mid-infrared laser absorption spectroscopy for...

MID-INFRARED LASER ABSORPTION SPECTROSCOPY

FOR CARBON OXIDES IN HARSH ENVIRONMENTS

A DISSERTATION

SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

R. Mitchell Spearrin

September 2014

c© Copyright by R. Mitchell Spearrin 2014

All Rights Reserved

ii

I certify that I have read this dissertation and that, in my opinion, it is fully adequate

in scope and quality as a dissertation for the degree of Doctor of Philosophy.

(Professor Ronald K. Hanson) Principal Adviser



(Dr. Jay B. Jeffries)



(Professor Mark A. Cappelli)

Approved for the University Committee on Graduate Studies

iii

Abstract

Advancements in measurement science are presented regarding in situ laser-based detection

of CO and CO2 in harsh combustion environments. Mid-infrared absorption sensing strategies,

utilizing transitions in the 2.7 µm and 4.3 µm vibrational bands for CO2 and the 4.8 µm vibrational

band for CO, were developed to enable sensitive measurements of temperature and carbon oxide

concentrations in high-temperature gases. These new strategies (1) extend the utility of carbon

oxide absorption sensing for hostile aeroengine applications and (2) offer significant improvements

to previous methods for shock tube kinetics studies. The recent maturation of mid-infrared diode and

quantum cascade lasers, combined with parallel progress in mid-infrared fiber optics, provides the

platform from which the field-deployable sensors were designed. State-of-the art signal processing

strategies, including calibration-free wavelength modulation spectroscopy, were implemented to

tackle the thermo-mechanically harsh environments of a pulse detonation combustor and direct-

connect scramjet. Time-resolved and spatially-resolved measurements of temperature and carbon

oxide species concentrations were demonstrated to provide an in situ metric to evaluate combustion

completion for engine development. For shock tube kinetics studies, a multi-band CO2 sensing

strategy was developed to provide ppm-level species detection and highly-sensitive measurements

of gas temperature with microsecond temporal resolution.

v

Acknowledgments

I dedicate this dissertation to my late friend and undergraduate roommate Joseph Hanzich. He

truly elevated my way of thinking and inspired me to maximize my personal potential. He was a

major catalyst in my decision to pursue a PhD and I continue to find motivation in memory of his

life.

I would further like to recognize a number of individuals who made my PhD possible. First, I am

indebted to my advisor, Professor Ron Hanson, for bringing me to Stanford and providing me ample

opportunities to learn and grow through research. He has made me a better experimentalist and

communicator, and I have truly been inspired by his passion for excellence. I would like to thank Dr.

Jay Jeffries for sharing his expertise and for guiding many of the critical technical decisions made

throughout the course of my PhD. He played a major role in my progression as a scientist, engineer,

and project manager. I would also like to thank the many staff and students (current and former) who

have provided me with day-to-day help in solving the numerous challenges which I encountered

during my research. Chris Goldenstein, Wei Ren, Rito Sur, Vic Miller, Matt Campbell, Ian Schultz,

and Marcel Nations amongst others have made contributions ranging from troubleshooting laser

problems over lunch to finishing an off-site measurement campaign on my behalf at the birth of my

first child. I am fortunate to have been surrounded daily by so many intelligent and gracious peers.

Additionally, I’d like to thank the collaborators from other universities who enabled the exciting

applications of my research. These include Professor Chris Brophy and Dave Dausen of the Naval

Postgraduate School, and Professor Chris Goyne, Bob Rockwell, and Brian Rice of the University

of Virginia. I must thank Professors Mark Capelli, Chris Edwards, and John Weyant of Stanford

for taking the time to serve on my dissertation reading and/or oral committee. I also would like to

recognize the primary sponsors of my PhD research: AFOSR and Dr. Chiping Li, NASA and Dr.

Rick Gaffney, and ISSI and Dr. John Hoke.

vi

Moreover, I would be remissed not to recognize the many folks outside of Stanford who have en-

abled me to pursue my doctoral degree. I would like to thank my parents, Raymond and Candience,

for their unconditional support and confidence in me since the day I left home as the first person in

my family to attend college. More importantly, I’d like to thank them for the principles they instilled

in me during the many years before that day which have helped me succeed. My father taught me

the value of honest, hard work and gave me a portion of his instinct for all things mechanical. My

mother created a home of love and security, giving me the confidence to take risks and reach for

lofty goals. I would like to thank the many other friends and family members who have provided

encouragement and helped me keep perspective on life. Most importantly, I am forever grateful

to my wife, Karen, who has supported my dreams and personal ambitions unwaveringly since we

were married just prior to our time at Stanford. During our 4 years here, she has been the primary

financial provider for our family while also dedicating more than her share of time both early morn-

ing and night towards raising our now 2 year-old daughter Rachel. I am deeply appreciative of her

patience during my many hours of work and her confidence in me to excel both professionally and

personally.

vii

Contents

Abstract v

Acknowledgments vi

1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 CO2 TDL sensing for harsh engine environments 32.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.2 Absorption spectroscopy theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.3 Sensor design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3.1 Wavelength selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3.2 Laser modulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3.3 Optical engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4 Sensor development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.4.1 Measurements of key spectroscopic parameters . . . . . . . . . . . . . . . 16

2.4.2 Modulation-induced wavelength shift . . . . . . . . . . . . . . . . . . . . 19

2.4.3 High gas density lineshape modeling . . . . . . . . . . . . . . . . . . . . 21

2.5 High-temperature shock tube validation . . . . . . . . . . . . . . . . . . . . . . . 22

2.6 CO2 measurements behind detonation waves . . . . . . . . . . . . . . . . . . . . . 23

2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3 CO sensing for detonation engines 263.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.2 Wavelength selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27


3.3.1 Modulation depth optimization . . . . . . . . . . . . . . . . . . . . . . . . 29

viii

3.3.2 Modulation-induced wavelength shift . . . . . . . . . . . . . . . . . . . . 32

3.3.3 Spectral modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.3.4 Optical engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.4 CO measurements behind detonation waves . . . . . . . . . . . . . . . . . . . . . 39

3.5 Multi-species measurements in a PDE . . . . . . . . . . . . . . . . . . . . . . . . 40

3.6 Summary of detonation engine research . . . . . . . . . . . . . . . . . . . . . . . 42

4 Ultra-sensitive CO2 diagnostic for kinetic studies 43

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.2 Spectroscopic framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.3 Line selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46


4.4.1 Light source selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.4.2 Spectral modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.4.3 Optical setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.5 Sensitive CO2 detection near 4.2 micron . . . . . . . . . . . . . . . . . . . . . . . 56

4.6 Cross-band CO2 thermometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

4.7 Kinetics Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4.7.1 Potential sensor improvements . . . . . . . . . . . . . . . . . . . . . . . . 60

5 Spatially-resolved CO and CO2 sensing for scramjets 62

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.2.1 Line Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.2.2 Laser absorption spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . 64

5.2.3 Optimization for non-uniform flows . . . . . . . . . . . . . . . . . . . . . 66

5.3 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5.3.1 Optical hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5.3.2 Facility interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

5.4.1 Time-resolved measurements . . . . . . . . . . . . . . . . . . . . . . . . 75

5.4.2 Spatially-resolved measurements . . . . . . . . . . . . . . . . . . . . . . . 76

5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

ix

6 Conclusion 80

6.1 Aeropropulsion research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

6.2 Shock tube kinetics research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

A Mid-infrared optics: practical issues 82

A.1 Laser output non-linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

A.2 Wavelength stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

A.3 Susceptibility to back reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

A.4 Beam spatial mode quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

A.5 Fiber-coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

A.6 Fiber transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

A.7 Multimodal dispersion in optical fibers . . . . . . . . . . . . . . . . . . . . . . . . 90

A.8 Mid-infrared optical materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

Bibliography 93

List of Tables

2.1 Collisional-broadening parameters for the R(26) CO2 line (E” = 273 cm−1 in com-

bustion exhaust gases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.1 Spectroscopic line assignments and collisional-broadening parameters for the CO

lines of interest. Uncertainties for broadening measurements made in this work (i.e.

CO–N2) are shown in parentheses. . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.1 Spectroscopic line assignments and modeling parameters for the CO2 lines of inter-

est. Parameters taken from HITEMP 2010 except where measured (γAr, nAr). . . . 54

5.1 Spectroscopic parameters of CO and CO2 transitions used in the scramjet sensor. . 64

A.1 Mid-infrared optical material properties . . . . . . . . . . . . . . . . . . . . . . . 92

x

List of Figures

2.1 Absorption line-strengths of CO2 and H2O at 2000 K based on the HITEMP

database. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Absorption line-strengths of ν1+ν3 CO2 combination bands at 1000 K (HITEMP). 8

2.3 CO2 and H2O absorbance spectra simulations (air bath gas) near the peak of the

R-branch in the ν1+ν3 band at an expected PDC temperature (1800 K), path-length

(L = 4 cm) and composition for (top) P = 1 atm and (bottom) P = 5 atm. . . . . . . 9

2.4 Pure CO2 absorbance spectra near 3733.48 cm−1 at T = 1000 K and P = 40 torr,

measured in a static cell and compared to a HITEMP 2010 simulation. . . . . . . . 10

2.5 Measured laser parameters (i0 and i2) as a function of modulation depth at f = 60

kHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.6 Simulated WMS-2f (background subtracted) at 3733.48 cm−1 as a function of mod-

ulation depth (a) for various pressures at 2000 K. . . . . . . . . . . . . . . . . . . 13

2.7 Simulated CO2 WMS spectra (2f, 1f, 2f /1f ) near 3733.48 cm−1 at an expected PDC

condition (analogous to figure 2.3). . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.8 Simplified light delivery and detection schematic. . . . . . . . . . . . . . . . . . . 15

2.9 Sensitivity analysis of the WMS-2f /1f mole fraction measurement to the line

strength and collisional broadening of the strongest CO2 transitions near 3733.48

cm−1 at an expected PDC condition (see figure 2.7). . . . . . . . . . . . . . . . . 17

2.10 Measured line-strength for the R(26), E” = 273 cm−1, and R(13), E” = 738 cm−1

lines from 600 to 1000 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.11 (top) Measured water-broadening coefficient for the R(26) line from 450 to 950 K,

including a best-fit power law equation and reference bounds using ±10% nH2O.

(bottom) Measured self-broadening coefficients for the R(26) line from 750 to 1000

K, including a best-fit power law equation and reference bounds using ±12% nself . 20

2.12 Measured CO2 in air WMS-2f /1f spectra overlaid with corresponding simulations

near 3733.48 cm−1 at T = 298 K, L = 23 cm, XCO2 = 0.015, P = 2-8 atm. . . . . . 21

xi

2.13 Example time-history of measured pressure and measured WMS-2f /1f for a shock

tube validation experiment. Time 1: initial condition. Time 2: between incident and

reflected shocks. Time 5: after reflected shock. Measurement 1 cm from end wall. . 22

2.14 Sensor validation data showing the measured CO2 mole fraction compared to the

known CO2 mole fraction over a range of conditions produced in a shock tube (σ ∼3.5%); L = 5 cm, XCO2 = 6− 9%. . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.15 Representative data from a pulse (20 Hz) detonation combustor operating on ethy-

lene/air; L = 3.55 cm; optical absorption ∼ 1 − 15%; pressure and WMS-2f /1f

measurements are coplanar in combustion chamber. . . . . . . . . . . . . . . . . . 24

3.1 Absorption line-strengths of CO, CO2, and H2O at 2000 K (HITEMP). . . . . . . . 28

3.2 Absorbance spectra simulations of equilibrium concentrations of CO, CO2, and

H2O (air bath gas) near 4.855 µm at expected PDC conditions (C2H4-air, φ ≈1); L

= 4 cm, (top) P = 10 atm, T = 2500 K and (bottom) P = 30 atm, T = 3000 K. . . . 30

3.3 Measured modulation depth at maximum injection-current amplitude (±120 mA)

as a function of laser modulation frequency. . . . . . . . . . . . . . . . . . . . . . 31

3.4 Measured laser intensity parameters (i0 and i2) as a function of modulation depth (f

= 50 kHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.5 Simulated WMS-2f (background subtracted) at 2059.91 cm−1 as a function of mod-

ulation depth for various pressures at 2000 K. . . . . . . . . . . . . . . . . . . . . 33

3.6 Simulated CO WMS spectra (2f, 1f, 2f /1f ) near 2059.9cm−1 at a typical PDC con-

dition; T = 2000 K, P = 20 atm, L = 4 cm, XCO = 1%. . . . . . . . . . . . . . . . 34

3.7 Measured absorbance spectra of CO in N2 near 2060 cm−1 with active P-branch

lines labeled P(v”,J”) and fit with the Voigt function; T = 804 K, P = 1 atm, L =

20.95 cm, XCO = 0.005. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.8 Sensitivity analysis of the WMS-2f /1f mole fraction measurement near 2059.91

cm−1 to the line-strength and collisional broadening parameters of the P(0,20) and

P(1,14) CO lines at a typical PDC condition (see figure 3.7). . . . . . . . . . . . . 36

3.9 Measured nitrogen-broadening coefficient for the P(0,20) line from 1100 to 2600 K

including a best-fit power law equation and comparison with broadening equations

from other sources; P = 10–50 atm. . . . . . . . . . . . . . . . . . . . . . . . . . . 37

xii

3.10 Sensor validation data showing the measured CO mole fraction compared to the

known mole fraction over a range of conditions produced in a shock tube (σ ≈ 3%);

L = 5 cm, XCO = 0.00495 in N2. . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.11 Simplified mid-infrared light delivery and detection schematic. . . . . . . . . . . . 38

3.12 Representative time-history data from a pulsed (20 Hz) detonation combustor op-

erating on ethylene/air; L = 7.62 cm; pressure and WMS-2f/2f measurements are

coplanar in combustion chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.13 Optical configuration for free-space fiber-coupling (left) and cross-section of pulse

detonation combustor showing remote light delivery and collection (right). . . . . . 40

3.14 Time-resolved species mole fraction data from a pulse detonation combustor oper-

ating on ethylene-air at 20 Hz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4.1 Absorption line-strengths of CO2 at 1000 K (HITEMP) . . . . . . . . . . . . . . . 44

4.2 Absorbance simulations for the CO2 ν3 band at T = 300 K and T = 1000 K; P = 1

atm, XCO2 = 400 ppm, L = 15 cm . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.3 Measured and simulated spectra in the far wings of the (top) P-branch and (bottom)

R-branch of the CO2 fundamental ν3 band; T = 700 K, P = 2 atm, L = 20.95 cm,

XCO2 = 1.5% in air; Simulated interference from ambient CO2 (400 ppm) at L = 1

m, T = 300 K also shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.4 Estimated detectability and temperature sensitivity of CO2 using fixed-wavelength

DA at line-center for candidate ν3 R-branch lines; T = 1400 K, P = 1.5 atm, L = 14

cm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.5 Line-strengths vs. temperature for transitions utilized in two-line CO2 thermometry

(past and present) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.6 Line-strength ratio and temperature sensitivity for cross-band line-pair compared to

previous intra-band line-pair (Farooq et al. 2008) . . . . . . . . . . . . . . . . . . 52

4.7 Scanned-wavelength direct absorption measurement of the R(76) CO2 transition

using a Daylight Solutions ECQCL (100 Hz); T = 550 K, P = 0.81 atm, L = 20.95

cm, XCO2 = 0.015 in air . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.8 Peak absorbance time-history for the R(76) line near 2390.52 cm−1 during a non-

reactive shock tube experiment of CO2 dilute in argon; XCO2 = 0.005, L = 14.1 cm,

T5 = 1540 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.9 Argon-broadening coefficients for the R(76) and R(96) CO2 lines. . . . . . . . . . 55

xiii

4.10 Optical setup for shock tube experiments. . . . . . . . . . . . . . . . . . . . . . . 56

4.11 Measured CO2 species time-histories during methyl butyrate oxidation (φ ≈ 1). . . 57

4.12 Pressure and CO2 species time-histories during methyl-butyrate pyrolysis at 1259

K with comparison to the kinetic model of Huynh et al. . . . . . . . . . . . . . . . 58

4.13 Example peak absorbance time-histories of the R(96) and R(28) lines in a non-

reactive shock tube experiment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4.14 Measured temperature based on cross-band thermometry for a non-reactive CO2-

Ar shock (same as Fig. 4.13) with comparison to calculations from the ideal shock

relations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.15 Comparison of cross-band thermometry with known temperature in the shock tube. 61

5.1 Absorption line-strengths for CO and CO2 from 4–5 µm; T = 2000 K; transitions

labeled as branch(v”,J”). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5.2 Line-strengths vs. temperature for the selected CO (blue) and CO2 (red) transitions. 67

5.3 Optical configuration for free-space fiber-coupling (left) and cross-section of scram-

jet combustor showing remote light delivery and collection (right). Flow direction

is out of the page. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5.4 UVaSCF supersonic combustor schematic; example optical lines of sight shown in

red. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.5 Measured carbon monoxide absorption from a single laser scan (6 kHz) shown as

(top) raw voltage signals versus time and (bottom) absorbance versus wavenumber;

T = 1725 K, P = 0.71 bar, L = 3.81 cm, XCO = 0.057; φ ≈ 0.15, measurement taken

2.18 cm downstream of cavity leading edge, 1 mm from cavity wall. . . . . . . . . 71

5.6 Measured carbon dioxide absorption from a single laser scan (100 Hz) shown as (a)

raw voltage signals versus time and (b) absorbance versus wavenumber; T = 1725

K, P = 0.71 bar, L = 3.81 cm, XCO2 = 0.062; φ ≈ 0.15, measurement taken 2.18

cm downstream of cavity leading edge, 1 mm from cavity wall (same as figure 5.5). 72

5.7 Measured carbon monoxide wavelength modulation scan (50 kHz, 200 Hz) shown

as (top) raw voltage signal versus time and (bottom) normalized second harmonic

signal versus wavenumber; T = 1690 K, P = 0.71 bar, L = 3.81 cm, XCO = 0.063;

φ ≈0.15, measurement taken 2.18 cm downstream of cavity leading edge, 1 mm

from cavity wall. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

xiv

5.8 Representative time-resolved CO absorbance areas and temperature at a fixed x-y

position in the combustor cavity; ethylene-air, φ ≈0.15; measurement taken 2.18

cm downstream of cavity leading edge, 1 mm from cavity wall. . . . . . . . . . . . 75

5.9 Representative time-resolved CO and CO2 mole fraction data at a fixed x-y posi-

tion in the combustor cavity; ethylene-air, φ ≈0.15; measurement taken 2.18 cm

downstream of cavity leading edge, 1 mm from cavity wall. . . . . . . . . . . . . . 76

5.10 Comparison of DA and WMS measurements of temperature and CO mole fraction

across the cavity plane at φ = 0.15. . . . . . . . . . . . . . . . . . . . . . . . . . . 77

5.11 Carbon oxide mole fraction measurements across the cavity plane for ethylene-air

combustion at (top) φ ≈0.15 and (bottom) φ ≈0.21. XH2O also shown. . . . . . . 78

A.1 Example non-linear output from a DFB-QCL with no absorption. . . . . . . . . . . 83

A.2 Example raw absorption scan fit with an emperical baseline after shifting to account

for thermal emission and scaling to account for variable window transmission at

elevated temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

A.3 Pyrocam images of the output beams from an external cavity QCL propogating the

TEM00 mode (left) and a DFB-QCL propogating the TEM01 mode. . . . . . . . . 86

A.4 Pyrocam images of the output beam from a DFB-QCL before (left) and after (right)

spatial filtering with an iris, which preceded fiber coupling. . . . . . . . . . . . . . 88

A.5 Attenuation versus wavelength for various mid-infrared fiber materials . . . . . . . 89

A.6 Beam intensity profiles at the output of a multimode fiber before and after mechan-

ical vibration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

A.7 Temperature dependent transmission of sapphire at a thickness of 1 cm, using the

empirical model by Thomas et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

xv

Chapter 1

Introduction

This dissertation describes recent advancements in mid-infrared laser absorption diagnostics for

harsh, high-temperature combustion systems. The design, development, and deployment of novel

carbon oxide (CO and CO2) gas sensors are presented in each chapter through the lens of unique

applications in aeropropulsion and shock tube kinetics. The chapters progress chronologically, in the

approximate order in which the research occurred. Some background is presented here, otherwise

each chapter is intended to be able to largely stand alone with modest referencing across chapters

to avoid redundancy. An appendix has also been included to provide greater detail on mid-infrared

optical engineering and common experimental issues.

1.1 Background

Combustion remains a vitally important research field as a majority of the world energy demand

is met via this energy conversion process. Hydrocarbon fuels are the dominant source of chemi-

cal energy for combustion, and the characterization of how hydrocarbons decompose and oxidize

enables the engineering of combustion devices that can maximize useful energy conversion and

minimize pollutant formation. Carbon monoxide (CO) and carbon dioxide (CO2) are major prod-

ucts of hydrocarbon combustion and accurate knowledge of the evolution of these species during a

hydrocarbon reaction process enhances understanding of the effects of competing flow field mecha-

nisms such as chemical kinetics and fluid mechanics. In a more direct sense, precise measurements

of CO and CO2 in a combustion engine can, in part, quantify how much chemical energy remains

unconverted to final products at a point in time or space.

1

2 CHAPTER 1. INTRODUCTION

Laser-based absorption diagnostics provide an ideal platform for in situ measurements of chem-

ical species in high-speed, reacting flow fields due to the inherent non-intrusiveness and high time

resolution capability. Due to these advantages, decades of research have been dedicated to laser

absorption spectroscopy (LAS) for species sensing in combustion gases [1–3]. Many advances in

both sensing techniques and species-detectability have come from parallel advancements in photon-

ics (ie. laser and detector technology). The current work is no exception. Recent technical progress

and commercial maturity in tunable, room-temperature semi-conductor lasers [4–6] and fiber optics

[7, 8] in the mid-infrared are leveraged to develop field-deployable laser absorption sensors for CO

and CO2 for application to aeropropulsion engines and shock tube kinetics.

Chapter 2

CO2 TDL sensing for harsh engineenvironments

The contents of this chapter have been published in the journal Measurement Science and Technol-

ogy under the full title ’Fiber-coupled 2.7 µm laser absorption sensor for CO2 in harsh combustion

environments’ [9]. Portions of the chapter’s content have also been presented at the 28th Interna-

tional Symposium on Shock Waves [10] and will be presented at the 35th International Symposium

on Combustion [11].

