced technologies bring a new host of challenges in interconnect processing: more metallization layers using novel materials, dual-inlaid metallization integrations for cost savings, as well as further shrinking of the feature dimensions.

Motorola’s HiPerMOS7 SOI product features nine levels of metal utilizing low-K dielectrics, implementing dual-inlaid processing at every layer.

Motorola’s HiPerMOS7 SOI product features nine levels of metal utilizing low-K dielectrics, implementing dual-inlaid processing at every layer.

Etch-Metallization Integration: Text-book view

Dual-inlaid structure with straight side-walls and well defined sidewall angles. Fill is as-expected.

Scaling for idealized cases is well defined as a function of via side-wall angle and aspect ratio.

BUT!

No resputter

sputter

BEOL Computational Integration

Less polymerizing chemistry vs More polymerizing chemistry

bow

Larger open area

Smalleropen area

BEOL Computational Integration

PhotoresistBARC

cap

dielectricetch stop

metalvia mask BARC etch via etch

trench mask Slug etch

trench etch ash

• See Next Slide

First Stab at BT Process from Vendor

• Slow etch that attacks PRdimension.

• Lost sidewall (CD) cannot be recovered in subsequent steps!

• 7 degrees of freedom.

Wide CD-190 nm before even breaking through.

Photoresist (PR)

BARC

TEOS (BT or Breakthrough)

Photoresist at 224 nm

SiCOH (Main Etch)

Subsequently ashed away

CuCap

Two Tools - Two Processes

5.0

0.0

Hei

ght (

cm)

Radius (cm)0.0 21.8

[CF2] (Max = 2.1 × 1011 cm-3)5.0

0.0

Hei

ght (

cm)

Radius (cm)0.0 21.8

[CF2] (Max = 2.1 × 1011 cm-3)

TEOS Etch in Tool A

21.70.00.0

12.0

Hei

ght (

cm)

Radius (cm)

[CF3] (Max = 9.2×1011 cm-3)

21.70.00.0

12.0

Hei

ght (

cm)

Radius (cm)

[CF3] (Max = 9.2×1011 cm-3)

TEOS Etch in Tool B

PR

BARC

TEOS

“Polymer”

O2 mix and pressure varied

Gas Mix, Power and Gap Varied

Suggested path

Minimum oxygenReplicates Tool B process best.

O2 in Process Critical to CD Control

TEOS etch rate slows PR eroded

Experimental Verification

Slow

Poor CDsSlowDemonstrates

O2 need

Needs Ar and more polymerOptimal

Summary

• Fundamentals based simulations can and do have an engineering impact in unit process and process integration development.

• The physics, chemistry and materials interactions problems remain complex but are tractable enough to warrant tackling theminside the semiconductor manufacturing industry.

• Developmental activities remain critical to making the leap to “predictive” simulation and fulfilling the promise of simulation adding value.