plasma assisted combustion: fundamental studies and ......i remember so vividly the night you agreed...
TRANSCRIPT
Plasma Assisted Combustion:
Fundamental Studies and Engine Applications
Joseph K. Lefkowitz
A DISSERTATION
PRESENTED TO THE FACULTY
OF PRINCETON UNIVERSITY
IN CANDIDACY FOR THE DEGREE
OF DOCTOR OF PHILOSOPHY
RECOMMENDED FOR ACCEPTANCE
BY THE DEPARTMENT OF
MECHANICAL AND AEROSPACE ENGINEERING
Adviser: Yiguang Ju
January 2016
© Copyright by Joseph K. Lefkowitz, 2015. All rights reserved.
iii
Abstract
Successful and efficient ignition in short residence time environments or ultra-lean mixtures is
a key technological challenge for the evolution of advanced combustion devices in terms of both
performance and efficiency. To meet this challenge, interest in plasma assisted combustion
(PAC) has expanded over the past 20 years. However, understanding of the underlying physical
processes of ignition by plasma discharge remains elementary. In order to shed light on the key
processes involved, two main thrusts of research were undertaken in this dissertation. First,
demonstration of the applicability of plasma discharges in engines and engine-like environments
was carried out using a microwave discharge and a nanosecond repetitively pulsed discharge in
an internal combustion engine and a pulsed detonation engine, respectively. Major conclusions
include the extension of lean ignition limits for both engines, significant reduction of ignition
time for mixtures with large minimum ignition energy, and the discovery of the inter-pulse
coupling effect of nanosecond repetitively pulsed (NRP) discharges at high frequency.
In order to understand the kinetic processes that led to these improvements, the second thrust
of research directly explored the chemical kinetic processes of plasma discharges with
hydrocarbon fuels. For this purpose, a low pressure flow reactor with a NRP dielectric barrier
discharge cell was assembled. The discharge cell was fitted with a Herriott type multipass mirror
arrangement, which allowed quantitative laser absorption spectroscopy to be performed in situ
during the plasma discharge. Experiments on methane and ethylene mixtures with oxygen,
argon, and helium revealed the importance of low temperature oxidation pathways in PAC. In
particular, oxygen addition reactions were shown to be of primary importance in the oxidation of
these small hydrocarbons in the temperature range of 300-600 K. Kinetic modeling tools,
including both a coupled plasma and combustion chemistry solver and appropriate reaction
iv
models, were developed and compared to the experimental results, revealing excellent agreement
for major fuel consumption pathways, but significant disagreement in the predictions of smaller
concentration products. The individual reactions responsible for the observed disagreements
were identified, and directions for further research are discussed.
v
Acknowledgements
Throughout my journey in graduate school, there have been many people who were there to
support me during difficult times and celebrate with me during good times. Excellent professors,
research scientists, and fellow graduate students have passed through the halls of Princeton
University during my time here, and have helped me build my understanding of science, life, and
friendship which have defined my graduate experience.
First and foremost, I would like to thank my adviser, Professor Yiguang Ju, for the years of
support, stimulating conversations, and encouragement to think deeper and work harder.
Although we did not always see eye to eye, I learned more from you than any other person in my
academic experience. You taught me how to think from a scientific perspective, to boldly
approach new research topics, and to never limit myself to what I know, but to push myself to
understand what I don’t know. Growth is painful, it requires hard work, long hours, dedication
despite doubts, mistakes made and corrected, and sometimes losing one’s way. In my years in
graduate school I have grown more than I ever thought I could. Looking back, I can only thank
you for believing in me and keeping me on track, even when I felt hopelessly behind.
Next, I want to thank my family, my parents Susan and Harry, my sisters Melissa and
Madeline, and my brother Jay. I have been so lucky to have you nearby these past years in New
Jersey. Thank you for taking the time to drive down to Princeton when I was too busy to drive
up. Every holiday, birthday, and spontaneous weekend when I could see you and relax, if only
for a few of days, meant so much to me. You have always supported me, believed in me, and
encouraged me to follow my dreams. Even as a young child playing with Legos in my room, you
knew that one day I would become an engineer, and you never let me turn back on that dream.
Thank you so much.
vi
The next person I need to acknowledge, as my constant mentor and compatriot, is Dr. Sang
Hee Won. From my first days in the laboratory you were there to guide me in my daily efforts,
give me honest advice, and let me know when I was being “hard working and stupid” or “not
hard working and smart.” Over many late nights in the lab, papers written and rewritten, and
bottles of beer at the lunch table, you helped me become an experimentalist and an honest
researcher. Thank you for a lasting friendship, and for being one of the “left behind,” because I
couldn’t have made it without you.
Next, I want to thank all of the professors who educated me at this excellent institution. First,
I would like to thank Professor Fred Dryer for the many hours of interesting and thought
provoking conversations during my first few years in graduate school. You stimulated me to
always think about the impact of my research, and to always take a step back and ask, “Is this
really the controlling phenomena I should be looking at?” I was terrified to face you in the
general exam, but in the end it was a rewarding experience and I am happy I had a chance to
show you who I was on the blackboard. Next, I want to thank Professor Richard Miles, who was
my research general examiner and thankfully let me go even though I knew so little. You were
there every year for committee meetings and at the yearly AIAA Aerospace Sciences Meetings,
watching my work progress and always knowing the next logical step for the following year.
Also, thanks for introducing me to optics and lasers in your class, which became so important to
the work in this thesis. I want to thank Professor Chung K. Law, for teaching thought provoking
combustion classes and working with me during my general exam interviews. I would like to
thank Professors Szymon Suckewer, Garry Brown, Howard Stone, and Phillip Holmes for their
excellent lectures and tough homework assignments. I have never learned so much in such a
short time as I did in your classes. Thank you for teaching me how to turn “doing my best” into
vii
“doing better than I ever imagined,” even though my grades didn’t always show it. Lastly, I
would like to thank the two professors who read and helped improve my thesis, Professors
Michael Mueller and Christophe Laux. Their helpful comments have made this a significantly
better body of work.
I would now like to thank all of the research scientists that taught me so much inside and
outside the laboratory. Dr. Pascal Diévart, more than someone to turn to for any chemistry
question, but also a great friend and the kindest person I have ever met. Dr. Mruthunjaya Uddi,
thanks for teaching me how to do absorption spectroscopy and think about measurement
methods. Professor Bret Windom, thank you for helping me with the most difficult time in my
graduate years, and always having a smile ready. Prof. Stephen Dooley, who always knew how
to bring people together and was a true inspiration for my first years of graduate studies. Dr.
Tomoya Wada, an excellent experimentalists and even better compatriot, thanks for being a good
friend. Dr. Brian Brumfield, for teaching me so much about spectroscopy, hard work, how to be
accurate in the field of diagnostics, and how to persevere in the world of science. Also, Dr. Peter
“Victor” Veloo, Tanvir, Prof. Tanvir “Farouk Farouk” Farouk, Dr. Hossam El-Asrag, Prof. Guo-
feng Lou, and Dr. Xueliang Yang, for your helpful advice over the years. Lastly, I would like to
the Dr. Timothy Ombrello, who invited me to work with him for a summer at the Air Force
Research Lab, an opportunity which opened the door to my next position working under his
advisement at Wright Paterson Air Force Base.
Next, I want to thank all of my fellow graduate students, without whom I wouldn’t have made
it past the first week of graduate school. My labmates Joshua Heyne, Jeff Santner, Hwanho Kim,
Wenting Sun, Mac Haas, Mike Burke, Weiqi Sun, Chris Reuter, Aric Rousso, Hao Zhao, Yann
Fenard, and Peng Guo. Also, my good friends Matthew Girardi, Tat Loon Chng, Chris Limbach,
viii
Fred Brasz, Jessica Shang, Brian Rosenberg, Leo Hallstrom, Tristen Hohman, Celine Stein, Peng
Zhao and many others who made graduate school fun, were there to work with on problem sets
or talk to about research, or who just wanted to kick back and drink a beer, play soccer, or climb
some rocks. I know I am forgetting a lot of you, but just know that this has been a great time, and
you have become some of my greatest and closest friends.
I would also like to thank the technicians and machinists without whom very little
experimental work would ever get done. Timothy Bennett, Joe Sivo, Larry McIntyre and Barry
Runner.
Finally, to Kristine, the love of my life, I owe you more thanks than I can ever say. In my first
year, when I had no car and too much work to take the train every day, you drove down and had
dinner with me at least twice a week. When you moved in with me the next year, on Humbert
Street, I had never been so happy. You sacrificed so much over the years, commuting at least two
hours every day so we could be together. I remember so vividly the night you agreed to marry
me, during my third year at Princeton. You made me the luckiest man on earth. You supported
me when I didn’t think I had a chance of making it through this degree. When I was up all night
trying to finish a paper or when I had no idea what was going on with an experiment, you always
told me I could do it. You have no idea how much that support has meant to me. Thank you
Kristine, for making me happier than I ever thought I could be.
This dissertation carries T3309 in the records of the Department of Mechanical and Aerospace
Engineering.
ix
To Kristine Lefkowitz, my soul mate, for your unwavering love and support.
x
Table of contents
Abstract ......................................................................................................................................... iii
Acknowledgements ....................................................................................................................... v
1 Introduction ........................................................................................................................... 1
1.1 Motivation ........................................................................................................................ 1
1.2 Plasma-assisted combustion concepts .............................................................................. 5
1.2.1 Types of plasmas....................................................................................................... 5
1.2.2 Electron collision processes ...................................................................................... 8
1.2.3 Reactions of plasma-produced species ................................................................... 10
1.2.4 Enhancement mechanism for plasma-assisted ignition .......................................... 15
1.3 Background of plasma-assisted combustion and ignition .............................................. 16
1.3.1 Early work ............................................................................................................... 17
1.3.2 Recent advances in PAC and PAI ........................................................................... 21
1.4 Technical questions remaining in PAC and PAI ............................................................ 29
1.5 Objectives of dissertation ............................................................................................... 31
1.6 Publication history.......................................................................................................... 33
1.6.1 Journal publications ................................................................................................ 33
1.6.2 Conference papers and presentations ...................................................................... 34
2 Plasma-assisted ignition for internal combustion engine and pulsed detonation engine
applications .................................................................................................................................. 37
2.1 Spark-assisted microwave discharge ignition in a small internal combustion engine ... 39
2.1.1 Introduction ............................................................................................................. 39
2.1.2 Experimental configuration .................................................................................... 41
2.1.2.1 Microwave enhanced ignition system ................................................................. 41
2.1.2.2 Engine platform ................................................................................................... 43
2.1.3 Results and discussion ............................................................................................ 46
2.1.4 Conclusions ............................................................................................................. 54
2.2 Nanosecond repetitively pulsed discharges: Inter-pulse coupling and ignition
enhancement in PDEs................................................................................................................ 56
2.2.1 Introduction ............................................................................................................. 56
2.2.2 Experimental setup.................................................................................................. 60
xi
2.2.2.1 Plasma igniter ..................................................................................................... 60
2.2.2.2 Pulsed detonation engine .................................................................................... 63
2.2.2.3 Flame-development visualization platform ......................................................... 64
2.2.2.4 Numerical modeling ............................................................................................ 67
2.2.3 Results and discussion ............................................................................................ 69
2.2.3.1 Flame-development visualization ........................................................................ 69
2.2.3.2 PDE testing ......................................................................................................... 86
2.2.4 Conclusions ............................................................................................................. 90
3 Low temperature dielectric barrier discharge reactor integrated with mid-infrared
TDLAS ......................................................................................................................................... 95
3.1 Non-equilibrium plasma reactor..................................................................................... 95
3.1.1 Reactor configuration.............................................................................................. 95
3.1.2 Discharge characteristics ........................................................................................ 99
3.1.3 Gas chromatography ............................................................................................. 104
3.2 Direct absorption spectroscopy .................................................................................... 106
3.2.1 Introduction ........................................................................................................... 106
3.2.2 Spectra of atoms and molecules ............................................................................ 107
3.2.3 Einstein theory of radiation ................................................................................... 112
3.2.4 The Beer-Lambert Law (or Beer’s Law) .............................................................. 115
3.2.5 Spectral lineshapes ................................................................................................ 122
3.2.5.1 Doppler broadening .......................................................................................... 122
3.2.5.2 Natural line broadening .................................................................................... 124
3.2.5.3 Collision broadening ......................................................................................... 125
3.2.5.4 Voigt profile....................................................................................................... 133
3.2.6 Fitting absorption profiles ..................................................................................... 136
3.2.7 TDLAS experimental configuration ..................................................................... 139
4 Numerical methods for plasma assisted combustion modeling ..................................... 146
4.1 Electron collision processes ......................................................................................... 146
4.2 Plasma-assisted Combustion Solver ............................................................................ 149
4.3 Plasma kinetic model ................................................................................................... 153
4.3.1 Ethylene model ..................................................................................................... 153
xii
4.3.2 Methane model...................................................................................................... 155
4.3.3 Methane and ethylene model for air plasmas ....................................................... 159
5 Kinetic study of dissociation and oxidation of ethylene in a NRP DBD flow reactor . 162
5.1 Introduction .................................................................................................................. 163
5.2 Results and discussion .................................................................................................. 165
5.2.1 Ethylene dissociation ............................................................................................ 165
5.2.2 Ethylene oxidation ................................................................................................ 170
5.3 Conclusions .................................................................................................................. 174
6 Kinetic study of methane oxidation in a NRP DBD flow reactor .................................. 176
6.1 Introduction .................................................................................................................. 177
6.2 Results and discussion .................................................................................................. 181
6.2.1 Transient measurements........................................................................................ 181
6.2.2 Steady-state measurements ................................................................................... 184
6.3 Conclusions .................................................................................................................. 195
7 Summary and future direction ......................................................................................... 198
7.1 Summary ...................................................................................................................... 198
7.2 Recommendations for future work ............................................................................... 203
7.2.1 Further exploration of PAC-related reactions ....................................................... 203
7.2.1.1 The reaction of O(1D) with hydrocarbons......................................................... 203
7.2.1.2 Electron collision cross-sections for large hydrocarbons ................................ 204
7.2.1.3 Energy transfer processes ................................................................................. 204
7.2.2 Further development of modeling approaches ...................................................... 206
7.2.3 Investigation of inter-pulse coupling .................................................................... 207
7.2.4 Advanced diagnostics for species and temperature measurements ...................... 209
7.2.5 Potential applications for PAC.............................................................................. 210
Appendix .................................................................................................................................... 213
A1. Solution of the Boltzmann Equation ................................................................................ 213
A2. Chemical kinetic study of tertiary-butanol in a flow reactor and a counterflow diffusion
flame ...................................................................................................................................... 223
A2.1 Introduction .............................................................................................................. 224
A2.2 Experimental methodology ....................................................................................... 226
xiii
A2.2.1 Variable Pressure Flow Reactor ........................................................................ 226
A2.2.2 Sample analysis ................................................................................................. 229
A2.2.3 Counterflow diffusion flame extinction and speciation .................................... 230
A2.3 Results and discussion .............................................................................................. 234
A2.3.1 t-Butanol oxidation in the Variable Pressure Flow Reactor .............................. 234
A2.3.2 Extinction limits and species profiles in the counterflow diffusion flame ........ 240
A2.3.2.1 Species profile measurements ...................................................................... 240
A2.3.2.2 Extinction limit measurements ..................................................................... 242
A2.3.3 Model comparison ............................................................................................. 244
A2.3.4 Sub-model analysis ........................................................................................... 248
A2.3.4.1 Acetone ......................................................................................................... 248
A2.3.4.2 Isobutene ..................................................................................................... 249
A2.4 Concluding remarks .................................................................................................. 251
A3. Uncertainty assessment of species measurements in acetone counterflow diffusion
flames ..................................................................................................................................... 253
A3.1 Introduction .............................................................................................................. 253
A3.2 Experiments and numerical simulations ................................................................... 255
A3.3 Results and discussion .............................................................................................. 258
A3.3.1 Discrepancy in species distribution measurements ........................................... 258
A3.3.2 Effects of separation distance ............................................................................ 259
A3.3.3 Quantification of the flow perturbation due to the sampling probe .................. 263
A3.3.4 Speciation profiles of intermediate species ....................................................... 267
A3.4 Concluding remarks .................................................................................................. 270
References .................................................................................................................................. 271
1
1 Introduction
1.1 Motivation
The motivation for plasma-assisted combustion (PAC) and plasma-assisted ignition (PAI)
studies is to enable advanced engine concepts. While this is quite a general motivating statement,
application of plasma devices have indeed been proposed for a wide variety of engine types, such
as internal combustion (IC) engines, gas turbines, scramjets, and pulsed and rotating detonation
engines. Several excellent review papers on plasma-assisted combustion are available
(Starikovskaia 2006, Starikovskii et al. 2006, Adamovich et al. 2009a, Starikovskiy et al. 2013,
Starikovskaia 2014, Ju et al. 2015b), which detail these applications and the fundamental science
that enables them. One well-known example of plasma-assisted combustion in use today is the
spark plug found in every spark ignition (SI) IC engine. Despite redesigns of the electronics of
this simple device over the one hundred years of automotive history (Heywood 1988), the type of
plasma and effect on combustion are the same now as they were in the earliest automotive
engines. This leads one to question why further study on plasma ignition devices for SI engines
is necessary at all. The answer is that plasma may take on many forms that differ greatly in their
effect on combustion processes, many of which have not been fully explored. The millisecond
duration equilibrium arc discharge used for SI engines may not be the optimal solution for
ignition of difficult to ignite mixtures and certainly provides very few options for control of the
ignition event (Maly 1984). As increasing demands are placed on combustion engineers to
produce devices that are more efficient and produce fewer emissions in light of the recent
concerns of climate change and public health, respectively, novel solutions for engine designs
are required that often depart from the traditional solutions.
2
The specific type of plasma device depends on the application, and the physics governing the
interaction between the plasma and the combustion processes can be drastically different for
each. A simple Venn diagram can illustrate the various interactions between plasma and
combustion, as shown in Figure 1.1.1.
Figure 1.1.1 Venn diagram of plasma and combustion interaction
The plasma discharge will produce three main effects that can interact with combustion gases:
the electric field, joule heating, and electron collision reactions. Depending on the type of
plasma, the dominance of each of these factors can be enhanced or repressed. The direct
interaction between plasma and chemical kinetics involves changes in the reaction pathways,
reaction rates, and ultimately the heat release rate. This can occur due to joule heating, which
increases the rate of Arrhenius reactions, or can occur due to electron collision reactions, which
Flame DynamicsChemical Kinetics
Plasma Physics
Extinction Ignition Flame speed Flame instability
Reaction pathways Reaction ratesHeat release rate
Electric field Joule heatingElectron collision reactions
Charged species Excited species
Plasmacombustion
studies
Ionic wind
Traditionalcombustion
studies
3
produce radicals, ions, and excited species that open new reaction pathways or accelerate
existing pathways by bypassing initiation steps (i.e. radical generation or fuel decomposition).
The direct interaction between flame dynamics and plasmas is most clearly demonstrated by the
“ionic wind” effect, in which ions produced by chemi-ionization in the flame are manipulated by
the electric field, influencing the position of the flame front. In general, the specific effects of
plasma on chemical kinetics and flame dynamics are difficult to separate, and most studies of
applications of plasma devices lies at the intersection of plasma physics, chemical kinetics, and
flame dynamics.
Due to the complexity of the multiple effects acting in tandem in plasma-assisted combustion,
scientific studies in this area often strive to isolate specific interactions. In order to do so, the
plasma type, gas mixture, and flow environment are manipulated to exaggerate the desired effect.
By isolating each effect, some of the interactions which define the full nature of PAC may be
lost. Unfortunately, until a greater understanding of the individual interactions of plasma with
kinetics and flame dynamics are achieved, the full interaction cannot be resolved. The subject of
primary interest for this thesis is the kinetic interaction between plasma produced species and the
gas molecules in combustible fuel/oxidizer mixtures. The method for separating the chemical
kinetic interactions from other phenomena will be described in detail in Chapters 4-6, and
involves the use of nanosecond duration plasma discharges, which produce high voltages but
only for a short duration, and are efficient in channeling energy into the electronic degrees of
freedom of gas molecules without significant joule heating. Chemical systems in both
combustion and plasma science are independently well developed fields; however, studies on
how plasma-produced species react with molecules typical of combustion systems, particularly
4
hydrocarbons, is a newly developed area with relatively few detailed studies available, most of
which are summarized in the recent review article by Starikovskaia (Starikovskaia 2014).
The interest in studying the chemical interaction between plasma and fuel mixtures stems
from the observation of significantly altered combustion properties when flames are exposed to
plasma discharges, which cannot be described by only considering the thermal contribution of
the discharge on the gas mixture. These include reduced ignition times in shock tube ignition
delay studies under the influence of low energy discharges (Bao et al. 2007, Lou et al. 2007,
Mintusov et al. 2009, Starikovskaya et al. 2009, Yin et al. 2011), increased flame speeds under
the influence of sub-critical microwave fields (Stockman et al. 2009, Michael et al. 2013),
increased extinction limits for counterflow flames using both gliding arc and nanosecond
repetitively pulsed (NRP) discharges (Ombrello et al. 2006, Ombrello et al. 2008a, Ombrello et
al. 2008b, Sun et al. 2011, Sun et al. 2012), and the “stretched” S-curve observed in counterflow
flames under the influence of an NRP discharge, which removes the hysteresis between ignition
and extinction of diffusion flames, creating a cool flame zone which can smoothly transition
between frozen and fully reacting mixtures (Sun et al. 2013, Sun et al. 2014). In all of these
studies, it is suggested that the kinetic interaction between plasma produced species and the
neutral gas molecules is primarily responsible for the observed effects. However, only several
studies have focused on the oxidation mechanisms at play and quantitative comparisons between
species produced in the PAC systems interactions, and kinetic modeling efforts are severely
lacking, especially at low temperature. In fact, in a recent paper describing the future direction of
PAC (Samukawa et al. 2012), quantitative species measurements and direct comparisons with
modeling efforts is identified as one of the necessary directions to continue the development of
PAC understanding. With the above considerations in mind, the purpose of this thesis will be to
5
provide quantitative data and detailed modeling to evaluate and expand the current understanding
of PAC kinetics.
1.2 Plasma-assisted combustion concepts
1.2.1 Types of plasmas
Due to the recent review papers available on the subject of PAC/PAI, mentioned in the
previous section, as well as excellent reference books (Raizer 1991, Capitelli et al. 2000,
Fridman 2008), only a brief description of the governing physical processes will be discussed
herein. In general, plasma discharge types can be separated into two categories: equilibrium
(thermal) plasmas or non-equilibrium plasmas. Equilibrium plasmas include arc discharges and
plasma torches while non-equilibrium discharges cover a wide variety of types of plasmas, but
can be typically characterized by large reduced electric fields and short time scales (or high
frequencies) that do not provide sufficient time for the internal energies of gas species
(rotational, vibrational, and electronic) to equilibrate before the applied field is changed. The
reduced electric field is the electric field E divided by the gas density N. This parameter is useful
for describing plasma energy because it is the controlling feature that governs the electron energy
(Fridman 2008), which in turn determines what types of interactions the electrons will have with
the surrounding gas. Recently, non-equilibrium plasmas have been the focus of PAC studies due
to their ability to channel energy into specific internal excitation modes of molecules beneficial
for active species generation without expending energy on gas and electrode heating
(Starikovskiy et al. 2013, Starikovskaia 2014, Ju et al. 2015b).
The non-equilibrium plasmas used for combustion studies are variously produced by dc, ac,
radio frequency, microwave, and NRP voltage waveforms, as well as laser induced breakdown.
6
They may take the form of corona discharges, glow discharges, spark discharges, gliding arc
discharges, dielectric barrier discharges, and sub-critical electric fields (below the breakdown
threshold for discharge formation). Each type can be used with different electrode
configurations, which have a strong influence on the plasma properties, with the main
considerations being the discharge gap distance, geometry of the electrodes (which determines
the localization of the discharge region), and material of construction. In addition, the gas
properties will also influence the discharge type, with the main properties being pressure,
temperature, internal energy configurations of the particular atoms and/or molecules in the gas,
and flow velocity. A summary of the types of plasmas and their respective electron temperature
(mean electron energy) and electron number density is presented in Figure 1.2.1. This wide range
of variables effecting the discharge properties has the double edged effect of not only producing
many opportunities for applications and scientific investigations of PAC but also often limiting
the conclusions of such investigations to a small range of plasmas similar to the ones used in
each particular investigation unless special care is taken to isolate specific effects. For example,
there have been a number of studies on scramjet ignition using various plasma devices. One
example is a study which used a plasma torch for ignition. Due to the high gas flow rate and gas
temperature associated with this type of plasma, a shock structure is set up in front of the
injection point, perturbing the flow field in the entire cavity. In another study using nanosecond
pulsed discharges, only a small perturbation was made in the flow, with relatively low gas
temperature. However, both devices were found to enhance scramjet ignition over capacitive
spark discharges, which are yet another type of plasma. Clearly these must have different
mechanisms for enhancing ignition. We wish to find out which plasma types are most effective
and energy efficient, and what physical processes determine their results.
7
Figure 1.2.1 Electron temperature and electron number density of various plasma types
typical in PAC studies (Ju et al. 2015b).
The studies presented herein will focus on a handful of plasma types. The applied ignition
studies of Chapter 2 will focus on the use of microwave discharges and NRP discharges in pin-
to-pin configurations, while Chapters 4-6 will use nanosecond repetitively pulsed discharges in a
double dielectric barrier discharge (DBD) configuration. The benefits of the NRP DBD
experiment for kinetic studies are that high E/N values (> 100 Td) can be obtained, which are
necessary for efficient active species generation, without significant energy being channeled into
joule heating. This allows separation of the kinetic and thermal enhancement effects. In addition,
due to the short duration of the discharge (~20 ns FWHM) and the relatively long intervals
between pulses (~10 µs) the influence of the electric field on bulk gas motion is negligible.
While isolating these effects is beneficial for the purposes of kinetic studies, it is undesirable for
practical applications, in which coupled thermal and kinetic enhancement is most effective. At
8
these conditions, the effect of plasma produced species on the combustion reaction mechanism is
targeted, while the effects of joule heating and electric field are minimized.
1.2.2 Electron collision processes
The main reactions during the discharge can be categorized as electron collision excitation,
dissociative excitation, ionization, and dissociative ionization reactions. After the plasma we
have quenching of excited states, recombination and dissociative recombination of ions and
electrons, as well as reactions between excited and ionized species with ground state species. All
of these occur simultaneously with the traditional combustion reactions of ground state species
with each other. Additional reaction processes certainly occur, such as ion-ion reactions and
electron attachment reactions. However, for the purpose of this study, these reactions are not of
primary importance due to the low ion concentration and the high energy of the electrons,
preventing significant attachment reactions from occurring. Further studies are needed to confirm
this assumption.
Electron collision processes are the first step in the chain of phenomena controlling the
interaction of plasma and combustion chemistry. In any gas there always exist a small number of
electrons and ions. When an electric field is applied, both of these particles accelerate along the
electric field lines. However, due to the small mass of electrons (5.489×10−4
atomic mass units),
these gain kinetic energy much more effectively than ions (Fridman 2008). When they collide
with other gas particles the collision will result in either elastic or inelastic interactions. Elastic
collisions contribute to gas heating, and are also important when calculating the mean energy of
the electrons (Raizer 1991), as will be discussed in Chapter 3. Inelastic collisions produce much
richer phenomena in terms of the possible reaction products. The nature of the inelastic collision
9
is dependent on the mean energy of the electrons, which is dependent on the E/N of the
experiment, as well as the available energy states of the collision partner. If the collision results
in ionization of the gas particle, an additional electron will be released. When a large enough
fraction of collisions result in ionization, we have an exponential growth in the electron (and ion)
concentration in the gas. This is called an electron cascade, and results in a Townsend
breakdown (Fridman 2008) in which some portion of the gas in the field is ionized. After the
breakdown, many electrons and ions are available to interact with neutral gas species.
In typical PAC mixtures, generally containing O2, hydrogen or hydrocarbon fuel, and some
diluent gas such as N2, Ar, or He, the dissipation of electron energy can be generally described as
a function of the E/N, although the details vary. The most commonly used unit for reduced
electric field (E/N) is Townsends (Td) which are equal to 10-17
V•cm2. For very low reduced
electric fields (< 1 Td) the majority of the electron energy is dissipated in the excitation of
rotational and translational states of neutral gas molecules. In the range of 1-100 Td, the majority
of energy dissipation switches to excitation of vibrational modes of molecules. Above 100 Td,
most energy is spent in exciting electronic states, some of which are dissociative, resulting in
high energy particles and radical generation (Flitti et al. 2009, Popov 2011, Starikovskiy et al.
2013). Above 1000 Td, most energy is spent on the ionization and dissociative ionization of gas
species.
To calculate the reaction rates of electron collision reactions, first the electron energy
distribution function (EEDF) must be calculated, as described in Chapter 3. The calculation
involves considering all of the possible energy gain and loss processes acting on electrons during
the discharge, and is generally solved using some approximation of the Boltzmann equation
(Hagelaar et al. 2005). Next, the specific probability of a reaction can be calculating by
10
considering the electron collision cross section for that reaction, provided those cross sections are
available in the literature. Electron collision cross sections are usually experimentally measured
in swarm experiments(Hayashi et al. 1990, Capitelli et al. 2000) , however, some computational
software for predicting the cross sections a priori does exist (Tennyson et al. 2007). Many of
these cross sections are tabulated and easily available (Pancheshnyi et al. 2008) for small
molecules, although large molecules remain a challenge (Starikovskaia 2014). Once the rates
have been computed, they can be incorporated into a larger kinetic scheme. This is the beginning
of the kinetic effect of the discharge on the overall reaction process. Once one knows the
products of all relevant electron collision reactions, the next question is how these products
interact with each other and the remaining reactants during the combustion process.
1.2.3 Reactions of plasma-produced species
For PAC studies using non-equilibrium plasmas, the E/N range of 100-1000 Td are of interest.