2.1 Introduction

Modern combustion engines often operate above atmospheric pressure due to advantages in

thermal efficiency. At such elevated pressures, combustion modeling is less reliable as a develop-

ment tool, and experimental research is generally needed to characterize combustion processes, and

ultimately engine performance. Moreover, the short time scales involved in combustion chemistry

and high-speed flow fields require diagnostics that can provide high-bandwidth response to resolve

important transient events and to track combustion progress both spatially and temporally. Non-

intrusive laser absorption diagnostics can provide high time resolution and have been employed

extensively in shock tube kinetics experiments (P∼1-20 atm), offering high potential for applica-

tion in high-pressure engine environments with some utilization to date [12–15]. Limiting factors

in designing effective tunable diode laser (TDL) absorption sensors for combustion engines can

be divided into two sets of challenges: high-pressure, high-temperature absorption spectroscopy

and harsh thermo-mechanical environments. The present work, conducted in a high-pressure shock

3

4 CHAPTER 2. CO2 TDL SENSING FOR HARSH ENGINE ENVIRONMENTS

tube and operating detonation combustor, addresses both sets of difficulties, with the objective of

developing time-resolved, in-situ carbon oxide concentration sensors for engine studies.

Carbon dioxide (CO2) is a major combustion product for hydrocarbon fuels, and the quantity of

CO2 provides an indication of combustion completion and local fuel-to-oxidizer ratio. Figure 2.1

shows the absorption spectra of CO2 in the near-infrared region, plotted as linestrengths at 2000 K

over a wavelength range of 1–3 µm. CO2 absorption-sensing technology in the 1.3–2.1 µm wave-

length region is well developed due to the availability and affordability of diode lasers and fiber-optic

components supporting the telecommunications industry. Therefore, most CO2 absorption-based

sensors have been designed to exploit the spectral bands near 1.55 µm (2ν1+2ν2+ν3) [16, 17] and

2.0 µm (ν1+2ν2+ν3) [18, 19]. However, relatively weak absorption at high temperatures (>1000

K) limits the utility of these bands for harsh combustion applications. The combination bands near

2.7 µm are 1000 and 50 times stronger, respectively, than the bands at 1.55 and 2.0 µm, and de-

velopments in diode-laser technology in recent years have extended the available wavelength range

beyond 3 µm [8], allowing access to these stronger vibrational bands and enabling more sensitive

CO2 detection. The current work explores the potential of the spectral bands at 2.7 µm (2ν2+ν3and ν1+ν3) for CO2 sensing in high-temperature, short-pathlength engine applications at which the

shorter wavelengths are insufficient.

Figure 2.1: Absorption line-strengths of CO2 and H2O at 2000 K based on the HITEMP database.

2.1. INTRODUCTION 5

The extension of absorption-based sensing techniques to elevated pressures is complicated by

the broadening and blending of discrete spectral lines at high gas density. Information unique to a

specific transition or line is convoluted by information from neighboring lines. This blending ef-

fect is amplified at high temperatures where more quantum energy states are populated and more

transitions are active, crowding the absorption spectra. Hence, a comprehensive spectral database

and line-shape model of all overlapping lines is required. Comparisons between measurements and

simulations using this model provide a means to infer gas properties, in the current case CO2 mole

fraction. In addition to the overlap of neighboring CO2 lines at elevated pressures, interference from

other species present in the gas flow must be considered. Water vapor (H2O) is another major com-

bustion product, and has strong absorption throughout the near-infrared and mid-infrared, as shown

in figure 2.1. Overlapping H2O lines must be included in the model of the absorption spectrum,

while wavelength selection aims to minimize such interference. The HITEMP database [20] in-

cludes line parameters required to model the combined CO2 and H2O spectra, although enough un-

certainty remains in this database that key parameters must be validated in the laboratory. The Voigt

line-shape function, used to model the absorption spectra, is dominated by the Lorentzian compo-

nent at elevated pressures. It is known that the impact approximation underlying the Lorentzian

line-shape model, which assumes instantaneous collisions, breaks down at high gas densities in a

manner that is species specific. However, previous research by Farooq et al. has shown that, for the

CO2 spectra at 2.7 µm, a wavelength-modulation-spectroscopy (WMS) measurement at line-center

is immune to these nonideal effects for gas densities up to 9 amg [21]. A WMS technique has been

employed for the current high-pressure sensor and validated against our spectroscopic model in the

laboratory using a high-pressure cell.

The current CO2 sensor is aimed ultimately at field application for combustion engines, which

require diagnostic methods that are insensitive to thermo-mechanical noise sources. The advan-

tages of using WMS as opposed to direct absorption spectroscopy (DAS) in this regard have been

well documented [22–25]. In addition, engine vibrations limit free-space coupling of sensitive laser

sources. Remote light delivery, via fiber optics, is therefore important to prevent potential damage.

At 2.7 µm, fiber optics are less robust than silicon fibers associated with telecom applications at

shorter wavelengths. Part of the current research involves the design of a single-mode fiber coupling

and infrared light delivery system to interface with a practical engine. High-temperature measure-

ments of CO2 concentration, using the fiber-coupled sensor, are carried out in shock-heated CO2-

air mixtures to validate sensor accuracy and precision over a range of pressures. Finally, the sensor

is demonstrated on a pulse detonation combustor (PDC), providing a time-resolved in-stream CO2


measurement up to 10 atm.

2.2 Absorption spectroscopy theory

The theory of absorption spectroscopy has been described thoroughly in other works [23, 26,

27], and is only briefly presented here to clarify notation. The Beer-Lambert law provides the

fundamental relation governing narrow-band laser absorption spectroscopy. This equation relates

the measurable quantities of incident laser light intensity and transmitted laser light intensity through

a uniform gas medium as

(It/I0)ν = exp(−αν) (2.1)

where αν represents the spectral absorbance at frequency ν. The spectral absorbance can be related

to specific gas properties, including mole fraction xabs, by

αν =∑j

PxabsSj(T )φ(ν)jL, (2.2)

where Sj(T ) (cm−2/atm) is the line-strength of a quantum transition j which varies only with

temperature, φ(ν)j (cm) is the line-shape function and L is the path-length. In equation 2.2, the

spectral absorbance is represented as a summation, which accounts for the overlap of neighboring

transitions, common at high pressures. It follows that species mole fraction can be directly inferred

from a measurement of the transmitted and incident light intensities at a specific wavelength, and

direct absorption spectroscopy is commonly employed for optical sensors due to its simplicity [3].

Wavelength modulation spectroscopy is a more complex technique than direct absorption spec-

troscopy in that additional hardware-related parameters must be included, but insensitivity to noise

in the transmitted light intensity makes the normalized (WMS-2f /1f ) technique favorable for harsh

engine applications [25]. The technique and notation of WMS adopted here are borrowed from pre-

vious works [24, 25, 28], and briefly reviewed. For WMS, the laser injection current is modulated

rapidly, producing simultaneous modulation in laser intensity and wavelength, or frequency. The

instantaneous frequency modulation (FM), ν(t), can be expressed as

ν(t) = ν + acos(ωt), (2.3)

where ν (cm−1) is the center laser frequency, a (cm−1) is the modulation depth and ω (sec−1) is the

angular modulation frequency (ω = 2πf ). The corresponding intensity modulation (IM), I0(t), is

2.3. SENSOR DESIGN 7

defined as

I0(t) = I0 [1 + i0cos(ωt+ ψ1) + i2cos(2ωt+ ψ2)] (2.4)

where I0 is the baseline laser intensity at ν, i0 is the normalized linear IM amplitude with phase

shift ψ1 and i2 is the normalized nonlinear IM amplitude with phase shift ψ2. The quantities i0, i2,

ψ1 and ψ2 are parameters related to the laser source and must be predetermined in the laboratory.

The harmonics (nf ) of the wavelength modulation relate directly to the absorbance spectra and

can be utilized to make measurements of unknown gas properties. The magnitude of the second

harmonic signal (WMS-2f ) is sensitive to absorption line-shape curvature and has been shown to

provide sensitive species detection when the absolute magnitude of absorption is low [24]. Non-

absorption transmission losses and baseline noise can be negated by normalizing the WMS- 2f

signal with the WMS-1f signal, thereby canceling out the absolute optical power term common to

both. This proves to be a major advantage of wavelength-modulation spectroscopy for harsh engine

applications, wherein nonabsorption loss and gain mechanisms such as scattering, beam steering,

window fouling, misalignment due to vibrations and emission are common problems that degrade

direct absorption sensing techniques. The 1f -normalized WMS-2f signal (WMS-2f /1f ) is a function

of predetermined laser parameters (a, i0, i2, ψ1, ψ2), and unknown gas properties (T, P, xabs). With

an independent, simultaneous measurement of temperature and pressure, this technique can be used

to measure mole fraction, or species concentration.

2.3 Sensor design

2.3.1 Wavelength selection

Selection of the sensor wavelength begins with a consideration of the engine conditions where

the sensor will be applied. The target application here is a hydrocarbon-air combustor with pressures

up to 10 atm or higher. Expected ranges of temperature, path-length and gas composition are based

on fuel selection, engine size and fuel-to-oxidizer ratio. Typical combustion temperatures range

from 1000–2500 K, with CO2 mole fraction ranging from 5% to 25% depending on the H:C ratio of

the fuel. The path length is engine-specific. In this work, our sensor is intended for demonstration

on a pulse detonation combustor with a path length of just under 4 cm. In detonation-based engines,

gas properties are very dynamic, necessitating a wavelength selection that maximizes sensitivity to

CO2 mole fraction, while minimizing sensitivity to other transient gas properties.

The feasibility of previously utilized CO2 lines in the ν1+ν3 band near 2.7 µm was initially


evaluated based on a review of prior TDL sensing research. Most relevant, Farooq et al. conducted

measurements of CO2 mole fraction in high-pressure kinetics experiments by probing transitions in

the P-branch of the ν1+ν3 band [21]. These experiments were conducted in a shock tube with low

combustion product concentrations and argon as the bath gas. Simulations and experiments of the

same lines at engine-like conditions, where the water concentration is much higher and the bath gas

is nitrogen, proved these lines less attractive due to substantial water interference and weaker peak

absorbance from the greater line-broadening effect of nitrogen compared to argon. An investigation

of all lines in the ν1+ν3 and 2ν2+ν3 bands followed.

Figure 2.2: Absorption line-strengths of ν1+ν3 CO2 combination bands at 1000 K (HITEMP).

Primary criteria for wavelength selection in this work include strong absorbance and minimal

overlapping water spectra. Figure 2.2 shows the strongest CO2 absorption bands near 2.7 µm at

1000 K. As can be seen, the ν1+ν3 band centered at 3715 cm−1 is at least 50 percent stronger than

all other neighboring absorption bands. The R-branch of this band is stronger than the P-branch,

and includes the strongest CO2 transitions near 2.7 µm, representing the most attractive wavelength

domain (3720–3750 cm−1) for maximizing absorbance.

At elevated pressures, the CO2 absorbance spectrum shows significant overlap between neigh-

boring lines. Moreover, H2O absorbance interferes with the CO2 spectra and is unavoidable near

2.7 µm due to the broadening of the strong fundamental water lines throughout the spectral domain.


Figure 2.3: CO2 and H2O absorbance spectra simulations (air bath gas) near the peak of the R-branch in the ν1+ν3 band at an expected PDC temperature (1800 K), path-length (L = 4 cm) andcomposition for (top) P = 1 atm and (bottom) P = 5 atm.


Figure 2.3 illustrates the overlapping absorbance spectra for both combustion products, simulated

at a typical temperature for ethylene-air combustion at 1 atm and 5 atm near the strongest spectral

region of the CO2 bands. The distinct absorption feature at 3733.48 cm−1 offers an attractive com-

bination of relatively strong absorbance and minimal H2O interference, and this wavelength (2.6785

µm) is selected for the work presented here.

The target CO2 absorption feature at 3733.48 cm−1 is composed of multiple overlapping CO2

lines in the ν1+ν3 band. The strongest individual line in this feature is the R(26) transition in

the ν(000→101) combination band at 3733.469 cm−1, which comes from the vibrational ground

states. While this line is dominant over the range of conditions considered, several other overlapping

lines from hot bands make a measurable contribution at elevated temperature and pressure, most

notably the R(13) line in the ν(010→111) band at 3733.499 cm−1. Figure 2.4 shows both the

simulated and measured absorbance for the R(26) line and its closest neighboring lines at 1000 K,

measured in a static cell at low pressure to allow discernment between lines that become blended at

higher pressures. Key line-strength and line-broadening parameters are measured in this work and

discussed in a later section.

Figure 2.4: Pure CO2 absorbance spectra near 3733.48 cm−1 at T = 1000 K and P = 40 torr,measured in a static cell and compared to a HITEMP 2010 simulation.


2.3.2 Laser modulation parameters

The magnitudes of the WMS harmonic signals vary directly with modulation depth (a), which

can be adjusted to optimize these signals for target gas conditions. The optimum modulation depth

is defined for this sensor as the value at which the WMS-2f signal at absorption line center is maxi-

mized. Though the first and second harmonic signals are typically not maximized at the same mod-

ulation depth, the WMS-2f signal is more critical for maximizing the signal-to-noise ratio (SNR) of

the normalized WMS-2f /1f measurement.

To accurately simulate the WMS-nf spectra, hardware-specific parameters unique to the laser

source (a, i0, i2, ψ1, ψ2) must be determined. The diode laser used in this work is a single-mode

distributed-feedback (DFB) laser with approximately 3 mW output power, centered in wavelength

at 2678 nm (Nanoplus GmbH). At a fixed modulation frequency, the modulation depth varies as a

function of the injection-current amplitude. Increasing the injection-current amplitude increases the

modulation depth, which is coupled with the linear and non-linear IM amplitudes (i0 and i2). The

relationship between these hardware-related parameters is shown in figure 2.5 at a fixed modulation

frequency of 60 kHz, sufficient to meet the bandwidth requirement of a pulse detonation combustor.

Polynomial fits are shown which define the functional relationships of the IM amplitudes to modula-

tion depth, and these functions are incorporated into the WMS model. The phase-shift components,

ψ1 and ψ2, are measured and found to be largely independent of modulation depth, with values of

1.31π radians and 1.07π radians, respectively. These phase shift components are a stronger func-

tion of modulation frequency. Further details regarding the methods used to determine these laser

parameters can be found in prior works [29, 30]. With the laser parameters fully defined at the

chosen modulation frequency, the WMS-2f signal at line-center can be computed as a function of

modulation depth for various expected test conditions, as shown in figure 2.6.

The WMS-2f signal is maximized at different modulation depths depending on test conditions,

which can vary both between different engine applications and within a single engine transient such

as with a detonation-based engine. Figure 2.6 provides the relative WMS-2f signal at linecenter

for conditions expected in the PDC application. While temperature and pressure are both expected

to be transient, the optimal modulation depth is most sensitive to pressure due to line broadening

and is shown here for various isobars at 2000 K. A modulation depth of 0.192 cm−1 is chosen for

this work, representing a compromise intended to provide sufficient signal over the largest range

of transient pressure. With the modulation depth prescribed, all hardware-related parameters are


Figure 2.5: Measured laser parameters (i0 and i2) as a function of modulation depth at f = 60 kHz.

known, yielding the normalized laser intensity for our 2678 nm laser,

I0(t)/I0 = 1 + 0.691 cos(2πft+ 1.31π) + 0.0125 cos(4πft+ 1.07π) (2.5)

where f is 60 kHz. A spectral simulation of the relevant WMS harmonics is illustrated in fig-

ure 2.7 at a typical PDC condition, consistent with the CO2 absorbance spectra shown in figure 2.3.

In practice, it is found that for this infrared DFB laser, the IM amplitudes (i0 and i2) as defined

in equation 2.5 may change slightly between the laboratory setting and a field setup, while the

phase-shift terms and modulation depth are insensitive to such change. Because the WMS-2f signal

exhibits a strong dependence on the linear IM amplitude (i0), especially when the 2f signal is small,

a calibration of this hardware-related parameter for each unique measurement setup improves the

measurement accuracy.

2.3.3 Optical engineering

To ensure that sensitive infrared laser equipment is isolated from the thermo-mechanically harsh

engine environment, optical fibers are employed for remote light delivery. Relatively few fiber op-

tions are commercially available that can transmit light beyond the∼2.3 µm threshold of traditional


Figure 2.6: Simulated WMS-2f (background subtracted) at 3733.48 cm−1 as a function of modula-tion depth (a) for various pressures at 2000 K.

silica-based fibers, and careful design considerations must be made to maximize light transmission

and minimize fiber-related noise when more exotic infrared fiber materials are utilized. Figure 2.8

illustrates a simplified schematic of the fiber delivery system developed for this sensor, composed

primarily of a fiber-coupling lens (1), an optical fiber (2) and a beam-collimating lens (3). The

following discussion outlines the process used to optimize these optical components.

For this sensor, the optical fiber is the central component to remote light delivery, and the ma-

terial and core diameter represent the most important design parameters of the fiber. Optical fibers

that transmit in the infrared are constructed from a number of materials including chalcogenide,

fluoride glass, sapphire and silver halide. ZBLAN, a multi-component fluoride glass material, is

chosen for this work due to its relatively high transmission at 2.7 µm (<0.05 dB m−1 loss), low

surface reflection losses and broad availability over a range of core diameters and lengths. The core

diameter of the optical fiber affects the ease and efficiency at which light can be coupled into the

fiber as well as the quality of the beam exiting the fiber. While large-diameter fibers are coupled

with the incident laser light more easily and efficiently, the light exiting the fiber can be difficult to

collimate and can suffer from an adverse phenomenon known as ’mode noise’. These issues corre-

late with the number of transversemodes supported by the fiber, which scales as N∼ D2, where D is


Figure 2.7: Simulated CO2 WMS spectra (2f, 1f, 2f /1f ) near 3733.48 cm−1 at an expected PDCcondition (analogous to figure 2.3).

the core diameter [31]. In prior research, multi-mode fluoride glass fibers (D = 200 µm) have been

used to deliver infrared light (∼3.39 µm) to a PDC environment, and were found to be the limiting

noise source for the optical diagnostic [13]. Advances in fiber technology in recent years have made

single-mode fluoride glass fibers available. Here,we utilize a 9 µm diameter single-mode ZBLAN

fiber (IR-Photonics S009S20FFP) with a numerical aperture of 0.2.

When focusing incident laser light into the single-mode fiber, a properly designed and sized lens

must be selected to realize adequate coupling efficiency. Efficient fiber-coupling is achieved when

the numerical aperture (NA) of the lens matches that of the fiber and when the minimum focal spot

size is less than the core diameter of the fiber. The diffraction-limited spot size (dmin) for a beam

with a single Gaussian transverse mode is defined as

dmin =4fλ

πd0(2.6)

where f (mm) is the focal length of the lens, d0 (mm) is the diameter of the incident beam and

λ (µm) is the wavelength [32]. A lens with a short focal length can then be used to achieve a

small focused spot size at the diffraction limit. Aspheric lenses are commonly used to focus and


Single-mode Fiber (2)

Collimation Optic (3)

PV Detector Test Medium

DFB TDL (2678 nm)

Coupling Lens (1)

BP Filter

Figure 2.8: Simplified light delivery and detection schematic.

collimate light without introducing spherical aberration, an anomaly which prevents diffraction-

limited performance. For this application, we select a zinc-selenide (ZnSe) aspheric lens with a

focal length of 6 mm and a numerical aperture of 0.25. With an incident beam diameter (1/e)

of ∼3 mm, the diffraction-limited spot size is ∼7 µm, which is less than the core diameter of

our single-mode fiber. The fiber is aligned with the lens using a five-axis stage (Newport FPR2-

C1A), and a coupling efficiency of∼50% is achieved, with losses attributable to non-Gaussian beam

characteristics, a slight mismatch of numerical apertures between the lens and fiber and imperfect

free-space alignment.

Though the coupling efficiency achieved here is lower than that which has been demonstrated

with larger diameter fluoride glass fibers, the quality of the single-mode beam exiting the fiber is

superior. In addition to a significant reduction in mode noise caused by vibrations of the fiber,

the single-mode beam can be more readily collimated. Here we use a ZnSe aspheric lens with

a 12 mm focal length to collimate the beam, achieving an ∼2 mm diameter beam at a working

distance of 50 cm. Analogous experiments with various multimode fluoride glass fibers showed that

a minimum beam diameter of no less than 10 mm could be achieved at the same working distance.

The collimation lens is integrated with an SMA connector to attach directly to the fiber termination,

and then mounted to a four-axis kinematic stage which directs the collimated beam through the

test medium. For this sensor, the benefits of improved collimation and stable single-mode light


propagation offset the lower coupling efficiency associated with the small diameter fiber.

Our optical setup is completed with an infrared photovoltaic detector (Vigo PVI-2TE-4) that

combines with a focusing lens and spectral filter to interface directly with the test hardware as shown

in figure 2.8. The detector bandwidth is 10 MHz. The light catch and detection components are

mounted together in a compact linear cage system (Thorlabs- 30 mm). The transmitted beam is fo-

cused by a 20 mm focal length CaF2 lens onto a 1 mm2 detection area (D = 2×1011cmHz1/2W−1).

A narrow-bandpass spectral filter (Northumbria SNB-2688-000288), centered at 2688 nm with a

spectral width of ∼50 nm, is situated between the lens and detector. Because the detector is sensi-

tive from ∼2 to 4 µm wavelength, a filter is critical for reducing broadband H2O emission at high

temperatures, typical in this infrared domain for combustion gases. A translational stage houses the

focusing lens to help achieve concentricity with the transmitted beam and to minimize sensitivity to

beam steering in the test medium. The length of the linear catch system, from window to detector

element, is less than 2 inches (50 mm).

2.4 Sensor development

2.4.1 Measurements of key spectroscopic parameters

The HITEMP 2010 spectroscopic database, a compilation of computed line parameters, pro-

vides the foundation for the CO2 spectral model. Because CO2 is a simple polyatomic molecule

(linear and symmetric), line parameters can be calculated with a high degree of accuracy [33, 34].

However, some line-broadening parameters that are important in combustion gases are not found

in the HITEMP database and must be determined experimentally. In addition, key linestrength val-

ues were validated. Figure 2.9 represents a sensitivity analysis of the WMS-based mole fraction

measurement to the strength and collisional broadening of the five most important CO2 lines near

3733.48 cm−1 (see figure 2.4) at a typical PDC condition. Such analysis helps identify which line

parameters deserve experimental study. Though the measured spectra at 3733.48 cm−1 represents a

blended contribution of several lines, the R(26) line dominates. In this work, we use a heated static

cell to validate the line-strength values in HITEMP for the R(26) and R(13) lines and to measure

water-broadening and self-broadening parameters for the R(26) line.

Line-strength measurements are conducted using a scanned direct-absorption technique, with

pure CO2 in a quartz static cell that is heated by a tube furnace. The details of this experimental setup

are discussed in a previous work [34]. The experiment provides precise control of temperature (±1

K) and pressure (±0.05 Torr), facilitating high-fidelity measurements of spectroscopic parameters

2.4. SENSOR DEVELOPMENT 17

Figure 2.9: Sensitivity analysis of the WMS-2f /1f mole fraction measurement to the line strengthand collisional broadening of the strongest CO2 transitions near 3733.48 cm−1 at an expected PDCcondition (see figure 2.7).

up to 1000 K. Each independent line-shape is fit by a Voigt function and superimposed. From this

Voigt line-shape solution, an integrated absorbance area, Ai, can be extracted for each line, which

is directly proportional to the line-strength, Si(T ), by

Ai = Si(T )PxabsL, (2.7)

where L is the path-length of the cell. Figure 2.10 shows the measured line strength as a function

of temperature for the R(26) and R(13) lines, compared to HITEMP calculations. The observed

agreement (σ < 3%) validates the line assignment (E”) and the line-strength values in the HITEMP

database.

Collisional line-broadening parameters are measured in a similar experiment, utilizing the di-

rectly measured absorbance spectra to solve for the unknown broadening coefficient, 2γ(T), asso-

ciated with each collisional partner. Here, we employ a somewhat less conventional technique to

tackle the difficulties associated with highly Doppler-broadened lines. The typical method for deter-

mining the collisional-broadening coefficient is to measure the collisional line-width, ∆νc (cm−1),


Figure 2.10: Measured line-strength for the R(26), E” = 273 cm−1, and R(13), E” = 738 cm−1 linesfrom 600 to 1000 K.

where

∆νc = P∑i

xabs2γj−i. (2.8)

At a known temperature and pressure, a Voigt fit to a measured line-shape yields two solutions: line-

strength, or integrated area, and collisional line-width. It is undesirable to measure the collisional

line-width in a low-pressure environment because the value can be small and thus susceptible to

small systematic error. However, in order to achieve a converging solution to the Voigt fit of each

line, low-pressure experiments are required to keep lines separable when neighboring lines are in

close proximity. A compromise, especially at high temperatures, can result in measured collisional

line-widths (∼0.001 cm−1) that are prone to errant noise. In this work, we fit our Voigt line-

shape model to the measured spectra after fixing the linestrength, Si(T ), according to the validated

HITEMP values, and the Doppler line-width, such that a Voigt fit has only one free parameter, the

collisional line-width, or 2γ(T ). This single free-parameter fitting method improved the scatter in

our measured broadening coefficients, which are shown in figure 2.11. A power law describes the

temperature dependence of collisional broadening as

2γ(T ) = 2γ(T0)

(T0T

)n(2.9)


where n is the temperature exponent. The HITEMP database includes air-broadening (CO2-air)

and self-broadening (CO2-CO2) coefficients at 296 K, as well as the temperature exponent for

air broadening (nair). To account for all major collisional partners in combustion gases, water-

broadening (CO2-H2O) parameters are measured, as well as the temperature exponent for self-

broadening (nself ) as in figure 2.11. Table 2.1 lists collisional-broadening parameters for the R(26)

line.

The broadening model used here does have some limitations and simplifications that deserve

explanation. For one, the model assigns the same water-broadening parameters and self-broadening

temperature exponent to the other CO2 lines surrounding the R(26) line. This represents an im-

provement over the original air-broadening assignment provided by HITEMP because broadening

for CO2 lines is less sensitive to the lower energy state (J”) than to the collisional partner. Addi-

tionally, the assumption that the bath gas is simply air for the CO2 and H2O combustion products is

incomplete because most of the oxygen (O2) in the air is consumed during combustion and carbon

monoxide (CO) may also be present in appreciable concentrations. Still, O2 and CO are observed to

have similar broadening effects as air, or nitrogen (N2). The net uncertainty in the broadening cal-

culation associated with these simplifications is estimated to be 2-3% depending on gas conditions,

and not limiting to the accuracy of the sensor.

2.4.2 Modulation-induced wavelength shift

High-frequency laser modulation can cause a shift in center laser wavelength away from the

unmodulated wavelength at the same mean current and temperature laser settings [21, 24]. This

wavelength shift results from ohmic heating of the laser cavity due to injection-current modulation,

and thus increases with modulation amplitude and frequency. To calculate the wavelength shift, the

measured WMS spectra is compared to simulations. The shift, ∆ν, for this 2678 nm DFB laser is

measured to be 0.102 cm−1 at our modulation settings (f = 60 kHz, a = 0.192 cm−1).

Table 2.1: Collisional-broadening parameters for the R(26) CO2 line (E” = 273 cm−1 in combustionexhaust gases.

Collisional partner 2γ(296K) [cm−1/atm] n

Air 0.143 [20] 0.68 [20]CO2 0.188 [20] 0.62±0.03H2O 0.25±0.02 0.81±0.03


Figure 2.11: (top) Measured water-broadening coefficient for the R(26) line from 450 to 950 K, in-cluding a best-fit power law equation and reference bounds using ±10% nH2O

. (bottom) Measuredself-broadening coefficients for the R(26) line from 750 to 1000 K, including a best-fit power lawequation and reference bounds using ±12% nself .


2.4.3 High gas density lineshape modeling

Previous researchers have observed that the Voigt line-shape model can fail to capture collisional

broadening effects when gas density is high [35–38]. Farooq et al. investigated these non-Lorentzian

effects for several CO2 transitions near 2.7 µm and concluded that a WMS-2f /1f measurement near

linecenter was insensitive to such effects at number densities below 9 amagats (amg) [21]. However,

most lines previously investigated were more isolated than those in this work, leading us to question

this assumption. A static cell experiment, similar to that just discussed, was conducted at elevated

number densities to validate the accuracy of our sensor at 3733.48 cm−1, which is at the effective

absorption peak of the R(26) and R(13) overlapping lines. Figure 2.12 shows the 2f /1f line-shape

measurements relative to simulations at room temperature, over a range of pressure from 2 atm

(1.8 amg) to 8 atm (7.3 amg), accounting for the wavelength shift due to modulation. While the

measurements deviate from the simulations away from the absorption peak, the measured values

at 3733.48 cm−1 agree well with simulations (σ ∼ 1-2%). These measurements indicate that at

number densities targeted for this sensor (< 6 amagats), the WMS- 2f /1f magnitude at the absorption

peak is not meaningfully influenced by non-Lorentzian effects.

Figure 2.12: Measured CO2 in air WMS-2f /1f spectra overlaid with corresponding simulations near3733.48 cm−1 at T = 298 K, L = 23 cm, XCO2 = 0.015, P = 2-8 atm.


2.5 High-temperature shock tube validation

Shock tubes can provide well-defined environments for short time periods (∼ms) that emulate

combustion conditions not achievable in a static cell. In this work, a high-pressure steel shock tube

is used to validate the sensor at temperatures up to 2500 K behind non-reactive reflected shocks

in CO2-air mixtures. In addition to a high-temperature examination, the short time scales involved

in shock tubes facilitate an assessment of the sensor bandwidth. The shock tube used for these

experiments has been discussed in past works [39, 40], and only limited detail is outlined here. The

optical setup is well represented by figure 2.8, where the collimated beam is transmitted through

wedged sapphire windows located approximately 1 cm from the shock tube end wall. The shock

tube inner diameter and effective path-length for our measurement is 5 cm. Gas properties across the

shock tube are assumed to be uniform in accordance with a planar shock. Temperature is calculated

from the shock jump relations, based on measured pre-shock conditions and shock velocity, while

pressure is measured with a high speed transducer (1 MHz). Typical steady-state test time for the

non-reactive shocks conducted here is 1 ms, and typical temperature uncertainty is ±1%.

Figure 2.13 provides an example time history of measured pressure and measured WMS-2f /1f

Figure 2.13: Example time-history of measured pressure and measured WMS-2f /1f for a shock tubevalidation experiment. Time 1: initial condition. Time 2: between incident and reflected shocks.Time 5: after reflected shock. Measurement 1 cm from end wall.

2.6. CO2 MEASUREMENTS BEHIND DETONATION WAVES 23

for a non-reactive shock tube experiment. As shown in the figure, the initial low pressure (< 400

Torr) test gas mixture is heated and compressed by the incident and then reflected shock waves.

Time zero signifies the passage of the reflected shock. The signals during the steady-state interval

(t∼0.1–2 ms) are utilized to determine a mean WMS-2f /1f magnitude and standard deviation (noise)

for a given frequency filter. The 2f and 1f signals are extracted here by applying a 40 kHz (bandpass)

frequency filter, yielding sufficient bandwidth to resolve changes on the order of 25 µs. CO2 mole

fraction is directly inferred from the measured WMS-2f /1f magnitude and compared to the known

mole fraction. Figure 2.14 shows this comparison for each test condition. Shock tube experiments

are conducted over a wide range of temperatures (1000–2500 K) at elevated pressures (3–12 atm)

to validate the accuracy (σ ∼ 3.5%) of the sensor.

Figure 2.14: Sensor validation data showing the measured CO2 mole fraction compared to theknown CO2 mole fraction over a range of conditions produced in a shock tube (σ ∼ 3.5%); L = 5cm, XCO2 = 6− 9%.

2.6 CO2 measurements behind detonation waves

Here we demonstrate the feasibility of our CO2 absorption-based sensor on a pulse detonation

combustor, highlighting the time-resolution and operability limits of the sensor. Figure 2.15 shows

representative co-planar measurements of pressure and WMS-2f /1f for two consecutive detonations

in the PDC produced by an ethylene-air mixture. The plot of facility-measured pressure reveals the

short time scale (∼ms) of the detonation event, which cycles every 50 ms (20 Hz). The measured

WMS-2f /1f signal illustrates the time-resolution capability of the sensor (∼10 kHz) in this harsh


Figure 2.15: Representative data from a pulse (20 Hz) detonation combustor operating on ethy-lene/air; L = 3.55 cm; optical absorption ∼ 1 − 15%; pressure and WMS-2f /1f measurements arecoplanar in combustion chamber.

environment. The signal-to-noise ratio (SNR) of the 2f /1f data exhibits a strong temporal depen-

dence, with higher fidelity measurements (SNR > 20) in the lower pressure (< 8 atm) tail of the

detonation cycle. Similar to the shock tube validation studies, measurement precision degrades

above 10 atm (SNR < 2). This results from a sharp decrease in the 2f signal when the absorption

spectrum broadens at higher pressures (illustrated in figure 2.12). Utilizing temperature data from

H2O thermometry [14], CO2 mole fraction is inferred directly from the measured WMS-2f /1f sig-

nal. Noise in the 2f /1f signal translates nearly linearly to noise in the mole fraction data, yielding

similar pressure limitations to the mole fraction measurement.

2.7 Summary

Most modern combustion devices operate at elevated pressures (>1 atm). Innovative diagnostic

methods are needed to provide quantitative, time-resolved measurements of combustion product

species in the flow fields of such devices. This work aimed to develop a non-intrusive, in situ

CO2 concentration sensor for engine application based on a fixed-wavelength WMS-2f /1f strategy.

The ν1 + ν3 vibrational absorption band near 2.7 µm was probed, offering approximately 50%

2.7. SUMMARY 25

stronger absorbance than the more commonly used near-infrared bands, and more sensitive detection

than prior high-pressure CO2 sensing near 2.7 µm [21]. Wavelength selection at 3733.48 cm−1

exploited the overlap of the R(26) and R(13) transitions, representing an exceptional combination

of cross-sectional strength and isolation from water absorption relative to alternative wavelengths

in the 2.7 µm CO2 bands. The modulation depth of a 2.68 µm DFB laser was optimized for

high-temperature application over a range of elevated pressures (2–12 atm). At higher pressures

(>12 atm), the 2f signal decreases sharply due to the broadening and blending of neighboring lines

near 3733.48 cm−1 that diminishes curvature in the absorption spectrum. The wavelength shift

due to modulation (at 60 kHz) was accounted for in calibrating the laser for the fixed-wavelength

approach. High gas density WMS-2f /1f measurements in a room-temperature static cell confirmed

that non-Lorentzian line-shape effects do not influence measurements up to 7 amagats, which is

the equivalent density to 35 atm at 1500 K. For remote light delivery, the laser was free-space

fiber-coupled to a single-mode ZBLAN fiber. The fiber-coupled sensor was validated on a shock

tube at combustion temperatures (up to 2500 K) over a range of pressures (3–12 atm). In situ

measurements of CO2 mole fraction were conducted on an ethylene-air pulse (20 Hz) detonation

combustor, demonstrating sensor bandwidth (∼10 kHz) and range (2–10 atm, 1000–2500 K) in a

harsh engine environment. Engine applications with longer path-lengths (>5 cm) and fuels of lower

H:C ratios (< 2) may offer higher pressure limits. Moreover, utilization of a similar technique, but

at a wavelength within the CO2 fundamental band near 4.3 µm may offer more sensitive detection

and better isolation from water interference, allowing extension to higher pressures. Initial engine

research at 4.3 µm is described in later chapters. To the author’s knowledge, this work represents

the first time-resolved, in-stream measurements of CO2 concentration in a detonation-based engine.

Chapter 3

CO sensing for detonation engines

The contents of this chapter have been published in the journal Applied Optics under the full ti-

tle ’Quantum cascade laser absorption sensor for carbon monoxide in high-pressure gases using

wavelength modulation spectroscopy’ [41]. Portions of the chapter’s content have also been pre-

sented at the 28th International Symposium on Shock Waves [10] and will be presented at the 35th

International Symposium on Combustion [11].

3.1 Introduction

While carbon dioxide indicates complete combustion, carbon monoxide (CO) provides a com-

plementary and sensitive indicator of incomplete combustion. Carbon monoxide is an intermediate

and sometimes major (> 1%) product of hydrocarbon combustion, serving as a metric of reaction

progress and, in equilibrium, local fuel-to-oxidizer ratio. Excess CO in combustion exhaust is unde-

sirable due to the correlation with poor engine efficiency and matters of health and safety. The need

to minimize CO concentrations in various combustion systems drives the demand for highly sensi-

tive diagnostics. The strong fundamental absorption band of CO near 4.7 µm offers potential for

very sensitive species detection, even at short path-lengths (< 10 cm). However, until recent years,

laser technology in the mid-infrared (> 3 µm) has typically required cryogenic cooling systems

that render the lasers impractical for most sensing applications beyond the laboratory. Due to read-

ily available room-temperature diode lasers and optical hardware in the near-infrared wavelength

domain (1–3 µm), most previous CO absorption-based sensors have been designed to probe the

overtone vibrational bands near 1.55 µm [16, 42–44] and 2.3 µm [30, 45, 46], which unfortunately

are more than four and two orders of magnitude weaker, respectively, than the fundamental band.

26

3.2. WAVELENGTH SELECTION 27

Weak absorption strength, combined with spectral interference from carbon dioxide and water, ren-

ders the near-infrared CO sensors insufficient for most short path-length combustion systems. The

recent maturation of the quantum cascade laser (QCL) has enabled access to mid-infrared wave-

lengths previously unavailable with room-temperature lasers [4, 5, 47], and researchers have since

exploited the fundamental CO absorption band near 4.7 µm for various applications requiring very

sensitive species measurements [48–50]. Here we utilized a distributed-feedback QCL near 4.85

µm to probe the fundamental band and extend CO absorption sensing to high pressures (> 40 atm)

in hydrocarbon-fueled engines.

As previously mentioned, field sensing in harsh combustion engines necessitates a measurement

technique which is robust against thermo-mechanical noise sources and conducive to spectral char-

acteristics associated with both high pressures and high temperatures (> 1000 K). A wavelength

modulation spectroscopy (WMS) technique was employed for the present sensor, similar to the

technique utilized for the diode laser sensor for CO2. With WMS, frequency filtering of harmonic

signals allows for substantial noise rejection. Moreover, the ratio of two harmonic signals (WMS-

nf /mf ), which can be related to the absorption spectra, is independent of absolute optical power or

intensity baseline [25]. These characteristics make WMS a well-suited spectroscopic technique for

harsh engine applications. High-temperature (1100–2600 K), high-pressure (10–40 atm) absorption

measurements, using this WMS sensor, were carried out in shock-heated CO−N2 mixtures in order

to validate the sensor and calibrate the CO spectral model. The QCL was then free-space coupled

to a single-mode fiber for remote light delivery, and the fiber-coupled sensor was demonstrated on

an operating pulse detonation combustor (PDC) at the Naval Postgraduate School in Monterey, CA,

yielding time-resolved, in-stream CO concentration measurements at conditions exceeding 40 atm

and 2700 K.

3.2 Wavelength selection

The primary criteria for wavelength selection include strong absorbance and minimal spectral

interference from other combustion products, namely water and carbon dioxide. Figure 3.1 shows

the fundamental CO absorption band from 4.4–5.5 µm, plotted as line-strengths at 2000 K, and

overlaid with the H2O and CO2 absorption lines within the same domain. The CO2 fundamental

band (∼4.3 µm) exhibits a very dense spectrum that interferes substantially with the R-branch of

the CO band under typical combustion conditions where the CO2:CO ratio is much greater than

unity. The P-branch also has significant interference from water, but the overlapping H2O spectrum

28 CHAPTER 3. CO SENSING FOR DETONATION ENGINES

is much less crowded compared to the interfering CO2 band, such that the larger spacing between

water lines yields some narrow regions that are nearly interference-free. Specifically, the spectral

domain near 2060 cm−1 provides an attractive combination of low relative interference and strong

absorption. Therefore, we select this wavelength region (4.855 µm) to detect carbon monoxide.

Figure 3.1: Absorption line-strengths of CO, CO2, and H2O at 2000 K (HITEMP).

Absorbance simulations at target engine conditions provide a more quantitative estimation of

signal strength and isolation from interfering species. Here the target application is a hydrocarbon-

air combustor with pressures ranging from 5 to 40 atm. Expected values of temperature, gas com-

position, and path-length are based on fuel selection, fuel-to-oxidizer ratio, and engine size. Typical

combustion temperatures range from 1000–3000 K. Carbon monoxide concentrations can vary dra-

matically from ppm levels to several percent (∼ 104 dynamic range) depending primarily on the

H:C ratio of the fuel, local equivalence ratio, and combustor residence time. This sensitivity en-

hances the value of the diagnostic for assessing combustion processes, but can make estimation

of expected mole fraction and sensor design challenging. The path-length is engine-specific. In

the current research, the CO sensor is intended for utilization on an ethylene-air pulse detonation

combustor with a path-length of slightly less than 4 cm. Assuming chemical equilibrium, the con-

centration of CO and other combustion products were estimated at several conditions [51], and the

absorbance spectra were simulated using these values.

Figure 3.2 shows absorbance simulations near 4.855 µm at two expected engine conditions by

utilization of the Voigt line-shape function coupled with the HITEMP spectroscopic database [20].

The high-pressure CO absorption spectrum near 2059.9 cm−1 is composed of multiple overlapping

transitions, but the P(20) line in the ν(0→1) fundamental band makes the dominant contribution.


The P(14) line at 2060.33 cm−1 and the P(8) line at 2059.21 cm−1, from the respective ν(1→2) and

ν(2→3) fundamental hot bands, also make a small contribution at high temperatures. Interference

from CO2 is negligible, and H2O interference is relatively minimal, especially at the line-center of

the P(20) line (2059.91 cm−1). The influence of the presence of CO2 and H2O on the CO absorption

measurement is further discussed in a later section.


3.3.1 Modulation depth optimization

The WMS harmonic signals vary directly with modulation depth (a), which can be adjusted to

optimize signal magnitudes for a given set of gas conditions. We define the optimum modulation

depth for this sensor as the value at which the WMS-2f signal is maximized, within constraints of

the laser. The laser used in this work is a single-mode distributed-feedback (DFB) QCL (ALPES),

centered in wavelength near 4855 nm, with a typical output power of 8 mW. The modulation depth

is primarily a function of two laser input parameters: modulation frequency (f ) and amplitude of the

modulated injection-current. The modulation depth increases approximately linearly with injection-

current amplitude, and decreases non-linearly with modulation frequency. A maximum modulation

depth at a given modulation frequency can therefore be measured at the upper limit of injection-

current amplitude, which is bound by the current threshold, maximum operating current, and mode-

hoping regions of the QCL. Figure 3.3 plots the maximum modulation depth for the QCL, when

tuned to the P(20) line-center, at modulation frequencies from 40 kHz to 100 kHz. A high mod-

ulation frequency is desirable due to a direct correlation with sensor bandwidth, but this comes at

expense of modulation depth.

Assessment of the WMS-2f signal as a function of modulation depth requires further knowledge

of the aforementioned characteristic laser parameters. The linear and non-linear IM amplitudes (i0and i2) are coupled with modulation depth, and the relationship between these hardware-related

parameters is depicted in figure 3.4. The phase-shift components, ψ1 and ψ2, are found to be

approximately independent of modulation depth, with measured values of 1.16π radians and 0.84π

radians, respectively. Details on the methods used to determine these parameters are found in other

works [29, 30]. With laser parameters defined, the relative magnitude of the WMS-2f signal at line-

center can be calculated as a function of modulation depth for various test conditions, as illustrated

in figure 3.5.

The maximum WMS-2f signal is highly dependent on gas conditions, which vary dramatically


Figure 3.2: Absorbance spectra simulations of equilibrium concentrations of CO, CO2, and H2O(air bath gas) near 4.855 µm at expected PDC conditions (C2H4-air, φ ≈1); L = 4 cm, (top) P = 10atm, T = 2500 K and (bottom) P = 30 atm, T = 3000 K.


Figure 3.3: Measured modulation depth at maximum injection-current amplitude (±120 mA) as afunction of laser modulation frequency.

in a detonation-based engine. Though temperature and pressure are both transient, the WMS-2f

signal is most sensitive to pressure due to sensitivity to line broadening. Figure 3.5 highlights this

sensitivity over a range of pressures expected in the PDC. As pressure increases, the optimum mod-

ulation depth increases, to the extent that maximizing the WMS-2f signal at pressures greater than

10 atm becomes impractical within the tuning limitations of the current laser, revealing a requisite

compromise between sensor bandwidth and signal magnitude or detectability. A modulation depth

and frequency of 0.235 cm−1 and 50 kHz, respectively, were chosen for this sensor to yield suf-

ficient signal over the broadest range of transient conditions, while maintaining the desired high

bandwidth capability. With modulation depth specified, all hardware parameters are established,

and the normalized laser intensity can be expressed as

I0(t)/I0 = 1 + 0.718 cos(2πft+ 1.16π) + 0.0106 cos(4πft+ 0.84π) (3.1)

where f is 50 kHz. Figure 3.6 shows a spectral simulation of the relevant WMS harmonic signals

at a typical PDC condition based on these selected laser settings. We note that in practice the

IM amplitudes, as defined in equation 3.1, vary slightly between experimental setups, whereas the

phase-shift terms and modulation depth are insensitive to such extrinsic changes. Due to the strong


Figure 3.4: Measured laser intensity parameters (i0 and i2) as a function of modulation depth (f =50 kHz).

dependence of the WMS-2f signal on the linear IM amplitude (i0), a calibration of this hardware-

related parameter for each unique optical setup improves measurement accuracy.