In this range, the energy expended on accelerating electrons is most efficiently deposited into the
generation of active particles, with little wasted on translational/rotational excitation
(Starikovskiy et al. 2013). Due to the short timescale of the discharges of primary interest to this
study (tens of nanoseconds), the electron collision reactions during the plasma can be considered
to be independent of the ground state reactions that typically operate in the microsecond to
millisecond range. This approximation helps simplify the numerical calculation of the discharge
by allowing us to consider each discharge pulse as a process independent of the ground state
reaction processes. In essence, the species created in the discharge are an initial condition for the
subsequent reaction processes. Thus, the calculation can be separated into the very short
timescale of the discharge and the much longer timescale of quenching and combustion
reactions, significantly increasing the overall timestep size for the numerical calculation. In
11
addition, low energy electron and ion processes, such as electron-ion recombination, and ion-
neutral charge exchange reactions also occur on a much larger timescale than electron-neutral
collision processes during the discharge, and can be included in the combustion mechanism. This
will be further discussed in Chapter 3.
The gas mixture immediately after the discharge will contain electronically and vibrationally
excited species, radicals resulting from the dissociation of molecules, and ions. A generalized
outline of the primary and secondary reactions important in the plasma is presented in Figure
1.2.2 to assist in the following discussion. Excited species, which may be excited states of the
reactant or diluent gas molecules or excited states of dissociation products, i.e. in the reaction of
e + O2 → e + O + O(1D), play an important role in PAC. Firstly, excited species typically have
faster reaction rates than the ground state reaction with the same reactants. A comparison of H-
abstraction rates from H2 and CH4 by O atom in the ground state and O atom in the excited
O(1D) state is presented in Figure 1.2.3. The reaction rates for the O + CH4 → OH + CH3 (Cohen
et al. 1991) reaction is as much as 7 orders of magnitude slower than the corresponding reaction
with O(1D) (DeMore et al. 1992), with the rates only converging above 2000 K. The same is true
of the reaction rate of O + H2 → H + OH (Baulch et al. 1992) as compared to the corresponding
reaction with O(1D) (DeMore et al. 1992). Both of these are important initiation and chain
branching steps in the combustion reaction mechanism of these fuels, and the significant increase
of the reaction rates with O(1D) demonstrates the importance of predicting accurate species
distributions in the discharge, as well as incorporating the rates of reactions of plasma produced
species. Another important point to mention is the opening of radical generation pathways not
available in the ground state reaction. For example, one of the major sources of O atoms in air
12
Figure 1.2.2 General diagram for the primary and secondary reactions important in PAC.
Figure 1.2.3 Ratio of H-abstraction rates by O(1D) compared to O(
3P) for H2 and CH4.
Initial collisionSecondary collision
Initial products
Secondary products
Reactants
Electrons
Active species
Fuel radicalTem
peratu
re
e- RH
RO2
R’+HX
O2O2
*/+
O+O*
M M*/+
H R’+ R”
R/R+
13
and air/fuel mixture plasmas is the reactions between excited states of N2 and ground state O2
resulting in the quenching of the excited states and the dissociation of the O2 molecule. In fact, it
has been shown in atmospheric pressure plasma discharges relevant to ignition applications that
this is the largest source of O atoms in the discharge (Rusterholtz et al. 2013).
The second major consideration involving excited states is that the energy contained in the
excited states will be released as thermal energy during quenching reactions, resulting in varying
rates of heat release depending on the excited states involved. For example, the so called “fast
gas heating” effect occurs microseconds after the plasma discharge primarily due to the
quenching of electronically excited states of N2 by O2 and of O(1D) by N2 (Popov 2011). In
addition, temperature rise can be observed from the relaxation of vibrationally excited states of
nitrogen occurring in milliseconds after the plasma discharge (Montello et al. 2013). This
temporally varying heat release must be accounted for in the prediction of PAC phenomena.
Moreover, the interchange of energy between excited states of the same molecule, i.e, electronic-
vibrational, vibrational-vibrational, and vibrational-translational relaxation rates, plays an
important role in depopulating the high level states into lower levels, which will have an effect
on the reaction pathways available during PAC, and thus must be considered.
The role of radical species created during the dissociation of gas molecules by electron
collision reactions may play the most important role in the PAC mechanisms. Above the
autoignition threshold temperature, even a small amount of initial radicals can change the
induction chemistry of the ignition process by bypassing the slow initial radical generation steps
(Ju et al. 2015b). Once the induction chemistry has been bypassed, chemical chain branching
reactions become the dominant radical production mechanism, and ignition will occur soon
thereafter (Starikovskaya et al. 2009). Therefore, it is vitally important to accurately predict the
14
degree of radical generation during the plasma when modeling PAI. Below the autoignition
threshold, radical generation in the plasma will oxidize the fuel even in the absence of chemical
chain branching mechanisms, as will be discussed in Chapter 4 and 5. This will lead to a
reforming of the fuel molecules, resulting in a mixture with significantly different properties than
the original mixture and that may have shorter ignition times, faster flame speeds, and extended
ignition delays as compared to the supplied fuel/air mixture (Kim et al. 2008, Ombrello et al.
2008a, Ombrello et al. 2008b, Kim et al. 2010b, Sun et al. 2011, Sun et al. 2012). It should also
be noted that the reformed mixture will have a different Lewis number than the original mixture,
which will affect the critical radius governing ignition dynamics for localized ignition in
quiescent environments (Chen et al. 2009b, Chen et al. 2011). Thus, the role of radical addition
and fuel dissociation due to electron impact reactions is an important consideration for PAC
kinetics.
Finally, the role of ions in the kinetic mechanism is also a significant factor, although the
reactions pathways and reaction rates still bear significant uncertainty. Ions are created by
electron impact ionization and dissociative ionization reactions, producing positive ions while
freeing additional electrons. Negative ions will also be formed in the plasma due to electron
attachment reactions, but in much lower concentrations (Capitelli et al. 2000). The positive ions
will then react with neutral molecules in charge exchange reactions or recombine with electrons
in recombination and dissociative recombination reactions. The recombination and dissociative
recombination reactions are a significant source of radicals and excited species in the plasma
(Florescu-Mitchell et al. 2006), and will contribute to the active species pool. There is significant
dependence on the electron temperature for these reactions, often prompting the need for two
15
temperatures, the electron temperature and the neutral gas temperature, to be considered by
modeling efforts (Pancheshnyi et al. 2008).
1.2.4 Enhancement mechanism for plasma-assisted ignition
Thus far, this short description has described the important plasma processes of interest in
PAC. How the plasma produced species interact with the combustion gas and change the
dynamics of the ignition process will be the main focus of this dissertation, and thus the
fundamental principles of PAI and some historical background are relevant.
For ignition of combustible mixtures, we can consider two types of events: homogenous
ignition and localized ignition. The first implies that the entire gas mixture simultaneously
experiences a thermal/radical runaway leading to a rapid temperature and pressure rise, based
solely on chemical kinetic processes involving radical chain branching reactions (Yetter et al.
2008). The second is the result of a localized heat/radical source initiating combustion in a small
region, which develops into a flame front and eventually propagates away from the initiation
region, as is typical in spark ignition. For practical devices operating at atmospheric pressure or
greater, plasma discharges generally form in localized regions. Even dielectric barrier discharges
produce many micro-streamers (Kogelschatz 2003), and only at very small gap distance do these
have a uniform distribution over time and space. Uniform volumetric discharges can be produced
at low pressure and are useful for investigation of plasma reactions because of the simplicity of
the 0-dimensional system, which allows for simple modeling calculations and are easy to probe
with both intrusive and non-intrusive diagnostic techniques. Through the course of this
dissertation, ignition using both localized and uniform plasma discharges will be discussed, and
16
knowledge gleaned from uniform discharge investigations will be used to help explain
phenomena observed in the more complex localized ignition case.
Localized ignition by plasma discharges, in the simples interpretation, can be considered to be
the deposition of heat into a small gas volume, initiating chain branching reactions that generate
heat and radicals travelling outwards in a wave, eventually transitioning into a propagating flame
(Maly 1981, Maly 1984, Yetter et al. 2008, Chen 2009). While this simple interpretation suffices
for ignition by equilibrium plasmas, it is not complex enough to describe non-equilibrium plasma
ignition. For example, efficient ignition devices have been demonstrated that do not significantly
raise the temperature above the ambient temperature but that are capable of igniting gas mixtures
of methane-air and ethylene-air (Cathey et al. 2008, Singleton et al. 2011), which clearly cannot
be modeled as a local heat source. Considering the fact that each type of plasma has different
E/N values, different electron number densities, different time durations, etc., it is necessary to
produce descriptions of plasma ignition dependent only on the fundamental physical parameters
of the discharge itself. Ideally, if the voltage waveform, electrode geometry, and local gas
conditions are known, it should be possible to model the ignition event. To do this, the
fundamental reactions and the heat release mechanism must be known. While this ideal is still far
from being realized, significant progress in this direction has been made over the last decade in
particular.
1.3 Background of plasma-assisted combustion and ignition
As mentioned in Section 1.1, there are a series of review papers on plasma-assisted
combustion and ignition published over the last three years (Starikovskiy et al. 2013,
17
Starikovskaia 2014, Ju et al. 2015a, Ju et al. 2015b). Only the relevant results pertaining to this
dissertation will be discussed.
1.3.1 Early work
Studies of spark ignition and the ensuing flame propagation process have long been an
important part of combustion research, especially considering the relevance of this subject to SI
engines. In the research community, there seems to be a periodic pattern of interest in the
optimization and fundamental explorations of electrical discharges for ignition. By the beginning
of the 20th
century, already a mass of literature on ignition by electric discharges existed due to
the recent invention of the internal combustion engine, much of which was summarized and
expanded upon by Paterson and Campbell (Paterson et al. 1918). The authors evaluated available
data on “induction sparks,” “condenser discharges,” and “break sparks”. The first type is
produced using an induction coil, the second by discharging a capacitor, and the final by rapidly
breaking a circuit in the combustible mixture. Paterson and Campbell realized that the potential
difference across the gap and the gap distance had an effect on the critical “intensity” for ignition
of a given mixture. Most importantly, it was noted that the total energy deposition by the spark
was not the controlling mechanism in determining the propensity of the spark to ignite a given
mixture, but was a more complex function of the discharge parameters. Following a thorough
study of condenser (or capacitor) discharges, several more conclusions were reached concerning
ignition in quiescent environments. First, in the case of multiple discharges, there was no benefit
of additional sparks if the first one was capable of igniting the gas mixture. Second, it was found
that the potential applied across the gap was the critical parameter in determining the ignition
propensity of a discharge. Third, it was found that electrodes with the smallest radius of
curvature are the most effective for ignition, but the material of the electrodes has a small effect.
18
Finally, it was concluded that the optimal ignition device will deposit the energy necessary for
ignition in the shortest time possible so as to limit the effect of heat loss to the gas and
electrodes. The underlying mechanism for these general observations were not yet resolved, but
it is interesting to note that these characteristics are still the driving force for research in spark
ignition today, and the question of the type of plasma needed for the optimal ignition device has
not yet been closed.
The next wave of research in this area occurred in the late 1940s when the development of jet
and ramjet engines was a major thrust of research among combustion engineers. At this time,
Blanc et al. (Blanc et al. 1947a) and Boyle and Llewellyn (Boyle et al. 1947) published two of
the first thorough studies of minimum ignition energy (MIE) using spark discharges. A
fundamental theory for minimum ignition energy, as well as minimum spherical flame volume
and quenching distance, were subsequently developed (Blanc et al. 1947a, Lewis et al. 1947,
Blanc et al. 1948). In addition, a review of all electric discharge ignition work up until that time
was published in the first International Symposium on Combustion (Bradford et al. 1948), in
which an attempt was made to answer the question of whether ignition is governed by kinetic
effects initiated by the electric discharge ionizing, exciting, and dissociating gas molecules or
thermal effects from the dissipation of electrical energy were the primary factors governing spark
ignition. Up until that time, thermal effects had been put forward as the driving force for electric
discharge ignition; however, Bradford and Finch argued otherwise based on available data for
carbon monoxide – hydrogen – air ignition timing, in which the ignition propensity of these
mixtures did not follow what would be the expected trends if gas heating was the only influence
on ignition propensity. Instead, they proposed that the specific excitation of certain gas species,
as yet unknown, were the primary driving force for ignition. This question of the dominance of
19
thermal or kinetic effects on localized ignition again arose in the 21st
century, as will be
discussed throughout this dissertation.
Following these early studies, many theories for predicting the probability of successful
ignition and kernel growth were proposed (Frank-Kamenetskii 1955, Ballal et al. 1975), most of
which were based on the concept of the ignition device as a heat source in a spherical geometry
providing some instantaneous amount of energy, and depending on chain branching reactions in
the gaseous mixture to sustain the ignition kernel growth. In addition, studies on flame
stabilization using plasma jets, gliding arcs, and stationary high frequency discharges for flame
speed or lean blowoff enhancement were published (Lawton et al. 1969, Harrison et al. 1971,
Weinberg et al. 1978, Orrin et al. 1981). It was found that the blow-off limits of flames could be
increased significantly by the use of a rotating (or gliding) arc (Harrison et al. 1971), and that
plasma torches could even be used to stabilize flames in supersonic combustion experiments
(Wagner et al. 1989). Again, it was concluded that excited species which engage in reactions
with a high propensity for radical generation must be at least in part responsible for the observed
enhancement effects.
In the 1980’s the next significant effort was made to clarify the mechanism of electric
discharge ignition and the optimal ignition device. This work was carried out by Maly and
coworkers in a series of investigations parametrically exploring discharge parameters in well
defined experimental platforms (Maly et al. 1979, Maly 1981, Maly et al. 1983, Ziegler et al.
1984, Ziegler et al. 1985a, Ziegler et al. 1985b). The results of these efforts were summarized by
Maly (Maly 1984), and suggestions for the optimal ignition device were recommended. It was
determined that any electrical discharge of sufficiently high potential could be described by three
distinct stages, always starting with the “breakdown” stage lasting nanoseconds, followed by the
20
arc discharge lasting milliseconds, and finally transition into the glow discharge which could be
sustained indefinitely. The breakdown phase was determined to be the most efficient time to
deposit energy, a large fraction of the discharge energy could be deposited in a short period of
time, producing a local “shell” of active particles at high temperature that would subsequently
propagate away from the discharge gap and eventually form the flame front. The following
stages were determined to primarily add heat to the ignition kernel as the potential across the gap
would necessarily drop after the discharge stage, forming a high current and low electric field
discharge. The later stages were determined to be of secondary importance, having no effect if
the breakdown phase occurs at a high enough potential such that it alone can ignite the mixture.
The steepness of the active species/temperature gradient initially produced by the discharge is
determined by the steepness of the voltage rise across the gap, which in turn determines the rate
of propagation of the species/temperature shell and thus the rate of formation of the flame front.
Therefore, the steepest possible initial potential gradient (and thus the highest initial potential) is
desired at the beginning of the breakdown phase, and the overall discharge time need not last
more than a few nanoseconds. As anticipated by the 1918 work of Paterson and Campbell
(Paterson et al. 1918), the ignition of a mixture is most efficiently achieved by depositing energy
in the shortest possible time using the greatest possible initial potential across the discharge gap.
This will produce a plasma discharge with a steep temperature/species gradient, large volume,
and minimal energy waste to heating not involved in the ignition kernel development. Still, the
specifics of the governing processes that result in these conclusions were limited, and predictions
of the discharge to ignition kernel development process, as well as the specific electron collision
and active species reactions involved, remained out of reach.
21
1.3.2 Recent advances in PAC and PAI
Over the past decade, a significant effort has been made to answer the open questions in
PAC/PAI, primarily due to renewed interest in scramjet engine development (Jacobsen et al.
2008). In addition, the increasing emphasis on environmental concerns has driven researchers to
seek ignition techniques to extend the lean limits in both IC and gas turbine engines (Cathey et
al. 2007, Pilla et al. 2008, Ikeda et al. 2009a, Ikeda et al. 2009b, Rapp et al. 2012, Lacoste et al.
2013b). Advancements in computational modeling as well as non-intrusive diagnostic techniques
have allowed for much more in depth probing of complex phenomena than possible in previous
decades (Ju et al. 2015b). Another important advancement was the advent of power supplies
capable nanosecond and even picosecond pulse durations (http://www.fidtechnology.com/).
These have the advantage of supplying high E/N values for short durations, favoring energy
addition into excited states and limiting gas heating (Macheret et al. 2006, Adamovich et al.
2009b, Starikovskaia 2014). Despite these advances, PAC has continued to provide challenges to
researchers, and open questions still remain in the fundamental phenomena governing the plasma
– combustion interactions.
Beginning in 1996, researchers at the Laboratory of Physics of Nonequilibrium Systems
at the Moscow Institute of Physics and Technology (MIPT) conducted a series of studies on the
ability of non-equilibrium nanosecond duration discharges to ignite gas mixtures in shock tube
experiments (Starikovskaia et al. 2001, Bozhenkov et al. 2003, Starikovskaia et al. 2004, Anikin
et al. 2006, Kosarev et al. 2008a, Kosarev et al. 2008b, Aleksandrov et al. 2009a, Aleksandrov et
al. 2009b, Kosarev et al. 2009, Starikovskaya et al. 2009). They made use of Fast Ionization
Wave (FIW) discharges produced by 100 kV peak voltage pulses lasting 40 – 60 ns. Mixtures of
hydrogen and n-alkanes from methane through pentane, along with oxygen and helium, argon, or
22
nitrogen, were utilized at pressures ranging from 0.2 to 2.0 atm in experimental and modeling
studies. One of the significant early conclusions from these studies were that, near the
autoignition threshold, excitation by a small amount of non-equilibrium energy addition (< 0.1
J/cm3), deposited into excited states and used to dissociate molecules as well as heat the gas,
could decrease the ignition delay time for H2/O2/N2 mixtures at atmospheric pressure and 1000 K
initial temperature more efficiently than by the same amount of energy deposited only as heat
(Bozhenkov et al. 2003). As larger hydrocarbon molecules were tested and modeling became
more complex (Kosarev et al. 2009), it was demonstrated that the inclusion of plasma chemical
reactions, including electron collision excitation, dissociation, and ionization reactions, as well as
excited species quenching reactions, were necessary to account for the observed order of
magnitude decrease in ignition delay times. In addition, the sensitivity of ignition delay to
induction chemistry of hydrocarbons of different chain lengths could be removed due to the non-
equilibrium supply of radicals, mainly O and H. This is because the production of initial radicals
is the slowest step in the autoignition process, but could be bypassed when non-equilibrium
plasma discharges were applied (Starikovskaia 2014). The initial radical generation steps are
highly fuel specific, and, therefore, by replacing the radical generation mechanism, the
sensitivity of ignition times on specific fuel molecules could also be removed.
While the ignition delay studies were informative, they lacked quantification of the important
species involved in the ignition process, limiting their usefulness in constraining the mechanism
of plasma-chemistry interaction. To address this deficiency, a series of experiments were
performed at the Non-Equilibrium Thermodynamics Laboratory at The Ohio State University
using advanced laser diagnostic techniques to quantify many of the important plasma-produced
species. These experiments were performed in low pressure flow reactors (P ≤ 100 Torr) at initial
23
temperatures of ≤ 500 K using bursts of nanosecond pulsed discharges of peak voltage up to 30
kV and repetition frequencies up to 50 kHz for continuous discharges or bursts of pulses of
varying durations. Initial experiments in continuously pulsed plasmas revealed that, in the cases
of methane and ethylene, the fuel was oxidized in the plasma discharge and converted primarily
into CO, CO2, and H2O, with plasma temperatures below 900 K as measured by optical emission
spectroscopy (Bao et al. 2007, Lou et al. 2007). Downstream of the plasma a flame could be
observed in ethylene – air mixtures for significantly high pulse repetition frequencies (≥ 30 kHz),
near stoichiometric equivalence ratios (0.8 – 1.2), and low flow velocities ≤ 25 m/s. Strikingly,
there was a smooth transition in terms of temperature and degree of fuel consumption between
non-igniting mixtures and igniting mixtures as the experimental parameters were varied,
indicating that ignition was not a sudden process, but a gradual transition.
Following these initial studies, measurements of O atom (Uddi et al. 2009b), NO (Uddi et al.
2009a), and OH (Yin et al. 2013a, Yin et al. 2013b) were performed using LIF, as well as the
distribution of the vibrationally excited states of N2 using coherent anti-Stokes Raman scattering
(CARS) (Montello et al. 2013). Measurement of the translation/rotational temperature of the gas
were measured and Rayleigh scattering and o[tical emission spectroscopy (OES) (Zuzeek et al.
2010, Yin et al. 2011, Zuzeek et al. 2011, Montello et al. 2013, Yin et al. 2013a, Yin et al.
2013b). These studies were accompanied by numerical modeling with detailed plasma and
combustion chemical reaction schemes, aimed at resolving the mechanism by which the
measured species and temperatures were produced. The measurements from this series of studies
provided an important step in the direction of accurate data for constraining model development
in PAC, and have helped to understand the radical generation pathways in the plasma. Important
conclusions include the understanding of the role of N2 excited states for both radical generation
24
and heat release (Montello et al. 2013), as well as the role of different fuels in the consumption
rates of specific radical species. This latter point was of particular interest. It was found that the
OH concentration in H2 – air and CH4 – air mixtures was at its maximum just after the plasma,
and were consumed within 1.0 ms. However, in C2H4 – air and C3H8 – air mixtures, the OH
concentration after the discharge was significantly lower (about a factor of 3), but began to rise
in the first 10 µs after the plasma, before being totally consumed by 100 µs. Modeling of the
reaction processes indicated that, for H2 and CH4 mixtures, the Konnov mechanism (Konnov
2000) predicted the reaction process satisfactorily, but for the two larger fuels disagreement was
significantly worse, indicating uncertainty in the complex kinetics of hydrocarbon fuels at low
temperature. These fuel specific results will be discussed in detail in Chapters 4 and 5. One
limitation of this series of studies is that, due to the limited number of species probed in the
apparatus, the specific pathways for fuel oxidation could not be resolved. Therefore, model
mismatches could not be traced to the reactions at the source of the disagreement.
To better model the discharges used in flow reactor experiments, a one-dimensional model of
low pressure PAC reactors has been developed by Nagaraja et al. (Nagaraja et al. 2013, Nagaraja
et al. 2014, Nagaraja et al. 2015b). The benefit of one-dimensional models is that the cathode
sheath layer and its contribution to the temperature and species profiles can be included in the
predictions. Another benefit is that the voltage waveform measured experimentally can be
directly implemented to predict the time-dependent E/N values. Previous modeling attempts
would vary this parameter to optimize the fit with experimental parameters, which is not a
strictly predictive modeling approach (Adamovich et al. 2015a). Initial results demonstrate
improved prediction as compared to 0-dimensional modeling of the same phenomena (Yang et
al. 2015a). Simulations with fuels as large as n–heptane have been performed (Nagaraja et al.
25
2015b), and results demonstrate the benefits of using nanosecond pulsed discharge for reduction
in ignition time of fuels with NTC regions due to reduction in the induction kinetics timescale.
While these modeling tools are valuable, due to their long computational time, they are not
optimal for use in modeling phenomena which can be approximated described by 0-dimensional
models, which are mainly used to analyze the kinetics of PAC.
Another approach to better understand PAC kinetics has been to isolate specific molecules
known to be produced in the plasma, and explore their reactions individually. A significant step
in this direction was achieved in the investigation of O2(a1Δg) and O3 effects on fuel chemistry by
Ombrello et al. (Ombrello et al. 2010a, Ombrello et al. 2010b). This work was inspired by
observations of in situ fuel reforming in gliding arc plasmas, resulting in NOx formation which
participated in lower activation energy reactions than the original mixture, and thus significantly
reduced the ignition temperature and increased the extinctions limits of counterflow flames
(Ombrello et al. 2006, Ombrello et al. 2008a, Ombrello et al. 2008b). The mechanism for this
reforming was, at the time, still unknown. Singlet oxygen molecules had previously been
proposed as one of the major species responsible for reducing ignition delays in PAC (Starik et
al. 2006, Starik et al. 2009, Starik et al. 2010) due to their long lifetime and high concentration in
air plasmas, along with reduced activation energy in reactions with hydrocarbons. The
experiments of Ombrello et al. proved that by supplying up to 5000 ppm of O2(a1Δg) liftoff
height of a lifted flame could be reduced by several percent in ethylene - air lifted flames.
Modeling work revealed the serious limitations in the knowledge of O2(a1Δg) quenching rates by
hydrocarbons, with the available estimates overestimating their reactivity. The effect of 1260
ppm addition of ozone on lifted flames resulted in a greater than 8% increase of flame speed in
propane - air mixtures. The reason for this enhancement was the dissociation of ozone
26
(beginning at ≈ 500 K), which supplied O atoms in only in close proximity to the flame front.
This resulted in an increased radical pool in the early flame zone, speeding the flame speed and
thus reducing the liftoff height. Further work on ozone effects on flame have since been carried
out (Wang et al. 2012, Vu et al. 2014, Won et al. 2015b). One of the most dramatic effects of
ozone addition on flame chemistry is the ability for ozone to initiate cool flames in fuel mixtures
exhibiting low temperature chemistry (Sohn et al. 2015, Won et al. 2015b). In this case, ozone is
able to exaggerate the middle branch of the S-curve apparent in fuels with negative temperature
coefficient (NTC) behavior, resulting in the stabilization of a flame (more precisely, fuel
reforming wave) at temperatures far below traditional flame temperature (600 – 800 K). This
was possible, again, due to the super-equilibrium supply of O atoms to the flame front.
The type of dramatic changes to flame structures demonstrated for ozone have also been
observed for plasmas applied directly to the flame. The experiments of Sun et al. (Sun et al.
2013, Sun et al. 2014), conducted in a counterflow burner using porous steel plugs as the burner
exits, applied a plasma discharge directly to the flame zone. A nanosecond pulsed plasma, with
12 ns FWHM, peak voltage of 7.6 kV, and pulse repetition rates up to 40 khz was applied to
mixtures of CH4/O2/He at 60 Torr. It was found that the pulsed plasma could ignite the flame,
and the extinction limits could be extended. Interestingly, at high oxygen loading conditions and
pulse frequencies of 24 kHz, it was found that a smooth transition between a non-reacting flow
and a fully reacting flow (flame) could be could be observed. This demonstrated that the plasma
could remove the hysteresis between ignition and extinction, termed the S-curve, allowing fuel
oxidation to occur at any temperature between 300 K and the flame temperature, even for fuels
that only display limited low temperature reactivity. Measurements of O atom using TALIF (Sun
et al. 2012), CH2O and OH using PLIF, temperature using Rayleigh scattering (Sun et al. 2013),
27
and stable species sampling using FTIR were performed (Sun et al. 2012). It was concluded that
excess radicals produced in the discharge, particularly O atoms, were responsible for initiating
and sustaining the chain branching reaction sequence, even at strain rates outside of regions
where the flame was self sustaining. Model comparisons revealed that even major species were
not well predicted, and it was concluded that further work on low temperature plasma – fuel
interactions was needed to improve the understanding of PAC kinetics.
Another paper in the same experimental platform observed similar results, this time using
dimethyl ether (DME) as the fuel (Sun et al. 2014). In this case, it was observed that the low
temperature oxidation reactions competed with the high temperature reactions for the plasma-
produced radicals. At pulse repetition frequencies of 24 kHz, this resulted in a gradual rise of
CH2O as the fuel mole fraction was increased, while the OH concentration remained low,
indicating that low temperature reactions were consuming the radicals, and that high temperature
ignition was not occurring. At high enough fuel mole fractions, a sudden decrease in the CH2O
concentration, accompanied by an increase in the OH concentration, indicated the onset of high
temperature reactions. When the pulse repetition frequency was increased to 34 kHz, the smooth
transition, observed for CH4 in the earlier paper, was recovered. The explanation for this
behavior was that the low temperature reactions occur at a rate proportional to the radical
concentration available, in this case from the plasma discharge, and are too slow to compete with
high temperature reactions when the radical concentration is too low. Therefore, at low pulse
repetition frequencies, there was a sudden switch between partially reacting and fully reacting
flows as the fuel mole fraction increased (thus increasing the available chemical enthalpy). When
more radicals were available at the higher pulse repetition frequencies, the low temperature
chemistry timescale became comparable to the high temperature timescale, and thus the S-curve
28
was again smoothed. It was noted that high temperature ignition is not very sensitive to
additional radical generation from the plasma, while low temperature reactions could be
dramatically accelerated in the discharge.
The major advances outlined in this short review are not by any means an exhaustive review
of the important works by the PAC community. Some other novel applications of PAC include
the observed increase in flame speed when a premixed flame is exposed to sub-critical
microwave discharges (Stockman et al. 2009, Michael et al. 2013) or to nanosecond pulsed
discharges at low pressure (Li et al. 2013, Nagaraja et al. 2015a). Increases in blow-off limits of
lifted flames in electric fields (Won et al. 2007, Won et al. 2008, Kim et al. 2010a) and turbulent
flames in nanosecond pulsed discharges (Pilla et al. 2006, Lacoste et al. 2013a), as well as the
stabilization of swirl burners (Pilla et al. 2008) have also been demonstrated, which are relevant
to jet engine combustion. In addition, there have been many recent studies on localized electric
ignition. Ignition devices at atmospheric pressure using microwave discharges (Ikeda et al.
2009a, Ikeda et al. 2009b, Lefkowitz et al. 2012b, Nishiyama et al. 2012, Rapp et al. 2012, Wang
et al. 2013a), nanosecond pulsed discharges (Wang et al. 2005, Pancheshnyi et al. 2006, Cathey
et al. 2007, Cathey et al. 2008, Shiraishi et al. 2009, Singleton et al. 2011, Rusterholtz et al.
2013), laser ignition (Morsy 2012) and laser-assisted spark ignition (Michael et al. 2010), and
gliding arcs (http://www.knite.com/) have all been demonstrated to improve on traditional spark
discharges. While these types of studies have often been conducted historically, the difference in
modern studies is the advancement of diagnostics and modeling capabilities. In Chapter 2, a
discussion of the significant advancements in the area of localized discharges will be presented.
It is now possible, with use of newly validated kinetic models and carefully measured thermal
and chemical properties in local discharges, to describe the process of ignition for a particular
29
discharge starting from the initiation of the electric field and proceeding all the way through to
the fully propagating flame.
1.4 Technical questions remaining in PAC and PAI
1) What type of discharge produces the shortest localized ignition event without excessive
energy deposition?
2) What is the reaction sequence for low temperature oxidation of hydrocarbon fuels, sustained
by an external active species source, and what deficiencies remain in the understanding of
these reactions?