3.3.2 Modulation-induced wavelength shift

High-frequency modulation of laser injection-current causes a shift in the center wavelength

away from the unmodulated laser wavelength at the same mean current and temperature settings.

This wavelength shift results from ohmic heating of the laser cavity that increases with modulation

frequency and injection-current amplitude [21]. To maintain line-center in our WMS measure-

ment, a corresponding wavelength correction is required. The magnitude of wavelength shift can

be determined by comparing measured WMS spectra using specific modulation settings to spectral

simulations. For the 4.85 µm DFB-QCL, the wavelength shift, ∆ν, was measured to be 0.032 cm−1

at the selected modulation settings (f = 50 kHz, a = 0.235 cm−1).

3.3.3 Spectral modeling

The current sensor employs a Voigt line-shape profile, coupled with the HITEMP 2010 database,

as the foundation for the CO spectral model. Figure 3.7 illustrates the suitability of the Voigt


Figure 3.5: Simulated WMS-2f (background subtracted) at 2059.91 cm−1 as a function of modula-tion depth for various pressures at 2000 K.

line-shape function for modeling the P(20) and P(14) CO lines near 2060 cm−1 based on direct-

absorption measurements in a heated static cell. Due to the simple diatomic structure of carbon

monoxide, the CO spectra can be modeled accurately using the Voigt line-shape, along with com-

puted HITEMP line parameters, over a broad range of conditions [30, 45]. At a given thermody-

namic condition, a Voigt line-shape profile is defined by two transition-specific spectroscopic pa-

rameters: line-strength, S(T ), and the collisional line-broadening coefficient, 2γ(T). The Doppler

line-width is also needed to fully describe the Voigt profile, but this parameter is a function of

wavelength and temperature only and not transition-specific. Calculated line-strength values from

HITEMP for the fundamental CO band have been validated previously, and shown to provide a

high degree of accuracy (σ ∼ 1%) [48]. Line-broadening parameters for N2, CO2, and H2O, the

dominant collision partners, are found in the literature and noted in Table 3.1 [52, 53]. Broaden-

ing uncertainties of less than 10% are reported for all perturbers. Figure 3.8 presents a sensitivity

analysis of the WMS-based mole fraction measurement to uncertainties in these key spectroscopic

parameters for the two most influential CO lines near 2059.91 cm−1 at a detonation condition.

Such analysis helps identify line parameters that deserve further experimental investigation. As pre-

viously discussed and underscored by the sensitivity analysis in figure 3.8, the P(20) line makes the

dominant contribution to our CO mole fraction measurement by more than a factor of ten relative


Figure 3.6: Simulated CO WMS spectra (2f, 1f, 2f /1f ) near 2059.9cm−1 at a typical PDC condition;T = 2000 K, P = 20 atm, L = 4 cm, XCO = 1%.

to the next most important contributor in the P(14) line of the ν(1→2) band. Moreover, the CO-

N2 broadening coefficient is highlighted as the most important broadening parameter, due to the

high concentration of diatomic nitrogen in hydrocarbon-air combustion exhaust. To improve sensor

accuracy and mitigate uncertainty, we calibrated the CO-N2 broadening coefficient for the P(20)

line at elevated temperatures, and validated the WMS measurement technique at detonation-like

conditions in a shock tube.

Table 3.1: Spectroscopic line assignments and collisional-broadening parameters for the CO linesof interest. Uncertainties for broadening measurements made in this work (i.e. CO–N2) are shownin parentheses.

Collisional Broadening (γ in 10−3×cm−1/atm)

P(v”,J”) ν0 E′′ S(296K) CO-N2 CO-H2O CO-CO2(cm−1) (cm−1) (cm−2/atm)

γ(1000 K) n γ(300 K) n γ(300 K) n

22.5 0.55P(0,20) 2059.91 806.4 87.6 ×10−2(3%) (3%)

119 0.72 51.8 0.50

P(1,14) 2060.33 2543.1 26.4 ×10−5 26.8 0.61 125 0.77 63.1 0.53


Figure 3.7: Measured absorbance spectra of CO in N2 near 2060 cm−1 with active P-branch lineslabeled P(v”,J”) and fit with the Voigt function; T = 804 K, P = 1 atm, L = 20.95 cm, XCO = 0.005.

Shock tubes can produce well-defined environments for short time periods (∼ms) at combus-

tion temperatures above that attainable in a static optical cell (> 1000 K). Here we used a high-

pressure, stainless-steel shock tube to heat non-reactive CO-N2 gas mixtures to a range of tem-

peratures (1000–2700 K) representative of the detonation combustor in order to calibrate collisional

broadening of the P(20) line by N2. The shock tube has been described in previous works [9, 39, 40],

hence only limited detail is outlined here. Measurements are made approximately 1 cm from the

shock tube end wall with an optical path-length of 5 cm. Gas properties are assumed uniform in

accordance with a planar shock, and temperature and pressure are calculated from normal shock

relations based on measured initial conditions and the measured incident shock velocity. Typical

temperature uncertainty is ±1% for steady-state test times of approximately 1 ms behind the re-

flected shock wave.

Assuming known thermodynamic variables in addition to line-strength, a WMS-2f /1f measure-

ment at the P(20) line-center can be used to infer the CO-N2 collisional broadening coefficient,

2γCO−N2(T ), the remaining free Voigt parameter for a dilute mixture of carbon monoxide in nitro-

gen. The broadening coefficient was iterated within the CO spectral model to match the simulated

WMS-2f /1f values to the measured values. Broadening coefficients, as measured in the shock tube,


Figure 3.8: Sensitivity analysis of the WMS-2f /1f mole fraction measurement near 2059.91 cm−1

to the line-strength and collisional broadening parameters of the P(0,20) and P(1,14) CO lines at atypical PDC condition (see figure 3.7).

are plotted in figure 3.9 against temperature. The temperature dependence of collisional line broad-

ening is typically modeled as a power law (see equation 2.9 where n is the temperature-dependence

exponent). Previous power-law characterizations of nitrogen or air broadening of the P(20) line are

shown along with the measured data in the figure [20, 52, 54, 55]. The HITEMP model exhibits

excellent agreement with the data at the lower temperature domain (∼1100 K), but under-predicts

broadening at higher temperatures (> 2000 K), where the empirical model by Varghese shows best

agreement. The disagreements between models and data may be attributed to the general insuf-

ficiency of the power law to capture temperature dependence of broadening over a large range of

temperature [56, 57], and the differing temperature domains for which these models were estab-

lished. Here we best-fit the measured broadening coefficients over the domain of 1100 K to 2600 K,

yielding a more refined broadening model for combustion temperatures. The reference broadening

coefficient and temperature-dependence exponent for this refined 2γCO−N2(T ) model are shown in

table 3.1. The authors recommend caution employing these parameters outside of the temperature

range of the measurements. With the calibrated spectral model, CO mole fraction was inferred from

the measured WMS-data produced in the shock tube and compared to the known mole fraction of

the test gas mixture. Figure 3.10 depicts the typical uncertainty of the spectral model (σ ∼ 3%) over

a wide range of both temperatures and pressures (10–40 atm) that emulate detonation conditions,

validating the accuracy of the CO concentration sensor for the intended PDC application.


Figure 3.9: Measured nitrogen-broadening coefficient for the P(0,20) line from 1100 to 2600 K in-cluding a best-fit power law equation and comparison with broadening equations from other sources;P = 10–50 atm.

3.3.4 Optical engineering

To isolate sensitive laser equipment from the thermo-mechanically harsh engine environment,

a remote light delivery system was developed. Figure 3.11 depicts a simplified schematic of the

optical setup as deployed in field experiments. An indium-fluoride (InF3) single-mode fiber (L = 5

m) with a core diameter of 17 µm was coupled to the 4.85 µm DFB-QCL using an aspheric zinc-

selenide (ZnSe) micro-lens with a focal length of 6 mm. The fiber is free-space aligned with the lens

using a five-axis alignment stage and has a characteristic loss of∼0.5 dB/m at the target wavelength.

The single-mode fiber mitigates mode noise associated with larger core-diameter multi-mode fibers

and facilitates improved beam collimation at the fiber output [31]. Here we collimated the output

beam with a 12 mm focal length ZnSe aspheric lens to achieve a∼2 mm diameter beam at a working

distance of 40 cm. The collimation lens is integrated with an SMA connector to attach directly to

the fiber termination, and a four-axis kinematic mount interfaces with the test hardware to direct the

beam through the engine. The light collection and detection components are mounted together in a

compact linear cage system (Thorlabs-30 mm). The transmitted light is focused by an anti-reflection

coated calcium fluoride lens (focal length = 20 mm) onto an infrared photo-voltaic detector (Vigo

PVI-4TE-5) with a 2 mm2 detection area and specific detectivity (D∗) of∼ 3×1011cmHz1/2W−1).

The detector bandwidth is 10 MHz. Two narrow-bandpass spectral filters are stacked between the


Figure 3.10: Sensor validation data showing the measured CO mole fraction compared to the knownmole fraction over a range of conditions produced in a shock tube (σ ≈ 3%); L = 5 cm, XCO =0.00495 in N2.

engine mount and collection lens to yield an effective spectral width of∼25 nm centered about 4860

nm. Due to the sensitivity of the infrared detector from ∼2 to 5 µm, spectral filtering was critical

for reducing thermal emission from H2O which is typical of combustion exhaust. An analogous, but

more detailed, discussion of optical system design for this application can be found in section 2.3.3.

Single-modeInF3 Fiber

CollimationOptic

PV Detector Test Medium

DFB QCL (4854 nm)

CouplingLens

AlignmentStage

BP Filter

Figure 3.11: Simplified mid-infrared light delivery and detection schematic.

3.4. CO MEASUREMENTS BEHIND DETONATION WAVES 39

3.4 CO measurements behind detonation waves

Here we demonstrate the feasibility of the fiber-coupled CO absorption sensor on an operat-

ing pulse detonation combustor, highlighting measurement time-resolution and sensor operability

range. Figure 3.12 exhibits representative measurements of pressure and WMS-2f /1f for two con-

secutive detonation cycles in the PDC chamber. Detonation waves were produced by ethylene-air

gas mixtures, spark-ignited every 50 ms (20 Hz), and the primary detonation event transpired within

just a few milliseconds as indicated by the pressure time history. Utilizing simultaneous H2O ther-

mometry along with pressure data, CO mole fraction was directly inferred from the WMS-2f /1f

measurement [58].

Figure 3.12: Representative time-history data from a pulsed (20 Hz) detonation combustor operatingon ethylene/air; L = 7.62 cm; pressure and WMS-2f/2f measurements are coplanar in combustionchamber

High-fidelity CO data was resolved over the full detonation cycle and entire range of operating

conditions. To reject noise, a 20 kHz low-pass filter was applied to the lock-in amplifier outputs (2f

and 1f signals). Such a wide filter, relative to the modulation frequency (50 kHz) was feasible due to

the strong harmonic signals associated with the probed CO spectra in the PDC environment. Tighter

filters offered minimal further noise reduction, while having the expense of lower measurement


bandwidth. The time-variable signal-to-noise ratio (SNR), as noted in the WMS-2f /1f time-history

in figure 3.12, resulted from the dramatically changing spectral structure of carbon monoxide during

a detonation event. Directly behind the detonation wave, at the most extreme conditions (P > 30

atm, T > 2500 K), a detection limit of approximately 0.5% carbon monoxide was inferred from an

SNR of ∼5 and 2.5% measured CO. In the more benign tail at later times of the detonation cycle

(P∼10 atm, T∼1200 K), a detection limit of ∼500 ppm was achieved (SNR∼10, XCO ∼ 0.5%).

The temporal resolution or measurement bandwidth is effectively equated to the width of the low-

pass filter on the harmonic signals, or 20 kHz. Other environments may require different frequency

filtering schemes to minimize noise, thus yielding different bandwidths.

CouplingLens

Fiber-CouplingAlignment Stage

Multi-modeZBLAN Fiber

Single-modeInF3 Fiber

Single-modeZBLAN Fiber

CollimationOptic

PV Detector for CO or CO2

InGaAs Detector for H2O

Test Medium

CO(4854 nm)

CO2(2678 nm)

H2O(2482 nm)

H2O(2474 nm)

Beam Splitter

Fibe

r Del

iver

y to

Eng

ine

Figure 3.13: Optical configuration for free-space fiber-coupling (left) and cross-section of pulse

detonation combustor showing remote light delivery and collection (right).

3.5 Multi-species measurements in a PDE

While the previous two sections outline sensor design and development for CO2 and CO sep-

arately, here the sensors are shown in combination as applied for assessment of a pulse detonation

engine (PDE). Detonation-based engines represent a challenging application for diagnostics due to

the wide range of thermodynamic conditions involved (T∼500-3000 K, P∼2-60 atm) and the short

time scales of change (∼ 10−6 to 10−4 sec) associated with such systems. The multi-species laser

absorption system was utilized on an operating PDE at the Naval Postgraduate School (Monterey,

3.5. MULTI-SPECIES MEASUREMENTS IN A PDE 41

CA USA). Figure 3.13 provides a graphical schematic of the optical interface with the test engine,

where co-planar measurements of temperature, pressure, and species mole fraction were made on

the combustion chamber. Figure 3.14 shows representative data from two consecutive detonations

produced in an ethylene-air mixture. Water concentration and temperature measurements are pro-

vided by a similar laser absorption sensor designed by Goldenstein et al [58].

The CO sensor is shown to resolve the most extreme initial detonation event (P∼50 atm,

T∼3000 K) during the first few milliseconds (SNR∼20), while quality data (SNR > 2) proved more

difficult to achieve with the CO2 sensor at early times, as expected, exacerbated by the low levels

of CO2 during this period resulting from incomplete combustion. SNR of all sensors improved (>

20) at later times in the detonation cycle when the pressure (< 10 atm) and temperature (< 2000

K) dropped in the combustor. An important aspect of the carbon oxide PDE measurements is

the prevailing inverse relationship of CO and CO2 mole fraction. Although water mole fraction

approaches equilibrium levels very quickly, the CO→CO2 oxidation appears much slower. As

such, a quantitative measure of the CO to CO2 ratio provides a sensitive measure of combustion

progress. In combination with other thermodynamic property measurements, these fundamental

time-resolved flow-field parameters can be used to determine important engine performance metrics

such as mass flow rate, enthalpy flux, and combustion completion.

0

20

40

60

0.00

0.05

0.0

0.1

0 20 60 800.0

0.1

P

P [a

tm]

T [K

/50]

XC

O

T

XH

2OX

CO

2

Time [ms]

Figure 3.14: Time-resolved species mole fraction data from a pulse detonation combustor operating

on ethylene-air at 20 Hz.


3.6 Summary of detonation engine research

As described in the last two chapters, mid-infrared laser absorption sensors were developed for

measuring carbon oxide mole fractions in detonation environments. Using rapidly tunable semi-

conductor lasers, a wavelength modulation spectroscopy technique was employed to reject thermo-

mechanical noise sources, and single-mode mid-infrared fibers were free-space coupled to remotely

deliver the light. A high-pressure shock tube was utilized to validate the range and accuracy of

each sensor, while sensor bandwidth (∼10 kHz) and practicality was demonstrated on an operat-

ing pulse detonation combustor. More broadly, this work extends the pressure capabilities (up to 50

atm) of laser absorption sensing for carbon oxides in harsh combustion environments, with potential

application to many high-pressure reacting flow fields.

Chapter 4

Ultra-sensitive CO2 diagnostic forkinetic studies

The contents of this chapter have been published in the journal Applied Physics B under the full title

’Multi-band infrared CO2 absorption sensor for sensitive temperature and species measurements in

high-temperature gases’ [59]. Portions of the chapter’s content have also been published in the

Journal of Physical Chemistry [60] and presented at the 8th National Combustion Meeting [61].

4.1 Introduction

Accurate determination of temperature and species concentration in high-temperature reacting

gas flows provides a basis for characterizing reaction pathways and chemical kinetic mechanisms.

Laser absorption diagnostics have been utilized extensively to measure gas temperature and species

concentration in various high-temperature gas systems in both laboratory and industrial settings

[3, 62, 63]. Non-intrusive, continuous-wave (CW) laser sensors provide the capability necessary

for resolving the short time scales involved in chemistry and high-speed flow fields. This chap-

ter presents a flexible, highly-sensitive temperature and CO2 concentration sensor based on mid-

infrared absorption using two CW lasers, aimed at high-temperature gas environments.

Carbon dioxide (CO2) is a major product of hydrocarbon combustion, a component in ambient

air, and yet is relatively inert, making the molecule a viable target for absorption measurements in

many high-temperature gas flows where it may be present naturally or can be added. Figure 4.1

shows the infrared absorption spectra of CO2 from 1–5 µm in wavelength, plotted as line-strengths

at 1000 K. Absorption-sensing technology in the 1.3–2.1 µm wavelength region is relatively mature

43

44 CHAPTER 4. ULTRA-SENSITIVE CO2 DIAGNOSTIC FOR KINETIC STUDIES

due to the affordability and availability of optical components supporting the telecommunications

industry. Therefore many laser-based CO2 absorption sensors have targeted the spectral combina-

tion bands near 1.5 µm (2ν1 +2ν2 +ν3) [16, 17] and 2.0 µm (ν1 +2ν2 +ν3) [18, 19, 37]. However,

these absorption bands are relatively weak at elevated temperatures (T > 800 K), and thus have lim-

ited range and applicability in high-temperature gas systems wherein carbon dioxide is not a major

component. More recently, progress in diode laser technology has enabled spectroscopic access to

the stronger vibrational band near 2.7 µm (ν1 + ν3), and researchers have developed temperature

and species concentration diagnostics for shock tubes and pulse detonation engines by probing mul-

tiple CO2 absorption transitions within this band [9, 34, 64], including that which was described in

chapter 2. The fundamental CO2 vibrational band near 4.3 µm (ν3) has approximately 70 and 3000

times stronger absorption, respectively, than the combination bands at 2.7 µm and 2.0 µm. This fun-

damental band offers potential for much more sensitive detection than established CO2 absorption

sensors, though a lack of available room-temperature tunable lasers at wavelengths of interest in this

band has mitigated such advancement for highly-quantitative spectroscopy at elevated gas temper-

atures. Here we have implemented an existing 2.7 µm diode laser with a newly developed tunable

external-cavity quantum cascade laser (ECQCL) that can access almost the entire ν3 fundamen-

tal band of CO2 (4.13–4.46 µm). The latter laser was utilized to advance spectroscopic strategies

which enable improved sensitivity for temperature and CO2 species concentration measurements in

high-temperature gases.

Figure 4.1: Absorption line-strengths of CO2 at 1000 K (HITEMP)

In the present work, mid-infrared CO2 sensing strategies were developed with an aim towards

4.2. SPECTROSCOPIC FRAMEWORK 45

application in shock tube kinetics studies. The objectives of this research are twofold: (1) develop

a highly sensitive (< 100 ppm) CO2 concentration diagnostic that is relatively insensitive to tem-

perature changes common to reactive (pyrolysis/oxidation) experiments, and (2) develop a sensitive

temperature diagnostic based on mid-infrared CO2 absorption which may be used for shock tube

experiments with nascent or seeded CO2. Two distinct strategies are presented to accomplish the

aforementioned objectives, highlighting the flexibility of the broadly tunable external-cavity QCL.

The sensor is validated and refined at high temperatures (700-1800 K) in shock-heated CO2-argon

mixtures, and demonstrated in various pyrolysis and oxidation studies.

4.2 Spectroscopic framework

Absorption spectroscopy theory is briefly recalled to clarify notation. Spectral absorbance is

related to gas properties, including temperature T [K] and species mole fraction xabs, by

αν = −ln (It/I0)ν = PxabsSi(T )φ(ν)iL, (4.1)

where Si(T ) [cm−2atm−1] is the line-strength of a quantum transition i which varies only with

temperature, φ(ν)i [cm] is the line-shape function and L [cm] is the path-length. It follows that gas

temperature can be directly inferred from the ratio of spectral absorbance at two different wave-

lengths, as expressed in equation 4.2:(αν1αν1

)=SA(T )φ(ν1)ASB(T )φ(ν1)B

. (4.2)

The absorbance ratio simplifies to the ratio of the respective line-strengths and line-shape functions

for each selected transition (A or B). With similar line-shape functions, temperature sensitivity of

the absorbance ratio is primarily driven by the line-strength ratio of the selected line-pair and can

be approximated as ∣∣∣∣dR/RdT/T

∣∣∣∣ ≈ (hck)|EA′′ − EB ′′|

T, (4.3)

where h [J-s] is Planck’s constant, c [cm/s] is the speed of light, k [J/K] is Boltzmann’s constant, and

E′′ [cm−1] is the lower state energy for the target absorption lines [65]. From equation 4.3, we note

that the difference in lower state energies between lines largely governs the temperature sensitivity

for a two-line absorption technique and serves as a guide for line selection. Additional spectral

parameters that affect the temperature dependence of the absorbance ratio include the difference in


center wavelengths and line-shapes between selected lines.

The spectral line-shape functions were modeled here using the Voigt profile, which captures both

collisional-broadening and temperature-broadening effects, characterized by a collisional width ∆νc

[cm−1] and Doppler width ∆νd [cm−1] respectively. Though the Doppler width scales simply with

line-center wavelength and gas temperature, the collisional width depends on gas composition, with

unique line-broadening effects for each molecular collision partner per

∆νc = P∑j

xabs2γj−abs. (4.4)

where P [atm] is total pressure and γj−abs [cm−1atm−1] is the broadening coefficient between per-

turbing species j and the absorbing species. The temperature dependence of collisional broadening

is often modeled as a power law,

γj(T ) = γj(T0)

(T0T

)n(4.5)

where n is the temperature-dependence exponent.

Two-wavelength thermometry is a common spectroscopic method for quantifying gas tempera-

ture based on laser absorption of a single species [50, 66, 67]. With temperature either measured or

known, species concentration can readily be determined per equation 4.1.

4.3 Line selection

Primary factors influencing line selection include the strength, isolation, and temperature sen-

sitivity of accessible absorption transitions. The fundamental CO2 asymmetric-stretch (ν3) band

centered near 4.26 µm (2350 cm−1) consists of many strong rovibrational absorption lines. This

ν3 band is also generally well isolated from interfering absorption when amongst other combus-

tion products and ambient species including water and carbon monoxide. However, due to the

great strength of the fundamental band, ambient CO2 (∼400 ppm) absorbs significantly at typical

path-lengths (∼1 m) associated with bench-top optical setups, creating issues of self-interference.

Figure 4.2 shows absorbance simulations of ambient (300 K) and high-temperature (1000 K) CO2 in

the wavelength domain of the ν3 band. As illustrated, even at a relatively short path-length (15 cm),

ambient CO2 can be optically thick for a large portion of the fundamental band, making utilization

of this portion of the band difficult and impractical for many applications. At high temperatures,

4.3. LINE SELECTION 47

the Boltzmann distribution forces a more dispersed population of molecules amongst energy states,

increasing the relative strength of absorption transitions associated with high lower-state energies.

This makes the far wings of the fundamental band at frequencies below 2300 cm−1 and above 2380

cm−1 attractive wavelength regions for high-temperature applications [36].

Figure 4.2: Absorbance simulations for the CO2 ν3 band at T = 300 K and T = 1000 K; P = 1 atm,XCO2 = 400 ppm, L = 15 cm

The far wings of the fundamental ν3 band contain many candidate lines with attractive

characteristics for quantitative laser absorption measurements in high-temperature environments.

The HITEMP database [20] provides a comprehensive set of CO2 line parameters including line-

strengths, lower-state energies, and broadening coefficients which facilitate a detailed assessment

and comparison of individual lines via simulated spectra at expected test conditions. An FTIR

spectrometer (∼0.1 cm−1 resolution) was also utilized in conjunction with a heated static cell to

provide empirical validation for HITEMP simulations and assist in line selection. Details of the

experimental setup involving the FTIR spectrometer and static cell can be found in a previous

work [13]. Figure 4.3 shows measured CO2 spectra in the far wings of the fundamental band and

comparison with HITEMP simulations at elevated temperature (700 K). The HITEMP simulation

exhibits excellent agreement with the FTIR measurement, providing confidence in the tabulated


line-strengths and air-broadening parameters. The figure also shows the interference from ambient

CO2 (400 ppm) at a one meter path-length and room temperature. Comparison of the respective

spectra in the far wings of the P-branch and R-branch reveals a substantial difference in relative

line isolation between the two domains. Individual spectral lines in the P-branch have extensive

interference with neighboring CO2 lines at elevated temperatures. This crowded P-branch spectrum

is due to an abundance of active transitions comprising the fundamental hot bands, including

ν(001→002), ν(002→003) and so on, where each progressive band center shifts to a lower

wavenumber due to the decreasing spacing of vibrational energy levels. Thereby the lines in the

far wing of the R-branch are, in contrast, relatively well-isolated, marking the spectral domain

around 2381–2397 cm−1 ideal for sensitive laser-based absorption sensing of CO2 at elevated

temperatures. Selection of specific lines within the targeted 4.17–4.2 µm window of the ν3

R-branch requires further consideration of the test conditions of interest and optical path-length of

the shock tube.

As outlined, the first objective of this work is to provide highly-sensitive detection (< 100 ppm)

of CO2 for temperature-insensitive species concentration measurements in a shock tube. With test

temperature known from the ideal shock relations, a single-line measurement approach can be taken.

Figure 4.4 shows the mole fraction of CO2 for 1% line-center absorbance (with argon bath gas) at a

typical test condition for each of 19 candidate lines in the R-branch labeled by their respective lower-

state rotational quantum number (J ′′). Prior experience suggests that absorbance greater than 1%

provides sufficient signal for quantitative detection (SNR > 10) with an ECQCL based on a ∼ 0.1%

absorbance detection limit [68]. Also plotted on the right axis is the temperature sensitivity of line-

center absorbance for each transition. We see that temperature sensitivity decreases with J ′′ as the

mole fraction for 1% absorbance increases. To provide a requisite sub-100 ppm detectability while

minimizing temperature sensitivity, the R(76) line near 2390.52 cm−1 is selected to compromise

these opposing demands.

Sensitive two-line thermometry represents our second objective, and for these experiments CO2

may be seeded in the shock tube test gas at modest concentrations (1–2%) without meaningfully

perturbing reaction chemistry. To achieve maximum temperature sensitivity, we primarily aim to

select two lines with a large difference in lower state energies (∆E′′) as defined by equation 4.3.

However, a large difference in lower-state energies is generally associated with a large difference

in line-strengths (∆S) and thus measured absorbance values when the two lines are selected from

within the same band. A large difference in absorbance values is non-ideal due to limitations in

4.3. LINE SELECTION 49

Figure 4.3: Measured and simulated spectra in the far wings of the (top) P-branch and (bottom)R-branch of the CO2 fundamental ν3 band; T = 700 K, P = 2 atm, L = 20.95 cm, XCO2 = 1.5% inair; Simulated interference from ambient CO2 (400 ppm) at L = 1 m, T = 300 K also shown.


Figure 4.4: Estimated detectability and temperature sensitivity of CO2 using fixed-wavelength DAat line-center for candidate ν3 R-branch lines; T = 1400 K, P = 1.5 atm, L = 14 cm

dynamic range, bound by the optically thin and thick limits. The result is a tradeoff between tem-

perature sensitivity and range when using conventional intra-band two-line thermometry. In this

work, we present a cross-band two-line CO2 thermometry technique to both increase temperature

sensitivity and expand the temperature range of utility compared to previous intra-band sensors. We

select two lines from different vibrational bands of CO2 (ν3 and ν1 + ν3), taking advantage of the

sizable difference in band strengths to select a line-pair with a large ∆E′′ and a relatively small ∆S.

Relaxing the detectability requirement (100 ppm→1%) for CO2 thermometry allows us to select

a line within the ν3 R-branch that has a much higher E′′ than the line selected to meet our first

objective. Moreover, because low E′′ transitions within the ν3 band are very strong, selection of a

line with a lower-state energy less than 1000 cm−1 leads to issues of self-interference with ambient

CO2 and a problematically large line-strength/absorbance ratio, R(1000 K) > 25, when paired to

yield ∆E′′ greater than 2000 cm−1. In combination, these intra-band issues constrain both range

and sensitivity for two-line thermometry. To address such difficulties, we look to the weaker ν1+ν3

band near 2.7 µm to choose the low E′′ line. The R(28) line near 3633.08 cm−1 is selected from

the ν1 + ν3 band to serve as the low E′′ line due to relative strength and isolation. This line has

been utilized previously for intra-band thermometry at 2.7 µm in similar applications [33, 34, 64].

We pair this line with the R(96) transition near 2395.13 cm−1 in the ν3 band to yield a ∆E′′ of

3305 cm−1. Figures 4.5 and 4.6 illustrate the expected temperature range and sensitivity of the new


cross-band thermometry sensor, with comparison to the previous intra-band sensor at 2.7 µm [34].

Table 4.1 lists relevant spectral parameters associated with each line used in this work.

Figure 4.5: Line-strengths vs. temperature for transitions utilized in two-line CO2 thermometry(past and present)


4.4.1 Light source selection

Two mid-infrared light sources are utilized for this sensor to access the three CO2 lines of

interest. Access to the R(76) and R(96) transitions near 4.2 µm is attained with a continuous-

wave, single-mode ECQCL from Daylight Solutions. The quantum cascade laser is broadly tunable

(2240–2420 cm−1) over the CO2 ν3 band using a motorized grating (∼Hz) and more precisely

tuned over a small range by a piezoelectric driver (∼100 Hz). Figure 4.7 shows an example piezo

scanned-wavelength measurement of the R(76) line using the ECQCL. Output power ranges from

30–100 mW depending on wavelength, with a typical power of 40 mW at 4.2 µm. The ECQC laser

linewidth is 5 MHz with a wavelength stability of ±0.002 cm−1 in a fixed-wavelength mode. The

R(28) transition of the ν1 + ν3 band is probed with a DFB diode laser from Nanoplus centered near

2.75 µm with an output power of ∼5 mW. The diode laser linewidth is ∼3 MHz with wavelength

stability of ±0.001 cm−1 and a tunability range of ±3 cm−1.


Figure 4.6: Line-strength ratio and temperature sensitivity for cross-band line-pair compared toprevious intra-band line-pair (Farooq et al. 2008)

4.4.2 Spectral modeling

Accurate fixed-wavelength direct-absorption (fixed-DA) measurements require a comprehen-

sive spectral model. The HITEMP database combined with the Voigt line-shape function has been

demonstrated for very accurate modeling of the infrared CO2 spectra in air, as shown in figure 4.3

and documented elsewhere [9, 21, 45]. However, shock tube kinetics studies are commonly carried

out with argon as the bath gas, for which collisional-broadening and collisional-shifting parameters

do not exist in the HITEMP database. For the R(28) line of the ν1 + ν3 band, these parameters

have been measured by previous researchers [34]. To account for argon broadening and shifting

parameters of the R(76) and R(96) lines of the ν3 band, static cell and non-reactive shock tube

measurements were made of CO2 dilute in argon in order to adjust our spectral model. First, mea-

surements in a heated static cell confirmed that pressure-shift in argon is small and negligibly dif-

ferent than pressure-shift in air over the conditions of interest. Second, in addition to the relatively

low-resolution FTIR spectrometer measurements, the ECQCL was used to validate HITEMP line-

strength values for spectral lines near 4.2 µm using a scanned-wavelength (100 Hz) direct absorption

method. In these experiments, a Voigt line-shape function was fit to the experimental data, as shown

in figure 4.7, from which an integrated absorbance area (Ai) could be extracted and related directly


Figure 4.7: Scanned-wavelength direct absorption measurement of the R(76) CO2 transition usinga Daylight Solutions ECQCL (100 Hz); T = 550 K, P = 0.81 atm, L = 20.95 cm, XCO2 = 0.015 inair

to line-strength by: Ai = PxabsSi(T )L . Measured line-strength values agreed with HITEMP tab-

ulated values within measurement uncertainty (< 2%). Using the Voigt function, the only remaining

parameter necessary to fully characterize the spectral line-shape is the argon-broadening coefficient.

Here we inferred argon-broadening coefficients at temperatures of interest (700-1800 K) by

making fixed-wavelength measurements at the absorbance peak of each line during non-reactive

shocks of CO2 dilute in argon. Experiments in the shock tube provide well-defined environments

for short time periods (∼ms) at temperatures not attainable in a static cell (> 1000 K). Figure 4.8

shows an example time-history of a non-reactive shock tube experiment assessing the R(76) line

near 2390.52 cm−1 at 1540 K. Based on the measured pressure and dilute nature of the gas mixture,

the broadening coefficient at a given temperature is effectively solved by equation 4.4, where peak

absorbance and collisional width are uniquely coupled in the spectral model. Such experiments are

critical for understanding the temperature dependence of collisional broadening, which is not well

defined by equation 4.5 over a large temperature range, and thus cannot be reliably inferred from

low temperature measurements alone [57, 69]. Measured broadening coefficients for the R(76) and

R(96) lines of the ν3 band are shown in Figure 4.9 over a range of temperatures (800–1900 K) and


each fit with a respective power law function. These measurements are consistent with other CO2-

Ar broadening measurements in an FTIR (550–800 K) by Thibault et al. [70]. Caution should be

exercised when extending the respective power law functions based on this data to a wider range of

temperature. Collisional-broadening parameters for all three lines used in this work, as well as other

relevant spectroscopic modeling parameters associated with the CO2 lines, are given in table 4.1.

Further detail on the experimental setup is discussed in the subsequent section.

Figure 4.8: Peak absorbance time-history for the R(76) line near 2390.52 cm−1 during a non-reactive shock tube experiment of CO2 dilute in argon; XCO2 = 0.005, L = 14.1 cm, T5 = 1540K.

Table 4.1: Spectroscopic line assignments and modeling parameters for the CO2 lines of interest.Parameters taken from HITEMP 2010 except where measured (γAr, nAr).

Line ν0 E′′ Band S(296K) γAr(1000 K) nArγself (296 K)

[cm−1] [cm−1] [cm−2/atm] [10−3×cm−1/atm] [10−3×cm−1/atm]

R(76) 2390.52 2278 ν3 10.1×10−3 18.3±0.6 0.60±0.02 65

R(96) 2395.14 3622 ν3 18.5×10−6 16.4±0.4 0.48±0.02 63

R(28) 3633.08 317 ν1 + ν3 59.5×10−2 26.7±0.7 0.51±0.02 93


Figure 4.9: Argon-broadening coefficients for the R(76) and R(96) CO2 lines.

4.4.3 Optical setup

Figure 4.10 shows a representative optical setup for the current shock tube experiments. The

lasers were used both independently (for spectral validation and ultra-sensitive species detection),

and in combination as shown (for thermometry). The shock tube has a 14.13 cm internal path-length

and barium-fluoride windows located approximately 2 cm from the end wall. The incident beam

from the quantum cascade laser (4.2 µm) is split with an anti-reflection coated calcium-fluoride

beam splitter to provide a reference signal for common-mode intensity noise rejection. The diode

laser (2.7 µm) exhibits intensity noise of less than 0.2% of absolute signal and the ECQCL can

achieve similar intensity stability after common-mode rejection. The beams from both lasers (QCL

and diode) are collimated and spatially conditioned with an iris to be ∼2 mm in diameter prior to

entering the test section at co-planar angles. Each laser beam is spectrally filtered with a bandpass

filter (∼60 nm) after passing through the test section, and focused onto a photovoltaic detector by

a calcium-fluoride plano-convex lens (20 mm f.l.). The photovoltaic detectors (Vigo PVI-X) are

aligned collinear with the beam path and are thermo-electrically cooled, having a 2 mm detection

area and 10 MHz bandwidth. Data from the detectors is collected at a sample rate of 1 MHz.

Between tests, a wavelength meter (Bristol 621B) is used to verify that each laser is centered at line-

center. More detailed information on the shock tube utilized for this work can be found in previous

publications [71, 72].


Test Medium

2.75 μmDFB Diode

BeamSplitter

4.18 μmECQCL

BP Filter

CaF2 Lens

PVDetector

Iris

Figure 4.10: Optical setup for shock tube experiments.

4.5 Sensitive CO2 detection near 4.2 micron

The first objective of this work was to provide the ability for ultra-sensitive (< 100 ppm) species

concentration measurements of CO2 with minimal temperature dependence. This strategy is tar-

geted for utilization in multi-species measurement campaigns in shock tubes to inform and improve

the modeling of chemical kinetic mechanisms. In previous research, limited resolution in CO2 con-

centration measurements (∼300 ppm) based on absorption at 2.7 µm required the utilization of

greater initial fuel concentrations (2-3%) than desired [73], which leads to larger changes in tem-

perature during the experiment that can meaningfully alter the kinetic interpretation. Additionally,

to achieve the aforementioned resolution in the 2.7 µm band, the relatively strong R(28) line was

used, which has a larger temperature sensitivity than lines at higher rotational quantum numbers.

To improve on these aspects, we probed the R(76) transition near 4.2 µm to achieve greater abso-

lute sensitivity and lower temperature dependence as discussed in section 4.3. Figure 4.11 shows

example species time-histories during oxidation of methyl butyrate (MB) at several temperatures to

exhibit the high fidelity data producible with the new sensor. The sensor demonstrates the ability

to resolve complex kinetic mechanisms on multiple time scales in a self-consistent manner at fuel

concentrations more than ten times lower than previous experiments. Figure 4.12 highlights the im-

proved detection limit of the new sensing strategy and the ability to resolve kinetic rate information

at early times, critical for anchoring chemical mechanisms [74]. Based on an observed signal-to-

noise ratio (SNR) of ∼10 at 50 ppm, the detection limit of the current sensor is approximately 5

4.6. CROSS-BAND CO2 THERMOMETRY 57

ppm with 1 MHz bandwidth at typical shock tube conditions. Results of the first kinetic studies

utilizing this new sensor were presented in a separate paper [61].

Figure 4.11: Measured CO2 species time-histories during methyl butyrate oxidation (φ ≈ 1).

4.6 Cross-band CO2 thermometry

The sensitive cross-band thermometry technique, utilizing the R(96) and R(28) lines, was vali-

dated in the same shock tube as discussed previously. Non-reactive CO2-Ar shocks were performed

over a range of temperatures (600–1800 K) and pressures (1.2–2.5 atm) to evaluate the accuracy of

the spectral model and the precision of the fixed-wavelength sensor. An example time-history of the

shock tube validation is shown in figure 4.13, wherein the strong inverse temperature dependence

can be observed between the two wavelengths as the experiment proceeds in time, and increases in

temperature. This strong inverse dependence yields the high level of temperature sensitivity in the

ratio of absorbances. With similar line-strengths, the concentration of CO2 in the bath gas can be

readily tailored to optimize the absorbance magnitudes between the optically thin and thick limits

(α ≈0.1–1.5) over a wide range of temperatures. As seen in the figure, both wavelengths provide

favorable absorbance levels with 2% CO2 seeded in the bath gas.

The ratio of absorbances was then converted to temperature, as depicted in figure 4.14, based on

the respective spectral models for each wavelength, using the measured pressure and known mole


Figure 4.12: Pressure and CO2 species time-histories during methyl-butyrate pyrolysis at 1259 Kwith comparison to the kinetic model of Huynh et al.

fraction as inputs. For each shock, temperature was also calculated from the measured incident

shock velocity using the ideal shock relations and known initial conditions in the shock tube. Typi-

cal uncertainty in temperature using this calculated method is approximately 0.6% for non-reactive

shocks, and treated as the known temperature here [75]. The known temperatures (pink) for region

2 (between incident and reflected shock) and region 5 (after reflected shock) are shown overlaid

with the measured temperature from cross-band CO2 thermometry, with excellent agreement dis-

played between the two methods. Looking more closely at the measured temperature behind the

reflected shock reveals the precision that can be achieved using the cross-band technique (±5 K or

less than 0.5%) over the test time (∼1 ms) of the non-reactive shock. Figure 4.15 provides an as-

sessment of the accuracy of the cross-band technique, comparing the measured temperatures to the

known temperatures from the ideal shock relations. Over the full range of conditions, the measured

temperature agreed with the known temperature within 1%, and the standard error was∼0.5%. Pre-

cision decreased with temperature such that noise in the temperature measurement reached ±2% at

the high end of the temperature range compared to a typical 0.5% as illustrated in figure 4.14.

4.7. KINETICS SUMMARY 59

Figure 4.13: Example peak absorbance time-histories of the R(96) and R(28) lines in a non-reactiveshock tube experiment.

4.7 Kinetics Summary

Accurate and precise determination of temperature and species concentration is important to

the assessment of reacting gas systems. In this work, we developed a highly sensitive laser-based

absorption diagnostic that probes the infrared spectra of carbon dioxide to measure gas temperature

and CO2 concentration at time scales (µs) and temperatures (600-1800 K) relevant for shock tube

kinetics studies. An ECQCL was utilized to exploit the far R-branch wing of the fundamental ν3band of CO2 near 4.2 µm, which is well isolated from potential interfering absorption. The R(76)

line was selected for sensitive (< 100 ppm) species detection with relatively weak temperature de-

pendence. Oxidation and pyrolysis experiments were performed on various methyl esters, wherein

the sensor exhibited a detection limit of better than 5 ppm at conditions of interest for combustion ki-

netics. This mid-infrared (4.2 µm) CO2 concentration sensing represents an improvement of greater

than 50 times the sensitivity achieved previously near 2.7 µm. A cross-band two-line thermometry

technique was also established using the R(96) line of the ν3 fundamental band and the R(28) line of

the ν1 + ν3 combination band. The cross-band spectroscopic approach facilitates high temperature

sensitivity (∆E′′ > 3000 cm−1) with similar ranges of absorbance between the two lines, which is


Figure 4.14: Measured temperature based on cross-band thermometry for a non-reactive CO2-Arshock (same as Fig. 4.13) with comparison to calculations from the ideal shock relations.

desirable in order to maximize the useful range of the sensor. Non-reactive CO2 in argon shocks

were carried out to validate the accuracy (∼0.5%), precision (0.2–2%), and range (600–1800 K)

of the cross-band thermometry technique. Compared to previous intra-band fixed-DA thermometry

methods, this cross-band method offers more than twice the sensitivity along with expanded useful

range to higher temperatures.

4.7.1 Potential sensor improvements

While the current sensor substantially improves upon prior CO2 absorption sensing techniques,

future work is recommended to further optimize the sensor and expand its applicability. Normalized

wavelength modulation techniques (WMS-nf /mf ) can be employed to desensitize measurements to

non-absorption light intensity fluctuations such as emission, beam steering, and vibrations that may

limit precision, especially at the high end of the temperature range. WMS methods can also desen-

sitize measurements to spectrally broad interfering absorption, which may be present with various

fuels. Additionally, the optical engineering of the current sensor may be improved to create a more

field-deployable diagnostic. Specifically, the two beams for thermometry may be combined onto a

single optical path, using a beam splitter or bifurcated optical fiber, to minimize the optical foot-

print of the sensor and mitigate the influence of spatial non-uniformities in the test gas, which are

more pronounced in engine applications. From a spectroscopic standpoint, the ECQCL may also

4.7. KINETICS SUMMARY 61

Figure 4.15: Comparison of cross-band thermometry with known temperature in the shock tube.

be utilized to probe any of the thirty well-isolated CO2 transitions in the ν3 R-branch between 2381

cm−1 and 2397 cm−1, which may be desirable for certain applications with slightly different test

conditions, detection requirements, and path-lengths. Here we have demonstrated two distinct appli-

cations, for which the R(76) and R(96) lines were used, but the current diagnostic offers additional

flexibility in line selection, and can be optimized for a broad range of high-temperature applications.

Chapter 5

Spatially-resolved CO and CO2 sensingfor scramjets

The contents of this chapter have been submitted and accepted to the journal Applied Physics B

under the full title ’Simultaneous sensing of temperature, CO, and CO2 in a scramjet combustor

using quantum cascade laser absorption spectroscopy’ [76]. Portion of this chapter’s content have

also been presented at the 52nd Aerospace Sciences Meeting (SciTech2014) [77] and submitted to

the Journal of Propulsion and Power [78].

5.1 Introduction

Advancements in air-breathing supersonic propulsion systems have led to the need for a new

generation of diagnostics to characterize the flow fields produced in these devices. While exten-

sive analytical research has been dedicated to hydrogen-fueled supersonic combustors, systems that

utilize more practical hydrocarbon fuels have received growing interest of late [79]. Previous laser-

based absorption diagnostics, typically probing water vapor, have proven valuable for assessing

supersonic reacting gas systems [80–83]. Here we extend absorption sensing in scramjets to the

other major hydrocarbon combustion species, CO and CO2. Previous measurements of CO2 have

been reported by Rieker et al. in a vitiated scramjet, though limited to low temperatures by relatively

weak near-infrared absorption above 600 K [25]. This paper describes the development of a new

mid-infrared quantum cascade laser absorption spectroscopy (QCLAS) sensor capable of provid-

ing non-intrusive, in situ measurements of temperature and carbon oxide species concentrations in

high-speed hydrocarbon combustion flows. The sensor is initially targeted for scramjet combustor

62

5.2. METHODS 63

environments with further applicability to other high-temperature, short path-length (< 5 cm) gas

systems.