3) Which reactions, originating from electron collision processes or reactions of species
produced in the plasma, are necessary to consider for predicting observable phenomena in
plasma-assisted combustion?
4) What are the electron collision cross sections and branching ratios of product species for
hydrocarbons?
5) Is it possible to fully validate a plasma-kinetic model? If so, what experiments, species
measurements, and time scale of measurements are required to produce such a mechanism?
The first question is quite general, and has been asked by many researchers over the past 150
years. However, without clear knowledge of the species produced by the plasma and their
reactions with fuel–air mixtures, it is still not possible to optimize the type of discharge that will
result in the most efficient ignition process. It is important to realize that the choice of optimal
ignition device is governed by the application of the device. The optimal device for an SI engine
may not be the optimal device for a scramjet engine or an augmenter in a military jet engine.
Optimally, once certain design criteria are known, such as the gas mixture, flow parameters, and
30
geometric limitations of the engine, it should be possible to prescribe the electrode geometry and
voltage waveform that will most efficiently and reliably ignite mixture. Another aspect of the
optimization process for certain applications is ensuring that the minimal harmful emissions,
such as NOx, are produced. Due to the many orders of magnitude in time and spatial scales
covered by the ignition process, building a predictive model for PAI is a significant challenge for
researchers. However, with the current knowledge of electron collision parameters, kinetics of
ensuing combustion reactions, heat transfer mechanism in the discharge, and flame kernel
development dynamics (at least in laminar flows), it should be possible to build the first iteration
of such a model. Once this task is accomplished, the optimal ignition device for a specific engine
should be possible to define.
Outside of spark or spark-like discharges, many other plasma types are available to
researchers, as discussed in Section 1.2.1. This opens new possibilities for engine types using
different ignition or flame holding techniques. Plasmas can be considered as another source for
reaction generation, which can be controlled by the discharge and gas parameters to tailor the
reactions rates for a desired purpose. As questions 2 – 4 address, there are many questions
remaining in how plasmas can be used to alter fuel oxidation and reforming processes. As more
knowledge is being gained from detailed studies of plasma kinetics, particularly for larger
hydrocarbons, the reaction mechanisms and benefits of different types of plasmas are becoming
clear. However, there are still many uncertainties in the PAC mechanism for small hydrocarbons,
which must be explored before the significantly more difficult kinetics of larger hydrocarbons
can be addressed. Along these lines, question 5 arises. How can we validate a kinetic model
which scales over times from nanoseconds to hundreds of milliseconds, which relies on
intermediate species which exist in many difficult to measure excited states, and which can never
31
truly be described as zero-dimensional or even one-dimensional? The answer to this question lies
in understanding what type of phenomena we wish to predict, which then determines how
detailed the model must be. This will be a major theme throughout this thesis.
Perhaps the broadest question facing the PAC/PAI research community is how the knowledge
learned in the past two decades can be applied in novel combustion applications to best leverage
the unique characteristics of different discharge types. Thus far, plasmas have only been used in
practical applications for local ignition, much like they were originally used 150 years ago. How
can the PAC/PAI community expand the application of plasma discharges to new types of
engines with significant promise for the future?
1.5 Objectives of dissertation
The overall objective of this dissertation is to explore the chemical processes of PAI in fully
coupled systems in which the active species in the plasma and the oxidation reactions of
hydrocarbon fuels interact directly. This knowledge will be used to understand localized ignition
using high E/N ignition sources for practical applications.
This broad objective is separated into two areas. The first area, presented in Chapter 2,
involves applications of PAI devices and proof of concept for two discharge types: spark-assisted
microwave discharges in SI engines and nanosecond repetitively pulsed (NRP) spark discharges
in pulsed detonation engines (PDEs). The second area, presented in Chapters 3 – 6, explores the
kinetics of nanosecond pulsed discharges in mixtures of CH4 and C2H4 (these fuels were chosen
as a first step in understanding PAC kinetics in real fuels) with specific emphasis on the
important species produced in the discharge, the low temperature (< 700 K) reactions that take
32
place after the discharge, and the remaining challenges in the understanding of fuel specific
kinetics for PAC.
In the applications section, the objective is to prove the superiority of the two plasma types
over conventional capacitive spark discharges. In the case of the spark-assisted microwave
ignition device, the operation limits of a small SI engine are shown to be extended to leaner
mixtures while maintaining acceptable cycle repeatability. In the case of NRP spark discharges,
it will also be shown that the lean limits of a PDE can be extended, and the ignition times for
near limit conditions can be reduced by up to 25% using the plasma device. It is also shown that,
with the same energy deposition, NRP ignition proceeds faster than multiple spark discharge
ignition using a capacitive ignition device. The underlying cause of the ignition improvement is
explored, and the concept of inter-pulse coupling is introduced as the mechanism by which
reduced ignition times can be observed, particularly in turbulent or high speed flows. The
objective here is to understand why NRP discharges are superior in terms of overall ignition time
reduction and efficiency as compared to capacitive discharges.
In the second section, an experimental platform is developed to isolate the kinetic effects of
NRP discharges on fuel oxidation using both in-situ tunable diode laser absorption spectroscopy
(TDLAS) and ex-situ sampling diagnostics to measure the temperature and intermediate/product
species quantitatively. This experimental platform is described in Chapter 3. In addition, kinetic
models of CH4 and C2H4 oxidation are assembled for comparisons to the collected data, which
are based on detailed plasma and combustion reaction models developed explicitly for PAI
predictions. The model theory and architecture is described in Chapter 4, and specific reaction
mechanisms are described in Chapters 5 and 6. The reaction pathways are assessed, and species
production rates are compared with experimental results to isolate the reaction sequences that
33
contribute to product species production. The contribution of plasma species to the low
temperature oxidation process is the center of this discussion, and the need for accurate reaction
rates in temperatures below 700 K are highlighted for C2H4 in Chapter 5 and CH4 in Chapter 6.
The objective is to understand the reactions that result in the continuous transition between non-
reacting and reacting flows observed in previous studies.
In Chapter 7, conclusions from the studies presented in this dissertation are discussed, and
recommendations for future work are made. In particular, the direction for applications of PAI,
the remaining need for advanced diagnostics, and the reactions requiring further exploration are
discussed.
1.6 Publication history
1.6.1 Journal publications
9) B. Brumfield, J. K. Lefkowitz, X. Yang, G. Wysocki, Y. Ju, “Simultaneous Quantification of OH
and HO2 in the Oxidation of Dimethyl Ether in a Laminar Flow Reactor,” Proc. Combust. Inst.,
submitted.
8) A. Rousso, S. Yang, J. K. Lefkowitz, W. Sun, Y. Ju, “Low Temperature Oxidation and Pyrolysis of
n-Heptane in Nanosecond-pulsed Plasma Discharges,” Proc. Combust. Inst., submitted.
7) S. Yang, J. K. Lefkowitz, S. Nagaraja, X. Gao, V. Yang, Y. Ju, W. Sun, “Numerical and
Experimental Investigation of Nanosecond Pulsed Plasma Activated C2H4/O2/Ar Mixtures in a Flow
Reactor,” J. Propul. Power, submitted.
6) Y. Ju, J. K. Lefkowitz, C. B. Reuter, S. H. Won, X. Yang, S. Yang, W. Sun, Z. Jiang, Q. Chen,
“Plasma Assisted Low Temperature Combustion,” Plasma Chemistry and Plasma Processing (2015)
1-21.
34
5) J. K. Lefkowitz, P. Guo, A. Rousso, Y. Ju, “Species and Temperature Measurements of Methane
Oxidation in a Nanosecond Repetitively Pulsed Discharge,” Phil. Trans. R. Soc. A 373
(2015) 20140333.
4) J. K. Lefkowitz, P. Gou, T. Ombrello, S. H. Won, C. Stevens, J. Hoke, F. Schauer, Y. Ju, “Schlieren
Imaging and Pulsed Detonation Engine Testing of Ignition by a Nanosecond Repetitively Pulsed
Discharge,” Combust. Flame 162 (2015) 2496-2507.
3) J. K. Lefkowitz, M. Uddi, B. C. Windom, G. Lou, Y. Ju, “In situ Species Diagnostics and Kinetic
Study of Plasma Activated Ethylene Dissociation and Oxidation in a Low Temperature Flow Reactor,”
Proc. Combust. Inst. 35 (2015) 3505-3512.
2) J. K. Lefkowitz, S. H. Won, Y. Fenard, Y. Ju, “Uncertainty Assessment of Species Measurements in
Acetone Counterflow Diffusion Flames,” Proc. Combust. Inst. 34 (2013) 813-820.
1) J. K. Lefkowitz, J. S. Heyne, S. H. Won, S. Dooley, H. H. Kim, F. M. Haas, S. Jahangirian, F.L. Dryer,
Y. Ju, “A Chemical Kinetic Study of tertiary-Butanol in a Flow Reactor and a Counterflow Diffusion
Flame,” Combust. Flame 159 (2012) 968-978.
1.6.2 Conference papers and presentations
17) J. K. Lefkowitz, “Kinetic study of low temperature methane oxidation in a nanosecond repetitively
pulsed dielectric barrier discharge”, 22nd
International Symposium on Plasma Chemistry, Antwerp,
Belguim (2015) presentation O-4-3.
16) J. K. Lefkowitz, A. Rousso, P. Guo, Y. Ju, “A kinetic study of low temperature methane oxidation in
a nanosecond repetitively pulsed discharge,” 9th U.S. National Combustion Meeting, Cincinnati, OH
(2015) paper 3B03.
15) X. Yang, J. K. Lefkowitz, B.E. Brumfield, Q. Chen, G. Wyscoki, Y. Ju, “Kinetics studies of
O3/O2/CH3OH/Ar mixtures in a photolysis flow reactor,” 9th U.S. National Combustion Meeting,
Cincinnati, OH (2015) paper 1A10.
35
14) S. Yang, S. Nagaraja, W. Sun, V. Yang, J. K. Lefkowitz, Y. Ju, “Numerical and experimental study
of pulsed nanosecond plasma discharges for C2H4/O2/Ar gas mixtures in a low temperature reactor,”
9th U.S. National Combustion Meeting, Cincinnati, OH (2015) paper 3B01.
13) J. K. Lefkowitz, P. Guo, A. Rousso, Y. Ju, “Low temperature oxidation of methane in a nanosecond
pulsed plasma discharge,” 53rd
AIAA Aerospace Sciences Meeting, Kissimmee, Florida (2015) AIAA
paper 2015-0665.
12) T. Wada, J. K. Lefkowitz, Y. Ju, “Plasma Assisted MILD Combustion,” 53rd
AIAA Aerospace
Sciences Meeting, Kissimmee, Florida (2015) AIAA paper 2015-0666.
11) S. Yang, S. Nagaraja, V. Yang, W. Sun, J. K. Lefkowitz, Y. Ju, “Numerical and Experimental
Investigation of Nanosecond-Pulsed Plasma Activated C2H4/O2/Ar Mixtures in a Low Temperature
Flow Reactor,” 53rd
AIAA Aerospace Sciences Meeting, Kissimmee, Florida (2015) AIAA paper
2015-1614.
10) Y. Ju, J. K. Lefkowitz, T. Wada, X. Yang, S. H. Won, W. Sun, “Plasma assisted combustion: new
combustion technology and kinetic studies,” 53rd
AIAA Aerospace Sciences Meeting, Kissimmee,
Florida (2015) AIAA paper 2015-0156.
9) B. Brumfield, N. Kurimoto, X. Yang, T. Wada, P. Diévart, J. Lefkowitz, G. Wysocki, Y. Ju, “Kinetic
Studies of Low and Intermediate Temperature Oxidation of Dimethyl Ether,” 248th ACS National
Meeting & Exposition, San Fransisco, California (2014).
8) B. Brumfield, X. Yang, J. Lefkowitz, Y. Ju, G. Wysocki, “Towards Simultaneous Measurement of
OH and HO2 in Combustion Using Faraday Rotation Spectroscopy,” CLEO:2014, San Jose,
California (2014) Paper SF21.4.
7) J. K. Lefkowitz, B. C. Windom, W. MacDonald, S. Adams, T. Chen, M. Uddi, Y. Ju, “Time
Dependent Measurements of Species Formation in Nanosecond-Pulsed Plasma Discharges in
36
C2H4/O2/Ar Mixtures” 52nd
AIAA Aerospace Sciences Meeting, National Harbor, Maryland (2014)
AIAA paper 2014-1179.
6) J. K. Lefkowitz, M. Uddi, B. C. Windom, Y. Ju, “In situ Mid-infrared Absorption Measurements in a
Nanosecond Pulsed Plasma Discharge,” The 2nd
International Education Forum on Environment and
Energy Science, Huntington Beach, California (2013).
5) J. K. Lefkowitz, Y. Ju, C. Stevens, T. Ombrello, F. Schauer, J. Hoke, "The Effects of Repetitively
Pulsed Nanosecond Discharges on Ignition Time in a Pulsed Detonation Engine," 49th
AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, San Jose, California (2013)
AIAA paper 2013-3719.
4) M. Uddi, J. K. Lefkowitz, B. Windom, Y. Ju, “Species Measurements of Ethylene Oxidation in a
Nanosecond-Pulsed Plasma Discharge Using QCL Absorption Spectroscopy Near 7.6µm,” 51st AIAA
Aerospace Sciences Meeting, Grapevine, Texas (2013) AIAA paper 2013-0435.
3) J. K. Lefkowitz, Y. Ju, R. Tsuruoka, Y. Ikeda, “A Study of Plasma-Assisted Ignition in a Small
Internal Combustion Engine,” 50th AIAA Aerospace Sciences Meeting, Nashville, Tennessee (2012)
AIAA paper 2012-1133.
2) J. K. Lefkowitz, J. S. Heyne, S. H. Won, S. Dooley, H. H. Kim, F. M. Haas, S. Jahangirian, F. L.
Dryer, and Y. Ju. “A Chemical Kinetic Study of the Alternative Transportation Fuel, tertiary-
Butanol,” 49th AIAA Aerospace Sciences Meeting, Orlando, Florida (2011) AIAA paper 2009-698.
1) J. Heyne, J. K. Lefkowitz, F. M. Haas, S. H. Won, S. Dooley, H. H. Kim, S. Jahangirian, F. L.
Dryer, Y. Ju, “Combustion Kinetics Study of t-Butanol,” 7th U.S. National Combustion Meeting,
Atlanta, Georgia (2011).
37
2 Plasma-assisted ignition for internal combustion engine
and pulsed detonation engine applications
This chapter investigates two different applications of PAI in practical systems and
demonstrates its efficacy for combustion control. The first application is of a spark-assisted
microwave discharge in a small internal combustion engine. The second application is of a NRP
discharge in the spark regime in a PDE. The underlying physics and mechanism of enhancement
are discussed to the extent that the experimental and numerical results allow. These experiments
are meant to be proofs of concept for each ignition device, and further research into the
fundamental physics and chemistry governing these systems is discussed in Chapters 3-6.
The results presented in Section 2.1 were published and presented in the AIAA Aerospace
Sciences Meeting in January of 2012 (Lefkowitz et al. 2012b):
J. K. Lefkowitz, Y. Ju, R. Tsuruoka, Y. Ikeda “A Study of Plasma-Assisted Ignition in a Small
Internal Combustion Engine,” 50th
AIAA Aerospace Sciences Meeting, Nashville, Tennessee
(2012) AIAA paper 2012-1133.
I was responsible for planning and execution of data collection, analysis of results, and the
writing and presenting of the paper mentioned above. Assistance in collecting the experimental
data was provided by Dr. Ryoji Tsuruoka, and the experimental platform including the plasma
ignition system was supplied by Dr. Yuji Ikeda through his company Imagineering. The initial
concept and guidance for the research was provided by Prof. Yiguang Ju.
38
The results presented in Section 2.2 were published in Combustion and Flame (Lefkowitz et
al. 2015a):
J. K. Lefkowitz, P. Gou, T. Ombrello, S. H. Won, C. Stevens, J. Hoke, F. Schauer, Y. Ju,
“Schlieren Imaging and Pulsed Detonation Engine Testing of Ignition by a Nanosecond
Repetitively Pulsed Discharge,” Combust. Flame 162 (2015) 2496-2507.
The results were presented by Dr. Christopher Stevens in the AIAA/ASME/SAE/ASEE Joint
Propulsion Conference and Exhibit in July of 2013 (Lefkowitz et al. 2013):
J. K. Lefkowitz, Y. Ju, C. Stevens, T. Ombrello, F. Schauer, J. Hoke, "The Effects of
Repetitively Pulsed Nanosecond Discharges on Ignition Time in a Pulsed Detonation
Engine," 49th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, San Jose,
California (2013) AIAA paper 2013-3719.
I was responsible for assembly and installation of the ignition device, planning and execution of
data collection, assembly of the kinetic model, analysis of the results, and writing the paper
mentioned above. Dr. Peng Gou was responsible for running the numerical simulations. Dr.
Christopher Stevens assisted in data collection, and the pulsed detonation engine platform was
provided by Dr. John Hoke and Dr. Frederick Schauer of the Air Force Research Laboratory
(AFRL). The initial concept for this study was provided by Dr. Timothy Ombrello, and guidance
in analysis of the results and writing of the paper was provided by Dr. Timothy Ombrello, Dr.
Sang Hee Won, and Prof. Yiguang Ju. This work was made possible by the Summer Faculty
Fellowship Program of the Air Force Office of Scientific Research (AFOSR), which provided
funding for this collaborative university/AFRL study.
39
2.1 Spark-assisted microwave discharge ignition in a small internal
combustion engine
2.1.1 Introduction
The challenge of operating internal combustion engines with low coefficient of variation
(COV) within the lean burn regime has become a particular concern for continued improvement
in energy efficiency and reduction of NOx to meet increasingly stringent emissions regulations.
This is also a concern for unmanned aerial vehicles (UAVs), which need to fly for long durations
in surveillance or combat operations, and lean burn operation can significantly improve the
engine specific fuel consumption, allowing longer flights with the same payload. The problems
associated with small engines used in UAVs relate to their high surface/volume ratio, which
quenches radicals and leads to increased heat transfer losses. The desire to use distillate fuels
similar to those already employed in other strategic energy conversion systems leads to
difficulties in achieving full vaporization and homogenous mixing, which in turn leads to
difficulties igniting the mixture with conventional spark plugs. Currently, small engines run at
stoichiometric or slightly rich conditions where the minimum ignition energy is smallest in order
to achieve steady combustion (Ju et al. 2011). Moreover, the ignition phenomena and initial
flame kernel growth have significant influence on total charge burn time. To improve fuel
efficiency, it is necessary to run at fuel lean conditions so as to reduce the maximum temperature
in the combustion chamber, minimizing heat transfer to the walls and increasing thermal
efficiency (Heywood 1988).
Conventional capacitive discharge spark plugs ignite using sparks typically lasting a few
milliseconds (Heywood 1988), and expend much of their supplied energy in heating the
40
electrodes in comparison to the energy delivered to the gaseous mixture to be ignited (Maly
1984). The discharge gap is typically small, and the spark location is stochastic, limiting the
amount of gas exposed to the discharge. These characteristics all lead to the formation of a small
initial flame kernel. Evolution of this kernel to rapid flame propagation can easily exceed 25% of
the total charge burning time (Heywood 1988). The duration of the ignition process and the
minimum ignition energy (MIE) both increase as the lean flammability limit of the fuel/air
mixture is approached (Blanc et al. 1947b, Lewis et al. 1947). In addition, large Lewis number
gases require higher ignition energy than lower Lewis number gaseous (i.e. light fuels) for the
ignition kernel to develop into a self-sustaining flame (Chen et al. 2007, Chen et al. 2009a). The
small mass diffusivity of large fuels results in high Lewis numbers, so the ignition energy is
large for fuels like gasoline and Jet A. In theory, the use of a large volume non-equilibrium
plasma discharge will deliver the ignition energy to the mixture more effectively and also have
the potential to break down some of the large molecules and decrease the Lewis number.
Previous studies have found that it is possible to use a non-equilibrium plasma device to achieve
consistent ignition in internal combustion engines that run near the flammability limits (Cathey
et al. 2007, Ikeda et al. 2009a, Ikeda et al. 2009b, Nishiyama et al. 2012, Rapp et al. 2012, Wolk
et al. 2013).
Imagineering, Inc. has developed a microwave-enhanced spark plug which is capable of
producing non-equilibrium plasmas at pressures of up to 2.0 MPa. This device has been tested in
larger engines and shown to enhance the lean limit of engine operation (Ikeda et al. 2009a, Ikeda
et al. 2009b, Nishiyama et al. 2012, Rapp et al. 2012). This study will show the effectiveness of
such a device to achieve similar results for a much smaller engine, in which the challenges of
quenching and heat loss are more significant. Microwave frequency plasmas are known to
41
increase the flame speed and decrease ignition time by producing excited species, active radicals,
and longer lifetime species such as ozone and singlet delta oxygen (Stockman et al. 2009,
Ombrello et al. 2010a, Ombrello et al. 2010b, Michael et al. 2013).
The goal of this study is to conduct experiments using a non-equilibrium plasma-assisted
ignition system as a means of operating a small scale engine at lean equivalence ratios over a
wide load/speed range to achieve improved specific fuel consumption. An engine test platform
that integrates the microwave ignition system with a small engine and dynamometer is
established. To quantify the ignition enhancement, experiments are performed to measure engine
stability (COVimep) and specific fuel consumption (sfc) at lean conditions over a variety of engine
load/speeds. In addition, the differences in the pressure profiles using a conventional ignition
device and the spark-assisted microwave device are compared to elucidate the differences in
ignition development and flame propagation time.
2.1.2 Experimental configuration
2.1.2.1 Microwave enhanced ignition system
Imagineering, Inc. has developed a novel microwave-enhanced spark plug for utilization in
automobile engines. This device operates by channeling a 2.45 GHz oscillating electric field
from a commercial magnetron into a non-resistor spark plug, as shown in Fig. 2.1.1. The two
electrodes of the spark plug act as the microwave antennas, so the geometry of the spark plug
need not be altered from its original form. The field produced by the magnetron is subcritical, so
to initiate the discharge a spark is provided by a standard ignition coil. The microwave energy is
absorbed by the electrons in the spark, which are raised to high electron temperature. These
electrons collide with gas molecules and transfer their energy to create excited species, ions, and
42
radicals, as well as increasing the gas temperature. In a previous work investigating this ignition
technology, the rotational temperature and N2 density were measured by rotational Raman
scattering in the region of the microwave discharge (ElSabbagh et al. 2011). The prior study
indicates that, in a 750 Torr mixture of helium and nitrogen (5:70 He:N2), the center of the
discharge reaches a maximum rotational temperature of 1500 K, and at 4.5 mm from the center
of the discharge the temperature reaches as high as 750 K. In addition, it was noted that the N2
concentration was lower than expected for the initial conditions measured just after the
discharge. This indicates that some of the N2 may have been dissociated into atomic nitrogen,
indicating that this type of plasma can effectively produce active radicals.
The microwave ignition system operates at subcritical conditions, necessitating the initiation
of ionization by another source, in this case a standard capacitive spark discharge. The
microwave is coupled to the electrons in the spark during a preset time period starting before and
ending after the capacitive spark discharge. The setup of the ignition system allows for control of
spark timing, microwave deposition timing, and microwave duration. An encoder attached to the
back side of the engine crankshaft outputs a signal of one pulse per revolution of the crankshaft,
set to occur at engine top dead center, which is fed into a BNC Model 575 pulse delay generator.
The delayed signal is sent to an Elmos Model AWG50 waveform generator, as shown in Fig.
2.1.1. A trigger signal is sent to the igniter power supply for the initiation of the standard spark in
the center electrode of the spark plug. The signal generator also outputs a series of pulses to the
magnetron power supply lasting for a total of 2 ms for all experimental conditions. The
magnetron output is channeled through a directional coupler and into an attenuation device for
impedance matching, which is adjusted to minimize the reflected energy from the spark plug
while in the engine. The reflected power is monitored by an Anritsu Model ML2488A power
43
meter. A total of 750 mJ of electrical energy is delivered to the outer electrode of the spark plug,
and 1/4 to 1/3 of that energy is reflected back, resulting in total energy deposition of 500 – 560
mJ. For comparison, the standard spark deposits ≤50 mJ of energy. The signals output by the
waveform generator and the magnetron power supply are monitored using a Tektronix Model
DPO2024 oscilloscope, and the ignition timing is recorded for every experiment.
2.1.2.2 Engine platform
A small scale engine, Fuji Imvac Model BF-34EI, is coupled to a Fuji Electric dynamometer,
and equipped with a 360 pulse per revolution encoder, an NTK air/fuel ratio sensor, a TSI air
flow sensor, and a Kistler in cylinder pressure transducer for quantitative measurement and
control of important engine parameters. A picture of this setup is displayed in Fig. 2.1.2 The Fuji
Imvac BF34-EI is a carbureted single cylinder, four-stroke engine with 34 cc of displacement (39
mm bore and 28 mm stroke length), weight of 2.6 kg, peak power output of 1.49 kW (2.0 hp) at
Figure 2.1.1 Diagram of microwave ignition system
44
7,500 rpm, and peak torque of 1.96 Nm (46.5 lbft) at 5,000 rpm. Equivalence ratio of the intake
charge can be adjusted via two knobs on the carburetor, one for high speed and one for low
speed. All tests are performed using gasoline with an octane number of 90.1. The engine is
coupled to a 7.5 kW (10 hp) Fuji Electric motor via a timing belt, allowing the engine to be
driven and maintained at a constant rotational velocity.
The cylinder pressure is measured using a Kistler Type 6052 pressure transducer installed
parallel to the spark plug in the engine head near the intake valve. The pressure signal is
amplified using a Kistler Type 5064 signal amplifier. An encoder is mounted to the back side of
the engine delivering 360 pulses per revolution and an additional signal of 1 pulse per revolution
for triggering the ignition system. Both of these devices output to an Onosokki Type DS2000
engine monitoring system, which is connected to a CPU for real time data monitoring and
recording. The air flow rate and temperature are monitored via a TSI Model 4021 mass flow
meter. It is accurate to 2% of the reading or 0.05 standard L/min, whichever is greater, and has a
response time of less than 4 msec. The fuel consumption is monitored by reading a graduated
cylinder and taking time measurements every 5 mL at 5000 RPM and 2 mL at 2000 RPM. In
addition, the A/F ratio is monitored with a Model ABM-10 NTK AF-Boost A/F ratio sensor
installed in the exhaust port of the motor just before the muffler, which was found to have an
absolute error of A/F = 1.2.
45
Figure 2.1.2 Diagram and picture of engine test setup.
Dynamometer
Fuji
Engine
Dynamometer Control
CPU
Cylinder Pressure
Air Flow Rate
Fuel Flow Rate
Outputs
Inputs
Throttle Position
Fuel Flow Needles
Ignition Timing
Timing Belt
A/F Ratio Sensor
Crank Angle
Fuel Flow
Knobs
Microwave Ignition
Device
Air Flow
Sensor
A/F SensorPressure
TransducerDC Motor
46
2.1.3 Results and discussion
Motored engine tests were conducted at 2000 and 5000 rpm at wide open throttle (WOT). The
maximum brake torque (MBT) timing for stoichiometric conditions was located and is listed in
Table 2.1.1 as “Timing 1.” The air/fuel mass ratio was varied from rich to increasingly lean
conditions until the engine reached the lean misfire limit, which in this case was judged to be the
point at which the COVimep>50%. When this limit occurred, the ignition timing was advanced
until the new MBT timing was established, and then the process of increasing the A/F ratio was
continued. In this process, the MBT timing was changed 4 times at 2000 rpm and 3 times at 5000
rpm, as listed in Table 2.1.1.
Each A/F ratio condition is tested with and without the use of the microwave-enhanced spark
plug. The plotted data points represent the average pressure data for 263 engine cycles,
maximum and minimum values of the A/F sensor readings over these cycles, one fuel
consumption measurement of a fixed volume (measured by timing the drop in fuel level in a
graduated cylinder, 4% standard deviation in measurement), and at least three engine cycles of
air flow sensor data per A/F ratio setting. Each A/F ratio is tested at least 5 times. It was found
that A/F ratio measured by the AF-Boost meter and that measured by the combination of fuel
Table 2.1.1 Timing of spark and MW discharge
Spark [C.A. ATDC]
MW Start [C.A. ATDC]
MW End [C.A. ATDC]
2000 rpm, Timing 1 17 -5 19
2000 rpm, Timing 2 -42 -69 -44
2000 rpm, Timing 3 -106 -128 -104
2000 rpm, Timing 4 -115 -141 -117
5000 rpm, Timing 1 -43 -96 -36
5000 rpm, Timing 2 -81 -148 -88
5000 rpm, Timing 3 -121 -188 -128
47
flow rate and air flow rate data show a 2% bias, and the combination air and fuel flow rate had a
standard deviation of 1%, while the A/F boost meter had a standard deviation of 2%. Therefore,
a conservative experimental error is ±5% of the reported A/F value, assuming there is no shared
bias between the devices. The A/F sensor is limited to a maximum A/F ratio of 30, and fails to
operate reliably when combustion is incomplete. Therefore, as the lean limit is approached, it is
necessary to rely only on the flow rate data.
Plotted in Fig. 2.1.3 is the COVimep as a function of the A/F ratio. The microwave-assisted
spark plug shows improvement to the COVimep at most A/F ratio conditions, and again is most
effective near the lean limit. Marked on this plot is the stable operating limit, defined as the point
where engine operation becomes unstable, and occurring at COVimep = 10% (Heywood 1988). At
2000 rpm, the engine can operate within the stable operating limit consistently up to A/F ≈ 21
with the microwave-enhanced spark plug, while only rarely can an A/F ≥17 be reached with the
standard spark. At MBT timing 3, the COVimep is just outside of the stable operating limit at A/F
= 28, which indicates that more stable conditions at A/F between 20 and 28 may exist if the
MBT is set between Timing 2 and 3, which was not attempted in this study. At 5000 rpm, the
stable operating limit is definitively extended from A/F ≈ 17 for the standard plug to A/F ≈ 22
for the microwave-enhanced plug. In addition, in the range of A/F near stoichiometric
conditions (A/F = 14.7) the COVimep shows less unstable outliers with the microwave-assisted
spark plug as compared to the standard spark plug. These results indicate that by using the
microwave ignition device, the flame kernel development and flame propagation processes can
be fully completed over a wider range of conditions than with the capacitive discharge.
The imep as a function of A/F ratio is plotted in Fig. 2.1.4. At near stoichiometric (A/F≈14.7)
conditions, the engine outputs nearly the same imep using both spark plugs. However, in the
48
Figure 2.1.3. Coefficient of variation of the indicated mean
effective pressure for standard and microwave-enhanced engine
operation, showing the limit of stable operating conditions. Top:
2000 rpm. Bottom: 5000 rpm.
0
10
20
30
40
50
12 16 20 24 28
CO
Vim
ep,
%
A/F Ratio
No MW, Timing 1MW, Timing 1No MW, Timing 2MW, Timing 2MW, Timing 3Stable Operating Limit
0
10
20
30
40
50
12 14 16 18 20 22 24
CO
Vim
ep,
%
A/F Ratio
No MW, Timing 1MW, Timing 1No MW, Timing 2MW, Timing 2Stable Operating Limit
49
Figure 2.1.4 Indicated mean effective pressure as a function of
air/fuel mass ratio. Top: 2000 rpm. Bottom: 5000 rpm
0
200
400
600
800
1000
1200
12 16 20 24 28
IME
P,
kP
a
A/F Ratio
No MW, Timing 1
MW, Timing 1
No MW, Timing 2
MW, Timing 2
MW, Timing 3
0
200
400
600
800
1000
1200
12 14 16 18 20 22 24
IME
P,
kP
a
A/F Ratio
No MW, Timing 1
MW, Timing 1
No MW, Timing 2
MW, Timing 2
50
range of A/F = 17-22, the microwave spark plug shows an average improvement in the imep of
9.8% at 2000 rpm and 6.1% at 5000 rpm. Particularly at the leanest equivalence ratios, there is a
clear improvement in the imep at both engine speeds. At 2000 rpm, above A/F=24 there is no
ignition for the standard spark cases, while the imep using the microwave spark plug can be
maintained at reasonable levels even up to A/F = 28. At 5000 rpm, we see more than a factor of 2
increase in the imep above A/F=20 using the microwave plug. Note that the microwave energy
coupled into the spark only accounts for 2–2.25% of the total energy production of the engine
(assuming the maximum engine output of 1.49 kW at 7500 rpm), so this increase in the imep is
due to improvements in the ignition kernel growth process and not due to overall heating of the
gas in the cylinder.
Figure 2.1.5 shows P-V diagrams just before the lean limit of normal spark plug operation for
the 2000 and 5000 rpm cases (A/F = 23.5 and 22.2, respectively). These cycles are typical for
their respective operating conditions. The integrated area under the P-V curve is clearly larger
for the microwave ignition cases, indicating greater power output using this igniter. This is likely
because the ignition kernel development and ensuing flame propagation happened faster in the
microwave ignition case, allowing heat release to occur at the proper time in the cycle.
The specific fuel consumption (sfc) is plotted as a function of A/F ratio in Fig. 2.1.6. For 2000
rpm, a minimum is reached between A/F = 17-20, and for 5000 rpm the minimum is between
A/F = 17-19. While there is little difference in the minimum sfc between the microwave-
enhanced spark plug and the standard spark plug, it should be noted that, in these A/F ranges, the
COVimep can be reduced by using the microwave-enhanced spark plug. These findings indicate
that, at lean conditions where the sfc is minimized, the improvement in the COVimep is significant
and should allow small engines to operate in the optimum range near A/F = 17-20. For fully
51
Figure 2.1.5 Pressure as a function of volume for standard and
microwave-enhanced engine operation. Top: 2000 rpm, A/F =
23.5. Bottom: 5000 rpm, A/F = 22.2
-500
0
500
1000
1500
2000
2500
0 10 20 30 40
Pre
ss
ure
, k
Pa
Volume, cm3
No MW
MW
-500
0
500
1000
1500
2000
2500
3000
3500
0 10 20 30 40
Pre
ss
ure
, k
Pa
Volume, cm3
No MW
MW
52
Figure 2.1.6 Specific fuel consumption as a function of A/F ratio.
Top: 2000 rpm, A/F = 23.5. Bottom: 5000 rpm, A/F = 22.2
0
0.05
0.1
0.15
12 16 20 24 28
SF
C, m
g/J
A/F Ratio
No MW, Timing 1
MW, Timing 1
No MW, Timing 2
MW, Timing 2
MW, Timing 3
0
0.025
0.05
0.075
0.1
0.125
0.15
12 14 16 18 20 22 24
SF
C [
mg
/J]
A/F Ratio
No MW, Timing 1
MW, Timing 1
MW, Timing 2
No MW, Timing 2
MW, Timing 3
53
optimized MBT timing, it can be expected for the sfc to be a continuous arc from the rich to the
lean equivalence ratios.
Another metric to observe engine ignition performance is the location of maximum pressure.
The maximum pressure as a function of the crank angle location of maximum pressure is
presented in Fig. 2.1.7 for the 5000 rpm operating condition. The top right edge of this dataset is
termed the “fast burn” limit (Heywood 1988). This indicates the region in which the maximum
pressure is largest, i.e, where the engine is converting all of the chemical enthalpy into work at
the fastest rate, and is generally for conditions far from the lean limit. The maximum pressure
attainable must decrease as the maximum pressure location is delayed past 0° ATC due to the
increasing cylinder volume The application of the microwave ignition system extends the fast
burn region to lower crank angles for the same ignition timing as the capacitive discharge,
indicating that microwave system can accelerate the ignition process to occur closer to top TDC
as compared to the capacitive igniter, even for strong mixtures. Note that, at these early crank
angles the range of maximum pressures is smallest, indicating that these operating conditions are
stable as compared to more delayed ignition events. Also note that the bottom branch indicates
misfires, in which the maximum pressure is always reached at TDC. The left hand side limit of
the dataset is the “slow burn” limit. These reduced pressure cycles indicate relatively weak
burning mixtures, i.e, very rich or very lean. Note the decrease (in crank angle) of the slow burn
limit using the microwave ignition system, dramatically extending the range of operating
conditions in which successful ignition can be achieved at crank angles near top dead center. The
capacitive ignition system can only completely burn lean mixtures at later crank angles, after the
chamber volume has increased, so the maximum pressure attained is necessarily lower.
54
2.1.4 Conclusions
A microwave-assisted ignition system has been successfully tested in a small gasoline engine
for two engine speed conditions. The results indicate that the microwave ignition system can
extend the lean burn limit by 20-30%, according to the COVimep in stable engine operating limits.
The results also demonstrate an improvement in the imep of 6-10% at near stoichiometric
operating conditions. The minimum sfc was found to occur at lean operating conditions of A/F =
17-20, where the COVimep is above the stable operating limit for the standard spark plug, yet the
microwave ignition system can successfully operate. Finally, the crank angle location of
Figure 2.1.7 Maximum pressure as a function of the location of maximum pressure.
Engine speed: 5000 rpm, A/F = 12-24.
1500
2000
2500
3000
3500
4000
4500
5000
5500
0 5 10 15 20 25 30
Ma
x P
res
su
re,
kP
a
Location of Max Pressure, Degrees ATC
Plasma Timing 1
Plasma Timing 2
Plasma Timing 3
No Plasma Timing 1
No Plasma Timing 2
55
maximum pressure can be reduced using the microwave ignition system, thus providing both
faster ignition at stoichiometric conditions as well as for lean mixtures. It is thus demonstrated
that using the microwave-assisted spark plug allows for a wider range of stable operating
conditions, which encompasses the region of greatest fuel conversion efficiency, and is not
achievable with the capacitive spark plug. Another important advantage of operating at leaner
conditions is the necessary reduction of flame temperature, which will result in less NOx
emissions and is a significant concern in automobile engines. Future studies should be aimed at
confirming the proposed emission reduction, and perhaps evaluate whether catalytic conversion
can be avoided using PAI.
While these results do conclude that ignition devices with higher energy and plasma volume
than capacitive sparks can more successfully ignite a small scale IC engine, they do not elucidate
the fundamental mechanism by which this is possible. To gain further insight into the ignition
process, direct imaging and detailed diagnostics of the combustion process with PAI are
necessary. The remainder of this dissertation will take a closer look at the ignition process for
high energy and large volume plasma ignition sources. However, only through works proving the
usefulness of PAI can the discharges that will significantly alter real combustor behavior be
identified. It is clear spark-assisted microwave plasmas can significantly improve engine
operation, and further research should be conducted to optimize this discharge type for use in
commercial SI engines.
56
2.2 Nanosecond repetitively pulsed discharges: Inter-pulse coupling and
ignition enhancement in PDEs
2.2.1 Introduction
The development of reliable ignition sources for aerospace combustion systems that operate
within restricted residence times is a major bottleneck for technologies such as scramjets and
pulsed detonation engines (PDEs) (Starikovskiy et al. 2013, Ju et al. 2015b). The ignition kernel
development time in quiescent systems is controlled by the mixture reactivity, temperature,
pressure, and ignition energy source (Maly 1984, Heywood 1988, Chen et al. 2011). In aerospace
engines, heterogeneous fuel/air mixtures, high bulk flow velocities, and local turbulent
fluctuations all affect the required energy for successful ignition kernel generation (Ballal et al.
1975, Peng et al. 2013). Because of these variations in required energy, the minimum ignition
energy (MIE) for fuel-air mixtures in demanding applications can be significantly greater than
the MIE determined in quiescent environments. In addition, for applications such as PDEs,
decreasing the ignition time directly affects the overall thrust of the engine.
In the work of Maly (Maly 1984), the optimal ignition device is described as one that (a)
deposits the energy needed for ignition in the shortest possible time period, (b) creates an
expanding plasma volume with the highest possible surface velocity, (c) reaches the largest
possible plasma radius, (d) does not supply additional energy after the flame has propagated
beyond the plasma radius, and (e) only deposits energy into a narrow spherical shell propagating
away from the electrodes. To fulfill these basic criteria, short duration (10-100 ns) plasma
discharges have been suggested as a method to reduce ignition times while increasing energy
deposition efficiency as compared to capacitor discharge ignition systems (CDI, typically 0.1-0.3
57
ms spark duration) or transistorized coil ignition systems (TCI, typically 1-5 ms duration) (Maly
1984, Raizer 1991). The effectiveness of this discharge type is due to the deposition of most of
the discharge energy into the breakdown phase of the gas discharge.
In order to form a gas discharge, a large energy input is required to ionize the initially neutral
gas molecules and induce an electron avalanche, which in turn dissociates, excites, and ionizes
the molecules and atoms in the discharge gap. After the initial breakdown phase, which occurs
on the order of a few nanoseconds at atmospheric pressure, the discharge can develop into a
longer lifetime glow, arc, or streamer discharge. The breakdown phase has a high peak current
(on the order of 100 A) as compared to arc (order of 1-10 A) and glow (order of 0.1-1 A)
discharges so has a faster energy deposition rate than these other plasma discharge types (Maly
1984, Raizer 1991). While all atmospheric pressure discharges must be initialized by the
breakdown phase, the proportion of energy into each phase can be controlled by the peak initial
voltage and duration of the discharge. After the breakdown stage, a steep temperature and active
species concentration gradient is set up around the inter-electrode region, which propagates
outwards from this region into the unburned mixture. Additional energy deposition after the
breakdown into the following arc or glow stages primarily serves to heat the gas between the
electrodes, and is largely ineffective from an ignition perspective if the initial breakdown has
enough energy to ignite the mixture. However, if the initial breakdown does not supply enough
energy to bring the ignition kernel to the critical radius and temperature for transition to a freely
propagating flame (Chen et al. 2011, Kim et al. 2013) or if the initial propagation rate is
sufficiently slow, then there is time to add energy into the ignition kernel and reduce its
development time. This can be accomplished by increasing the duration of the discharge pulse,
or by depositing multiple pulses into a single ignition kernel.
58
To take advantage of the efficient energy deposition in the breakdown phase, nanosecond
duration discharges have recently been investigated (Starikovskaia 2006, Starikovskii et al. 2006,
Starikovskiy et al. 2013, Ju et al. 2015b). Notably, single pulse ignition using a transient plasma
igniter (TPI) has been shown to reduce ignition time compared to conventional spark discharges
in both engines and quiescent environments (Wang et al. 2005, Busby et al. 2007, Cathey et al.
2008, Shiraishi et al. 2009, Singleton et al. 2011). The faster ignition times in methane-air and
ethylene-air mixtures were attributed to the multiple simultaneous streamer type discharge
channels heating a larger gas volume than the conventional spark discharge. Also, a contribution
from kinetic enhancement via reactions of excited species and radicals created during the
breakdown due to the high reduced electric field (E/N) was reported. It was also found that the
ignition time was independent of the duration of the discharge in the range of 10-50 ns, and was
only significantly affected by the pulse rise rate and the peak E/N (i.e., peak voltage for constant
discharge gap and gas density). Pancheshnyi et al. (Pancheshnyi et al. 2006) explored ignition
behavior using a nanosecond repetitively pulsed (NRP) plasma in quiescent propane-air mixtures
using a pin-to-pin configuration. Each pulse was below the breakdown threshold for the gas
mixture, but it was found that multiple pulses caused significant pre-ionization and eventually
breakdown, leading to temperatures in excess of 3000 K and significant dissociation and
ionization. The ignition time could be decreased with the application of additional discharge
pulses beyond the minimum amount necessary to ignite the mixture if the MIE for the mixture
was greater than the per pulse energy of the discharge. This was attributed to a building up of
neutral radicals in the ignition kernel before the flame propagated away from the discharge
region. In subsequent studies on the same discharge in air (Rusterholtz et al. 2013),
measurements of rotational temperature, vibrational temperature, N2(B), N2(C), O, and electron
59
number were carried out. These experiments support the conclusion that the plasma indeed does
produce a large amount of O atoms (up to 50% dissociation) in the discharge center, which is
mostly produced by the relaxation of excited states of N2. The quenching of excited states is also
responsible for the so called “fast gas heating” (Mintoussov et al. 2011, Popov 2011) occurring
within nanoseconds of the discharge onset, and heating the discharge region to temperatures in
excess of 2000 K.
The purpose of the present study is to extend the understanding of NRP discharges as an
ignition source to flowing environments and practical aerospace fuels, as well as to explore the
inter-pulse coupling effect and the fuel kinetic effects contributing to changes in the ignition
time. For this purpose, a 12 ns duration plasma discharge with a repetition frequency of up to 40
kHz is used with a pin-to-pin electrode configuration for ignition in two experimental platforms.
First, the dynamics of kernel development and subsequent flame propagation for methane-air
mixtures in a turbulent flow field are observed in a flow tube using schlieren imaging. Second,
NRP discharge ignition is tested in a PDE and compared to conventional capacitor discharge
ignition (CDI) in ethylene-air and aviation gasoline (avgas)-air mixtures. Testing of NRP
discharges in the glow and corona regimes in PDE engines has been previously shown to reduce
ignition time compared to conventional ignition (Busby et al. 2007, Rakitin et al. 2008, Zhukov
et al. 2008, Starikovskiy et al. 2012). This study aims to extend the fundamental understanding
of the observed ignition enhancement by determining the controlling parameters leading to
reduced ignition times using NRP discharges in the spark regime.
60
2.2.2 Experimental setup
2.2.2.1 Plasma igniter
Both a NRP discharge and a multiple spark discharge (MSD, P/N 62152) system were tested
parametrically in this study, as outlined in Table 2.2.1. The NRP power supply (FID GmbH FPG
30- 50MC4) produces peak pulse amplitudes of 22 kV, pulse duration of 12 ns FWHM, and
pulse repetition frequencies of up to 40 kHz. Three values were used for the peak open-circuit
voltage in the present experiments (9 kV, 15 kV, and 22 kV), while the frequency is varied in the
range of 0.869-40 kHz and the number of pulses is varied from 2-200. The MSD igniter is an
aftermarket automotive CDI device, which produces a series of sparks (generally 3 for the
current set of experiments) at a fixed pulse repetition frequency of 0.869 kHz. The peak voltage
of the MSD igniter was measured to be 6.5 kV, and after breakdown produced a peak current of
15 A for a power pulse of approximately 1 μs followed by a current pulse continuing for 50 µs
thereafter (Busby et al. 2007). This type of plasma can be characterized as an arc discharge. The
MSD power supply provides between 105 and 115 mJ; however, values between 5-10 mJ were
measured in these experiments due to the poor efficiency of arc discharge energy deposition
(Maly 1984, Raizer 1991). The same non-resistive spark plug (Autolite Racing AR3911) and
cable were used in all experiments. The ground strap of the spark plug was removed and
replaced with a sharpened steel electrode, and the center electrode was sharpened to a point to
produce a pin-to-pin configuration. The spark gap was fixed at 1.4 mm, which was found to be
the maximum gap distance permissible to ensure the discharge occurred only between the two
electrodes. The voltage and current traces were measured using a LeCroy high voltage probe
(PPE20KV) and a Pearson Coil (Model 6585), respectively, on the 3 m cable from the power
supply to the spark plug. In the case of the NRP discharge, a 300 Ω resistor was connected in
61
parallel with the power supply to supply a constant load and damp reflections in the voltage
pulse.
Plasma energy measurements were made following the method outlined by Pancheshnyi et al.
(Pancheshnyi et al. 2006). The displacement current is computed as
, where the
measured capacitance (C) is 18 pF. The conduction current is the difference between the
measured current and the displacement current. The total energy was computed by integrating
power over the initial pulse using and . Analogous to the
results of Pancheshnyi et al. (Pancheshnyi et al. 2006), the peak current and voltage were not
constant, but rather the voltage was higher in the first few pulses, while the current was lower in
the following pulses. This corresponds to the pre-breakdown and streamer to spark transition,
which increases the conductivity of the gas until all pulses are in the spark regime (Pancheshnyi
et al. 2006, Pai et al. 2010b, Pai et al. 2010a). The first pulse was never sufficient to cause gas
breakdown, but did create significant pre-ionization for breakdown to occur on the second pulse
for all of the pulse amplitudes tested. After the first three pulses, the voltage and current
Table 2.2.1: Discharge parameters for both MSD and NRP igniters. Parameters in MSD
are fixed.
Igniter
Average
Energy/pulse
[mJ]
Pulse
Duration
[ns]
Pulse
Frequency
[kHz]
Number
of pulses
Open-
circuit
Voltage
[kV]
Max
Current
[A]
NRP 0.8, 1.9, 3.2 12 0.869-40 2-200 9, 15, 22
kV 20-50
MSD 5.7 1000 0.869 3 6.5 kV 15
62
waveforms of the discharge are repeated for all subsequent pulses. In addition, pulse frequency
had only a small effect on the pulse energy. The accuracy and uncertainty of the measurements
of ignition energy deposition are highly sensitive to the bandwidths of both the voltage probe
(100 MHz, 3.5 ns rise time) and the Pearson Coil (200 MHz, 1.5 ns voltage rise time), which are
close to the ≈ 5 ns rise time of the NRP discharge pulse. This causes relatively large uncertainties
considering that the entire waveform is needed to calculate the energy, not just the peak voltage
or current. The estimated total uncertainty of the NRP discharge energy measurements is 50% of
the total energy. This was due to the time resolution of the measurement devices, interference
from reflected pulses which could not be eliminated in this system, pulse-to-pulse variation in
energy deposition, changes in electrode geometry due to wear, and unaccounted parasitic losses.
Energy measurements of NRP discharges using the measurement devices herein, as well as with
similar devices, are discussed in further detail in other works (Pancheshnyi et al. 2006, Pai et al.
2010b, Pai et al. 2010a, Rusterholtz et al. 2013).
NRP discharges in the pin-to-pin configuration have been shown to operate in three regimes:
corona, glow, and spark. Pai et al. (Pai et al. 2010b) have outlined the conditions in which these
discharges exist in terms of applied voltage, pulse frequency, inter-electrode distance, and
ambient gas temperature. The different regimes are characterized by increasing discharge current
and emission intensity in the visible range, particularly from electronically excited and ionized
nitrogen (N2 C-B (0,0) and N2+ (B-X) (0,0) transitions). Pai et al. reported that only corona and
spark discharges existed in their system below 700 K gas temperature at a 5 mm inter-electrode
distance, 30 kHz pulse repetition frequency, and 1-10 kV max voltage. Based on these findings,
and visible observation, the discharge used in this study is in the spark regime, for the lowest
voltage used is 9 kV, the inter-electrode gap is 1.4 mm, and the ambient temperature is at
63
maximum 400 K, precluding the possibility of corona or glow discharges. In spark discharges,
rapid temperature increase in the narrow discharge channel leads to a fast pressure rise, and a
shock wave propagates outwardly from the discharge after the initial breakdown pulse. The
presence of this shock wave is apparent in the Schlieren imaging, which is further evidence of
the spark regime.
2.2.2.2 Pulsed detonation engine
The PDE facility is described in detail in other works (Schauer et al. 2001, Busby et al. 2007,
Tucker et al. 2008). The engine consists of four detonation tubes, two of which were fired in this
study, one as the experimental platform and the other one as a dummy ballast tube. The system is
built on a General Motors Quad-4 Dual Overhead Cam (DOHC) 4-cylinder engine head. In place
of the cylinder and piston assemblies, steel detonation tubes 152 cm in length and 5.08 cm in
diameter are bolted to the engine head. The experiment tube is fitted with a Shchelkin-like spiral
to enhance turbulence generation and facilitate the deflagration-to-detonation transition (DDT),
the length of which was 91.4 cm for ethylene experiments and 121.9 cm for avgas experiments.
The PDE works on a three-phase (fill, fire, and purge) cycle. The fill phase of the cycle
consists of filling the detonation tube with premixed flammable gases through the intake valves
of the engine head, after which the intake valves are closed and the fire phase begins. The fire
phase is subdivided into three steps. First, a user determined “ignition delay” is applied to allow
pressure transients to dissipate. Second is the “ignition time,” initiated by the triggering of the
igniter, and consisting of the time for the ignition kernel to develop into a self-propagating flame
and undergo DDT. Third is detonation wave propagation, which produces a rapid pressure rise,
accelerating the burned gas through the detonation tube, and producing thrust. The purge phase
64
follows, in which a half-tube full of air is pumped through the exhaust valves of the engine to
purge the burned gas, and the cycle is then repeated. All tests were conducted at an engine cycle
rate of 10 Hz.
The PDE diagnostic system consists of a dynamic pressure transducer in the engine head and
nine ion probes located at equal intervals of 15.24 cm along the detonation tube, starting 25.72
cm from the engine head. The ignition time is determined when the signal from the pressure
transducer reaches a slope of 5V/second. The position and speed of the deflagration or
detonation wave is monitored by the ion probes. When two consecutive wave speed
measurements first result in the same measured value, the wave speed is determined to be the C-J
velocity. To determine the DDT distance, the wave speed is linearly interpolated between the
measurements just before and after the C-J velocity is reached. The DDT time is determined by
calculating the time when the wave reaches the DDT distance.
2.2.2.3 Flame-development visualization platform
For the flame-development visualization experiments, the detonation tube is replaced by a
steel tube 261.6 cm in length and fitted with a 91.4 cm long, 5.08 cm x 5.08 cm square test
section with polycarbonate walls, as depicted in Fig. 2.2.1 (Stevens et al. 2011). The Schlieren
imaging setup consists of an LED light source, collimating mirrors, a knife edge, and a Phantom
v711 high-frame-rate Complementary Metal-Oxide Semiconductor (CMOS) camera. The light
emitted from the LED source is collimated by one of the mirrors and sent through the
polycarbonate walls of the Schlieren test section. A mirror setup identical to the collimating
mirrors focuses the light onto the CMOS chip of the camera, with part of the light blocked by the
knife edge. Images were collected beginning just prior to the ignition event and continued until
65
the combustion wave had fully propagated out of the view of the camera. The frame rate was
fixed at 100 kfps with 2 µs exposure time.
To process these images, the frames before the ignition event were averaged to provide a
background image that all subsequent images were divided by. For the images provided in the
following section, a color look-up table was applied for increased color contrast and hence better
visual identification of the flame front. A flame tracking program developed in MATLAB®
using its built-in edge finding function, employing the Canny method (Canny 1986), was used to
monitor the evolution of the flame front by finding the local maxima in pixel intensity gradients
and tracking the evolution of these maxima throughout the images of each ignition event
(Santner et al. 2013). Due to the turbulent flow environment, the ignition kernel was far from
symmetric and the flame front was not always sharp in the images. In addition, lubrication oil
which is mixed with the fuel/air mixture to lubricate the intake valves continuously collects on
the window inner surface, again reducing the sharpness of the Schleiren imaging. As a result,
the edge tracking program cannot reliably be used to measure the evolution of the flame front
and extract the flame speed as is done in laminar spherically propagating flames (Kim et al.
2013, Santner et al. 2013). This is a major drawback of this particular experimental setup, and
should be improved upon in future studies. Instead of tracking the entire flame area, the leftmost
and rightmost edges are tracked, and the difference in pixels between them is referred to as the
area of the kernel. The rightmost pixel in the ignition kernel, before it develops into a self-
propagating flame, is used to determine the bulk gas velocity. The average initial velocity was
found to be 10 ± 2.5 m/s in all flow visualization experiments. Because these experiments were
performed in a modified PDE, the flow velocity is changing throughout the cycle, and is also
66
subject to cycle-to-cycle variations typical of engine experiments. Therefore, the exact velocity
throughout the flame-development process cannot be accurately determined.
Figure 2.2.1 Schematic of flame-development visualization experiment
(Stevens et al. 2011).
PDE Head
Te
st g
eo
me
try
Sp
he
ric
al
mir
ror
Sp
he
ric
al
mir
ror
Ca
me
ra ta
ble
Po
lyc
arb
on
ate
w
alls
De
ton
ati
on
tu
be
Kn
ife
ed
ge
Fo
cu
sin
g l
en
s
LE
D li
gh
t s
ou
rce
CM
OS
C
am
era
50
mm
F
L le
ns
15
0 m
m F
L
len
sA
dju
sta
ble
slit
Du
mm
y tu
be
Sp
ark
Plu
g
67
High-frame-rate imaging and optical emission spectrometry (OES) measurements of the NRP
discharge were also performed in quiescent laboratory air. The high-frame-rate images were
performed at 390 kfps with 2 µs exposure time using a Phantom v711 CMOS camera. The
camera was focused in a region of ≈ 1 cm2 containing the electrode gap and surrounding air. The
emission intensity ranging from 400 – 950 nm was extracted as a function of time to determine
the lifetime of the plasma discharge. To quantify the emission intensity, the average pixel value
in the image domain was calculated for each image. As the onset of emission generally saturates
the CMOS sensor for the 2 µs exposure time used, the peak intensity cannot be extracted
accurately. To truly measure peak emission intensity, a shorter exposure time or a neutral density
filter could be used to bring down the total photon count on the CMOS sensor. However, the aim
of the present measurement is to monitor the inter-pulse emission, which is significantly less
intense than the peak emission, so no effort was made to decrease the light intensity. The optical
emission spectroscopy measurements were conducted using an Ocean Optics USB2000+
spectrometer that operates in the range of 190-874 nm with an optical resolution of ≈ 1.4 nm
FWHM, and has been calibrated for a flat response across the operating wavelength range. The
light is collected through an optical fiber focused on the center of the electrode gap, and
averaged in time over the entire NRP discharge burst.
2.2.2.4 Numerical modeling
To approximate the plasma processes in NRP discharges, ZDPlasKin (Pancheshnyi et al.
2008), a plasma chemistry reaction solver incorporating BOLSIG+ (Hagelaar et al. 2005) for
calculation of electron-collision reactions, and the SENKIN package of CHEMKIN II (Lutz et al.
1988) are coupled together to efficiently model species evolution in 0-dimensional plasma-
assisted combustion (PAC). A full description of the model is provided in Chapter 4. Gas heating
68
during the plasma is calculated in the method of Flitti and Pancheshnyi (Flitti et al. 2009), which
accounts for energy coupled into the translational temperature of electrons and of the bulk gas
separately. The simplified methane-air plasma model of Aleksandrov et al. (Aleksandrov et al.
2009b) is used as the basis for the plasma chemistry, and ethylene reactions from Lefkowitz et al.
(Lefkowitz et al. 2015d) are added to complete the model for ethylene-air. The plasma species
considered follows the example of Aleksandrov et al. (Aleksandrov et al. 2009b), and includes
; ,
,
(all grouped as N2(B));
,
(all grouped as N2(a’));
; ;
,
,
(all grouped
as O2(4.5eV)); ;
;
;
;
. Energy level
diagrams for N2 and O2 are provided in Chapter 4. The electron-collision cross-sections for O2,
N2, and CH4 are downloaded from the LXCat online database (Hayashi , Napartovich et al. ,
Phelps), and cross-sections for C2H4 are computed in the method described by Janev and Reiter
(Janev et al. 2004). Quenching reactions from the plasma model are added to the combustion
chemistry model USC-Mech II (Wang et al.) to create a simplified reaction scheme for methane-
air and ethylene-air PAC at elevated temperatures (> 800 K) to operate during the inter-pulse
time. The aim of this model is not to include every relevant species in the plasma, but only the
ones responsible for principal generation of radicals and heat release. Because the Poisson
equation for the electric field distribution is not solved, and thus charge separation and sheath
formation are not considered, the reduced electric field (E/N) cannot be explicitly calculated
from the measured applied voltage. Thus, an estimated E/N value of 320 Td was assumed in
order to mimic the experimental results for emission intensity. Specifically, a large enough value
of E/N such that the radical concentration did not fully decay between pulses for the 40 kHz case
was selected. The degree of radical production was manipulated via the duration of the plasma
69
pulse in the model. Note that spark discharges can only truly be modeled by a multi-dimensional
scheme; however, this 0-dimensional model is found to be sufficient to give an estimate of which
radicals are created by the plasma and the lifetimes of these radicals.
2.2.3 Results and discussion
2.2.3.1 Flame-development visualization
To gain an understanding of the effects of peak voltage, pulse repetition frequency, and the
number of pulses in a burst on flame-development dynamics, each of these parameters were
varied independently in both the flow visualization and PDE testing platforms. The ignition
visualization experiments were carried out for stoichiometric methane-air and ethylene-air
mixtures. Two timescales were identified for comparison of ignition events: the flame-
development time, which represents the time it takes for the kernel to reach 10% of the total area
of the observation domain, and the rapid burning time, which is the time it takes for the flame to
go from 10% to 90% of the observation domain (Heywood 1988). In the ethylene-air
experiments, no significant improvement of flame-development time was found; however, for
methane-air mixtures the flame-development time was found to be a function of total ignition
energy and pulse frequency. The rapid burning time was found to be insensitive to the ignition
device or discharge parameters for both fuels.