A scanned-wavelength direct absorption technique was primarily utilized, targeting rovibra-

tional carbon monoxide (CO) and carbon dioxide (CO2) transitions, near 4854 nm and 4176 nm

respectively. Temperature measurements were attained from the ratio of two CO transitions, and the

thermometry range (∼800–2400 K) was enhanced in harsh environments by calibration-free second

harmonic detection using a wavelength modulation spectroscopy (WMS) technique. Two quantum

cascade lasers are used in the sensor. The respective outputs of each laser were combined by free-

space coupling to a bifurcated hollow-core fiber for remote light delivery. The fiber-coupled sensor

was utilized to simultaneously measure temperature and species mole fractions in ethylene-air com-

bustion conditions at multiple planes in the flow of a direct-connect model scramjet combustor at

the University of Virginia. The diagnostic provides enhanced sensitivity to hydrocarbon combustion

progress compared to analogous absorption sensing of water vapor.

5.2 Methods

5.2.1 Line Selection

Recent availability of tunable, room-temperature quantum cascade lasers enable access to the

strong mid-infrared absorption bands of CO and CO2 centered near 4.6 µm and 4.3 µm, respec-

tively [47]. These bands offer orders of magnitude greater absorption and more sensitive detection

compared to the next strongest near-infrared bands for each species. Wavelength selection within

each band is primarily influenced by the isolation, strength, and temperature sensitivity of the dis-

crete rovibrational lines that compose the bands, as well as laser availability. Figure 5.1 shows the

fundamental absorption bands for CO and CO2 plotted as line-strengths at 2000 K [20]. At com-

bustion temperatures, we note the P-branch of CO2 and the R-branch of CO interfere spectrally, and

this overlapping domain was avoided for line selection. Two neighboring lines are chosen from the

more isolated P-branch of the fundamental carbon monoxide band near 2059.9 cm−1 and 2060.3

cm−1 to infer temperature and CO mole fraction. The proximity of these wavelengths (∼0.4 cm−1)

allows for a two-line measurement with a single laser. Though not shown in the figure, water vapor

interference further constrained line selection for CO, and the chosen line-pair exhibits excellent

isolation relative to other candidate transitions [41]. Carbon dioxide mole fraction is measured by

probing a temperature-insensitive R-branch transition in the fundamental band near 2394.4 cm−1.

64 CHAPTER 5. SPATIALLY-RESOLVED CO AND CO2 SENSING FOR SCRAMJETS

The selected CO2 line is near the fundamental bandhead, which involves tight line-spacing as evi-

denced by the dense (solid red) spectra in figure 5.1, and benefits from excellent spectral isolation

from water vapor and CO2 hot bands [59]. Relevant spectroscopic data is presented in Table 5.1 for

the three selected transitions. Line-strength and lower-state energy values for CO and CO2 transi-

tions, especially in the fundamental bands, are generally well-known due to the simple structure of

the molecules [54, 84], and taken here from the HITEMP 2010 database [20].

Figure 5.1: Absorption line-strengths for CO and CO2 from 4–5 µm; T = 2000 K; transitions labeledas branch(v”,J”).

5.2.2 Laser absorption spectroscopy

Scanned-wavelength absorption spectroscopy techniques were used in this work to determine

thermodynamic properties of interest from the measured absorption spectra. A brief discussion

of the theoretical framework is provided here to define units and nomenclature. Recall the Beer-

Lambert law, given by equation 5.1, provides the fundamental relation governing narrow-band light

Table 5.1: Spectroscopic parameters of CO and CO2 transitions used in the scramjet sensor.

Line Assignment Species Wavelength Frequency E′′ S(296K)

Branch(v”,J”) [nm] [cm−1] [cm−1] [cm−2/atm]

P(0,20) CO 4855 2059.91 806.4 87.6×10−2

P(1,14) CO 4854 2060.33 2543.1 26.4×10−5

R(0,92) CO2 4176 2394.42 3329.0 73.9×10−6

5.2. METHODS 65

absorption across a uniform gas medium.

αν = −ln (It/I0)ν = PxabsSi(T )φ(ν)iL, (5.1)

Measured quantities of incident and transmitted light intensities define the spectral absorbance, αν ,

at frequency ν, which is further related to the product of spectroscopic line parameters (Sj , φν),

the partial pressure of the absorbing species, and the optical path-length. By scanning the lasers

in wavelength across each discrete transition j, the often complex dependence on the spectral line-

shape, φν , can be eliminated, simplifying the measurement to an integrated absorbance area, Aj ,

which is only reliant on a single spectroscopic parameter, the line-strength, Sj(T ), as expressed in

equation 5.2.

Aj =

+∞∫−∞

ανdν = Sj(T )PxabsL (5.2)

The ratio of integrated absorbance of two transitions further simplifies to a ratio of line-strengths R,

which is a function of temperature only. With the lower-state energy and line-strength at a reference

temperature known for each transition (see Table 5.1), the simultaneous measure of two lines of

a single species facilitates direct inference of gas temperature. The temperature sensitivity of the

line-strength ratio, (dR/R)/(dT/T ), for a given pair of lines (A and B) can be expressed as∣∣∣∣dR/RdT/T

∣∣∣∣ ≈ (hck)|EA′′ − EB ′′|

T, (5.3)

where h [J-s] is Planck’s constant, c [cm/s] is the speed of light, k [J/K] is Boltzmann’s constant, and

E′′ [cm−1] is the lower state energy for the target absorption lines [65]. As defined in equation 5.3,

the difference in lower-state energies between the two selected transitions governs the temperature

sensitivity for two-line thermometry. The selected CO line-pair has a lower-state energy difference

of ∆E′′= 1737 cm−1, which yields a temperature sensitivity of greater than unity up to 2500 K.

With temperature measured, species mole fraction can be obtained from the absorption of either

transition with concurrent knowledge of pressure and path-length per equation 5.2. Integrated ab-

sorbance areas for each line were determined by Voigt line-shape fitting methods. The primary tech-

nique, scanned-wavelength direction absorption (DA), involved directly fitting the Voigt function to

measured absorbance profiles, which is a relatively simple and well-established method [26]. In ad-

dition, a recently developed scanned-wavelength modulation spectroscopy (WMS) technique was

employed to extract the absorbance area by fitting a 1f -normalized second harmonic (2f ) profile,


extending the dynamic range of the CO temperature measurement by enhanced noise suppression.

Details on this technique are found in a separate publication by Goldenstein et al. [85]. These laser

techniques and spectroscopic parameters were validated in a heated static optical cell over a range

of temperatures (500–800 K) and pressures (0.5–2 atm) prior to field deployment of the sensor, with

typical measurement uncertainty of ±2% for each species.

5.2.3 Optimization for non-uniform flows

In the scramjet combustor, as in most propulsion applications, the path-length or line-of-sight

over which the measurement is performed may involve gradients in gas properties. Though some-

times these gradients can be negligible, in many cases flow non-uniformities can convolute line-of-

sight measurement schemes and distort quantitative results. Here we briefly describe the strategic

approach for these line-of-sight measurements in the scramjet environment that minimizes sensitiv-

ity to non-uniformities. A previous paper [86] has been dedicated to this topic, so here we constrain

the discussion to our particular application.

The species-specific nature of laser absorption spectroscopy lends to species-weighted measure-

ments, implying that measurements of gas properties such as temperature will be weighted towards

regions of the flow wherein the species being probed has a higher density and absorption strength.

Previous research has been aimed at quantifying non-uniformities for line-of-sight measurements

[87–89]. The goal here was to obtain a species-weighted path-average measurement that serves

as a metric that can be readily compared to computational modeling. To obtain a path-average

measurement, our strategy is to probe a spectroscopic quantity that has a linear dependence on the

non-uniform gas properties, namely temperature, pressure, and mole fraction. As defined in equa-

tion 5.2, the integrated absorbance area provides an ideal parameter for non-uniform gases due to

the simple linear relationship with pressure, mole fraction, path-length, and line-strength. The inte-

grated area is more easily interpreted compared to other line-fitting parameters such as line-width,

which has a highly non-linear dependence on thermodynamic properties. The scanned-wavelength

techniques described in the previous section (DA and WMS) are thus both designed to yield the

integrated absorbance area via a Voigt fitting routine. Further, to obtain a linear dependence on

temperature, transitions should be selected to have a line-strength which is linear with temperature

over the target temperature range as discussed by Goldenstein et al. [86]. Figure 5.2 shows the three

lines selected for this sensor plotted as a function of temperature (800–2400 K). The P(0,20) and

P(1,14) lines of CO exhibit a nearly linear and inverse relationship over the temperature range of

1500–2000 K, the primary range of interest. The R(0,92) line of the CO2 fundamental band was

5.3. EXPERIMENTAL SETUP 67

chosen due to insensitivity to temperature over a broad range of high temperature. Line-strength for

the CO2 transition varies less than ±10% from 1200–1850 K. Such insensitivity is advantageous

for the CO2 concentration measurement to negate the effect of temperature non-uniformity and also

because the temperature being used to infer concentration here is that measured from CO, which

may be slightly different than the CO2 temperature due to variable distributions of the molecules

along the line of sight. Such careful consideration of the laser technique and line selection enhances

the quantitative value of these line-of-sight absorption measurements.

1 0 0 0 1 5 0 0 2 0 0 00

5

1 0

1 5

P ( 1 , 1 4 )

Line-s

treng

th [cm

/mole

cule]

*10-20

T e m p e r a t u r e [ K ]

P ( 0 , 2 0 )

R ( 0 , 9 2 )

T C O L i n e - P a i r

Figure 5.2: Line-strengths vs. temperature for the selected CO (blue) and CO2 (red) transitions.

5.3 Experimental setup

5.3.1 Optical hardware

Figure 5.3 provides a graphical depiction of the optical configuration for the multi-species laser

absorption sensor. For carbon monoxide, a distributed-feedback (DFB) quantum cascade laser

(ALPES) near 4860 nm provides a single-mode light source with ∼10 mW output power. An

external-cavity quantum cascade laser (Daylight Solutions) centered near 4250 nm provides single-

mode light at 30 mW nominal output power for probing the carbon dioxide spectra. Each incident

beam is free-space coupled into a 1.25 m bifurcated hollow-core fiber (OKSI, d=300 µm), which

combines the beams onto a parallel and near co-linear path (300 µm core diameter separation) for


pitching across the test medium. Net transmission through the fiber is better than 50% for each

wavelength. The fiber output is then re-collimated using a plano-convex silicon lens (f.l. = 40 mm)

and transmitted through a wedged sapphire window. The combined beam, after passing through

the test medium, is de-multiplexed using an anti-reflection coated CaF2 beam splitter, and spec-

trally bandpass filtered (∼50 nm) for each respective wavelength, followed by collection on two

infrared thermo-electrically cooled photovoltaic detectors (Vigo PVI-4TE-5). The detectors have a

bandwidth of 10 MHz and 2 mm2 detection area (D∗ ≥ 3× 1011cmHz1/2W−1).

Fiber-Coupling Alignment Stage

BifurcatedHollow-core

Fiber

PV DetectorsCO

(4854 nm)

CO2(4184 nm)

Beam Splitter

Fibe

r Del

iver

y to

Eng

ineCoupling

LensTranslation

Stage

TOP VIEW

CollimationOptic Sapphire/ ZnSe

Window

Test Medium

Figure 5.3: Optical configuration for free-space fiber-coupling (left) and cross-section of scramjetcombustor showing remote light delivery and collection (right). Flow direction is out of the page.

For direct absorption measurements, the CO DFB laser is centered at 2060.2 cm−1 and scanned

at 6 kHz (sawtooth) with an injection current amplitude of 80 mA, yielding a scan range of approx-

imately 0.9 cm−1 to capture both the P(0,20) and P(1,14) carbon monoxide transitions in a single

laser scan. The external-cavity CO2 QCL is centered at 2394.4 cm−1 and piezo-electrically scanned

at 100 Hz (sine wave), yielding a scan range of approximately 1 cm−1. Prior to and between tests,

the respective beams are redirected to a wavemeter (Bristol 621B) to confirm center wavelength.

A germanium Fabry-Perot etalon, with a free-spectral range (FSR) of 0.016 cm−1, enabled data

conversion from the time to wavelength domain.

In wavelength modulation spectroscopy measurements, the DFB QCL for carbon monoxide was

modulated rapidly at 50 kHz and simultaneously scanned more slowly (sine wave) at 200 Hz to cap-

ture the two P-branch lines. A modulation depth of 0.057 cm−1 was chosen to maximize the peak

second harmonic signal for the conditions expected in the scramjet. The lock-in amplifier outputs,

5.3. EXPERIMENTAL SETUP 69

namely the first and second harmonic signals, were low-pass filtered at 10 kHz. To eliminate depen-

dence on the absolute optical power, the 2f signal is normalized by the 1f signal. This normalization

mitigates baseline uncertainties resulting from scattering, window fouling, and beam steering [9].

5.3.2 Facility interface

Initial field measurements using this sensor were carried out at the University of Virginia Super-

sonic Combustion Facility (UVaSCF) [90]. The direct-connect combustor was oriented vertically,

with continuous air flow provided by a compressor and underground heater system. Air out of the

heater had a total temperature of 1200 K, and was expanded by a nozzle to Mach 2 conditions prior

to entering the combustor. Heated ethylene was injected through five ports located approximately

2.5 cm upstream of a cavity flame holder [91, 92]. The optical line-of-sight (LOS) was transverse to

the injector with a path-length of 3.81 cm corresponding to the cross-sectional width of the combus-

tor. Position of the optical LOS in the x-y plane was controlled by a set of high-precision translation

stages (Zaber). Figure 5.4 illustrates a basic schematic of the combustor with representative optical

lines-of-sight across a transverse measurement plane. Demonstration data shown in the subsequent

section was taken in the combustor cavity plane approximately 2.18 cm downstream of the cavity

leading edge.

TranslatingLOS

FuelInjection

RecirculationCavity

ReactionZone

MeasurementPlane

Air Flow

YX

Figure 5.4: UVaSCF supersonic combustor schematic; example optical lines of sight shown in red.

During combustor operation, a number of harsh thermo-mechanical phenomena were intro-

duced to the ambient environment that required mitigation for successful sensor employment. To


counter mechanical vibrations and acoustic perturbations, the lasers were mounted to a honeycomb

vibration-dampening breadboard. Due to radiation emitted by the combustor, the photovoltaic de-

tectors were mounted to water-cooled plates to prevent saturation or damage from overheating.

Heating in the room also led to elevated levels of water vapor in the humid ambient air, which could

spectrally interfere with the absorption measurements and condense on cooled optical equipment.

A nitrogen purge of the optical breadboard was constructed to minimize this water vapor. The N2

purge was extended to the hollow-core fiber and optical detection components as well to encompass

the entire optical path of each laser output.

5.4 Results

The present mid-infrared sensor aimed to measure gas temperature, carbon monoxide, and car-

bon dioxide simultaneously with both temporal and spatial resolution in the harsh combustion en-

vironment. Raw data quality for direct absorption measurements can be examined by inspection

of the laser scans, exhibited in the top plots of figures 5.5 and 5.6, for CO and CO2 respectively.

Transmitted laser light intensity, in detected volts, is shown along with the baseline intensity, which

was captured after flame extinction. For the DFB QCL dedicated to CO measurements, the saw-

tooth scan (see fig. 5.5) was deliberately set to go below the threshold current of the laser for a short

time, during which thermal emission could be measured and subtracted from the adjacent signal.

For the ECQCL dedicated to CO2, the piezo-scan simply tunes a mechanical grating such that laser

output intensity is nearly constant across the laser scan (see fig. 5.6). Therefore, to account for

thermal emission, a beam chopper (∼12.5 Hz) was mounted at the output of the ECQCL to inter-

mittently block the laser after every eight scans for a period of ∼5 ms wherein thermal emission

was measured and subtracted in a similar manner as described above. Due to the temperature- and

wavelength-dependent transmission of the sapphire windows, the baseline, which was measured at

a lower temperature after combustion ceased, was scaled to match the non-absorbing regions of the

transmitted intensity scan. Both thermal emission and window losses were noted to change on suf-

ficiently slow time scales that these effects were effectively constant during a single laser scan. The

bottom plots of figures 5.5 and 5.6 show the conversion of the raw intensity signals to absorbance

and transposed to the wavenumber domain, from which the Voigt line-shape function could be fit to

each transition to yield the integrated absorbance areas.

Since line-strengths are well known for the selected transitions, the quality of the baseline and

5.4. RESULTS 71

0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 00 . 0 0

0 . 0 5

0 . 1 0

0 . 1 5

I tInten

sity [V

]

R e l a t i v e T i m e [ m s ]

I 0

C O

0 . 0

0 . 5

1 . 0

1 . 5

2 0 5 9 . 8 2 0 6 0 . 0 2 0 6 0 . 2 2 0 6 0 . 4- 404

P ( 1 , 1 4 )Abso

rbanc

e

M e a s u r e m e n t V o i g t F i tP ( 0 , 2 0 )

Resid

ual [%

]

W a v e n u m b e r [ c m - 1 ]

Figure 5.5: Measured carbon monoxide absorption from a single laser scan (6 kHz) shown as (top)raw voltage signals versus time and (bottom) absorbance versus wavenumber; T = 1725 K, P =0.71 bar, L = 3.81 cm, XCO = 0.057; φ ≈ 0.15, measurement taken 2.18 cm downstream of cavityleading edge, 1 mm from cavity wall.


0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 00 . 2

0 . 4

0 . 6

C O 2Int

ensity

[V]


I 0

I t

0 . 0

0 . 2

0 . 4

0 . 6

2 3 9 4 . 3 5 2 3 9 4 . 4 0 2 3 9 4 . 4 5 2 3 9 4 . 5 0- 606

Abso

rbanc

e

M e a s u r e m e n t V o i g t F i t

R ( 0 , 9 2 )

Resid

ual [%

]

W a v e n u m b e r [ c m - 1 ]

Figure 5.6: Measured carbon dioxide absorption from a single laser scan (100 Hz) shown as (a) rawvoltage signals versus time and (b) absorbance versus wavenumber; T = 1725 K, P = 0.71 bar, L =3.81 cm, XCO2 = 0.062; φ ≈ 0.15, measurement taken 2.18 cm downstream of cavity leading edge,1 mm from cavity wall (same as figure 5.5).

5.4. RESULTS 73

Voigt fits typically limit measurement uncertainty for scanned-wavelength direct absorption tech-

niques in harsh, turbulent environments. Non-Gaussian noise sources in the baseline may be mani-

fested by systematic drift of detected signal during the laser scan period, such as beam steering or

other fluctuations at a similar frequency as the laser scan frequency. With such large absorbance

in this application, these uncertainties were relatively small in most measurement locations, and

can be partly quantified by inspection of the residual to the Voigt fit, displayed in figures 5.5 and

5.6 as percent of peak absorbance. For CO detection, both lines as depicted in the bottom plot of

figure 5.5 exhibit SNR > 30 with minimal systematic baseline error (i.e. < 1% baseline drift and <

1% residual noise in non-absorbing regions). The baseline shifting and scaling process was highly

repeatable scan to scan and less than 3% overall measurement uncertainty was introduced for the de-

termination of integrated areas at nominal concentrations in the combustor cavity. For the two-line

CO temperature measurement, this translates to less than 3% temperature uncertainty below 2500

K at a nominal mole fraction (XCO ∼0.05). The SNR was observed to degrade nearly linearly with

mole fraction, yielding a CO detection limit of∼1000 ppm at 6 kHz using the stronger P(0,20) line.

The carbon dioxide measurements were more prone to baseline distortion resulting from random

fluctuations in laser output power (∼2%) due to sensitivity of the ECQCL to mechanical vibrations.

This noise source, notable in the residual plot in figure 5.6, appeared larger in amplitude compared

to beam steering, which was independently observed in the shared line-of-sight with the CO beam,

and also dominates interference absorption (< 1%) from neighboring CO2 lines. Although such

vibration-induced intensity fluctuations were random, the baseline fitting uncertainty limited over-

all CO2 measurement uncertainty to 5%. Based on noise alone, a detection limit of ∼1200 ppm

CO2 was achieved (SNR∼1).

The scanned-wavelength modulation spectroscopy technique, implemented with the DFB-QCL

for CO, further suppressed measurement noise. The top plot in figure 5.7 shows an example scan of

the raw WMS data for CO, taken at a similar operating condition and spatial location as the data pre-

sented in figure 5.5. The absorption from each of the two P-branch lines can be observed, centered

in the sinusoidal half-scan (200 Hz). With the modulation depth predetermined in the laboratory,

the normalized second harmonic can be plotted against wavenumber as shown in the bottom plot of

figure 5.7. As previously mentioned, the Voigt fitting routine to the background WMS signal ren-

ders the integrated absorbance areas when matched to the measured data. The fitted 2f line-shape

for each transition can be seen in the figure compared to the measured line-shapes. This WMS tech-

nique was implemented to enhance the dynamic range of the temperature measurement, which is

limited by the weaker P(1,14) line. Although CO mole fraction can be calculated with an estimated


0 1 2 30 . 2 5

0 . 3 0

0 . 3 5

Inten

sity [V

]


C O

f 1 = 2 0 0 H zf 2 = 5 0 k H z

0 . 0

0 . 5

1 . 0

2 0 5 9 . 8 2 0 6 0 . 0 2 0 6 0 . 2 2 0 6 0 . 4- 404

P ( 1 , 1 4 )

WMS 2

f - No

rmaliz

ed

M e a s u r e m e n t V o i g t F i tP ( 0 , 2 0 )

Resid

ual [%

]

W a v e n u m b e r [ c m - 1 ]

Figure 5.7: Measured carbon monoxide wavelength modulation scan (50 kHz, 200 Hz) shownas (top) raw voltage signal versus time and (bottom) normalized second harmonic signal versuswavenumber; T = 1690 K, P = 0.71 bar, L = 3.81 cm, XCO = 0.063; φ ≈0.15, measurement taken2.18 cm downstream of cavity leading edge, 1 mm from cavity wall.

5.4. RESULTS 75

temperature and the P(0,20) line alone, temperature requires the line-pair. Utilizing the WMS line-

shape fitting method, temperature could be determined at CO concentrations approximately 3 times

lower (∼1000 ppm limit for TCO, ∼350 ppm for XCO) than with direct absorption, determined by

the noise floor at 200 Hz of the normalized 2f signal (SNR∼1).

5.4.1 Time-resolved measurements

The time resolution of each species measurement is equivalent to the scan rate of each QCL that

comprises the sensor: 6 kHz or 200 Hz for CO, and 100 Hz for CO2. Such bandwidths, especially for

the faster CO DFB laser, can provide valuable time-resolved information about flow field properties.

Figure 5.8 shows the integrated absorbance areas for each of the two CO transitions, measured at

a fixed x-y location in the combustor cavity during ethylene-air combustion conditions (φ ≈0.15).