In Fig. 2.2.2, images of the ignition event at 4 ms after the initial discharge in a stoichiometric
methane-air mixture visually represent the effects of pulse energy and pulse number on the
flame-development time. The pulse repetition frequency was fixed at 40 kHz for all images. For
fixed energy per pulse, the growth rate of the ignition kernel increases with the number of pulses
up to 50 pulses, which corresponds to 1.25 ms after the initial pulse. Further increases in pulse
70
number produces negligible reduction in the flame-development time since the kernel surface has
already moved away from the inter-electrode region and is transitioning into a self-propagating
flame. For a fixed number of pulses, there is little difference between the 1.9 mJ/pulse and 3.2
mJ/ pulse conditions; however, the 0.8 mJ/pulse condition exhibits a significantly slower flame-
development process, indicating some fundamental change in the discharge properties if the
pulse energy is too low. The MSD ignition is comparable to the three pulse NRP experiments at
1.9 and 3.2 mJ/pulse, but is significantly slower than the 20 pulse NRP experiments. The flame-
development time as a function of total ignition energy for all methane-air visualization
experiments is summarized in Fig. 2.2.3. The total energy (average energy per pulse multiplied
Figure 2.2.2 Schlieren images of stoichiometric methane-air ignition 4 ms after initial
discharge for both NRP and MSD ignition systems. NRP pulse frequency = 40 kHz. Gas
flow = 10 m/s from left to right.
50 Pulses
1.9 mJ/pulse0.8 mJ/pulse 3.2 mJ/pulse3 Pulses
20 Pulses
No Ignition
MSD, 5.7 mJ/pulse, 3 pulses
Inc
rea
sin
g n
um
ber o
f p
uls
es
Increasing energy per pulse
71
by number of pulses) of the discharge has a strong effect on the flame-development time, as
evidenced by the best fit to the high frequency data. In addition, at pulse repetition frequencies
below 10 kHz the flame-development time is significantly increased. A closer look at low and
high frequency ignition events reveals why this is the case.
Figure 2.2.4 compares two ignition events, both with 16 mJ total energy deposition (5 pulses
at 3.2 mJ/pulse), but one with a pulse repetition frequency of 2 kHz (one pulse every 0.5 ms,
total time of 2.5 ms) and the other at 40 kHz (one pulse every 0.025 ms, total time of 0.125 ms).
The 2 kHz condition is displayed in the left panels, with each image corresponding to a new
voltage pulse for the first three images. With a time interval of 0.5 ms between pulses, the bulk
Figure 2.2.3 Flame-development time extracted from schlieren imaging of
stoichiometric methane-air ignition events for varying total energy
deposition and pulse frequency.
0
1
2
3
4
5
6
7
1 10 100 1000
Fla
me
-De
ve
lop
me
nt
Tim
e [
ms
]
Total Energy [mJ]
MSD
NRP discharge, f = 1-5 kHz
NRP discharge, f = 10-40 kHz
72
gas velocity of 10 ± 2.5 m/s transports each kernel 4 - 6 mm from the inter-electrode region
before the next voltage pulse arrives, which is far enough to prevent additional energy from the
discharge from entering into the ignition kernel. Every subsequent pulse creates a new ignition
kernel, which combines with previous ignition kernels downstream of the inter-electrode region,
in this case forming two large distinct flame regions visible by 2 ms after the initial discharge.
By 4 ms, the two large kernels have separated further, and by 6 ms both kernels have begun to
transition into self propagating flames. At even lower frequencies, the kernels would never
interact downstream of the electrodes, which often resulted in quenching of the individual
kernels and unsuccessful ignition. The 40 kHz condition is displayed in the right panels. Here, all
of the discharge pulses are deposited into a single large ignition kernel since the gas only travels
0.2 – 0.3 mm between discharge pulses. By the 0.5 ms image, the five pulses have already
completed (0.375 ms earlier), and a large single ignition kernel has formed, which is then
transported away from the electrode region, where it proceeds to transition into a self-
propagating flame and by 6 ms has already expanded into most of the observation domain.
Figure 2.2.5 shows the trend of decreasing flame-development time with increasing repetition
frequency for fixed energy inputs. The three energy conditions show similar trends of decreasing
flame-development time with increasing pulse frequency.
Note that for the same repetition frequency (0.869 kHz) and total energy input (≈ 16 mJ), the
MSD ignition has a shorter flame-development time than the NRP discharge. This can be
attributed to the substantial difference of pulse duration between NRP discharges (≈ 20 ns) and
the MSD (≈ 1 µs), resulting in a higher per pulse energy for the MSD igniter, as summarized in
Table 2.2.1. At the MSD frequency (0.869 kHz), only one discharge pulse can be deposited into
an ignition kernel before it moves away from the electrode region, preventing the inter-pulse
73
coupling effects observed for high-frequency NRP discharges. Therefore, for low frequency
pulsed discharges, the flame-development time is not a function of the total energy deposition,
only the per-pulse energy deposition. The flame-development time of the NRP discharge
becomes shorter than that of the MSD at pulse repetition frequencies greater than 5 kHz,
indicating the activation of inter-pulse coupling effects. The details of these effects will be
discussed in the following sections. Also note that in Figs. 2.2.3 and 2.2.5 there is significant
scatter in the collected data. This is largely due to variations in the flow velocity, which has a
particularly large effect when the inter-pulse time is on the same order as the ignition kernel
residence time, so small flow variations can alter the number of pulses deposited into a single
ignition kernel.
To visualize the plasma characteristics in more detail, high frequency images of NRP
discharges in air were taken at a rate of 390 kfps. The plasma emission intensity is plotted for
three different pulse frequencies at 22 kV pulse amplitude (3.2 mJ/pulse) in Fig. 2.2.6a. The
plasma emission is completely quenched between pulses for 5 kHz and 10 kHz pulse repetition
frequencies, while for 40 kHz repetition frequency the emission is always visible and the
minimum emission intensity between the pulses reaches a steady state by the third pulse. In Fig.
2.2.6b, the emission intensity at three different pulse amplitudes and 40 kHz pulse repetition
frequency is presented. At 9 kV pulse amplitude, little residual emission builds up between
pulses, while for 15 kV pulse amplitude the minimum emission between pulses is continuously
building up, and for 22 kV pulse amplitude the minimum emission again rapidly builds up to a
steady state. These images indicate that active species remain in the inter-electrode region long
after the voltage pulse has ended, and their concentration is large enough to effect subsequent
pulses in the burst. To identify the species contributing to the emission, OES measurements were
74
Figure 2.2.4 Ignition kernel development for 5 pulses of 3.2 mJ per pulse. Left images:
pulse repetition frequency of 2 kHz. Right images: pulse repetition frequency of 40 kHz
for stoichiometric methane-air mixtures.
0.01 ms
0.5 ms
1.0 ms
2.0 ms
4.0 ms
6.0 ms
2 kHz 40 kHz
75
conducted, also in air. The results are presented in Fig. 2.2.7 for a burst of 50 pulses at 40 kHz
repetition frequency and 22 kV pulse amplitude, which represent the highest energy discharge
used for this study.
The emission lines were compared with the NIST atomic spectral database (Kramida et al.
2014) and the book of Pearse and Gaydon (Pearse et al. 1965). Clearly visible are lines from the
N2 2nd
positive system ( ) in the 300-400 nm range, O+ and N
+ ions in the 400-700
nm range, and O and N atoms in the 700-900 nm range. There are also two strong Cr lines at 426
Figure 2.2.5 Flame-development time extracted from schlieren imaging of stoichiometric
methane-air ignition for varying frequency at fixed total energy deposition.
0
1
2
3
4
5
6
7
0.1 1 10 100
Fla
me
-De
ve
lop
me
nt
Tim
e (
ms
)
Pulse Frequency (kHz)
MSD (3 pulses, 17 mJ)
NRP discharge (5 pulses,16 mJ)
NRP discharge (10 pulses, 32 mJ)
NRP discharge (100 pulses, 320 mJ)
76
Figure 2.2.6 a) Plasma emission intensity in air for three different pulse frequencies at 22
kV pulse amplitude (3.2 mJ/pulse). b) Plasma emission intensity for three different pulse
amplitudes at 40 kHz pulse repetition frequency.
0
20
40
60
80
100
120
140
160
180
200
220
240
0 0.05 0.1 0.15 0.2 0.25
Inte
ns
ity (
arb
. u
nit
s)
Time (ms)
5 kHz
10 kHz
40 kHz
a.
0
20
40
60
80
100
120
140
160
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Inte
ns
ity (
arb
. u
nit
s)
Time (ms)
9 kV
15 kV
22 kV
b.
77
nm and 521 nm from the stainless steel ground electrode and below 300 nm there are strong lines
from other metal ions, including Fe, Cu, and Ni (not shown). The emission lines are in good
agreement with the OES measurements of Pancheshnyi et al (Pancheshnyi et al. 2006) in air
under similar conditions. Therefore, the discharge is a source of radicals, ions, and excited
species, and, for sufficiently high pulse frequency and peak amplitude, this source can
continuously supply active species to the ignition kernel. The emission images explain why the
high voltage cases (15 kV and 22 kV) are more effective than the low voltage case (9 kV), and
also why increasing frequencies continually decrease the flame-development time. The
Figure 2.2.7 Emission spectrum in air from a burst of 50 pulses at 22 kV pulse amplitude
and 40 kHz repetition frequency.
O+
(2P
o-2
P)
O+
(4F
-4D
o)
Cr
O+
(4P
-4D
o)
N+
(3P
o-
3P
) N+
(3D
-3F
o)
Cr
N+
(3P
o-
3D
)
N+
(3P
-3D
o)
O (
5P
-5D
o)
N+
(1P
o-
1P
)N
+ (
3P
o-
3P
)
N (
4P
-4S
o)
O (
5S
o-
5P
)
N (
4P
-4P
o)
O (
3S
o-
3P
)
N(4
P -
4D
o)
0
0.2
0.4
0.6
0.8
1
300 350 400 450 500 550 600 650 700 750 800 850
Inte
ns
ity (
arb
. u
nit
s)
Wavelength (nm)
N2 2nd Positive
78
maintenance of a constant radical pool in the center of the ignition kernel is the critical factor in
decreasing the flame-development time for multiple pulse ignition devices.
To fully connect the observation of a constant radical pool formed by the plasma to the
decrease in flame-development time, the concept of critical radius can be utilized. The process of
kernel development from an ignition energy source to a self-propagating flame has been
investigated both theoretically (Chen et al. 2007, Chen 2009, Chen et al. 2011) and
experimentally (Maly 1984, Chen 2009, Kim et al. 2013) for laminar flames. The development
process can be divided into three regimes: spark assisted ignition kernel propagation (Regime I),
unsteady transition from spark ignition to laminar flame propagation (Regime II), and laminar
flame propagation (Regime III). The first two regimes are sensitive to the initial energy addition,
while the final regime is only a function of fuel chemistry. Therefore, the effects of plasma at
most extend to Regimes I and II, beyond which energy addition in the kernel cannot effect the
steady-state flame propagation. This chapter focuses on both Regimes I and II, which are
difficult to isolate in turbulent flow conditions. Chapters 5 and 6 will focus on the chemistry
occurring in Regime I. Increased energy addition can decrease the time needed to get to the
critical radius, but cannot reduce the magnitude of the critical radius, unless the initial ignition
volume becomes comparable to the critical radius. At a given initial energy, the time to the
critical radius is controlled solely by the induction chemistry of the mixture, thus, to understand
the effect of heat and radical addition on flame-development time, we must understand their
effect on the induction chemistry. The emission spectrum showed the presence of excited N2, N,
O, N+, and O
+. After the discharge, these species will react with the surrounding gas and be
partially quenched before the next pulse arrives. To help understand this reaction process,
particularly for the case when fuel is present, numerical modeling of the plasma and subsequent
79
combustion processes for stoichiometric methane-air was conducted. The model predicts that the
main radicals produced in the plasma discharge are O atoms resulting from dissociation of O2 by
electron collision reactions and by quenching reactions with excited N2, which is in agreement
with previous NRP discharge studies (Popov 2011, Rusterholtz et al. 2013).
The dissociation of O2 is the dominant radical production channel, as opposed to the
dissociation of N2, since O2 has a bond dissociation energy (BDE) of 119 kcal/mole as compared
to the 226 kcal/mole BDE of N2. The production pathways for O atom formation in and after the
plasma are presented in Fig. 2.2.8 for a single pulse of 320 Td peak E/N, initial temperature of
850 K, and pressure of 1 atm. The main production reactions of O atoms are:
e + O2 → e + O + O (R1)
N2 (A,B,a’,C) + O2 → N2 + O + O (R2-R5)
N + O2 → NO + O (R6)
H + O2 → O + OH (R7)
HO2 + H → O + H2O (R8)
The consumption reactions are:
CH4 + O → CH3 +OH (R9)
CH3 + O → CH2O + H (R10)
CH2O + O → HCO + OH (R11)
HO2 + O → OH + O2 (R12)
During the plasma, the largest formation route is through electron impact dissociative excitation
of O2. The next four major routes are the quenching reactions of the electronically excited states
80
of N2. After the discharge pulse, quenching of the excited N2(A,B,a’,C) states becomes the
dominant pathway until these states are depopulated, which is in accordance with previous
experimental measurements (Stancu et al. 2010, Rusterholtz et al. 2013). A diagram of the
energy levels for N2 is presented in Chapter 4. N atoms also produced in the discharge by
electron collision reactions or through the Zeldovich mechanism (i.e. O + N2 → NO + N) will
react with O2 in R6 until the N atoms are consumed. At this point the typical combustion
reactions R7 and R8 take over as the dominant radical production processes. The H atoms
participating in R7 and R8 are produced in the discharge by electron collision reactions with
methane and from combustion reactions such as R10 that are typical of high temperature
oxidation. For all of the conditions simulated in this work, reactions R1-R5 were responsible for
at least 70% of O atom production, while R7 and R8 reactions never accounted for more than
20%, even when the mixture was close to ignition. Consumption of O atoms is through reaction
with the fuel and fuel intermediates, with a negligible amount recombining into O2. Based on the
model results, the main effect of the plasma is to produce active species, which quench to
produce O atoms and release heat. The O atoms go on to initiate the fuel oxidation process,
which results in chain propagating and branching reactions which maintain the population of O
atoms and other radicals until the next pulse arrives. We will now numerically investigate how
the plasma produced O atoms and heat release contributes to the observed inter-pulse coupling
effects.
First we explore the effect of varying pulse repetition frequency. In Fig. 2.2.9a, pulses each
producing 150 ppm of O atoms are applied at three different pulse repetition frequencies. Similar
81
to the emission intensity measurements from Fig. 2.2.6a, increased pulse frequencies result in
higher concentrations of O atoms in the inter-pulse region, and also an accumulation of O atoms
in the 40 kHz case. Joule heating by the plasma, as well as quenching of active species and
extraction of chemical enthalpy from the fuel, incrementally increases the temperature of the
mixture after each pulse. This in turn accelerates the rate of fuel oxidation by O atoms (R9-11),
which results in the activation of chain branching reactions, particularly R7. Therefore, a coupled
cycle of heat and radical production in the plasma, subsequent fuel oxidation, further generation
of radicals in the combustion process, and then another pulse to restart the cycle at an increased
temperature is set up in the repetitively pulsed system. This numerical experiment demonstrates
Figure 2.2.8 O atom production rates for a single pulse of 100 ppm of
O atom generation in a stoichiometric methane-air mixture at 850 K
initial temperature and 1 atm pressure.
Time (ms)
Ato
mic
Op
rod
uc
tio
nra
tes
(mo
lc
m-3s
-1)
0 0.005 0.01 0.015 0.02 0.02510
-7
10-6
10-5
10-4
10-3
10-2
10-1
100
101
8
5
6 7
1, e+O2=>e+O+O
2, N2(B)+O
2=>N
2+O+O
3, N2(C)+O
2=>N
2+O+O
4, N2(a')+O
2=>N
2+O+O
5, N2(A)+O
2=>N
2+O+O
6, N+O2=>NO+O
7, H+O2=>O+OH
8, HO2+H=>O+H
2O
0 2E-06 4E-0610
-7
10-6
10-5
10-4
10-3
10-2
10-1
100
101
1, 2, 3, 4, 5
82
that the timescale of the radical consumption reactions is long enough for repetitive pulses to
build up and sustain a radical pool, provided the temperature is sufficiently high so that chain
branching chemistry can occur.
Figure 2.2.9b presents the effect of different degrees of O atom generation (Case I: 50 ppm,
Case II: 100 ppm, Case II: 150 ppm) at a fixed pulse frequency of 40 kHz. Similar to Fig. 2.2.6b,
the O atom concentration and temperature in Cases II and III builds up throughout the pulse
burst, while there is negligible build up of O atoms for Case I since the consumption reactions
are fast enough to return the O atom concentration back to the original value, and the
temperature is not high enough for chain branching combustion reactions to increase the radical
concentration. Higher pulse energy allows for greater increases in temperature, driving faster fuel
oxidation reactions, more radical generation in the combustion process, and thus stronger pulse-
to-pulse coupling. In the actual spark discharge, the radicals and heat produced in the narrow
discharge region will quickly diffuse outwards from the discharge center, setting up a radical
gradient on the border of the discharge region where reaction with the unburned fuel-air mixture
begins to occur. The larger the radical pool in the discharge center, the greater the radical flux to
the unburned mixture, and thus the faster the flame-development process will proceed. The
flame-development time is then sensitive to how quickly a flame can be formed in this
radical/temperature gradient, which is a function of the induction chemistry of the particular fuel.
Similar numerical predictions performed by Do et al. (Do et al. 2010a) modeled hydrogen-air
ignition using NRP discharges, and also found that the temperature and radical pool build up is a
function of pulse frequency and pulse energy. It was found that high frequency and low pulse
83
Figure 2.2.9 Computed atomic oxygen concentration and
temperature as a function of time for stoichiometric methane-air
mixtures at 850 K initial temperature and 1 atm pressure with a)
150 ppm O atom production repeated at 5 kHz, 10 kHz, and 40
kHz frequencies and b) with 50 ppm (Case I), 100 ppm (Case II),
and 150 ppm (Case III) O atom production repeated at 40 kHz
frequency.
0
200
400
600
800
1000
1200
1400
1E-2
1E+0
1E+2
1E+4
1E+6
0 0.05 0.1 0.15 0.2 0.25
Te
mp
era
ture
(K
)
Mo
le F
rac
tio
n (
pp
m)
Time (ms)
O Atom, 5 kHz O Atom, 10 kHz O Atom, 40 kHz
T, 5 kHz T, 10 kHz T, 40 kHz
a.
0
200
400
600
800
1000
1200
1400
1E-2
1E+0
1E+2
1E+4
1E+6
0 0.05 0.1 0.15 0.2 0.25
Te
mp
era
ture
(K
)
Mo
le F
rac
tio
n (
pp
m)
Time (ms)
O Atom, Case I O Atom, Case II O Atom, Case IIIT, Case I T, Case II T, Case III
84
energy discharges were more effective than low frequency and high pulse energy discharges
(using the same overall power deposition) only if the radical consumption rates were faster than
production rates, which was true up to 800 K. In the case of methane-air, radical consumption is
faster than production even up to 1300 K, as shown in Fig. 2.2.9, due to the significantly slower
rates of chain branching reactions in methane compared to hydrogen. Following this result,
investigation of the differences in induction chemistry between methane and ethylene will lend
insight as to why there is a significant reduction in flame-development time for stoichiometric
methane-air, but not for ethylene-air.
Pancheshnyi et al. (Pancheshnyi et al. 2006) observed that repetitive pulses only had a strong
effect on ignition time for mixtures with MIE larger than the individual pulse energy.
Stoichiometric ethylene-air has an MIE of 0.096 mJ, which is a factor of five smaller than the
0.49 mJ MIE of stoichiometric methane-air (Fenn 1951). The MIE increases with flow velocity
and turbulent intensity (Ballal et al. 1975, Peng et al. 2013), thus, it is reasonable to consider
methane-air mixtures in the present experiments to have a comparable MIE to the individual
pulse energy of the plasma, especially for the 0.8 mJ/pulse condition, while the ethylene-air MIE
is less than the individual pulse energy. As mentioned earlier, the critical radius is a useful
parameter for describing the ignition kernel to self-propagating flame dynamics. Calculations to
find the critical flame radius for stoichiometric methane-air and ethylene-air mixtures were
performed using the Adaptive Simulation of Unsteady Reacting Flow (A-SURF) (Chen 2009)
code, again using USC-Mech II (Wang et al.) for the combustion model. The critical radii were
found to be 0.573 cm and 0.462 cm for methane-air and ethylene-air, respectively. Indeed, MIE
can be correlated to the cube of the critical radius (Chen et al. 2011). Although the computed
critical radii for both mixtures are comparable, MIE for ethylene is estimated to be less than 52%
85
of that methane, which is qualitatively consistent with the measured MIE discussed above.
However, it is also important to consider the time to reach the critical radius for both mixtures,
which is very sensitive to the induction chemistry of the mixture. The methane-air mixture took
2.34 ms to reach to critical radius, while the ethylene-air mixture took only 0.586 ms, both using
the same initial hot spot temperature of 1600 K and radius of 2 mm. This result indicates that the
induction chemistry time scale is significantly different for the two fuels.
Once again extending the analogy of ignition delays in homogenous mixtures, Fig. 2.2.10
shows the ignition delay time for one pulse and five pulses of plasma, each generating 100 ppm
Figure 2.2.10 Computed ignition delay time as a function of initial
temperature for stoichiometric methane-air and ethylene-air mixtures
with 1 pulse and 5 pulses of 100 ppm O atom production repeated at 40
kHz and at 1 atm pressure.
1E-5
1E-4
1E-3
1E-2
1E-1
1E+0
1E+1
800 900 1000 1100 1200 1300
Ign
itio
n D
ela
y (
s)
T0 (K)
Methane-air with 1 pulseMethane-air with 5 pulsesEthylene-air with 1 pulseEthylene-air with 5 pulsesInter-pulse timeTime for 5 pulses
86
of O atoms repeated at 40 kHz, for both ethylene-air and methane-air stoichiometric mixtures.
The ignition delay for methane-air is two orders of magnitude longer than for ethylene-air with
one pulse, and is as much as three orders of magnitude longer for five pulses. For both fuels,
increasing the pulse number from one to five results in a decrease in the ignition delay; however,
the absolute time reduction for methane-air is two orders of magnitude greater than for ethylene-
air. Because the induction chemistry of ethylene is so fast with just a single pulse, the effect of
additional pulses is not noticeable on the time scale of the present experiments. In addition, note
that the ignition delay of ethylene air for 1 pulse and 5 pulses converge at temperatures above
1000 K. This is due to the ignition delay time approaching the inter-pulse time, resulting in
discharge pulses being deposited after ignition has commenced in the 5 pulse case. Do et al. (Do
et al. 2010a) found a similar convergence between autoignition and plasma-assisted ignition for
hydrogen-air at temperatures above 1200 K. This illustrates the competition between heat and
radical generation from chemical chain branching reactions or from the plasma discharge. If the
induction chemistry produces heat and radicals faster than the plasma discharge, no reduction in
the ignition delay can be observed by depositing additional energy. From the standpoint of
ignition enhancement, it is far more beneficial to apply multiple discharge pulses to a mixture
with a long ignition delay time, i.e. large MIE, as compared to a fuel which is easily ignited and
thus can only be marginally improved.
2.2.3.2 PDE testing
The flow visualization experiments demonstrated that the flame-development time could be
reduced using NRP discharges as an ignition source. This section demonstrates the usefulness of
this result in a pulsed detonation engine. The thrust output of a PDE is dependent on the
frequency of detonation tube usage, which is limited by how quickly each tube can be ignited,
87
transition to a detonation, and exhaust the burned gas to produce thrust. Within this cycle, the
ignition process typically consumes the longest period of time (5 – 20 ms compared to 2 ms for
DDT (Busby et al. 2007)), and is thus the limiting process determining the firing frequency of
each tube (Busby et al. 2007). For a PDE, there are two basic methods to successfully produce a
detonation wave: direct initiation or indirect initiation. Direct initiation requires an ignition
source with enough energy to produce a strong shock wave that directly produces a coupled
auto-ignition wave. The amount of energy required for direct initiation is beyond the capability
of practical ignition systems. Indirect ignition occurs when a turbulent deflagration wave is
formed in a confined volume such that sufficient heat release is produced to thermally choke the
flow, which results in subsequent pressure wave and ignition coupling and leads to detonation.
The detonation velocity is solely a property of the gas mixture, so the only opportunity to
decrease the duration of the firing phase of the PDE cycle is by decreasing the ignition time and
DDT time. Therefore, PDEs are an ideal platform to test new ignition methods in an application
in which the flame-development time is the limiting factor.
The effects of pulse frequency, pulse amplitude, and pulse number on PDE ignition time were
studied by varying each of these parameters independently for two different fuels over a wide
range of equivalence ratios. Each condition was run for five ignition events, and the plotted data
is the average of these data points. The discharge characteristics used for testing are listed in
Table 2.2.1. For all of the conditions tested, the ignition method had no effect on the DDT
distance or time (measured independent of the ignition time). Therefore, all of the results
presented below focus on the ignition time (analogous to the flame-development time in the flow
visualization study) as measured by the pressure rise at the engine head.
88
Figure 2.2.11 Ignition time in a PDE as a function of equivalence ratio
for MSD ignition and NRP discharges at 22 kV peak amplitude, 40 kHz
repetition frequency. a) Ethylene-air mixtures, b) Avgas-air mixtures.
0
5
10
15
0.5 0.7 0.9 1.1 1.3 1.5 1.7
Ign
itio
n T
ime
(m
s)
Equivalence Ratio
MSD (3 pulses, 17 mJ)
NRP Discharge (2 pulses, 6.4 mJ)
NRP Discharge (5 pulses, 16 mJ)
NRP Discharge (10 pulses, 32 mJ)
a.
7
9
11
13
15
0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Ign
itio
n T
ime
(m
s)
Equivalence Ratio
MSD (3 pulses, 17 mJ)
NRP Discharge (2 pulses, 6.4 mJ)
NRP Discharge (5 pulses, 16 mJ)
NRP Discharge (10 pulses, 32 mJ)
NRP Discharge (15 pulses, 48 mJ)
NRP Discharge (20 pulses, 64 mJ)
b.
89
To quantify the maximum reduction in ignition time for varying equivalence ratios, the NRP
power supply was held constant at its peak pulse amplitude (22 kV) and peak repetition
frequency (40 kHz), while the number of pulses was varied from 2-20 for avgas-air and 2-10 for
ethylene-air mixtures for a range of equivalence ratios. The ignition times as a function of
equivalence ratio for ethylene-air and avgas-air mixtures are plotted in Fig. 2.2.11. For ethylene-
air mixtures, there are negligible differences in ignition time between the NRP discharge and
MSD, exhibiting only slightly shorter ignition times using the NRP discharge for very lean
mixtures (φ ≤ 0.8). However, for avgas-air, the ignition time is found to be significantly reduced
by the NRP discharge when more than two pulses are applied. This pronounced reduction of
ignition time for avgas-air over ethylene-air mixtures can be explained by the significant
difference of MIE between the two tested fuels, as discussed in the previous section. By
assuming iso-octane as an avgas surrogate, the MIE of a stoichiometric mixture is estimated to
be 1.35 mJ (Fenn 1951). This is more than a factor of two greater than stoichiometric methane-
air, and more than an order of magnitude greater than stoichiometric ethylene-air. Consequently,
the ignition time of stoichiometric avgas-air can be reduced by up to 25% using NRP ignition as
compared to MSD ignition. Near the lean and rich limits, the MIE increases dramatically (Yetter
et al. 2008), so, the 10-20 pulse NRP discharge most significantly reduces ignition time in these
regions. For avgas-air, the NRP discharge could also extend the lean limit to an equivalence ratio
of 0.9 and the rich limit to 1.4, representing a 0.1 improvement in equivalence ratio at both
extremes.
To explore the effects of specific discharge characteristics, the equivalence ratio for both
ethylene-air and avgas-air mixtures were held constant while the plasma parameters were varied.
For ethylene-air, φ=0.8 was selected, while, for avgas-air, φ=1.0 was selected, for these
90
equivalence ratios represent cases in which the NRP discharge produced significantly shorter
ignition times compared to the MSD discharge, while both systems regularly achieved successful
ignition and DDT. The plasma frequency, pulse amplitude, and number of pulses were varied
individually. The results of these tests are plotted in Fig. 2.2.12. Similar to the flow visualization
experiments, two trends emerged in the data. The first is the decrease in ignition time with
increase in total energy, as illustrated by the best fit trend line to the 10-40 kHz data. The number
of pulses and pulse amplitude, when varied independently, had similar effects on ignition time,
provided the pulses were applied within a time scale much shorter than the overall ignition time.
The second trend observed is the decrease of ignition time with increased pulse frequency, even
at fixed total energy deposition. This trend is significantly more apparent in the avgas-air data, in
which all of the ignition times using less than 10 kHz pulse repetition frequency are above the
average of the 10-40 kHz data. Therefore, the same results from the flow visualization study are
attainable in PDE engine conditions.
2.2.4 Conclusions
The effect of nanosecond repetitively pulsed plasma discharges on ignition has been studied in
both flame-development visualization and PDE testing platforms. Clear decreases in the
observed flame-development time and PDE ignition time were demonstrated for the NRP
discharge as compared to conventional spark ignition for a variety of fuels, equivalence ratios,
and discharge parameters. In addition, both leaner and richer ignition could be achieved for
avgas-air mixtures in the PDE experiments as compared to MSD ignition. The experimental
observations are: 1) reduction in flame-development time (or ignition time in the PDE) using
NRP discharges was observed only for methane-air and avgas-air mixtures, indicating that a
91
Figure 2.2.12 Ignition time in PDE engine as a function of total energy
deposition. a) Ethylene/air, φ=0.8. b) Avgas-air , φ=1.0.
5
6
7
8
9
1 10 100
Ign
itio
n T
ime
(m
s)
Total Energy (mJ)
MSD
NRP discharge, f = 1-5 kHz
NRP discharge, f =10-40 kHz
a.
8
10
12
14
16
18
1 10 100
Ign
itio
n T
ime
(m
s)
Total Energy (mJ)
MSD
NRP discharge, f = 1-5 kHz
NRP discharge, f =10-40 kHz
b.
92
significant impact can be achieved for fuels with slow induction chemistry; 2) flame-
development time is decreased as the total ignition energy is increased beyond the mixture’s
MIE, either by higher pulse amplitudes or by increased pulse number; 3) high frequency pulses
(in this case 10-40 kHz) allow the opportunity for multiple pulses to be coupled into an enlarged
single flame kernel, resulting in a shorter flame-development time compared to the formation of
individual non-interacting flame kernels at lower frequencies; and 4) high frequency (> 10 kHz)
and high pulse amplitude (> 9 kV) discharges in air generate excited and ionized species in the
inter-electrode region, the reaction timescales of which are longer than the inter-pulse time,
which results in a plasma kernel sustained at a larger volume and for a longer time than lower
amplitude or frequency discharges. Therefore, inter-pulse coupling between subsequent plasma
pulses is an important mechanism to enhance ignition due to an increase in the kernel radical
pool and temperature and a reduction in the induction chemistry timescale.
Numerical modeling of homogenous fuel-air mixtures revealed O atoms as the active species
with the highest concentration generated by the plasma discharge, which are mainly generated by
electron collision reactions with O2 and quenching of N2(A,B,a’,C) by O2. The main
consumption pathway for O atoms is via reaction with the fuel and its primary intermediates. In
addition, heat is generated in and after each discharge pulse, increasing the rate of O atom
reaction with the fuel and chain branching reactions in the combustion process. Therefrore, a
positive feedback loop is established by the NRP discharge, consisting of: 1) heat and active
species production during the plasma discharge; 2) production of O atoms via quenching
reactions just after the discharge; 3) initiation of fuel oxidation by O atom reactions with the fuel
and its primary intermediates; 4) radical chain branching reactions sustaining or building the
radical pool during the inter-pulse time; and 5) further building of the radical pool and gas
93
temperature by subsequent discharge pulses. This positive feedback loop was particularly
effective for high pulse frequencies and pulse amplitudes, in which the minimum O atom
concentration could be continually increased during the inter-pulse time for every subsequent
discharge pulse. This inter-pulse coupling resulted in more than an order of magnitude decrease
in the computed ignition delay going from one pulse to five pulses at 40 kHz repetition
frequency and 100 ppm O atom generation per pulse.
Additionally, it was shown that the induction chemistry of fuels with high MIEs, which
correspond to large critical radii and long induction chemistry timescales, are much more
sensitive to increased energy deposition than are fuels with smaller MIEs. Accordingly, methane-
air mixtures exhibited significantly greater decreases in the flame-development time as compared
to ethylene-air mixtures subjected to the same degree of plasma energy deposition during the
flame-development visualization experiments. This result was also observed in the PDE testing
experiments, in which NRP discharge ignition improved ignition time for stoichiometric avgas-
air mixtures by 25%, and increased the equivalence ratio range by 50%, as compared to MSD
ignition. However, there was only a slight improvement in the ignition time of ethylene-air
mixtures, and no expansion of the equivalence ratio operating limits. Based on the observations
in these experiments, further improvement in flame-development times can be achieved by using
NRP plasma discharges with greater pulse repetition frequency, which would increase the
buildup of excited species in the flame kernel and allow more pulses to be applied before the
flame front leaves the inter-electrode region.
Although this study has demonstrated the usefulness of current kinetic modeling capabilities
in describing the phenomenological processed in NRP ignition, quantitative modeling of the
complex and 3-dimensional ignition kernel formation and development process remains out of
94
range. To truly reveal the reaction processes occurring in PAC, a much simpler experimental
platform must be utilized. Preferably, a quasi 0-dimensional system would provide the best
experimental facility for measurement of species and temperature, as well as for the testing of
kinetic models. In the following chapters, the development of an experimental and numerical
platform for just this purpose will be described, as well as evaluations of the kinetic processes
involved in ethylene and methane oxidation.
95
3 Low temperature dielectric barrier discharge reactor
integrated with mid-infrared TDLAS
3.1 Non-equilibrium plasma reactor
To isolate the kinetic effects of PAC from transport and thermal effects, a plane to plane
double dielectric barrier discharge has been adopted to produce a uniform plasma discharge at
low pressure (60 Torr). Details of the experimental setup have been published for several
variations of the same reactor (Uddi et al. 2013, Lefkowitz et al. 2014, Lefkowitz et al. 2015c,
Lefkowitz et al. 2015b, Lefkowitz et al. 2015d). The reactor used in Chapters 5 & 6 differs
slightly in the applied voltage waveforms and the configuration of the TDLAS measurement
system, but is otherwise identical.
3.1.1 Reactor configuration
Figure 3.1.1 provides a schematic, and Figures 3.1.2 and 3.1.3 provides images of the
experimental setup. The reactor is constructed primarily of quartz and Macor, with a stainless
steel inlet gas flow manifold and stainless steel brackets to hold the walls in place. One sidewall
is quartz, which allowed observation into the cell for laser alignment, while the other is Macor
ceramic, used for easy machining. Each of the two 45 mm x 45 mm stainless steel electrodes are
sandwiched between a quartz plate and a Macor plate that make up the top and bottom of the
reactor, forming a plane-to-plane double DBD with quartz as the dielectric material and barrier
thickness of 1.6 mm. The inner dimensions of the rectangular quartz reactor section are 14 mm in
height, 45 mm in width, and 152 mm in length. The quartz cell is contained inside a stainless
steel vacuum chamber of approximate dimensions 152 x 102 x 305 mm. There is a wedge
96
shaped 25.4 mm OD calcium fluoride window in the sidewall of the vacuum chamber to allow
the laser beam to pass through the chamber wall and into the Herriott cell for absorption
measurements. There is another UV fused silica window in the rear of the vacuum chamber, as
shown in Fig 3.1.2, to observe the plasma in the direction opposite to the incoming flow, and
which was also used as an observation port to collect short gated intensified images to ascertain
plasma uniformity as well as spectroscopic measurements.
The fuel, oxygen, and diluent gases are metered using mass flow controllers (Brooks
instruments), and are premixed prior to entering the plasma reactor. The manifold delivers the
fuel/oxidizer/diluents mixture, and uses stainless steel ball bearings followed by a ceramic
honeycomb to ensure the flow is well distributed and flowing in the axial direction. A nitrogen
coflow is supplied around the quartz section inside the vacuum chamber to ensure exhaust gases
do not affect the absorption measurements. The flow velocity inside the quartz section is
maintained at 0.4 m/s in Chapter 5, and 0.2 m/s in Chapter 6. The different velocities were
selected due to differences in the discharge configuration, discussed in sub-section 3.1.2. All
experiments were conducted at a total pressure of 60 Torr and initial temperature of 296 K.
97
Figure 3.1.1 Schematic of experimental setup for DBD flow reactor with TDLAS
measurement system. In Chapter 5, the DFB-QCL laser is not included.
Ge Etalon
Reactor
Collimating
Lenses
Mirror
Flip Mirror
Quartz
Wall
Macor
Wall
Vacuum
Chamber
Beam Splitter
Electrode
98
Figure 3.1.2 Image of experimental setup for DBD flow reactor with TDLAS
measurement system.
Vacuum
chamber
Detector
Electrode
connectionQCLN2 purge boxN2 purge tube
Flow to vacuumObservation
window
Pressure gauge
99
Figure 3.1.3 Image of DBD plasma reactor a) side view b) top view
3.1.2 Discharge characteristics
The plasma discharge is produced using a nanosecond pulsed power supply (FID GmbH FPG
30-50MC4), which produces voltage waveforms with 12 ns FWHM, peak voltage of 32 kV, and
maximum pulse repetition frequency of 50 kHz. The discharge is operated in either continuous
mode or burst mode, and is triggered by a digital pulse generator (SRS DG535).
The current during the nanosecond discharge is measured using a Pearson Coil current probe
(Model 6585), and the voltage externally applied to the discharge is measured using a Lecroy
Laser coupling
hole
Adjustable Herriott
cell mirror
Electrode
connection
Fixed Herriott cell
mirror
a
Macor top plate
Macor side wall
Quartz side wall
Kinematic
mirror mount
b
100
high voltage probe (PPE20KV). Plots of the applied voltage and current measurements can be
found in Chapter 5 and 6. The applied electric field is computed using:
(3.1.1)
where Eapp is the applied electric field, Vapp is the measured peak applied voltage, L is the gap
distance (14 mm), l is the dielectric layer thickness (1.6mm), and ε is the dielectric constant (4.3
for quartz) (Adamovich et al. 2009b, Takashima et al. 2013). The average peak voltage of all
experiments is 8.7 ± 0.2 kV, which results in a peak applied electric field of 5.94 ± 0.2 kV/cm
and a reduced electric field (E/N) of ≈ 300 Td at the initial gas composition and temperature.
The deposited energy was determined by integrating the product of the measured voltage and
current over the entire pulse duration, resulting in a total energy input of to be 1.5 ± 0.5 mJ.
Uniform discharge without filamentary structures was confirmed using single shot ICCD
imaging (100 ns gate width), as presented in Fig 3.1.4 for various discharge frequencies in
mixtures of 0.0625 C2H4/0.1875 O2/0.75 Ar. The electrodes are at the top and bottom of the
images. The flow is outward perpendicular to the page. There is some charge concentration near
the edges, but it is clear that there are no intensely bright regions, which would indicate streamer
formation. This indicates that the product formation, while not perfectly uniform, is not spatially
concentrated in any single region. The relative intensity and relative variation in intensity
integrated over the transverse direction for single pulse ICCD images are plotted in Fig. 3.1.5
and 3.1.6, respectively. The emission intensity remains within 40% of the average intensity for
all conditions. Figure 3.1.7 shows a direct image of the discharge taken with a digital camera
with a 2 second exposure time and a repetition frequency of 1 kHz. The intensity variation from
this image is presented in Fig 3.1.8. Clearly, the average intensity is much more uniform than the
single pulse intensity. It is found that the deviation in intensity is within 15% for the averaged
101
image. Therefore, the non-equilibrium plasma does not have hot spots, and can be used in the
present measurements as representative of a quasi 0-dimensional plasma discharge.
Figure 3.1.4 Single shot ICCD images of 0.0625/0.1875/0.75 C2H4/O2/Ar plasma at
different frequencies. Peak voltage: 8.7 kV, gate time: 100 ns.
200 Hz
500 Hz
1000Hz
1500 Hz
2000 Hz
2500 Hz
3000 Hz
102
Figure 3.1.5 Intensity of emission of 0.0625/0.1875/0.75 C2H4/O2/Ar plasma at different
frequencies based on single shot ICCD images.
Figure 3.1.6 Variation of emission data from Fig. 3.1.5
Figure 3.1.7 Two second exposure of 0.0625/0.1875/0.75 C2H4/O2/Ar plasma at pulse
repletion frequency of 1000 Hz.
-40%
-20%
0%
20%
40%
-30 -20 -10 0 10 20 30
Devia
tio
n
Position, mm
200 Hz
1500 Hz
2000 Hz
103
Figure 3.1.8 Average emission of 2000 pulses at 1000 Hz repetition rate.
The plasma emission is primarily from excited Ar in the range of 300-400nm and 700-800nm.
Figure 3.1.9 presents the emission spectrum for pure Ar and C2H4/O2/Ar plasma recorded using
an Ocean Optics USB2000 spectrometer. When O2 and C2H4 are added to the mixture, the
emission intensity for excited Ar (denoted Ar(I) in Fig 3.1.9) and singly ionized Ar (denoted
Ar(II) in Fig 3.1.9) are decreased. This is due to the difference in the electron energy distribution,
branching ratio of electron collision reactions, and the lifetime of excited/ionized Ar states in the
presence of other collision partners. More in depth analysis of the emission of Ar will be
discussed in the Appendix, while the kinetic details will be discussed in Chapter 5.
104
Figure 3.1.9 Measured emission spectrum from Ar and 0.0625/0.1875/0.75 C2H4/O2/Ar
plasma at 1000 Hz repetition frequency, 8.7 kVpp.
3.1.3 Gas chromatography
To validate the in situ laser diagnostics and provide species information of other hydrocarbons
and oxygenated intermediate species, gas chromatograph with thermal conductivity detection
(GC-TCD, Inficon 3000) measurements were taken to identify and quantify product species
resulting from a continuously pulsed plasma discharge in the range of 100 – 5000 Hz for
Ar/C2H4 and Ar/O2/C2H4 plasmas (Lefkowitz et al. 2015d) and from 100 – 30,000 Hz for
He/O2/CH4 mixtures (Lefkowitz et al. 2015b). Species sampling was performed using a quartz
probe placed at the exit of the reactor section, which was fed into the rear of the vacuum
chamber by replacing the rear window with a 1/8 inch vacuum feed through. Sample gas was
105
drawn out of the reactor and through a heated line into a heated stainless steel sample vessel
using a vacuum pump. The sample vessel is fitted with a piston such that the pressure can be
increased to 1 atm, which is required for injection to the GC. The sensitivity for all species other
than H2 is 10 ppm, while H2 interferes with the large He peak from the GC carrier gas, limiting
the sensitivity to about 1000 ppm in Chapter 6 when CH4/O2/He mixtures are used. The
uncertainty for GC-TCD quantification is assumed to be ±5% for all species, not including
uncertainties from the sampling procedure. To ensure that the plasma products have reached
steady state, several downstream positions were attempted and no variation in the product
distribution was noticed as long as the probe position was downstream of the visible plasma
region.
106
3.2 Direct absorption spectroscopy
3.2.1 Introduction
This brief description of absorption spectroscopy introduces the theory and numerical
methods used to make the molecular spectroscopy measurements in Chapter 5 and 6. For more in
depth discussions, the reader is directed to a number of detailed references: (Herzberg 1950,
Miles 2010, Hanson 2011, Demtröder 2013, Hanson 2013).
Every atom or molecule will absorb electromagnetic radiation at particular, discrete
wavelengths corresponding to electronic, vibrational, and/or rotational energy level transitions.
Atoms only exhibit electronic transitions, while molecules also have vibrational and rotational
mode transitions. The total internal energy of a molecule can be described by:
( 3.2.1 )
where E is the energy contained in each internal energy mode: electronic, vibrational, and
rotational. When an atom or molecule undergoes a transition by absorbing or emitting a photon,
or by collisions with other species in the gas, the energy contained in one or more of these modes
changes. In the case of photon absorption or emission, the nature and degree of that change can
be determined by the wavelength of light that is absorbed or emitted.
For many small molecules, extensive research has been conducted to determine the positions
of absorption/emission lines, and have been tabulated in the HITRAN database (Rothman et al.
2009). The separation between two different energy states can be defined using Planck’s Law:
Δ
( 3.2.2 )
where E is energy in Joules, h = 6.626×10−34
J s is Planck’s constant, and ν [s-1
] is the frequency
of light being absorbed. The frequency and wavelength of light are related by:
( 3.2.3 )
107
where c is the speed of light in a vacuum, c = 3×1010 cm/s, and λ is the wavelength of light in cm.
Another unit, known as the wavenumber, is often used in spectroscopy and defined as:
( 3.2.4 )
the units of which are usually given in cm-1
. So Planck’s Law can be rewritten as:
Δ ( 3.2.5 )
Therefore, by knowing at what frequency light is absorbed by an atom or molecule, one can find
how much energy is involved in a particular transfer between two energy states.
3.2.2 Spectra of atoms and molecules
In general, electronic transitions occur in the ultraviolet and visible frequency ranges.
Vibrational transitions contain a factor of 100 lower energy, and are in the near- and mid-
infrared ranges, and rotational transitions are at least another order of magnitude smaller, and
occur in the microwave range. A general diagram of potential wells, which depict the energy
levels of a molecule as a function of the internuclear distance, is given in Figure 3.2.1. In the
ground state, a molecule will exist at the bottom of the potential well. Energy must be provided,
either by collisions with other species or by absorption of photons, to raise the molecule out of
the ground state into a higher electronic, vibrational, or rotational state. In addition, the molecule
can reach antibonding states, which have no minimum and thus the two nuclei cannot form a
stable bond. The energy difference between the location at the bottom of a potential energy curve
and the level at which the internuclear distance can go to infinity (i.e. the top of the “potential
well”) is the dissociation energy.
108
Figure 3.2.1. Energy diagram of a molecule showing bonding and antibonding states.
The different electronically excited states for atoms and molecules correspond to different
electronic levels of the valence electrons of the molecule, which are characterized by the
probability of finding the electrons at specific distances from the parent nuclei. Since bonds are
created by interactions of the Coulomb forces between the nuclei and electrons, different
electronically excited states alter the shape of the molecule, so each electronic state has different
vibrational and rotational energy levels.
The vibrational states of a molecule correspond to internuclear motion, which can be
characterized by stretching or bending motions. In order for a transition to take place, the dipole
moment of the molecule must be moved. The dipole moment is classically described as the
product of the magnitude of charges and the separation between them. In a molecule, it is
described by:
Internuclear Separation
En
erg
y
Electronic
Levels
bonding
antibonding
Vibrational
LevelsRotational
Levels
Dissociation
Energy
109
( 3.2.6 )
where is the dipole moment, qi is the charge of particle i, and is the position vector of the
particle defined from the center of mass. In a diatomic molecule there is a center of charge (and
mass) somewhere along the internuclear axis, the magnitude of which is determined by the size
of each nucleus (which determines the charge of each nucleus). In a diatomic molecule with two
identical atoms (such as O2 and N2) vibrational motion does not cause the dipole moment to
move since the center of charge remains fixed due to molecular symmetry. The same is true of
rotational transitions, since the center of charge is not changing in space as the molecule rotates.
Only heteronuclear diatomic molecules exhibit rotational and vibrational transitions, since they
possess a permanent dipole moment due to the unequal charges of the two nuclei. For polyatomic
molecules, there is always a vibrational spectrum since not only stretching but bending modes
can exist. However, similar to diatomic molecules, some types of polyatomic molecules do not
exhibit pure rotational spectra (i.e. linear symmetric molecules).
There exist a number of rules, which result from the solution of the Schrodinger equation in
the presence of an external field, that determine the allowed transitions between energy states.
Different sets of rules exists for specific classes of molecules, and determines the structure and
position of the absorption bands. Detailed discussions of the spectra for different molecules can
be found in the classic books by Herzberg (Herzberg 1950). In general, the structure of ro-
vibrational spectra is separated into groups of transitions (or “lines”) corresponding to specific
vibrational transitions. These groups are called bands. Each band has many individual lines, each
one corresponding to a transition between specific rotational states within the upper and lower
vibrational state. Figure 3.2.2 presents the absorption spectra of a few molecules relevant to
combustion studies (Rothman et al. 2009).
110
Figure 3.2.2. Spectra of combustion-related species in the near to mid-IR
As can be seen in Fig 3.2.2, many ro-vibrational absorption bands exist in the 1000 – 8000
cm-1
range (1 - 10 µm, 3×1013
– 3×1014
Hz). Only by studying the spectra of the molecule of
interest and the system in which a measurement is to be made can one properly choose the
absorption feature that will provide an interference-free measurement. In addition, different
regions of the spectrum have different line strengths (this concept will be discussed later). For
diatomic molecules there is only one very strong band (i.e, NO, CO, OH in Fig 3.2.2). This is
called the “fundamental” band, and is for transitions in which the vibrational quantum number
“v” is changed by 1. At twice the wavenumber of the strong band will appear a weaker band,
which represents transitions in which the vibrational quantum number is changed by ±2. This is
called the “first overtone”. There will be progressively weaker bands at every multiple of the
fundamental, which are also called overtone bands. Polyatomic species have more complex
spectra involving multiple vibrational modes so multiple fundamental bands. Any combination
of these bands and the overtones thereof will be part of the spectrum. For molecules such as
H2O, this means there will be more absorption features than for a diatomic molecule.
1.E-22
1.E-21
1.E-20
1.E-19
1.E-18
1.E-17L
ine
Str
en
gth
(cm
/mo
lec
ule
)H2O
CO2
CO
NO
1.E-22
1.E-21
1.E-20
1.E-19
1.E-18
1.E-17
1000 2000 3000 4000 5000 6000 7000 8000
Lin
e S
tre
ng
th
(cm
/mo
lec
ule
)
Wavenumber (cm-1)
CH2O
CH4
C2H2
C2H4
OH
111
Figure 3.2.3. Absorption features in the fundamental CO band.
Figure 3.2.3 presents the absorption spectra of CO for the fundamental stretch mode
(Rothman et al. 2009). Note that the band is split into a right side and left side, called “branches”.
There is a nomenclature for the sides of the bands: the right side (higher energy) is the “R
branch”, the left side (lower energy) is the “P branch”. This is due to a selection rule for this
molecule which states that a molecular transition must involve a ±1 change in the rotational
quantum number “J”. Each line in the fundamental band represents a Δv = ±1 vibrational number
change and a ΔJ = ±1 rotational number change. If the ΔJ=-1, it is in the P-branch, and ΔJ=+1 is
in the R-branch. Each line in a particular vibrational band (i.e, v = 0 →1) represents a transition
from some rotational state J in the v = 0 state to a J value on the v = 1 state. In terms of energy,
0
2000 2050 2100 2150 2200 2250 2300
Wavenumber (cm-1)
CO
RP
112
the rotational levels are not evenly distributed in each vibrational state, and actually go as J2.
Thus, a transition from J = 0→1 is close to the center of the band than a transition from J = 1→2,
and so on. Polyatomic molecules (and some diatomic molecules), also allow for ΔJ = 0
transitions. This results in another branch, called the “Q-branch” in the center of the other two
branches. Again, much more detail on this subject can be found in Herzberg’s book (Herzberg
1950).
The strength of a particular line is determined by the fraction of population in the lower state,
which can be determined by the Boltzmann distribution:
( 3.2.7 )
Where ni [cm-3
] is the number density of species in state i, gi is the degeneracy of state i (or
number of states with the same energy), [J] is the energy of state i, kB is the Boltzmann
constant (kB = 1.38×10−23
J/K), T is temperature, and Q is the partition function. The partition
function is defined as:
( 3.2.8 )
From this equation, it is clear that the line strengths are dependent on temperature, with
transitions from higher energy lower states becoming stronger as temperature is increased. Next,
we will have a look at the processes which can cause transitions when light interacts with matter.
3.2.3 Einstein theory of radiation
If one considers a system with an upper state “2”, with population density n2 and a lower state
“1” with population density n1, the processes that can bring a particle from one state to the other
can be simply described. When in the upper state, a particle can relax spontaneously
( 3.2.9 )
113
where A21 [Hz] is the Einstein coefficient for spontaneous emission. The particle can be excited
from state 1 to state 2 by absorbing a photon:
( 3.2.10 )
where B12 [cm3Hz] is the Einstein coefficient for stimulated absorption, ρ(ν) [J/cm
3Hz] is the
spectral energy density at frequency ν, and ν12 is the light frequency corresponding to the energy
difference between the states. Lastly, similar to stimulated absorption, there is stimulated
emission of a photon, described by:
( 3.2.11 )
where B21 [cm3Hz] is the Einstein coefficient for stimulated emission, which occurs when a
photon equal in energy to the difference between states 1 and 2 is absorbed by a particle in state
2, and as a result two photons of the same wavelength are released. The combination of these
three processes describes the interaction of light with any atom or molecule:
( 3.2.12 )
Figure 3.2.4. Light interactions with matter
For a system in equilibrium, the concept of detailed balance can be applied, in which
( 3.2.13 )
From here, relative concentrations of n1 and n2 can be related by using the Boltzmann equation
and Planck’s law.:
hν21
2
1
A21 B21ρ21 B12ρ21
E1, n1
E2, n2
114
( 3.2.14 )
The Planck blackbody distribution can then be used to get the equilibrium spectral density:
( 3.2.15 )
Therefore, at equilibrium,
( 3.2.16 )
plugging Eq. 3.2.14 into 3.2.16:
( 3.2.17 )
and therefore:
( 3.2.18 )
The above equation in equilibrium must hold true for any temperature, so:
( 3.2.19 )
and
( 3.2.20 )
therefore,
( 3.2.21 )
and
( 3.2.22 )
Finally, the spectral density for collimated light can be written as:
( 3.2.23 )
115
where np [cm-3
s-1
] is the number of photons per unit volume and at a particular frequency. We
can define the intensity of light at frequency ν as Iν
:
( 3.2.24 )
Therefore,
( 3.2.25 )
At this point, the interactions of light with matter have been covered in a simple fashion, and
have been able to relate spontaneous emission, stimulated emission, and stimulated absorption.
We will now see how these concepts can be used to quantitatively measure species
concentrations in a gas.
3.2.4 The Beer-Lambert Law (or Beer’s Law)
Absorption spectroscopy is generally performed with a laser that can be tuned in terms of its
wavelength. The light is transmitted through the sample and is focused onto a detector, as in
Figure 3.2.5.
Figure 3.2.5. Light transmission through a gas
The processes that occur must, as always, conserve energy. Therefore, we must account for the
loss (or gain) of intensity I (in units of (W/cm2)/cm
-1 or (W/cm
2)/Hz) at wavelength v. The
possible processes are absorption, reflection, scattering, and transmission, which must sum to
Iν Iν + dIνTunable Laser
Detector
Gas Sample
dx
116
unity. For our purposes, we will assume no light is reflected or scattered, which are reasonable
assumptions for a gas and relatively low intensity light. In addition, we will assume the spectral
emission from the sample is much lower in intensity than the transmitted light from the laser, so
we can ignore emission. We then have:
( 3.2.26 )
where α [no units] denotes the spectral absorbance and τ [no units] indicates the spectral
transmissivity at laser frequency ν. We can then define the absorbance as:
( 3.2.27 )
Or:
( 3.2.28 )
where kν [cm-1
] is the spectral absorption coefficient, which is a function of the particular
conditions of the absorbing species, gas properties, and absorption line characteristics. The
change in intensity (dIν) is a combination of emission and absorption of light, which cannot be
discerned in the measurement. However, if we assume that the initial intensity of the light is
much greater than the intensity of emission, we can say that the change in intensity is completely
due to absorption. To solve for Iν we can integrate this equation over the path length, called L
[cm]:
( 3.2.29 )
( 3.2.30 )
( 3.2.31 )
where Iν,0 is the initial intensity of light before entering the absorbing gas. Equation 3.2.31 is
known as the Beer-Lambert Law, or just Beer’s Law. The detector will measure a dip in intensity
117
when the light is scanned through a transition of some molecule in the gas. The initial intensity
of light can be measured in the same system when absorbers are not present, or it can be
approximated by a baseline prediction following the trend in the laser intensity before and after
the absorption feature.
The spectral absorption coefficient is the product of a number of parameters specific to the
molecule of interest, the transition of interest, and the experimental conditions. Another factor
that must be considered is the shape of the absorption line, which is not simply a square wave in
some small but finite region, but actually takes on a specific shape. We will use a generic
lineshape factor ϕ(ν) [s] to describe this shape. This factor is normalized by the integrated
absorption coefficient as follows:
( 3.2.32 )
Such that:
( 3.2.33 )
Figure 3.2.6. Lineshape function
We can now redefine the Einstein coefficients in a portion of frequency space ν → ν + dν by
multiplying each coefficient by the lineshape function at that wavelength. We will also redefine
our equation for spectral absorption and emission in a differential wavelength region.
ɸ(ν)
νν0
118
Figure 3.2.7. Absorption in the frequency rand ν → ν + dν
Next, we consider the possible processes which contribute to the change in intensity dIν.
There is induced emission, spontaneous emission, and induced absorption. We will assume again
that spontaneous emission is negligible.
( 3.2.34 )
Where we have used Eq. 3.2.25 to replace spectral density. The term nidx is the number of
molecules in state i per unit area; the terms B21ρ(ν) and B12ρ(ν)are the probabilities for emission
and absorption, respectively; and hν is the energy per photon. This can be rearranged to find the
absorption coefficient:
( 3.2.35 )
using Eqs. 3.2.14 and 3.2.22 we can simplify this to find:
( 3.2.36 )
if we integrate over the entire line, we can now find the integrated absorption, more commonly
known as the “line strength,” which is usually denoted S12 [cm-1
s-1
]:
Iνdν Iνdν + (dIν) d νTunable
LaserDetector
Gas Sample
dx
119
( 3.2.37 )
which can also be written:
( 3.2.38 )
We see that the line strength is a function of the incident light frequency, number density of the
absorber, temperature, and the Einstein coefficient for stimulated absorption at the transition of
interest. The line strength is not a function of wavelength of the lineshape but is specific for each
transition. The Beer-Lambert law can finally be written as:
( 3.2.39 )
Note that if we take the logarithm of this function and integrate over the entire line we no longer
need to actually know the form of the lineshape to quantify the line strength:
( 3.2.40 )
Depending on what we know in our experiment, we can quantify other parameters. Most
commonly, if we know the total pressure, temperature, and path length, we can quantify the mole
fraction of the absorber in state 1 (usually ground state) using the ideal gas law:
( 3.2.41 )
( 3.2.42 )
Clearly, the line strength is dependent on temperature. The dependence can be related to some
reference line strength in the following manner:
( 3.2.43 )
120
where E’’ is the energy of the lower state of the transition in cm-1
. The partition function can be
approximated by a polynomial function:
( 3.2.44 )
where the coefficients a, b, c, and d are tabulated in the HITRAN database, as well as the lower
state energy levels and integrated line strengths. In the case that a measurement is made in a gas
in which both the absorbing species mole fraction and temperature are unknown, the temperature
can be ascertained by comparing the integrated absorption profiles for two absorption lines, say
for lines 1 and 2.
( 3.2.45 )
where the subscript i represents the species and the subscripts 1 and 2 represent the two different
absorption lines. If the two absorption lines are nearby in frequency space (i.e. ), the
final two terms of Eq. 3.2.45 cancel, and the temperature is closely approximated by:
( 3.2.46 )
This will be referred to as the “two-line” method of temperature measurement (Farooq et al.
2008). Once the temperature is known, either one of the absorption features can be used to
quantify the species concentration.