Each data point represents the result from a single laser scan as described above. The integrated

areas are observed to vary dramatically (> 25%) during the acquired test time (1 sec). The positively

correlated trends over time between the P(0,20) and P(1,14) areas supports the physical natural of

this fluctuation. We further note that the physical oscillations in temperature at a fixed position

in the combustor are much smaller (< 10%) than the fluctuations in mole fraction, which scales

0 1 0 2 0 3 00

1 0 0 0

2 0 0 00 . 0 0

0 . 0 4

0 . 0 0

0 . 0 8

T [K]

T i m e [ m s ]

± 1 4 0 K

P ( 0 , 2 0 )

A 2 [cm-1 ]

P ( 1 , 1 4 )

A 1[cm-1 ]

Figure 5.8: Representative time-resolved CO absorbance areas and temperature at a fixed x-y po-sition in the combustor cavity; ethylene-air, φ ≈0.15; measurement taken 2.18 cm downstream ofcavity leading edge, 1 mm from cavity wall.


linearly with absorbance area and is shown on a longer time scale, along with CO2 mole fraction,

in Figure 5.9. Further, it is noted that the variation in CO appears much larger than the variation in

CO2 at the respective bandwidths of each sensor, but when applying a moving average to the CO

mole fraction data to yield a 100 Hz bandwidth, the fluctuations become very similar (∼15%). For

both temperature and species concentrations, these physical variations are greater than measurement

uncertainties described in the previous section. Similar fluctuations in both temperature and mole

fraction were noted in independently conducted water vapor absorption measurements [93]. The

time-resolved carbon oxide data provides a basis to evaluate combustion stability and gas dynamic

events.

0 1 0 0 2 0 0 3 0 00 . 0 0

0 . 0 5

0 . 1 0

0 . 0 0

0 . 0 5

0 . 1 0

1 0 0 H z E C Q C L

X CO2

T i m e [ m s ]

1 0 0 H z M o v i n g A v e r a g e6 k H z D F B - Q C L

X CO

Figure 5.9: Representative time-resolved CO and CO2 mole fraction data at a fixed x-y positionin the combustor cavity; ethylene-air, φ ≈0.15; measurement taken 2.18 cm downstream of cavityleading edge, 1 mm from cavity wall.

5.4.2 Spatially-resolved measurements

In addition to high-bandwidth measurements, the current experimental setup facilitates carbon

oxide sensing with two-dimensional spatial resolution. The translation stages can position the opti-

cal line of sight along a plane transverse to the bulk flow (x-axis) as illustrated in figure 5.4 and at

multiple planes along the flow direction (y-axis). Figure 5.10 provides an example of the spatially-

resolved CO data across a transverse plane in the model scramjet combustor at an axial location in

the flame-holding cavity (see fig. 5.4). All data points are time-averaged over approximately one

5.4. RESULTS 77

second, and the error bars represent the standard deviation (physical oscillations) over this time in-

terval. The data illustrate the relative range and agreement between the DA and WMS techniques

for measurements of temperature and CO mole fraction at the same operating condition (φ = 0.15).

Overall, the two techniques were observed to generally agree within the time-variable uncertainty

of the measurements due to the unsteady nature of the flame, especially considering that the mea-

surements were taken several minutes apart. It can be noted however that the agreement outside of

the turbulent flame-holding cavity tends to be better for both temperature and species mole fraction.

Moreover, we can observe that the WMS technique enhances the measurement range, especially for

temperature, further into the free-stream where less CO is present.

1 . 5 2 . 0 2 . 5 3 . 0 3 . 50

5 0 0

1 0 0 0

1 5 0 0

2 0 0 0

2 5 0 0

C a v i t yB o t t o m

Mole Fraction

T C O ( D A ) T C O ( W M S )

Temp

eratur

e [K]

D i s t a n c e f r o m W a l l O p p o s i t e I n j e c t o r [ c m ]

I n j e c t o r W a l l

0 . 0 0

0 . 0 4

0 . 0 8

0 . 1 2

0 . 1 6 X C O ( D A ) X C O ( W M S )

Figure 5.10: Comparison of DA and WMS measurements of temperature and CO mole fractionacross the cavity plane at φ = 0.15.

Of particular importance, spatially-resolved species measurements in the combustor facilitate

assessment of flame structure and combustion progress. Figure 5.11 shows data of CO and CO2

mole fraction measured across the transverse cavity plane. The two plots represent a snapshot of

the same measurement plane under different fuel to air mass flow ratios (φtop = 0.15 and φbottom= 0.21). All data points are similarly time-averaged as previously mentioned. Water concentration,

which was measured independently at the same test conditions [93], is also shown for reference

Though combustor analysis is reserved for another publication in the aerospace literature [78],


1 2 3 40 . 0 0

0 . 0 5

0 . 1 0

0 . 1 5 C a v i t y B o t t o m C O C O 2 C O + C O 2 H 2 O

Mole

Fractio

n


I n j e c t o r W a l l

1 2 3 40 . 0 0

0 . 0 5

0 . 1 0

0 . 1 5 C O C O 2 C O + C O 2 H 2 O

Mole

Fractio

n


I n j e c t o r W a l lC a v i t y B o t t o m

Figure 5.11: Carbon oxide mole fraction measurements across the cavity plane for ethylene-aircombustion at (top) φ ≈0.15 and (bottom) φ ≈0.21. XH2O also shown.

these example data highlight the value of this in situ carbon oxide sensor, and a few observations

deserve mention. We observe that conditions in the cavity appear near stoichiometric or rich, while

carbon oxides become quickly diluted once entering the free stream, both trends that would be

expected. Comparing the aggregate carbon oxide concentration with values of independently mea-

sured H2O, we find excellent consistency with C2H4 atom balance (XCO+XCO2 = XH2O) for both

conditions, even while relative carbon oxide levels vary considerably between equivalence ratios.

These observations provide physical validation to the measurements, but also point to the strength

and sensitivity of the diagnostic approach. At φ = 0.15, peak CO and CO2 mole fractions are both

near 6%, while at φ = 0.21, the peak values are approximately 10% and 3%, respectively. Simi-

larly comparing the H2O measurements across equivalence ratios, we note that mole fractions peak

around 12% (a near stoichiometric level) for both conditions, suggesting that the CO and CO2 mea-

surements are each a more sensitive measure of combustion progress for ethylene-air. Moreover,

5.5. SUMMARY 79

due to the inverse relationship of CO and CO2 with combustion progress, the combined measure-

ment (CO:CO2 ratio) yields even greater sensitivity. These observations reinforce the strength of

the two-species carbon oxide sensing strategy for assessing hydrocarbon combustion.

5.5 Summary

A mid-infrared quantum cascade laser absorption sensor was developed for simultaneous mea-

surements of carbon monoxide, carbon dioxide, and gas temperature in supersonic propulsion flows.

Two QCLs comprised the sensor, coupled with a bifurcated hollow-core fiber for remote light deliv-

ery. The sensor was successfully demonstrated at the University of Virginia’s direct-connect scram-

jet combustor facility for time-resolved detection (up to 6 kHz) at multiple fuel-air flow ratios, with

detection limits of∼1000 ppm for each species at the∼4 cm path-length. Utilization of wavelength

modulation spectroscopy enhanced noise rejection compared to direct absorption and extended the

useful range of the temperature measurement. Spatially-resolved species measurements across the

combustor highlighted the sensitivity of a combined CO/CO2 sensor to hydrocarbon combustion

progress. To the authors’ knowledge, these measurements represent the first combined, in situ de-

tection of CO and CO2 in a scramjet combustor. Furthermore, the proven utilization of novel mid-

infrared light sources and fiber optics in a harsh environment provides a sensor design framework

by which similar mid-infrared diagnostics may be developed for future studies.

Chapter 6

Conclusion

The research presented here demonstrates an advancement in mid-infrared laser absorption diag-

nostics for carbon oxides (CO, CO2) in harsh, high-temperature combustion systems. The work has

leveraged recent advancements in quantum cascade lasers and mid-infrared fiber optics to success-

fully implement field-deployable sensors for novel gas sensing in aeropropulsion facilities. Specif-

ically, diagnostics were designed, developed, and demonstrated for first-ever sensing of CO and

CO2 in a pulse detonation engine and direct-connect scramjet combustor, both burning ethylene in

air. An improved, more sensitive CO2 diagnostic has also been developed for shock tube kinetics

studies. In total, primary contributions from the author to the field of laser absorption diagnostics

include new wavelength selection for improved sensing strategies, mid-infared optical engineering

(which is described further in Appendix A), implementation of state-of-the-art signal processing

techniques, and additions to high-temperature spectroscopic databases. Combined, these contribu-

tions enabled novel applications for gas sensing in the case of aeropropulsion engines and improved

techniques in the application to shock tube kinetics. Detailed summaries for each project have been

offered at the end of the respective preceding chapters, and here a broader view is offered with an

outlook on future research.

6.1 Aeropropulsion research

As advanced propulsion technologies such as detonation-based engines and scramjets progress,

a transition from hydrogen to more practical hydrocarbon fuels should further drive the demand

for in situ carbon oxide diagnostics such as those presented here. A key, consistent finding from

both the pulse detonation engine and scramjet research was the spatial and temporal delay of the

80

6.2. SHOCK TUBE KINETICS RESEARCH 81

formation of CO2 relative to the formation of H2O in hydrocarbon combustion flow fields. This

results from the relatively slow chemical step of CO oxidation relative to that of the more reactive

intermediates(OH, etc.) that precede H2O formation. As such, a majority of the chemical energy in

the downstream tail of a hydrocarbon reaction zone, which directly precedes complete combustion,

likely remains in the form of unburnt CO. And thus, temporal and spatial resolution of carbon oxide

concentrations within the combustor provide critical insight for assessing engine efficiency. The

work conducted here extends the capability of laser-based, in situ detection of CO and CO2 to harsh

high-temperature environments and also to high pressure environments, which should enable many

future aeropropulsion studies over a large range of operating conditions.

6.2 Shock tube kinetics research

Determination of chemical kinetic parameters (ie. rate constants) rests heavily on highly con-

trolled shock tube studies. Perhaps the most important control parameter for a shock tube experi-

ment is gas temperature due to its dominant effect on reaction rates. Shock tubes generally allow

for accurate determination (< 1%) of initial temperature based on the time-tested conversion of

measured incident shock velocity to temperature using the known initial conditions and ideal shock

relations. However, temperature changes during exothermic or endothermic experiments are not as

easily prescribed for many reactant mixtures, and may distort measurement results. In this context,

the current mid-infrared CO2 absorption sensor improves upon prior diagnostics in two ways. First,

by increasing the sensitivity to CO2 by greater than 50 times over previous sensors and lowering the

detection limit to ∼5 ppm, lower fuel concentrations may be used to yield similar signal levels for

CO2. Lower fuel concentrations translate to much lower temperature changes during experiments

and thus more reliable results. Second, by utilization of a highly sensitive temperature sensing strat-

egy, as introduced here with cross-band CO2 thermometry, small temperature changes during an

experiment may be more accurately measured. It is expected that future work will be dedicated to

refining the CO2 thermometry technique for reactive experiments, combining the superior spectro-

scopic sensitivity with more advanced and robust signal processing techniques as laser technology

matures near 4.2 µm.

Appendix A

Mid-infrared optics: practical issues

This appendix is intended to cover a number of practical issues with gas-phase absorption spec-

troscopy in the mid-infrared that are mostly omitted from the preceding chapters. As has been

mentioned, laser diodes and fiber optics have largely been developed as demanded by the telecom-

munications industry, which has been rapidly replacing incumbent electrical transmission systems

with optical transmission systems due to a number of advantages in speed and signal quality. Be-

cause of the enormity of the telecom market relative to that of other end-use markets, the commercial

availability of lasers, fibers, and other optical equipment used for spectroscopy has been heavily bi-

ased towards telecom wavelengths (1.3–1.65 µm). In turn, optical hardware in the mid-infrared has

experienced much less development, and many issues which are solved with inexpensive, off-the-

shelf equipment in the near-infrared (NIR) remain problematic in the mid-infrared (MIR). Some of

these issues include: non-linear laser output intensity, wavelength stability, susceptibility to back

reflections, beam spatial mode quality, laser-to-fiber coupling, fiber transmission, multi-mode fiber

dispersion, beam collimation, high-temperature window transmission, opto-mechanical hardware

integration, and detector thermal management. As manufacturers mature their processes, it is ex-

pected that these issues will diminish over time, but at present they are problematic. These factors

in part have limited the applicability of spectroscopic gas sensing in the mid-infrared (> 2.5 µm),

and required significant attention during the development of the aforementioned carbon oxide diag-

nostics. A few of the issues and some practical solutions are discussed here.

82

A.1. LASER OUTPUT NON-LINEARITY 83

A.1 Laser output non-linearity

Quantum cascade distributed-feedback and external cavity lasers commonly exhibit some de-

gree of non-linear output intensity, whether attributable to imperfect material properties of the laser

chip or etalons in the lasing cavity. Etalons evident in the laser output may also result from the laser

mounting configuration between surfaces such as the collimation lens and window port. Figure A.1

shows an example of the non-linear output from a DFB-QCL with a linear injection current ramp.

Referencing the dashed line provides comparison to a linear output.

3600 3800 4000 4200 4400

1

1.2

1.4

1.6

1.8

Data Points

Sig

nal [

volts

]

DFB-QCL Output Intensityw/ Linear Current Ramp

Baseline (no absorption)

Linear Ramp

Figure A.1: Example non-linear output from a DFB-QCL with no absorption.

Such non-linearities are problematic in fitting baselines for scanned-wavelength direct absorp-

tion (scanned-DA) and in modeling wavelength modulation spectroscopy (WMS) background sig-

nals analytically. For scanned-DA, utilizing a linear or nearly linear output, the baseline is often

established by fitting the non-absorbing wings of the laser scan with a linear or polynomial func-

tion. With even modest non-linearities, this practice quickly becomes arbitrary and susceptible to

post-processor bias, especially when using higher order polynomial fits. Ideally, it is recommended

to use the laser output when no absorption is present as the baseline signal. Under conditions

wherein window fouling, thermal emission, or other sources cause a drift or low-frequency fluctu-

ation in the real baseline, it is recommended that the non-absorption baseline be simply scaled (for

84 APPENDIX A. MID-INFRARED OPTICS: PRACTICAL ISSUES

attenuation correction) or shifted (for offset correction) to match the non-absorbing wings of the

scan as exhibited in figure A.2. Further adjustments such as tilting may also be utilized, but the

idea is that by starting with a baseline signal which already contains significant information that is

representative of the real baseline, fewer and less arbitrary adjustments are required, which should

provide a more physical and accurate result.

0 1 2 3 4x 10-4

1

1.2

1.4

1.6

1.8

Time [s]

Sig

nals

- S

cale

d [v

olts

] Empirical Baseline

Transmitted Light

Figure A.2: Example raw absorption scan fit with an emperical baseline after shifting to account forthermal emission and scaling to account for variable window transmission at elevated temperature.

A similar approach can be taken when attaining emperical background signals for wavelength

modulation spectroscopy. Utilization of the baseline signal under a condition of no absorption is rec-

ommended compared to analytic simulation of the background when the laser output is non-linear.

An assessment of this strategy for implementation to scanned-wavelength WMS measurements has

recently been pioneered by Sun et al [94] and Goldenstein et al [85].

A.2 Wavelength stability

Mid-infrared quantum and interband cascade lasers often exhibit a high frequency shift, or jitter,

in wavelength over time. This may be attributable to the higher absolute currents (∼1 A) required

and thus higher current noise associated with the mid-IR laser drivers or perhaps results from poor

reproducibility scan-to-scan in laser response to the input function. In either case, wavelength jitter

A.3. SUSCEPTIBILITY TO BACK REFLECTIONS 85

may confound certain averaging techniques, especially techniques wherein a quantitative measure

of linewidth is important. Therefore, it is recommended to generally avoid averaging raw scan

signals to avoid an artificial broadening of the transition lineshape. If averaging is desired, it is

recommended to average the output parameters (area, collision width, etc) from a lineshape fitting

routine after assessing each raw scan individually.

A.3 Susceptibility to back reflections

Reflections of the incident laser light back into the laser cavity can cause undesirable instability

and oscillation in laser output intensity [95]. External cavity quantum cascade lasers have been

observed to be particularly susceptible to back reflection noise. In the near-infrared, optical isolators

are commonly used to serve as one-way valves for light that prevent any such back reflections from

occurring [96, 97]. Commercial availability and affordability of isolators in the mid-infrared is quite

limited, therefore additional precaution must be taken to avoid spurious back reflections. Common

best practices include (1) using wedged or angled transmissive surfaces (i.e. windows, filters, beam

splitters), (2) using anti-reflection coated optics whenever possible, (3) using a pair of irises near

the laser output to eliminate any reflections that are not co-linear with the laser beam, and (4)

coupling a single-mode fiber to the laser output beam [98]. When using irises, it is important not to

overconstrain the beam diameter as to avoid diffraction.

A.4 Beam spatial mode quality

As the naming convention implies, the relationships and equations defined in Gaussian beam

optics assume that the intensity distribution of the laser output beam is a Gaussian distribution. For

example, the equations used to determine a minimum beam waist diameter when focusing light or

the proper depth of focus when collimating light hinge on the assumption of a Gaussian beam. How-

ever, most laser output beams are not perfectly Gaussian, and thus these equations are imperfect. In

the near-IR, and with some of the more mature mid-IR manufacturers, lasers are reliably produced

with near-Gaussian output beams (although the beams are often slightly elliptical). Still, many QC

lasers are sold at present with highly non-Gaussian beam profiles which can severely complicate

downstream propagation.

The manner in which spatial mode quality is typically classified is by the type of transverse

electromagnetic modes (TEMs) which are being propagated. The TEM00 mode is the fundamental


mode and has a Gaussian form. Other types of TEMs are non-Gaussian and poorly suited for colli-

mation or fiber-coupling, and can amplify detected intensity fluctuations from other issues such as

beam steering or mechanical vibration. A common non-ideal mode observed for DFB-QCLs is the

two-lobed TEM01 mode, shown in comparison to the TEM00 mode in figure A.3. The overall inten-

sity profile for a given beam, as displayed here with a pyroelectric array camera (Spiricon - Pyrocam

III), is a superposition of the allowable modes in the laser cavity. Weakly allowed modes can yield

a rather complicated spatial intensity distribution in the outer fringe diameter of the beams, and it

is recommended to use an iris to spatially filter this fringe region. To best optimize downstream

optics, it is further recommended to specify the TEM00 mode during laser procurement.

TEM00 TEM01

TEM00 TEM01

Figure A.3: Pyrocam images of the output beams from an external cavity QCL propogating theTEM00 mode (left) and a DFB-QCL propogating the TEM01 mode.

A.5. FIBER-COUPLING 87

A.5 Fiber-coupling

Laser-to-fiber coupling in the mid-infrared has already been discussed in some detail in chap-

ter 2, and here a portion of that discussion is summarized and a few additional considerations are

noted. As suggested in the above section regarding spatial mode quality, fiber-coupling is most

efficient when the laser beam has a Gaussian distribution. Further, high coupling efficiencies are

achieved when the numerical aperture (NA) of the lens matches that of the fiber and when the min-

imum focal spot size is less than the core diameter of the fiber. Recall the diffraction-limited spot

size (dmin) for a beam with a single Gaussian transverse mode is defined as

dmin =4fλ

πd0(A.1)

where f (mm) is the focal length of the lens, d0 (mm) is the diameter of the incident beam and λ

(µm) is the wavelength [32]. A lens with a short focal length can then be used to achieve a small

focused spot size at the diffraction limit. Aspheric lenses are commonly used to focus and collimate

light without introducing spherical aberration, an anomaly which prevents diffraction-limited per-

formance [99]. As suggested, the shortness of the focal length of the lens must be balanced against

the acceptance angle or numerical aperture of the fiber. Both the lens and the fiber have respective

characteristic numerical apertures. The numerical aperture of the lens can simply be defined as

(NA)lens =nDlens

2f(A.2)

where n is the index of refraction of the medium in which the lens is focusing (nair ≈ 1), Dlens

is the diameter of the lens, and again f is the focal length. The numerical aperture of the fiber is

determined by the respective indices of refraction of the fiber core and the cladding material, and

approximated by

(NA)fiber =√n2core − n2clad. (A.3)

Generally, it is recommended to design the fiber-coupling system such that (NA)lens ≤ (NA)fiber,

ensuring the focused angle is equal to or more shallow than the acceptance angle of the fiber to

maximize the collection of the focused light. The fiber core may accept light at angles steeper than

the acceptance angle, but the rays will not totally reflect internally.

Coupling of a non-Gaussian beam into a fiber can be very difficult in practice. The multiple

spatial lobes that are present with any transverse electromagnetic mode that is not TEM00 have

been observed to severely reduce coupling efficiency, especially into a small-diameter single-mode


fiber. In this research, an iris was aligned with the help of a pyrocam to block one of the spatial

lobes of beam intensity prior to focusing to mitigate apparent destructive interference with the two

focused lobes of the TEM01 mode. Figure A.4 shows the beam profile before and after spatial

filtering, which allowed for more repeatable and efficient fiber-coupling. With the permanently iris

fixed in relation to the laser after careful alignment, fiber-coupling was performed readily during

field campaigns.

TEM01Incident

TEM01w/ Spatial Filter

Figure A.4: Pyrocam images of the output beam from a DFB-QCL before (left) and after (right)spatial filtering with an iris, which preceded fiber coupling.

Another simple strategy for fiber-coupling beams with poor spatial mode quality is to use a

larger core diameter multi-mode fiber. However, this can lead to significant noise in the transmitted

light due to modal dispersion, which is discussed in a subsequent section.

A.6. FIBER TRANSMISSION 89

A.6 Fiber transmission

Optical fibers for the near-infrared are generally silica-based, and do not transmit light beyond

∼2.3 µm. Fiber materials for the mid-infrared are more expensive, generally less robust mechani-

cally, and have higher attenuation thus limiting length for meaningful transmission. With increasing

demand for mid-infrared fibers, a number of materials have matured commercially in the last ten

years or so [7, 8, 100]. At present, several types of mid-infrared fibers are available for use in harsh

gas sensing environments to enable remote light delivery. Figure A.5 highlights some of these fiber

materials and their respective transmission spectra in the infrared.

Loss

(dB

/m)

1

2

3ZBLAN

ChalcogenideInF3

Hollow Glass

Si

2 3 4 5 61Wavelength (µm)

Figure A.5: Attenuation versus wavelength for various mid-infrared fiber materials

Two of the more attractive fiber types are fluoride glasses (ZBLAN and InF3) and hollow-core

waveguides. Both fiber types are mechanically more robust than chalcogenide, which is known to

be brittle [101] and together offer transmission across a broad wavelength range from ∼2 to 12

µm. The fluoride glass fibers offer lower attenuation than the hollow-core waveguides in the 2–5

µm wavelength domain and are available in greater lengths (> 5 m). The fluoride glass fibers are

also available as single-mode and were successfully utilized for multiple projects in this research.