There are many alternate forms of line strength and corresponding ways to write Beer’s law.
Throughout the derivation up to this point, the frequency of light has been expressed in units of
s-1
, however, units of wavenumber [cm-1
] could have just as easily been used. A few alternate
forms of the absorption coefficient and line strength are listed here for convenience. Recall that
ν = c/λ = c . Changing to wavenumber units:
121
( 3.2.47 )
( 3.2.48 )
( 3.2.49 )
we can also normalize by partial pressure or number density:
( 3.2.50 )
( 3.2.51 )
the last of which is the form used by HITRAN, which is convenient because it gives the line
strength only as a function of temperature. This can then be multiplied by the number density
and path length to find the absorbance of a mixture. Finally, we have a number of final versions
of Beer’s Law:
( 3.2.52 )
( 3.2.53 )
( 3.2.54 )
( 3.2.55 )
With these versions of Beer’s Law, it is straightforward to calculate parameters such as
number density or partial pressure (if the temperature is known) by using the integrated
absorption profile. If a laser is tuned through an absorption line, and a background signal is
collected, then it is simply a matter of numerically integrating. However, often it is more
accurate to fit a curve to the absorption signal. This can be accomplished by numerically
computing the lineshape function. Additionally, if the lineshape is well known, it is not
necessary to tune a laser through an entire absorption profile, and the laser can be parked at the
peak of the signal and monitored continuously.
122
3.2.5 Spectral lineshapes
While there are many broadening mechanisms possible in absorption spectroscopy, the ones
of particular concern when probing ro-vibrational lines are natural broadening, collisional (or
pressure) broadening, and Doppler broadening. Both natural and collisional broadening take on
Lorentzian lineshapes, while Doppler broadening takes on a Gaussian lineshape. The
convolution of these profiles results in the Voigt profile.
3.2.5.1 Doppler broadening
Doppler (or thermal) broadening is caused by the movement of gas molecules as they interact
with light. Not all gas particles move at exactly the same velocity, so different particles will
exhibit different degrees of broadening. Doppler broadening is thus referred to as
inhomogeneous. The Maxwell-Boltzmann distribution describes the probability of a gas
molecule to be moving at a particular velocity (in one dimension) to be:
( 3.2.56 )
where υx [m/s] is the velocity of a gas particle in the x direction (which is chosen to be in the
direction of propagation of the incident radiation), and m is the mass of the particle. As with all
probability functions:
( 3.2.57 )
multiplying both sides by N, the number density of absorbing species, gives us
( 3.2.58 )
and so the total number of particles with a velocity between υx and υx + Δ υx is
( 3.2.59 )
123
If the absorbing particle is moving away from the incident radiation, it effectively sees a different
frequency of light as compared to a particle standing still. This is called the Doppler effect. The
effective frequency, ν’ seen by the particle is:
( 3.2.60 )
If we assume that the particle will only absorb light exactly at the resonant frequency of its
transition, then we can solve for the external light frequency which will induce a transition in a
particle travelling at velocity υx:
( 3.2.61 )
which can be approximated using a first order Taylor expansion about υx/c = 0, since c>> υx, as:
( 3.2.62 )
If we rewrite Eq. 3.2.58 in terms of frequency, we can get:
( 3.2.63 )
where the differential element has been changed from dυx to dν, which will be reflected by a
factor of c/ν0 when we substitute Eq. 3.2.62 into Eq. 3.2.59 to get
( 3.2.64 )
The Gaussian distribution is defined as:
( 3.2.65 )
where σ is the standard deviation and μ is the mean. Therefore, the Doppler broadening
probability function follows a Gaussian profile. The line shape function is written as:
Δ
Δ
( 3.2.66 )
124
where the FWHM ( ) is:
Δ
( 3.2.67 )
is the FWHM of the curve. From equation 3.2.66, we can see that Doppler broadening is only a
function of temperature and the mass of the absorbing molecule.
3.2.5.2 Natural line broadening
Natural line broadening arises from the Heisenberg uncertainty principle:
( 3.2.68 )
which can be applied to the uncertainty in the energy of the upper and lower states. This
propagates to line broadening through Planck’s Law. Again considering the two state system
with lower state 1 and upper state 2, we have:
( 3.2.69 )
and
( 3.2.70 )
The terms Δt1 and Δt2 can be interpreted as the uncertainty in the time spent in states 1 and 2, or
the radiative lifetime of the molecule in that state:
( 3.2.71 )
Thus, we have equation:
( 3.2.72 )
Since state 1 is usually the ground state, the radiative lifetime is infinite, and only the lifetime in
the upper state contributes to natural line broadening:
( 3.2.73 )
125
3.2.5.3 Collision broadening
There are two types of collision broadening: depopulating and dephasing. Depopulating
collisions occur due to other molecules in the gas perturbing the energy state of the absorbing
particle, which results in a loss of energy of the excited particle, and limits the lifetime of the
dipole in the excited state. Dephasing collisions lead to anharmonicity in the motion of the
particle, so off-resonance absorption can occur, and the absolute strength of the absorption is
decreased at the center wavelength. First, we will consider collisions in general, and then discuss
how they affect the lineshape individually.
Considering collisions in general, we can describe a system using the hard sphere model of
particles, which assumes particles are spheres with a diameter of influence called the optical
collision diameter, σ. The optical collision cross-section is then
( 3.2.74 )
where
( 3.2.75 )
Figure 3.2.8: Collision cross-section
126
A good way to visualize particle motion through a gas of number density N is to imagine a
cylinder, with cross-sectional area equal to the collision cross-section, and with length l.
Consider particle B, which after passing through this cylinder, has collided with another particle
in the system, let’s say A, which has a particle density of NA in this gas mixture. This
configuration is presented in Figure 3.2.8. The product of NA*σ*l is unity when the probability of
such a collision is 1. In this case, l is called the mean free path and is defined as
( 3.2.76 )
The frequency of collisions of a single B particle with all particles A (ZBA) is then:
( 3.2.77 )
According to the Maxwell-Boltzmann distribution, the relative velocity can be defined as
( 3.2.78 )
( 3.2.79 )
which is based on the Maxwell-Boltzmann distribution. The total collisions of particle B with all
other particles in the system is then
( 3.2.80 )
If we consider the possibility of both the upper and lower states being perturbed by collisions, we
have:
( 3.2.81 )
which assumes the same collisional de-excitation rate for both the upper and lower states. This
can be represented as:
( 3.2.82 )
127
where [cm-1
/atm] is the collisional halfwidth for species A, defined for each collision of the
absorbing species with all other particles in the gas, including itself. Keep in mind that the
collisional halfwidth is a function of temperature via the velocity distribution. Therefore, we
must consider the variation of γA with temperature to quantitatively model the lineshape. This can
be done with the simple approximation:
( 3.2.83 )
where n is a temperature coefficient, equal to ½ in the hard sphere model.
So far, we have discussed that a molecule in an excited state has two ways of relaxing back
down to the ground state: natural relaxation and collisional relaxation. The time scales of each
for a particular state can be jointly represented as:
( 3.2.84 )
where T1 is the collisional time scale and τ is the natural relaxation time scale. To see how these
perturbations result in the Lorentzian lineshape, we will make an analogy between an oscillating
dipole moment and a damped harmonic oscillator. This is appropriate because the interaction of
the electromagnetic field with the molecular dipole moment forces the dipoles to oscillate at the
frequency of the field, and the natural and collisional broadening mechanisms act as dampers in
the system. This is called the classical approach, and can be represented by:
( 3.2.85 )
where γ is the damping coefficient, and x(t) is the position at time t. The natural frequency, ω0, is
defined as:
( 3.2.86 )
where k is the spring constant and m is mass. The damping coefficient can be thought of as
anything preventing the dipole from oscillating in the electromagnetic field. Any perturbation to
128
the energy of the upper state of the molecule will result in the dipole not contributing to the
overall absorption, so both natural and collisional broadening contribute in the same way to the
damping coefficient. The solution has the form:
( 3.2.87 )
When ω0 > γ/2 the system is called under-damped. In general, for atoms and molecules, ω0 >>
γ/2, so we can ignore the γ2/4 term in the radical. The solution becomes:
( 3.2.88 )
using Euler’s formula, this can be written as:
( 3.2.89 )
Taking only the real component, we have:
( 3.2.90 )
A plot of two waveforms of this type are presented in Figures 3.2.9 and Figure 3.2.10, with a
factor of 5 difference in their damping coefficients. From these figures, it is clear that the
damping coefficient controls the amplitude of oscillation. The damping coefficient can be
thought of as the limiting factor in the strength of the response of the oscillator to a driving field.
In this example, no driving field is present, but, to accurately represent the response of a dipole
to an electromagnetic field, a full classical derivation would include a harmonic driving field that
interacts with the dipole. In this case, the amplitude of oscillation would continue to grow
infinitely if no damping coefficient is present. The magnitude of the damping coefficient
determines the strength of the dipole’s response to the driving field. Continuing with our non-
driven oscillator example, we can learn still more about the nature of the dipole response.
129
Figure 3.2.9. Waveforms of damped harmonic oscillators with factor of five differential
in damping coefficient
Figure 3.2.10. Fourier transforms of the waveforms presented in Fig. 4.1.9
If the damping coefficient is 0, then clearly the dipole will only oscillate at ω0, but when the
damping coefficient is present it has an effect on the vibration frequency. We can see this by
taking the Fourier transform of the damped oscillator:
( 3.2.91 )
by taking advantage of the convolution theorem, we can separate the Fourier transform of the
function:
exp(-γt/2)
130
( 3.2.92 )
where the symbol “*” represents the convolution of the two functions. The first part of the
function can be transformed to get:
( 3.2.93 )
The second part can be transformed to get:
( 3.2.94 )
where δ is the Dirac delta function. In order to compute the convolution of these two functions,
we make note of the “sifting property” of the convolution of a function with a “shifted” Dirac
delta function:
( 3.2.95 )
Using this property, we find the convolution of Equations 3.2.93 and 3.2.94 to be:
( 3.2.96 )
We can assume that the driving frequency is nearly the resonant frequency, as in the case of
absorption spectroscopy, and drop the ω + ω0 term. The energy contained in the oscillator at a
particular frequency is the square of X(ω), which is analogous with the absorption strength of the
transition:
( 3.2.97 )
which is the form of the Lorentzian lineshape. The broadening coefficient is comprised of the
inverse of the two relaxation timescales, as in equation 3.2.84:
( 3.2.98 )
where γ is the HWHM of the Lorentzian lineshape.
131
The final form of broadening is collisional broadening due to dephasing collisions, which
require far less energy transfer per collision than depopulating collisions and thus occur far more
frequently. We can think of dephasing by imaging a dipole oscillating in the electromagnetic
field. If the dipole motion is interrupted by a collision with another molecule, but is not totally
relaxed from the vibration state, it may just be knocked out of phase with the other dipoles
oscillating in the field. Shortly thereafter, other collisions will knock it back into phase. The
actual waveform is then the sum of all the waveforms in which the oscillations are unperturbed.
The time that the dipole is unperturbed may be from between –τ/2 and τ/2, when the movement
of the dipole is described as:
( 3.2.99 )
which neglects, for the moment, depopulating and natural broadening. Outside of this time, we
can assume x(t) = 0 (since this dipole does not contribute to the absorption). This can be
described as the multiplication of a cosinusoidal waveform with a rectangular wave:
( 3.2.100 )
Again, to find how this waveform behaves in frequency space, we can take the Fourier transform
and use the convolution theorem:
( 3.2.101 )
The transform for the cosine function is the same as Equation 3.2.94. The Fourier transform
for the rectangular function is:
( 3.2.102 )
Finally, similar to equation 3.2.96, we have
( 3.2.103 )
132
Again neglecting the ω + ω0 term and taking the square of the magnitude, we have:
( 3.2.104 )
This represents a single oscillator. To consider the sum of all such oscillators in the absorbing
medium, we need to integrate all waveforms and use a weighting factor representing the
probability of the oscillator being unperturbed. This probability can be described as:
( 3.2.105 )
where T2 is the mean time between dephasing collisions. The probability of unperturbed
oscillation, starting from t = 0 just after a collision, is 1. As t approaches T2 the probability of a
collision approaches 1. Integrating over the weighted Fourier transform, we arrive at:
( 3.2.106 )
which is, again, Lorentzian in form. Note that in this case, the broadening coefficient is:
( 3.2.107 )
in order for the equation to take the same form as for depopulating and natural broadening. Thus,
the final form of the broadening coefficient is then:
( 3.2.108 )
In general, the natural line broadening is much narrower than the collisional broadening, and we
can just call this the collisional line broadening term. Normalizing the Lorentzian profile such
that:
( 3.2.109 )
We arrive at:
133
( 3.2.110 )
Or
( 3.2.111 )
Noting that 2πν=ω. Figure 3.2.11 presents the Lorentzian and Gaussian lineshapes for an
arbitrary molecule with an imaginary linecenter at 7 µm (1428.57 cm-1
). Both lineshapes have
the same broadening coefficient Δν (which is the FWHM of the lineshape).
Fig. 3.2.11. Comparison of Lorentzian and Gaussian lineshapes with equivalent
broadening coefficients (FWHM).
3.2.5.4 Voigt profile
To find the final form of the lineshape, we must use some combination of the Lorentzian and
Gaussian profiles. In order to do this, we must make the assumption that we can consider every
point on a collision (or natural) broadened lineshape is further broadened by Doppler effects.
Essentially, we can think of many collisionally broadened profiles. Each profile is separated into
groups of absorbing species, with each group having its own velocity. In this way, we can
134
decouple the Lorentzian and Gaussian lineshapes. The final profile, known as the “Voigt
profile”, is the convolution of the two lineshapes:
Δ
Δ
( 3.2.112)
where we have plugged Eqs. 3.2.66 and 3.2.111 for the Doppler and Lorentzian lineshapes,
respectively. Simplifying:
Δ
Δ
Δ
( 3.2.113 )
where u is the shifted convolution variable replacing ν - ν0. We can write this equation as:
( 3.2.114 )
where is the peak of the Doppler profile:
Δ ( 3.2.115 )
The rest of Eq. 3.2.113 is the “Voigt function” V(x,y). In this case, the parameters t, x, and y are:
Δ ( 3.2.116 )
Δ ( 3.2.117 )
Δ ( 3.2.118 )
To compute the Voigt profile, we introduce the Fadeeva function, w (Armstrong 1967):
( 3.2.119 )
135
For y > 0, this becomes:
( 3.2.120 )
where erfc is the complex error function. The Voigt function can be represented as the real part
of the Fadeeva function, denoted w:
( 3.2.121 )
where L is the imaginary part of the Fadeeva function. Therefore, if we solve the Fadeeva
function and take the real component, we will have the final lineshape. This can be accomplished
using the Matlab program written by Weideman (Weideman 1994), which uses a series
expansion to approximate the Fadeeva function using only a few lines of code. If x and y are
known, then computation of the Voigt profile using this code is relatively straightforward. Figure
3.2.12 shows a comparison of the normalized Doppler, collisional, and Voigt broadened profiles.
Here, the FWHM of all three lineshapes are set to be equal.
Fig. 3.2.11. Comparison of Lorentzian, Gaussian, and Voigt lineshapes with equivalent
FWHM.
136
Clearly, the Voigt profile contains characteristics of both the Gaussian and Lorentzian
lineshapes, with the wide wings of the Lorenztian and the wide peak of the Gaussian. In Figure
3.2.12, the Voigt profile is shown with the Lorentzian and Gaussian profiles again. However, in
this case the Lorentzian and Gaussian have the same FWHM, while they are used as inputs for
computing the Voigt profile. Note that the combination of both profiles results in an even wider
lineshape than either the Lorentzian or Gaussian, while the peak is lower.
Fig. 3.2.12. Comparison of Lorentzian, Gaussian, and Voigt lineshapes using the same
FWHM for Lorentzian and Gaussian lineshapes, and using these as inputs for the Voigt
profile.
3.2.6 Fitting absorption profiles
The final form of the Beer-Lambert law can be rewritten as
( 3.2.122 )
137
where the HITRAN formulation has been utilized. The HITRAN database (Rothman et al. 2009)
contains enough information to fit the absorption profiles of many small species. Among other
relevant parameters, HITRAN provides the line centers [cm-1
], integrated absorption strengths
S* [cm/molec.], collisional halfwidth for self broadening and broadening by air γA [cm-1
/atm] as
used in Eq. 3.2.82, reproduced here:
as well as the temperature coefficient n for scaling the halfwidths for use in 3.2.83, reproduced
here:
In addition, lower state energies E” [cm-1
] for each transition are tabulated, which can be used to
calculate temperature using Eq. 3.2.46. Although not mentioned herein, there is also a coefficient
for a line center shift, which accounts for a shifting in the line center that scales with pressure.
This term arises out of the discussion for collision broadening in the full derivation, and can be
computed for a gas mixture using the following formula:
( 3.2.123 )
where [cm-1] is the total collisional line shift, PA [atm] is the partial pressure of species A,
and [cm-1
/atm] is the line shift due to each collision partner. This is analogous with the
equation for total collision broadening.
In order to obtain an optimal fitting based on a measured absorption profile, a least squares
non-linear fitting algorithm (such as the Matlab function “lsqnonlin”) can be utilized. The input
parameters are varied, and the Voigt profile is computed until the measured absorption profile is
fit to the predicted one. There are many ways to implement this type of fitting depending on what
the known parameters of the experiment are. In the experimental results presented in the
138
following chapters, only the total pressure and path lengths are known. Temperature can be
computed by the two-line method using the integrated line shapes of two absorption lines (say
line 1 and line 2) for the same species at the same concentration and temperature, and using Eq.
3.2.46:
This is done by initially assuming some temperature, which sets the line strength S*(T) and the
Doppler broadening FWHM Δ , and then computing the line shape using the mole fraction
(calculating the concentration using the ideal gas law) , collision broadening FWHM , and
collision shift as the variable parameters . The line shapes are then integrated, and the
temperature recomputed using Eq. 3.2.46. This is done iteratively until the mole fraction used to
compute each line shape converges. The temperature can then be used as an input into the fitting
algorithm to compute the concentrations of other species using different absorption lines
measured at the same condition.
Note that the collision broadening FWHM and collision shift should be known if the partial
pressure of the absorbing species is known and the diluent gas is air. However, the diluents used
throughout the next chapters are not air, so these parameters are varied independently from the
partial pressures of the gas species. Again, depending on the experimental configuration, any of
the other inputs (or all of them) can be varied as part of the fitting algorithm to obtain
information about the system.
139
3.2.7 TDLAS experimental configuration
A Continuous Wave External Cavity Mode Hop Free (CW-EC-MHF) quantum cascade laser
(referred to as the EC-QCL) from Daylight Solutions and a distributed feedback quantum
cascade laser (referred to as the DFB-QCL) from Alpes Lasers (sbcw3176) are used in the
present in situ species and temperature measurements. The EC-QCL has a high output power of
100–220 mW and a line width resolution less than 0.001 cm-1
. It also has a wide tunability range
of 1302–1350 cm-1
covering the fundamental rovibrational absorption bands of species of
interest in ethylene oxidation such as H2O, CH4, and C2H2 which are 5-100 times stronger than
the absorption bands in the near infrared. The DFB-QCL has a maximum power output of 12.5
mW, and a scan range of only 1725.5 – 1727.8 cm-1
but targets a CH2O absorption line free from
interference with other major species in combustion, as is clearly visible in Fig. 3.2.2, which is a
difficulty in most CH2O measurements (Wang et al. 2013b).
In Chapter 5, only the EC-QCL was used, and the absorption lines used for water and
temperature quantification are located at 1338.55 cm-1
and 1339.15 cm-1
; the line for C2H2
quantification is at 1342.35 cm-1
; and the line for CH4 quantification is at 1341.32 cm-1
. In
Chapter 6, the absorption lines used for CH4 and temperature quantification with the EC-QCL
are located at 1343.56 cm-1
and 1343.63 cm-1
, and the line for CH2O quantification with the
DFB-QCL is at 1726.79 cm-1
. The two lines used for temperature measurement were selected
following previous guidelines (Farooq et al. 2008). As presented in Fig. 3.1.1, the two lasers are
co-aligned and coupled into a 24-pass Herriott cell such that either laser can be used individually
by flipping a mirror, allowing convenient measurements without realignment of the cell. The
Herriott cell is comprised of two opposed 12.7 mm OD, 50 mm focal length protected gold
concave mirrors, which are located 30 mm downstream of the front edge of the electrodes. The
140
laser light is coupled into the cell through a 2 mm hole in a stationary mirror in the quartz wall,
and cell alignment is accomplished with the help of a 2-dimensional tilt stage (Thorlabs KMS)
fitted with a screw for axial translation, which is necessary for proper alignment of the Herriot
cell, fixed to the Macor wall, as shown in Fig. 3.1.3. Images of the beam profile of the Herriott
cell are provided in Figure 3.2.13 for different pass numbers. The effective laser path length
through the plasma is 1.08 m (24 x 0.045 m). The transmitted output beams are incident on an
MCT detector (Vigo PVM-2TE10.6). In order to eliminate atmospheric water absorption outside
the test section, the laser path is purged with nitrogen.
Figure 3.2.13 Images of beam spots on Herriot cell mirror. The top left image is the beam
profile used for all measurements.
141
Figure 3.2.14 Collected signals in EC-QCL experiment with calculated laser scan rate.
“Pulse trigger” refers to the signal sent to trigger the plasma discharge, “signal” is the
absorption signal (of a water line), “etalon fringes” are the peaks and troughs measured
after the laser is passed through the Ge etalon, “scan rate” is the rate of laser tuning
calculated using the etalon fringes, “piezo scan” is the signal sent to the piezo-electric
controller to tune the laser, and “background” is the signal from the EC-QCL in the
absence of absorber.
The EC-QCL laser is scanned through the absorption lines by a 100 Hz sinusoidal signal sent
to the controller (Thorlabs MDT694A) for the piezo-electric actuator in the laser cavity using a
function generator (SRS DS345). In the time-dependent measurements, the laser scan was
synchronized with the discharge trigger such that the absorption peak occurred at a controllable
-10000
-8000
-6000
-4000
-2000
0
0
0.5
1
1.5
2
0.005 0.0055 0.006 0.0065
Sc
an
Ra
te (
cm
-1/s
)
Sig
na
l (V
)
Time (s)
Pulse Trigger Piezo Scan/10Signal BackgroundEtalon Fringes Etalon PeaksScan Rate
142
delay after the first plasma discharge pulse. The absorption signal used for quantification was
averaged over 30 - 50 plasma burst experiments. Figure 3.2.14 displays all of the signals
collected for a single EC-QCL experiment, along with the calculated laser scan rate. To scan the
DFB-QCL, the laser temperature is held constant while the current is swept by a sawtooth
waveform at a rate of 15,000 Hz. This allowed all time dependent CH2O data points to be
collected in a single scan, with an absorption measurement taken between every plasma pulse,
and averaged over 100 experiments. The fitting of the CH2O waveform is presented in Figure
3.2.15. The current controller for the DFB-QCL is a Wavelength Electronics QCL1000; the
thermoelectric temperature controller is an ILX Lightwave LDT-5980; and the laser is housed in
an ILX Lightwave LDM-4872 QCL mount. The wavelength scan rate of both lasers is monitored
using a 50.8 mm Germanium etalon (FSR ≈ 0.74 GHz at 1345 cm-1
). Due to inconsistency in
scan rates for the EC-QCL, the etalon signal is collected for every experiment. All signals are
recorded on a Tektronix DPO 7104C oscilloscope at a sampling rate of 20 Mega-samples per
second.
The direct absorption profile at different conditions of temperature and pressure can be
modeled using data from the HITRAN database (Rothman et al. 2009), as described in Section
3.2.6. Due to a lack of available measurements for the line broadening properties of helium or
argon, the pressure broadening coefficient was treated as variable parameter, along with the
absorber concentration, in a least squares nonlinear fitting algorithm for calculating the Voigt
function (Armstrong 1967, Weideman 1994). The temperature in Chapter 5 was calculated by
scanning over two different absorption lines of water at the same delay after the pulse burst,
while in Chapter 6 two methane lines were used.
143
Gases of known temperatures and concentrations of the absorbing species have been used to
validate the measurement technique, and a standard deviations of 3 ppm for methane and 1 ppm
for acetylene in the range of 5 – 100 ppm were observed. Temperature measurements have a
standard deviation of 5 K in the range of 300-500 K. Plots of these measurements are provided in
Figures 3.2.16 – 3.2.18.
Figure 3.2.15 Example of fitting to a CH2O absorption line. Blue stars are the raw data,
and red line is the fitting calculated for this condition.
144
Figure 3.2.16 Acetylene mole fraction measured by LAS vs. prepared mixture
Figure 3.2.17 Acetylene mole fraction measured by LAS vs. prepared mixture
y = 1.0036xR² = 0.9996
0
20
40
60
80
100
120
0 20 40 60 80 100 120
C2H
2M
ea
su
red
(p
pm
)
C2H2 Set (ppm)
y = 1.0508xR² = 0.9987
0
20
40
60
80
100
120
0 20 40 60 80 100 120
Me
as
ure
d C
H4
(pp
m)
CH4 Set (ppm)
145
Figure 3.2.18 Temperature measured by LAS vs. temperature measured by thermocouple
y = 1.0015xR² = 0.9942
250
300
350
400
450
500
550
250 300 350 400 450 500 550
Te
mp
era
ture
Me
as
ure
d (
K)
Temperature Set (K)
146
4 Numerical methods for plasma assisted combustion
modeling
The numerical methods used herein are based on a combination of theory from both
combustion and plasma physics. These methods are used to calculate the simulations presented
in Chapters 2, 5, and 6. Specific attention will be paid to the determination of electron collision
reaction rates, since the simulation of these reactions are significantly different from traditional
Arrhenius type reaction descriptions.
4.1 Electron collision processes
In addition to the kinetics of ground state species generally of interest in combustion science,
plasma-assisted combustion kinetics also includes the reactions of electrons, ions, and
electronically or vibrationally excited species, which are only of minor importance in traditional
combustion kinetics. The creation of these species is initiated by the application of an electric
field, which accelerates electrons present in small concentrations in the gas. These electrons
collide with neutral particles and sometimes ionize them, freeing additional electrons in a chain
branching process. If the probability of an ionization reaction is great enough, this results in an
exponential growth of electrons (and ions). This is called an “avalanche”, and the combination of
many local avalanches will result in an “electric breakdown” after which the gas is partially or
fully ionized.
In the plasma, electrons accelerate in the electric field until they collide with another species
present in the gas. The rates of the reactions that occur are dependent on the energy of the
electrons and the “cross section” of each particular reaction. Therefore, in order to predict the
147
reaction processes in a plasma, we must be able to predict the electron energy. This is done using
the Boltzmann Equation (Hagelaar et al. 2005):
( 4.1.1 )
Which can be expanded in six dimensional phase space to give:
( 4.1.2 )
where D is the total derivative, f is the electron distribution function, is the velocity vector, e is
the elementary charge (1.602 × 10-19
coulombs), m is the mass of an electron (9.109 × 10-31
kg),
E is the electric field, is the velocity-gradient operator, and C is the rate of change of the
electron distribution due to collisions. We can think of f as the mean number of particles existing
at time t located between r and dr and with a velocity between v and dv, i.e, in the element of
volume in phase space d3r d
3v. The total derivative then evaluates how f changes in time (1
st term
in Eq. 4.1.2) due to electrons entering (or leaving) the location between r and dr (2nd
term in Eq.
4.1.2), from electrons entering (or leaving) the velocity phase space between v and dv (3rd
term in
Eq. 4.1.2) due to acceleration by some force (in this case the electric field, E), and electrons
being scattered into the volume/phase space due to collisions or being created in the
volume/phase space due to reactions, which is the source term on the right hand side of Eq. 4.1.2
(Reif 2009).
The Boltzmann equation, including all of the transport and collision terms possible, is far too
complicated to solve without making some simplifications. A detailed description of the
derivation of the Boltzmann Equation is included in the Appendix. We follow the example of
Hagelaar and Pitchford (Hagelaar et al. 2005), for their code (BOLSIG+) is implemented to
solve the Boltzmann Equation in the computational code described later in this chapter, which
148
allows us to find the rates of electron collision reactions as a function of the reduced electric
field, or the electric field divided by the gas number density (E/N). The simplified version of the
Boltzmann equation, assuming a uniform field in both space and velocity phase space, can be
reformulated as:
( 4.1.3 )
which is the same form as a convection-diffusion continuity equation. Here, ε is the electron
energy, F0(ε) is the electron energy distribution function (EEDF), and we have defined the
coefficients and source terms as:
( 4.1.4 )
( 4.1.5 )
( 4.1.6 )
( 4.1.7 )
( 4.1.8 )
( 4.1.9 )
where is the collisional cross section of species k with electrons, Mk is the mass of species k,
xk is the mole fraction of species k, and uk is the threshold energy for the formation of excited or
149
ionized species k. We can think of all the terms in as sources of electrons coming into the
location and velocity phase space. The part of that represents growth or depletion of electrons
is termed G, which includes a normalization term . Eq. 4.1.3 can be solved numerically, as
described in (Hagelaar et al. 2005).
Once the electron energy distribution function is known, it can be used to calculate many
processes of interest in plasma science, including electron transport, energy and momentum flux
in plasmas, etc. Of interest to the current work are the rates of reactions of the various species
created or consumed in electron collision reactions. These rates can be calculated as described
by:
( 4.1.10 )
This is the final output from the solution of the Boltzmann Equation. The required input
parameters are the number density and temperature of the bulk gas, mole fractions of the gas
components, the magnitude of the electric field, and the cross sections (as a function of energy)
and threshold energies for the reactions of interest.
Conveniently, the cross sections for momentum transfer, excitation, de-excitation, attachment,
and ionization of many atoms and small molecules have been compiled in the LXCat database
(Pancheshnyi et al. 2010). Specific databases used for computations are described in the
following chapters.
4.2 Plasma-assisted Combustion Solver
Numerical computations have been utilized to investigate the influence of non-equilibrium
plasma generated by NRP discharges on the homogeneous ignition and related kinetics of
different combustible mixtures. In addition to ground state species, electrons, ions, and
150
electronically excited levels of molecules are included in the detailed chemical model. These
species are generated by electron impact excitation, dissociation, and ionization processes, and
react in quenching reactions (for excited states), electron-ion recombination reactions, and
charge exchange reactions. The code couples the commercially available zero-dimensional
plasma kinetics solver ZDPlasKin (Pancheshnyi et al. 2008) with the CHEMKIN (Lutz et al.