The hollow-core waveguides offer transmission over a broader wavelength range compared to the

fluoride glasses and also typically come in larger diameters for ease of coupling which is especially

needed for non-Gaussian beams. Additionally, because there is no solid core to the fiber, there is

no risk of spurious back reflections that often occur during fiber coupling. The hollow-core fibers

do tend to induce substantial light attenuation when bent and may require purging to avoid ambient

spectral interference of species like CO2 and H2O as was described in chapter 5.


A.7 Multimodal dispersion in optical fibers

When multiple transverse modes are transmitted through an optical fiber, the propagation ve-

locity is not the same for all modes, and this leads to a distortion of the optical signal known as

modal dispersion [102]. Multimodal dispersion may be manifested by unstable intensity transmit-

ted through the fiber. Such noise is typically amplified when moving or vibrating the fiber, thus

is highly undesirable for gas sensing in harsh environments. Figure A.6 illustrates how the spatial

beam profile changes upon mechanical vibration.

Before Fiber Vibration

After FiberVibration

Figure A.6: Beam intensity profiles at the output of a multimode fiber before and after mechanicalvibration.

A single-mode fiber should be used for remote light delivery whenever possible. The number

of modes in a step-index fiber is related to the fiber’s normalized frequency or V-number, which is

defined as

V =2πDfiber

λ(NA)fiber (A.4)

where Dfiber is the core diameter of the fiber and λ is the wavelength of transmitted light. A

fiber with a V-number less than 2.405 supports only a single transverse mode. In some scenarios

collecting light onto a single-mode fiber is not practical, such as catching light after passing through

an engine. If remote light collection is desired for these scenarios, it is recommended to use a

large diameter multi-mode fiber (> 500 µm) with sufficient length to allow an equilibrium mode

distribution. Coiling or bending multimode fibers can also help scramble the modes such that the

equilibrium mode distribution, which outputs a more stable signal, can be achieved in a shorter

length of fiber. Mode scramblers, which clamp and bend fibers in a more systematic manner to

achieve the equilibrium mode distribution, are commercially available. Hollow-core fibers have

also been demonstrated as effective spatial mode filters [103]. Based on observation, the most

A.8. MID-INFRARED OPTICAL MATERIALS 91

problematic fibers with regards to multimodal dispersion noise tend to be intermediate diameter

solid-core fibers (Dfiber ∼50–200 µm) which are too large in diameter to be single-mode, but too

small in diameter to allow the equilibrium mode distribution to be established over a reasonable

length of fiber.

A.8 Mid-infrared optical materials

Table A.1 lists a number of common optical materials that transmit in the infrared. Quartz, or

fused silica (SiO2), is the most common and inexpensive optical material, but it only transmits up

to ∼3.5 µm, thus it is not sufficient for most species in the mid-infrared which have fundamental

vibrational bands at longer wavelengths. Sapphire (Al2O3) is an excellent material for harsh en-

vironments due to its superior strength and high melting point, which makes it suitable for most

high-pressure combustion applications. Sapphire also has relatively high thermal conductivity and

low thermal expansion, making the material highly resistance to thermal shock. For many cases,

sapphire is the ideal material selection for optical ports/windows in high-temperature gas flows.

However, sapphire transmission degrades around 5 µm and the transmission cutoff has a strong

temperature dependence which is a very important consideration in the design of mid-infrared opti-

cal systems aimed at sensing certain species such as CO2 (∼4.3 µm), CO (∼4.8 µm), or NO (∼5.2

µm). An empirical model of the temperature-dependent transmission of sapphire was developed

by Thomas et al [104], and shown in figure A.7 over a range of 300–2100 K for a window thick-

ness of 1 cm. If probing a wavelength in the ∼3.5–6 µm domain, it is important to understand the

time-varying transmission that may occur if deploying the diagnostic to a high-temperature system.

Transmission cutoffs aside, quartz and sapphire are the best infrared materials for spectroscopic

windows on harsh combustion systems and should be used when feasible.

Other classes of mid-infrared optical materials are less robust than quartz or sapphire but may

be needed for transmission beyond 5 µm. The fluoride crystals including BaF2, CaF2, and MgF2

have broad transmission from the ultra-violet up to 12 µm (in the case of BaF2) and low indices

of refaction (n) which helps mitigate back reflections and etalons, but they are somewhat weak and

highly susceptible to thermal shock due to large thermal expansion coefficients. Furthermore, when

heated above 800 K the fluoride crystals react with water vapor. Zinc Selenide (ZnSe) and Zinc

Sulfide (ZnS) offer even broader transmission into the far-infrared (up to 20 µm). They are a bit

stronger than the fluoride crystals, and offer greater resistance to thermal shock, but are relatively

brittle. Perhaps most problematic for high-temperature applications, these zinc materials oxidize


2 3 4 5 60

0.2

0.4

0.6

0.8

1

Tran

smis

sion

[%]

Wavelength [um]

300K500K700K900K1100K1500K2100K

L = 1 cm

Figure A.7: Temperature dependent transmission of sapphire at a thickness of 1 cm, using theempirical model by Thomas et al.

above 500 K. The surface reaction issues with both the fluoride and zinc optical materials may

be resolved with high-temperature coatings. Germanium and Silicon are also good infrared optical

materials at room temperature, but these materials rapidly become opaque as temperature is elevated

[105]. It should be noted that for shock tube applications, the windows are only exposed to a high

temperature gas for a very short time period (∼ms), thus many of these temperature-dependent

issues are less important, since minimal heat transfer occurs during that time.

Table A.1: Mid-infrared optical material propertiesMaterial λlow [µm] λhigh [µm] Tmelt [K] n Erupture [ksi]

SiO2 0.16 3.5 1980 1.5 7Al2O3 0.17 5 2300 1.75 65BaF2 0.15 12 1620 1.45 3.9CaF2 0.15 8 1700 1.4 5.3MgF2 0.13 7 1520 1.35 7.2ZnSe 0.55 20 1800 2.4 8ZnS 0.38 14 1450 2.5 8.7Ge 1.8 21 1200 4 7Si 1.2 10 1690 3.5 18.1

Bibliography

[1] Hanson, R. K., Kuntz, P. A., and Kruger, C. H. Applied optics 16(8), 2045–8 August (1977).

[2] Sulzmann, K., Lowder, J., and Penner, S. Combustion and Flame 20(2), 177–191 April

(1973).

[3] Hanson, R. K. Proceedings of the Combustion Institute 33(1), 1–40 (2011).

[4] Faist, J., Gmachl, C., Capasso, F., Sirtori, C., Sivco, D. L., Baillargeon, J. N., and Cho, A. Y.

Applied Physics Letters 70(20), 2670 May (1997).

[5] Wysocki, G., Lewicki, R., Curl, R., Tittel, F., Diehl, L., Capasso, F., Troccoli, M., Hofler, G.,

Bour, D., Corzine, S., Maulini, R., Giovannini, M., and Faist, J. Applied Physics B 92(3),

305–311 May (2008).

[6] Hildebrandt, L., Stry, S., Knispel, R., Sacher, J. R., Beyer, T., Braun, M., Lambrecht, A.,

Gensty, T., Elsäß er, W., Mann, C., and Fuchs, F. In Laser Applications to Chemical and

Environmental Analysis, TuF5 (Optical Society of America, Annapolis, Maryland, 2004).

[7] Saad, M. In SPIE OPTO, Epstein, R. I. and Sheik-Bahae, M., editors, 82750D–82750D–6.

International Society for Optics and Photonics, February (2012).

[8] Kriesel, J. M., Gat, N., Bernacki, B. E., Erikson, R. L., Cannon, B. D., Myers, T. L., Bledt,

C. M., and Harrington, J. A. In SPIE Defense, Security, and Sensing, 80180V. International

Society for Optics and Photonics, May (2011).

[9] Spearrin, R. M., Goldenstein, C. S., Jeffries, J. B., and Hanson, R. K. Measurement Science

and Technology 24(5), 055107 May (2013).

[10] Spearrin, R. M., Goldenstein, C. S., Jeffries, J. B., and Hanson, R. K. In 29th International

Symposium on Shock Waves (, Madison, WI, 2013).

93

94 BIBLIOGRAPHY

[11] Goldenstein, C. S., Spearrin, R. M., Jeffries, J. B., and Hanson, R. K. In 35th International

Symposium on Combustion, (2014).

[12] Mattison, D. W., Brophy, C. M., Sanders, S. T., Lin, M., Hinckley, K. M., Jeffries, J. B., and

Hanson, R. K. Journal of propulsion and power 19(4), 568–572 (2003).

[13] Klingbeil, A. E., Jeffries, J. B., and Hanson, R. K. Journal of Quantitative Spectroscopy and

Radiative Transfer 107(3), 407–420 October (2007).

[14] Goldenstein, C. S., Schultz, I. A., Spearrin, R. M., Jeffries, J. B., and Hanson, R. K. In

ICDERS, 24th, (2013).

[15] Caswell, A. W., Roy, S., An, X., Sanders, S. T., Hoke, J., Schauer, F., and Gord, J. R. In 50th

AIAA Aerospace Sciences Meeting, number January, 1–10, (2012).

[16] Mihalcea, R. M., Baer, D. S., and Hanson, R. K. Measurement Science and Technology 9(3),

327–338 March (1998).

[17] Sonnenfroh, D. M. and Allen, M. G. Applied Optics 36(15), 3298 May (1997).

[18] Mihalcea, R. M., Baer, D. S., and Hanson, R. K. Applied Optics 37(36), 8341 December

(1998).

[19] Webber, M. E., Claps, R., Englich, F. V., Tittel, F. K., Jeffries, J. B., and Hanson, R. K.

Applied Optics 40(24), 4395 August (2001).

[20] Rothman, L., Gordon, I., Barber, R., Dothe, H., Gamache, R., Goldman, a., Perevalov, V.,

Tashkun, S., and Tennyson, J. Journal of Quantitative Spectroscopy and Radiative Transfer

111(15), 2139–2150 October (2010).

[21] Farooq, A., Jeffries, J. B., and Hanson, R. K. Applied optics 48(35), 6740–53 December

(2009).

[22] Philippe, L. C. and Hanson, R. K. Optics Letters 16(24), 2002 December (1991).

[23] Philippe, L. C. and Hanson, R. K. Applied optics 32(30), 6090–103 October (1993).

[24] Farooq, A., Jeffries, J. B., and Hanson, R. K. Applied Physics B 96(1), 161–173 March

(2009).

BIBLIOGRAPHY 95

[25] Rieker, G. B., Jeffries, J. B., and Hanson, R. K. Applied Optics 48(29), 5546 October (2009).

[26] Platt, U. and Stutz, J. Differential Optical Absorption Spectroscopy: Principles and Applica-

tions. Springer, (2008).

[27] Kluczynski, P. and Axner, O. Applied Optics 38(27), 5803 September (1999).

[28] Bomse, D. S., Stanton, A. C., and Silver, J. A. Applied Optics 31(6), 718 February (1992).

[29] Li, H., Rieker, G. B., Liu, X., Jeffries, J. B., and Hanson, R. K. Applied Optics 45(5), 1052

February (2006).

[30] Chao, X., Jeffries, J. B., and Hanson, R. K. Measurement Science and Technology 20(11),

115201 November (2009).

[31] Hecht, J. Understanding fiber optics. Pearson/Prentice Hall, (2006).

[32] Siegman, A. E. Lasers. University Science Books, (1986).

[33] Farooq, A., Jeffries, J. B., and Hanson, R. K. Applied Physics B 96(1), 161–173 March

(2009).

[34] Farooq, A., Jeffries, J., and Hanson, R. Applied Physics B 90(3-4), 619–628 January (2008).

[35] Rieker, G., Liu, X., Li, H., Jeffries, J., and Hanson, R. Applied Physics B 87(1), 169–178

December (2006).

[36] Perrin, M. and Hartmann, J. Journal of Quantitative Spectroscopy and Radiative Transfer

42(4), 311–317 October (1989).

[37] Rieker, G. B., Jeffries, J. B., and Hanson, R. K. Applied Physics B 94(1), 51–63 November

(2008).

[38] Farooq, A., Jeffries, J., and Hanson, R. Journal of Quantitative Spectroscopy and Radiative

Transfer 111(7-8), 949–960 May (2010).

[39] Petersen, E. L., Röhrig, M., Davidson, D. F., Hanson, R. K., and Bowman, C. T. Symposium

(International) on Combustion 26(1), 799–806 January (1996).

[40] Davidson, D., Petersen, E., Hanson, R., and Bates, R. International Journal of Thermo-

physics 19(6), 1585–1594 November (1998).

96 BIBLIOGRAPHY

[41] Spearrin, R. M., Goldenstein, C. S., Jeffries, J. B., and Hanson, R. K. Applied Optics 53(9),

1938–1946 (2014).

[42] Nguyen, Q., Edgar, B., Dibble, R., and Gulati, A. Combustion and Flame 100(3), 395–406

(1995).

[43] Upschulte, B. L., Sonnenfroh, D. M., and Allen, M. G. Applied Optics 38(9), 1506 March

(1999).

[44] Cassidy, D. T. and Bonnell, L. J. Applied optics 27(13), 2688–93 July (1988).

[45] Sur, R., Sun, K., Jeffries, J. B., and Hanson, R. K. Applied Physics B July (2013).

[46] Wang, J., Maiorov, M., Baer, D. S., Garbuzov, D. Z., Connolly, J. C., and Hanson, R. K.

Applied Optics 39(30), 5579 October (2000).

[47] Namjou, K., Cai, S., Whittaker, E. A., Faist, J., Gmachl, C., Capasso, F., Sivco, D. L., and

Cho, A. Y. Optics Letters 23(3), 219 February (1998).

[48] Ren, W., Farooq, A., Davidson, D. F., and Hanson, R. K. Applied Physics B 107(3), 849–860

May (2012).

[49] Stefanski, P., Lewicki, R., Tarka, J., Ma, Y., Jahjah, M., and Tittel, F. K. In CLEO: 2013,

JW2A.68 (OSA, Washington, D.C., 2013).

[50] Vanderover, J., Wang, W., and Oehlschlaeger, M. a. Applied Physics B 103(4), 959–966 May

(2011).

[51] Reynolds, W. C. Technical report, Stanford University, (1986).

[52] Predoi-Cross, A., Luo, C., Sinclair, P., Drummond, J., and May, A. Journal of molecular

spectroscopy 198(2), 291–303 December (1999).

[53] Hartmann, J. M., Rosenmann, L., Perrin, M. Y., and Taine, J. Applied optics 27(15), 3063–5

August (1988).

[54] Varghese, P. and Hanson, R. Journal of Quantitative Spectroscopy and Radiative Transfer

24(6), 479–489 (1980).

[55] Varghese, P. and Hanson, R. Journal of Quantitative Spectroscopy and Radiative Transfer

26(4), 339–347 (1981).

BIBLIOGRAPHY 97

[56] Bonamy, J., Robert, D., and Boulet, C. Journal of Quantitative Spectroscopy and Radiative

Transfer 31(1), 23–34 (1984).

[57] Pack, R. T. The Journal of Chemical Physics 70(7), 3424 April (1979).

[58] Goldenstein, C. S., Spearrin, R. M., Jeffries, J. B., and Hanson, R. K. Applied Physics B

January (2014).

[59] Spearrin, R. M., Ren, W., Jeffries, J. B., and Hanson, R. K. Applied Physics B (2014).

[60] Ren, W., Spearrin, R. M., Davidson, D. F., and Hanson, R. K. The journal of physical

chemistry. A 118(10), 1785–98 March (2014).

[61] Ren, W., Spearrin, R. M., Davidson, D. F., and Hanson, R. K. In The 8th National Combustion

Mtg, 11 (, Salt Lake City, Utah, 2013).

[62] Teichert, H., Fernholz, T., and Ebert, V. Applied Optics 42(12), 2043 April (2003).

[63] Hodgkinson, J. and Tatam, R. P. Measurement Science and Technology 24(1), 012004 Jan-

uary (2013).

[64] Ren, W., Jeffries, J. B., and Hanson, R. K. Measurement Science and Technology 21(10),

105603 October (2010).

[65] Zhou, X., Liu, X., Jeffries, J. B., and Hanson, R. K. Measurement Science and Technology

14(8), 1459–1468 August (2003).

[66] Schultz, I. A., Goldenstein, C. S., Jeffries, J. B., and Hanson, R. K. In 28th Aerodynamic

Measurement Technology,Ground Testing, and Flight Testing Conference, (2012).

[67] Pyun, S. H., Porter, J. M., Jeffries, J. B., Hanson, R. K., Montoya, J. C., Allen, M. G., and

Sholes, K. R. Applied optics 48(33), 6492–500 November (2009).

[68] Chao, X., Jeffries, J. B., and Hanson, R. K. Proceedings of the Combustion Institute 33(1),

725–733 (2011).

[69] Rosenmann, L., Hartmann, J. M., Perrin, M. Y., and Taine, J. Applied optics 27(18), 3902–6

September (1988).

[70] Thibault, F., Calil, B., Buldyreva, J., Chrysos, M., Hartmann, J.-M., and Bouanich, J.-P.

Physical Chemistry Chemical Physics 3(18), 3924–3933 January (2001).

98 BIBLIOGRAPHY

[71] Oehlschlaeger, M. A., Davidson, D. F., and Hanson, R. K. The Journal of Physical Chemistry

A 108(19), 4247–4253 May (2004).

[72] Stranic, I., Chase, D. P., Harmon, J. T., Yang, S., Davidson, D. F., and Hanson, R. K. Com-

bustion and Flame 159(2), 516–527 (2012).

[73] Farooq, A., Davidson, D., Hanson, R., Huynh, L., and Violi, A. Proceedings of the Combus-

tion Institute 32(1), 247–253 (2009).

[74] Huynh, L. K., Lin, K. C., and Violi, A. The journal of physical chemistry. A 112(51), 13470–

80 December (2008).

[75] Stranic, I., Davidson, D. F., and Hanson, R. K. Chemical Physics Letters 584, 18–23 (2013).

[76] Spearrin, R. M., Goldenstein, C. S., Schultz, I. A., Jeffries, J. B., and Hanson, R. K. Applied

Physics B (2014).

[77] Spearrin, R. M., Jeffries, J. B., and Hanson, R. K. In 52nd Aerospace Sciences Meeting

(SciTech 2014) (, National Harbor, MD, 2014).

[78] Schultz, I. A., Goldenstein, C. S., Spearrin, R. M., Jeffries, J. B., and Hanson, R. K. Journal

of Propulsion and Power (2014).

[79] Tishkoff, J. M., Drummond, J. P., Edwards, T., and Nejad, A. In 35th Aerospace Sciences

Meeting and Exhibit, number January (, Reno, NV, 1997).

[80] Rieker, G., Jeffries, J., Hanson, R., Mathur, T., Gruber, M., and Carter, C. Proceedings of the

Combustion Institute 32(1), 831–838 (2009).

[81] Liu, J. C., Rieker, G., Jeffries, J., Gruber, M., Carter, C., Mathur, T., and Hanson, R. Applied

Optics 44(31), 6701 November (2005).

[82] Li, F., Yu, X., Gu, H., Li, Z., Zhao, Y., Ma, L., Chen, L., and Chang, X. Applied Optics

50(36), 6697–707 December (2011).

[83] Griffiths, A. and Houwing, A. Applied Optics 44, 6653–6659 (2005).

[84] Rosenmann, L., Langlois, S., Delaye, C., and Taine, J. Journal of Molecular Spectroscopy

149(1), 167–184 September (1991).

BIBLIOGRAPHY 99

[85] Goldenstein, C. S., Strand, C. L., Schultz, I. A., Sun, K., Jeffries, J. B., and Hanson, R. K.

Applied Optics 53(3), 356–367 (2014).

[86] Goldenstein, C. S., Schultz, I. A., Jeffries, J. B., and Hanson, R. K. Applied optics 52(33),

7950–62 November (2013).

[87] Ouyang, X. and Varghese, P. L. Applied optics 28(18), 3979–84 September (1989).

[88] Sanders, S. T., Wang, J., Jeffries, J. B., and Hanson, R. K. Applied Optics 40(24), 4404

(2001).

[89] Ma, L., Li, X., Sanders, S. T., Caswell, A. W., Roy, S., Plemmons, D. H., and Gord, J. R.

Optics express 21(1), 1152–62 January (2013).

[90] McDaniel, J. C., Goyne, C. P., Bryner, E. B., Le, D. B., T, S. C., and Krauss, R. H. In AIP

Conference Proceedings, volume 762, 1277–1282. AIP, May (2005).

[91] Gruber, M. and Donbar, J. Journal of Propulsion and Power 20, 769–778 (2004).

[92] Ben-Yakar, A. and Hanson, R. Journal of Propulsion and Power 17, 869–877 (2001).

[93] Goldenstein, C. S., Schultz, I. A., Spearrin, R. M., Jeffries, J. B., and Hanson, R. K. Applied

Physics B (2013).

[94] Sun, K., Chao, X., Sur, R., Goldenstein, C. S., Jeffries, J. B., and Hanson, R. K. Measurement

Science and Technology 24(12), 125203 December (2013).

[95] Henry, C. and Kazarinov, R. IEEE Journal of Quantum Electronics 22(2), 294–301 February

(1986).

[96] Shibukawa, A., Katsui, A., Iwamura, H., and Hayashi, S. Electronics Letters 13(24), 721

November (1977).

[97] DeShazer, L. G. Review of Scientific Instruments 38(2), 248 December (1967).

[98] Kelly, J. F., Sams, R. L., Blake, T. A., and Kriesel, J. M. In SPIE OPTO, Razeghi, M.,

Tournié, E., and Brown, G. J., editors, 89931O. International Society for Optics and Photon-

ics, December (2013).

[99] Sands, P. J. Journal of the Optical Society of America 60(11), 1436 November (1970).

100 BIBLIOGRAPHY

[100] Kanamori, T., Terunuma, Y., Takahashi, S., and Miyashita, T. Journal of Lightwave Technol-

ogy 2(5), 607–613 October (1984).

[101] Nishii, J., Morimoto, S., Inagawa, I., Iizuka, R., Yamashita, T., and Yamagishi, T. Journal of

Non-Crystalline Solids 140, 199–208 January (1992).

[102] Cheng, J., Pedersen, M. E. V., Wang, K., Xu, C., Grüner-Nielsen, L., and Jakobsen, D. Optics

letters 37(3), 347–9 February (2012).

[103] Kriesel, J. M., Hagglund, G. M., Gat, N., Spagnolo, V., and Patimisco, P. In SPIE OPTO,

Razeghi, M., Tournié, E., and Brown, G. J., editors, 89930V. International Society for Optics

and Photonics, December (2013).

[104] Thomas, M. E., Joseph, R., and Tropf, W. J. Applied Optics 27(2), 239–245 (1988).

[105] Harris, T. R. Optical Properties of SI, GE, GAAS, GASB, INAS, and INP at elevated temper-

atures. PhD thesis, Air Force Institute of Technology, (2010).

mid-infrared laser absorption spectroscopy for...

Documents