1988) package, as depicted in Fig. 4.2.1. During the discharge processes, ZDPlasKin employs a
Boltzmann equation solver (Hagelaar et al. 2005) to calculate the electron energy distribution
function and reaction rates based on the input list of electron impact reaction cross sections. The
cross section data can be downloaded from the online database LXCat . After each discharge
pulse, CHEMKIN is used to calculate the thermal properties and chemical reaction rates at the
conditions output by ZDPlasKin, and the species and temperature equation are integrated using a
VODE (Brown et al. 1989) solver. At each interface between the two processes, a matrix
including the values of time, temperature and species concentrations is used to transfer
information.
The ZDPlasKin model considers a non-equilibrium system with electron temperature much
higher than gas temperature, while ion temperature is not considered. During the discharge
process, the time evolution of species i through imax are computed using Equation 4.2.1
(Pancheshnyi et al. 2008):
( 4.2.1 )
Where [Ni] represents the number density of species i, Qij represents the source rates of species i
corresponding to different reactions j, and imax and jmax are the total number of the species and
151
Figure 4.2.1. Schematic representation of the numerical modeling approach, along with
governing species and energy equations for each portion of the computation.
reactions considered in the model, respectively. Each reaction proceeds at some reaction rate,
defined as:
( 4.2.2 )
Where kj is the rate constant for reaction j, [Nl] is the concentration of species l, lmax is the
number of reactants in reaction j, and alr is the reactant coefficient of species l (number of times
species l appears on the reactant side of the reaction). The source term Qij is then represented by:
( 4.2.3 )
…
E/N
Time0
ZDPlasKin CHEMKIN II - SENKIN
1
1
( )
=
𝑌
=
Qij : production rate of species i in reaction j
γ : ratio of specific heats
kB : Boltzmann constant
P: Power
ρ : density
Yk : mass fractions of species k
ωk: integrated reaction rates for species k
Wk : molecular weight of species k
ek : internal energy of species kPancheshnyi et al. 2008
http://www.zdplaskin.laplace.univ-tlse.fr
20 ns > 25 µs
152
Where aip is the product coefficient of species i. The rate constants are either provided as inputs
or calculated using BOLSIG+ with the reaction cross section provided as the input.
Gas heating in the discharge is computed using the method of Flitti and Pancheshnyi (Flitti et
al. 2009), which distributes the external power deposited to gas, denoted Pext, into three
channels: the translational degree of freedom of electrons Pelec, the translational degree of
freedom of the gas Pgas, and the internal degree of freedom of gas Pchem:
( 4.2.4 )
The gas temperature can then be computed considering energy deposition into all three
channels using Equations 4.2.5 – 4.2.8 (Pancheshnyi et al. 2008):
( 4.2.5 )
( 4.2.6 )
( 4.2.7 )
( 4.2.8 )
Where γ is the specific heat ratio, kB is the Boltzmann constant, Tgas is the gas temperature, e
is the elementary charge, Ne is the electron number density, υe is the electron drift velocity, E is
the electric field, Te is the effective electron temperature, and Ei is the internal energy of species
i. During the inter-pulse time, CHEMKIN computes species evolution and heat addition for the
constant volume system using:
( 4.2.9 )
( 4.2.10 )
Where is gas density, Yi is the mass fraction of species i, t is time, i is the production or
consumption rate of species i, Wi is the molecular weight of species i, CV is the specific heat of
153
the gas at constant volume, and Ei is the internal energy for species i in mass units. In addition to
gas heating, a heat loss term must be included to accurately predict the measured temperature
profiles. Adamovich et al. (Adamovich et al. 2009b) reported that the energy loss is primarily
due to conduction to the quartz channel walls. In this 0-D model, the heat loss is described by
incorporating a conduction heat transfer term into the energy equation:
( 4.2.11 )
Where λ(Tgas) is the thermal conductivity as a function of gas temperature, Twall is the wall
temperature, L is the channel height, and L/π is the spatial scale for conduction heat transfer with
uniform generation in the rectangular geometry.
4.3 Plasma kinetic model
4.3.1 Ethylene model
The kinetic model employed for discharges in ethylene/oxygen/argon mixtures is a
combination of an air plasma model (Uddi et al. 2009b) and two different combustion chemistry
models for ethylene dissociation and oxidation: USC Mech II (Wang et al.) and the Princeton
HP-Mech. The HP-Mech includes a high pressure sub-mechanism for H2 (Burke et al. 2012),
CO/methanol (Dievart et al. 2014), and C2-C6 low temperature oxidation kinetics (Yang et al.
2013). Additional reactions not supplied as part of the air plasma model were assembled and
added to the plasma model to account for the presence of the fuel and its intermediates and are
listed in Table 4.1. The entire model is available here: (Lefkowitz et al. 2015d). The reaction
rates of electron impact ionization, dissociative ionization, excitation, dissociative excitation, and
154
Table 4.1: Plasma kinetic model for C2H4/O2/Ar mixtures
Reaction Rate Constant (cm3/s) Ref.
Electron impact dissociative excitation
E1 C2H4 + e = C2H3 + H + e σ (Janev et al. 2004)
E2 C2H4 + e = C2H2 + H2 + e σ (Janev et al. 2004)
E3 C2H4 + e = C2H2 + H + H + e σ (Janev et al. 2004)
E4 C2H4 + e = C2H + H2 + H + e σ (Janev et al. 2004)
Electron impact ionization and dissociative ionization
E5 C2H4 + e = C2H4
+ + e +e σ (Janev et al. 2004)
E6 C2H4 + e = C2H3+ + H + e + e σ (Janev et al. 2004)
E7 C2H4 + e = C2H2+ + H + H + e + e σ (Janev et al. 2004)
E8 C2H4 + e = C2H2+ + H2 + e + e σ (Janev et al. 2004)
Quenching of Excited Ar
E9 Ar* + C2H4 = Ar + C2H2 + H + H 4.39 × 10-10 (Velazco et al. 1978)
E10 Ar* + C2H4 = Ar + C2H2 + H2 4.88 × 10-11 (Velazco et al. 1978)
E11 Ar* + C2H4 = Ar + CH2 + CH2 4.88 × 10-11 (Velazco et al. 1978)
E12 Ar* + C2H4 = Ar + C2H3 + H 4.88 × 10-11 (Velazco et al. 1978)
Electron-ion Recombination
E13
E
E22
e + C2H2+ = C2H + H 2.34 x 10-6 (300/T)0.5 (Mitchell 1990)
E14 e + C2H2+ = C2 + H2 9.35 x 10-8 (300/T)0.5 (Mitchell 1990)
E15 e + C2H2+ = C2 + H + H 1.40 x 10-6 (300/T)0.5 (Mitchell 1990)
E16 e + C2H2+ = CH2 + C 2.34 x 10-6 (300/T)0.5 (Mitchell 1990)
E17 e + C2H2+ = CH + CH 6.08 x 10-7 (300/T)0.5 (Mitchell 1990)
E18 e + C2H3+ = C2H2 + H 2.26 x 10-6 (300/T)0.5 (Mitchell 1990)
E19 e + C2H3+ = C2H + H2 4.68 x 10-7 (300/T)0.5 (Mitchell 1990)
E20 e + C2H3+ = C2H + H + H 4.60 x 10-6 (300/T)0.5 (Mitchell 1990)
E21 e + C2H3+ = C2 + H2 + H 2.34 x 10-7 (300/T)0.5 (Mitchell 1990)
E22 e + C2H3+ = CH3 + C 4.68 x 10-8 (300/T)0.5 (Mitchell 1990)
E23 e + C2H3+ = CH2 + CH 2.34 x 10-7 (300/T)0.5 (Mitchell 1990)
Charge Exchange
E24 Ar+ + C2H4= Ar + C2H4
+ 4.40 x 10-11 (Tsuji et al. 1993)
E25 Ar+ + C2H4= Ar + C2H3+ + H 8.36 x 10-10 (Tsuji et al. 1993)
E26 Ar+ + C2H4= Ar + C2H2+ + H + H 2.20 x 10-10 (Tsuji et al. 1993)
dissociative recombination reactions were computed from their electron collision cross-
sections, which were obtained from the LXCat database (Pancheshnyi et al. 2010), specifically
from (Hayashi et al. 1990, Janev et al. 2004, Itikawa 2005). It should be noted that the plasma
model is somewhat simplistic, and neglects some reaction pathways with the fuel and plasma-
produced species such as O(1D) and
Δ , which have been shown to be important
reactions in other modeling efforts (Starikovskaia 2014, Lefkowitz et al. 2015b).
155
In the modeling for Chapter 5, the pulse shape is assumed to be a square wave with 12 ns
width. The reduced electric field (E/N) was considered as an adjustable parameter to fit the
measurement of C2H2 in C2H4 dissociation because of the strong sensitivity of this pathway to
electron energy. The resulting reduced electric field is 350 Td, which is 16% greater than the
measured value based on the peak applied electric field. This discrepancy may be caused by a
number of factors, and deserves further attention in future modeling efforts. It should be noted
that none of the electron collision reaction rates change by more than 10% due to the increased
E/N.
4.3.2 Methane model
Again, the kinetic model can be separated into two parts: the combustion kinetic model and
the plasma kinetic model. While excited species quenching reactions, electron-ion recombination
reactions, and charge exchange reactions are needed in both models, the excitation and ionization
electron collision reactions are limited to the plasma model, while the combustion chemistry
reactions of neutral ground state species are mostly limited to the combustion model. The
combustion model employed here is HP-Mech, previously described above and available in
(Lefkowitz et al. 2015d), which includes hydrocarbon reactions for up to four carbon chain
length molecules, and is specifically designed for high pressure and low temperature (< 800 K)
chemistry. The plasma model has been reassembled primarily from an air plasma model
(Capitelli et al. 2000, Pancheshnyi et al. 2008), with helium reactions from (Stafford et al. 2004,
Sun et al. 2012) and methane reactions from (Kosarev et al. 2008a, Sun et al. 2012). Included
excited and ionized species are:
, Δ ,
(all grouped as O2(4.5eV) by
analogy with (Kosarev et al. 2008a)); Δ ;
;
; O(1D); O(
1S); O
+; He(2s
1S);
; and
. The potential well diagram of the excited levels of O2 are provided in Figure
156
4.3.1 for reference (Farooq et al. 2014). Vibrationally excited species are neglected since their
energy is insufficient to lead to dissociation of molecules and are not of significant kinetic
importance in the present mixture. In terms of energy addition, at the reduced electric fields of
this study, energy coupling into the plasma is strongly branched toward electronic excitation and
ionization, with a relatively small amount being coupled into the vibrational states of molecules
(Flitti et al. 2009). therefore, from both a kinetic and thermal perspective, it is reasonable to
neglect the vibrational levels of O2 and CH4. In addition, negative ions and complex positive ions
are neglected due to their relatively low concentrations and short lifetimes in the present plasma.
The electron-collision cross-sections for O2 and He are downloaded from the LXCat online
database (Pancheshnyi et al. 2010) from the Phelps (Phelps database) and Biagi databases,
respectively (Biagi database), and cross-sections for CH4 are computed in the method described
by Janev and Reiter (Janev et al. 2002). Comparisons of the electron collision cross sections
among the databases available from LXCat are in agreement for total excitation and ionization of
He and O2 (Alves et al. 2013). However, there is some discrepancy for cross sections into
particular excitation levels of O2. The Phelps database was selected in this model due to its use
as the basis of many of the other datasets in the literature. Electron-collision reactions are not
considered for any intermediate species due to their low concentration as compared to the
reactants and helium. The entire plasma model is listed in Table 4.2, and is published in
(Lefkowitz et al. 2015b), along with the full combustion model. Due to the significant changes
and additions made to previous models, it is appropriate to list all reactions used in the table.
157
Table 4.2: Plasma kinetic model for CH4/O2/He mixtures
Reaction Rate Constant (cm3/s) Ref.
Electron impact excitation
M1 e + O2 => e +
Δ σ (Phelps database)
M2 e + O2 => e +
σ (Phelps database)
M3 e + O2 => e + O2(4.5eV) σ (Phelps database)
M4 e + He => e+ He(2s 1S) σ (Phelps database)
Electron impact dissociative excitation
M5 e + O2 => e + O + O σ (Phelps database)
M6 e + O2 => e + O + O(1D) σ (Phelps database)
M7 e + O2 => e + O + O(1S) σ (Phelps database)
M8 e + CH4 => e + CH3 + H σ (Janev et al. 2002)
M9 e + CH4 => e + CH2 + H2 σ (Janev et al. 2002)
M10 e + CH4 => e + CH + H2 + H σ (Janev et al. 2002)
M11 e + CH4 => e + C + 2 H2 σ (Janev et al. 2002)
Electron impact ionization and dissociative ionization
M12 e + O2 => e + e + O2
+ σ (Phelps database)
M13 e + He => e + e + He+ σ (Biagi database)
M14 e + CH4 => e + e + CH4+ σ (Janev et al. 2002)
M15 e + CH4 => e + e + CH3+ + H σ (Janev et al. 2002)
Quenching of Excited O2
M16
Δ + CH4 => O2 + CH4 1.40 × 10-18 (Becker et al. 1971)
M17 Δ + CH4 => CH3 + HO2 6.14 × 10-12 exp(-17900/T) (Mayer et al. 1968)
M18
+ CH4 => Δ + CH4 1.08 × 10-13 (Dunlea et al. 2005)
M19 O2(4.5eV) + CH4 =>
+ CH4 1.08 × 10-13 estimate
M20 Δ + He => O2 + He 8.00 × 10-21 (T/300)0.5 (Stafford et al. 2004)
M21
+ He => Δ + He 1.00 × 10-17 (T/300) 0.5 (Stafford et al. 2004)
M22 O2(4.5eV) + He => Δ + He 1.00 × 10-13 (Hicks et al. 2005)
M23 O2(4.5eV) + He =>
+ He 9.00 × 10-13 (Hicks et al. 2005)
M24 Δ + O + He => O2 + O + He 6.30 × 10-33 (Hicks et al. 2005)
M25 Δ + O + O2 => O2 + O + O2 1.00 × 10-32 (Hicks et al. 2005)
M26 Δ + O => O2 + O 7.00 × 10-16 (Capitelli et al. 2000)
M27 Δ + O2 => O2 + O2 3.80 × 10-18 exp(-205/T) (Capitelli et al. 2000)
M28 Δ + O3 => O2 + O2 + O(1D) 5.20 × 10-11 exp(-2840/T) (Capitelli et al. 2000)
M29 Δ +
Δ => O2 +
7.00 × 10-28 T3.8 exp(700/T) (Capitelli et al. 2000)
M30
+ O => Δ + O 8.10 × 10-14 (Capitelli et al. 2000)
M31
+ O => O2 + O(1D) 3.40 × 10-11 (300/T) 0.1 exp(-
4200/T)
(Capitelli et al. 2000)
M32
+ O2 => Δ + O2
4.30 × 10-22 T2.4 exp(-
281/T)
(Capitelli et al. 2000)
M33
+ O3 => O2 + O2 + O 2.20 × 10-11 (Capitelli et al. 2000)
M34 O2(4.5eV) + O => O2 + O(1S) 9.00 × 10-12 (Capitelli et al. 2000)
M35 O2(4.5eV) + O2 =>
+
3.00 × 10-13 (Capitelli et al. 2000)
Quenching of Excited O
M36 O(1D) + CH4 => CH3 + OH 1.13 × 10-10 (Sander et al. 2006)
M37 O(1D) + CH4 => CH2OH + H 3.00 × 10-11 (Sander et al. 2006)
M38 O(1D) + CH4 => CH2O + H2 7.50 × 10-12 (Sander et al. 2006)
M39 O(1D) + He => O + He 1.00 × 10-13 (Stafford et al. 2004)
158
M40 O(1D) + O => O + O 8.00 × 10-12 (Capitelli et al. 2000)
M41 O(1D) + O2 => O + O2 6.40 × 10-12 exp(67/T) (Capitelli et al. 2000)
M42 O(1D) + O2 => O + Δ 1.00 × 10-12 (Capitelli et al. 2000)
M43 O(1D) + O2 => O +
2.60 × 10-11 exp(67/T) (Capitelli et al. 2000)
M44 O(1D) + O3 => O2 + O + O 1.20 × 10-10 (Capitelli et al. 2000)
M45 O(1D) + O3 => O2 + O2 1.20 × 10-10 (Capitelli et al. 2000)
M46 O(1S) + O => O(1D) + O 5.00 × 10-11 exp(-300/T) (Kossyi et al. 1992)
M47 O(1S) + O2 => O(1D) + O2 1.30 × 10-12 exp(-850/T) (Kossyi et al. 1992)
M48 O(1S) + O2 => O + O + O 3.00 × 10-12 exp(-850/T) (Kossyi et al. 1992)
M49 O(1S) + Δ => O + O2(4.5eV) 1.10 × 10-10 (Kossyi et al. 1992)
M50 O(1S) + Δ => O(1D) +
2.90 × 10-11 (Kossyi et al. 1992)
M51 O(1S) + Δ => O + O + O 3.20 × 10-11 (Kossyi et al. 1992)
M52 O(1S) + O3 => O2 + O2 2.90 × 10-10 (Kossyi et al. 1992)
M53 O(1S) + O3 => O2 + O + O(1D) 2.90 × 10-10 (Kossyi et al. 1992)
Quenching of Excited He
M54 He(2s 1S) + CH4 => He + CH + H2 + H 5.60 × 10-13 (Sun et al. 2012)
M55 He(2s 1S) + CH4 => He + CH4+ + e 7.90 × 10-12 (Tsuji et al. 1991)
M56 He(2s 1S) + CH4 => He + CH3+ + H + e 8.30 × 10-12 (Tsuji et al. 1991)
M57 He(2s 1S) + CH4 => He + CH2+ + H2 + e 7.10 × 10-13 (Tsuji et al. 1991)
M58 He(2s 1S) + He(2s 1S) => He + He+ + e 1.00 × 10-9 (T/300)0.5 (Stafford et al. 2004)
M59 He(2s 1S) + O2 => O2+ + He + e 2.54 × 10-10 (T/300) 0.5 (Stafford et al. 2004)
M60 He(2s 1S) + O3 => O2+ + O + He + e 2.54 × 10-10 (T/300) 0.5 (Stafford et al. 2004)
M61 He(2s 1S) + Δ => O2
+ + He + e 2.54 × 10-10 (T/300) 0.5 (Stafford et al. 2004)
M62 He(2s 1S) + O => O+ + He + e 2.54 × 10-10 (T/300) 0.5 (Stafford et al. 2004)
M63 He(2s 1S) + O(1D) => O+ + He + e 2.54 × 10-10 (T/300) 0.5 (Stafford et al. 2004)
M64 He(2s 1S) + O(1S) => O+ + He + e 2.54 × 10-10 (T/300) 0.5 (Stafford et al. 2004)
Electron-ion Recombination
M65 e + O2
+ => O + O 5.11 × 10-8 (300/Te)0.7 (Florescu-Mitchell et al. 2006)
M66 e + O2+ => O + O(1D) 8.48 × 10-8 (300/Te)0.7 (Florescu-Mitchell et al. 2006)
M67 e + O2+ => O(1D) + O(1D) 5.02 × 10-8 (300/Te)0.7 (Florescu-Mitchell et al. 2006)
M68 e + O2+ => O(1D) + O(1S) 7.90 × 10-9 (300/Te)0.7 (Florescu-Mitchell et al. 2006)
M69 e + O+ + e => O+ e 7.00 × 10-20 (300/Te)4.5 (Capitelli et al. 2000)
M70 e + O+ + M => O + M 6.00 × 10-27 (300/Te)1.5 (Capitelli et al. 2000)
M71 e + He+ => He(2s 1S) 6.76 × 10-13 Te-0.5 (Stafford et al. 2004)
M72 e + e + He+ => He(2s 1S) + e 5.12 × 10-27 Te-4.5 (Stafford et al. 2004)
M73 e + CH4+ => CH2 + H + H 1.70 × 10-7 (300/Te)0.5 (Florescu-Mitchell et al. 2006)
M74 e + CH4+ => CH3 + H 1.70 × 10-7 (300/Te)0.5 (Florescu-Mitchell et al. 2006)
M75 e + CH3+ => CH2 + H 3.50 × 10-7 (300/Te)0.5 (Florescu-Mitchell et al. 2006)
M76 e + CH2+ => CH + H 2.50 × 10-7 (300/Te)0.5 (Florescu-Mitchell et al. 2006)
Charge Exchange
M77 O+ + O2 => O2
+ + O 2.00 × 10-11 (300/T) 0.5 (Capitelli et al. 2000)
M78 O+ + O3 => O2+ + O2 1.00 × 10-10 (Capitelli et al. 2000)
M79 He+ + O2 => O+ + O + He 1.07 × 10-9 (T/300) 0.5 (Stafford et al. 2004)
M80 He+ + O3 => O+ + O2 + He 1.07 × 10-9 (T/300) 0.5 (Stafford et al. 2004)
M81 He+ + O2 => O2+ + He 3.30 × 10-11 (T/300) 0.5 (Stafford et al. 2004)
M82 He+ +
=> O+ + O + He 1.07 × 10-9 (T/300) 0.5 (Stafford et al. 2004)
M83 He+ +
=> O2+ + He 3.30 × 10-11 (T/300) 0.5 (Stafford et al. 2004)
M84 He+ + O => O+ + He 5.00 × 10-11 (T/300) 0.5 (Stafford et al. 2004)
M85 He+ + O(1D) => O+ + He 5.00 × 10-11 (T/300) 0.5 (Stafford et al. 2004)
159
M86 He+ + O(1S) => O+ + He 5.00 × 10-11 (T/300) 0.5 (Stafford et al. 2004)
M87 CH4+ + O2 => CH4 + O2
+ 5.00 × 10-10 (Kosarev et al. 2008a)
Optical transitions and predissociation
M88 Δ => O2 2.60 × 10-4 (Capitelli et al. 2000)
M89
=> Δ 1.50 × 10-3 (Capitelli et al. 2000)
M90
=> O2 8.50 × 10-2 (Capitelli et al. 2000)
M91 O2(4.5eV) => O2 1.10 × 101 (Capitelli et al. 2000)
O3 reactions
M92 O3 + O2 => O2 + O + O2 2.51 × 10-10 exp(-11600/T) (Capitelli et al. 2000)
M93 O3 + O => O2 + O + O 4.15 × 10-10 exp(-11430/T) (Capitelli et al. 2000)
M94 O3 + O => O2 + Δ 1.00 × 10-11 exp(-2300/T) (Capitelli et al. 2000)
M95 O + O2 + O2 => O3 + O2 7.60 × 10-34 (300/T)1.9 (Capitelli et al. 2000)
M96 O + O2 + O => O3 + O 3.90 × 10-33 (300/T)1.9 (Capitelli et al. 2000)
M97 O + O2 + He => O3 + He 3.40 × 10-34 (300/T)1.2 (Stafford et al. 2004)
4.3.3 Methane and ethylene model for air plasmas
The model employed in Chapter 2.2 used a modified version of the model described in the previous
sections. The main difference is the inclusion of N2 electron collision reactions. The simplified
methane-air plasma model of Aleksandrov et al. (Aleksandrov et al. 2009b) is used as the basis
for the plasma chemistry, and ethylene reactions from Section 4.3.1 are added to complete the
model for ethylene-air. The plasma species considered follows the example of Aleksandrov et al.
(Aleksandrov et al. 2009b), and includes
; ,
,
(all
grouped as N2(B));
,
(all grouped as N2(a’)); ;
;
, ,
(all grouped as O2(4.5eV));
;
; ;
;
. Energy level diagrams for O2 and N2 are provided in Figs. 4.3.1
(Farooq et al. 2014) and 4.3.2 (Lofthus et al. 1977), respectively. The electron-collision cross-
sections for O2, N2, and CH4 are downloaded from the LXCat online database (Hayashi ,
Napartovich et al. , Phelps), and cross-sections for C2H4 are computed in the method described
by Janev and Reiter (Janev et al. 2004). Quenching reactions from the plasma model are added to
the combustion chemistry model USC-Mech II (Wang et al.) to create a simplified reaction
160
scheme for methane-air and ethylene-air PAC at elevated temperatures (> 800 K) to operate
during the inter-pulse time.
Figure 4.3.1. Potential energy diagram of O2 electronically excited states (Farooq et al.
2014).
161
Figure 4.3.2. Potential energy diagram of N2 electronically excited states (Lofthus et al.
1977).
162
5 Kinetic study of dissociation and oxidation of ethylene in
a NRP DBD flow reactor
The results presented in this chapter were published and presented in:
J. K. Lefkowitz, M. Uddi, B. C. Windom, G. Lou, Y. Ju, “In situ Species Diagnostics and
Kinetic Study of Plasma Activated Ethylene Dissociation and Oxidation in a Low Temperature
Flow Reactor,” Proc. Combust. Inst. 35 (2015) 3505-3512.
Some of the results were also presented by Mruthunjaya Uddi in the following conference:
M. Uddi, J. K. Lefkowitz, B. Windom, Y. Ju, “Species Measurements of Ethylene Oxidation in a
Nanosecond-Pulsed Plasma Discharge Using QCL Absorption Spectroscopy Near 7.6µm,” 51st
AIAA Aerospace Sciences Meeting, Grapevine, Texas (2013) AIAA paper 2013-0435.
I also presented the results in the following conferences:
J. K. Lefkowitz, M. Uddi, B. C. Windom, Y. Ju, “In situ Mid-infrared Absorption Measurements
in a Nanosecond Pulsed Plasma Discharge,” The 2nd
International Education Forum on
Environment and Energy Science, Huntington Beach, California (2013).
J. K. Lefkowitz, B. C. Windom, W. MacDonald, S. Adams, T. Chen, M. Uddi, Y. Ju, “Time
Dependent Measurements of Species Formation in Nanosecond-Pulsed Plasma Discharges in
C2H4/O2/Ar Mixtures” 52nd
AIAA Aerospace Sciences Meeting, National Harbor, Maryland
(2014) AIAA paper 2014-1179.
163
I was responsible for construction of the experimental apparatus, assembly of the kinetic model,
data collection and processing, analysis of the experimental and numerical results, and writing of
the publication mentioned above, and presentation of the work at two of the above conference
presentations. Prof. Mruthunjaya Uddi shared responsibility for construction of the experimental
apparatus and presented the work at one of the conferences mentioned above. Prof. Bret Windom
assisted in data collection, Prof. Guofeng Lou was responsible for maintenance of the numerical
code. Mr. William MacDonald, Ms. Sarah Adams, and Mr. Tony Chen assisted in data collection
and data processing. The initial concept and guidance for the research was provided by Prof.
Yiguang Ju.
5.1 Introduction
As discussed in the introductory chapter, it remains difficult to understand the fundamental
mechanism of the observed combustion enhancement of PAC due to the complicated thermal
and kinetic coupling between plasma and combustion. In order to isolate these effects, several
investigators have developed simplified plasma-assisted combustion experiments by using
nanosecond pulsed discharges in simple geometries (Lou et al. 2007, Kosarev et al. 2008a,
Kosarev et al. 2009, Mintusov et al. 2009, Uddi et al. 2009a, Uddi et al. 2009b, Sun et al. 2011,
Yin et al. 2011, Sun et al. 2012, Sun et al. 2013, Yin et al. 2013a, Yin et al. 2013b, Sun et al.
2014, Yin et al. 2015). The plasma in these experiments produce short lifetime metastable
species that activate the oxidation of fuel on shortened time scales and lowered temperatures at
which normal combustion cannot occur. Unfortunately, strong interference between the plasma
and probes typically applied to investigate combustion systems render conventional ex situ
diagnostics such as GC and FTIR inapplicable.
164
Laser induced fluorescence (LIF) has been used recently for quantification of OH (Cathey et
al. 2008, Stockman et al. 2009, Lacoste et al. 2013b, Sun et al. 2013, Yin et al. 2013b), NO
(Uddi et al. 2009a), O (Uddi et al. 2009b, Stancu et al. 2010, Sun et al. 2011, Sun et al. 2013), H
(Ganguly et al. 2004), CH2O (Sun et al. 2013, Sun et al. 2014), and temperature (Yin et al. 2011,
Yin et al. 2013a, Yin et al. 2015). However, LIF is limited to small species and quantitative
measurements are difficult. Cavity ring down spectroscopy (CRDS) has also recently been used
for quantification of N2(A) (Stancu et al. 2010), N2+ (Yalin et al. 2002a, Yalin et al. 2002b) O3,
and O2 (a1Δg) (Ombrello et al. 2010a, Ombrello et al. 2010b), and is an excellent technique for
specific species if the proper wavelength can be accessed. Infrared laser absorption spectroscopy
(IR LAS) has been used as a combustion diagnostic in applications such as shock tubes, flames,
rapid compression machines, scramjets, and IC engines (Hanson 2011, Uddi et al. 2012). The
advantage of IR-LAS over other in situ techniques is the ability to make quantitative, sensitive
(sub-ppm level), time-dependent measurements of multiple species and temperature for small
hydrocarbons using a single laser. However, implementation of IR-LAS diagnostics in non-
equilibrium, low temperature plasma is challenging due to the small dimensional restraints
required to sustain uniform plasma. The value of simultaneous multi-species measurements is
evidenced by the uncertainty in combustion kinetics at temperatures below 900 K. The widely
used USC-Mech II (Wang et al.) in PAC modeling does not include low temperature chemistry
to appropriately model low temperature plasmas. Therefore, kinetics of plasma activated low
temperature fuel oxidation between 300 and 900 K are poorly understood, and an appropriate
kinetic model for PAC is lacking (Sun et al. 2014).
Therefore, the goals of this chapter are to: 1) develop an in situ mid-IR laser absorption
spectroscopy method to obtain time dependent species concentrations in C2H4/Ar dissociation
165
and C2H4/O2/Ar oxidation activated by a nanosecond-duration repetitively pulsed plasma, and 2)
to gain fundamental understanding of the kinetics of plasma activated low temperature fuel
oxidation via kinetic model development and experimental validation. This will be accomplished
using the experiment described in Chapter 3 and the numerical modeling tools described in
Chapter 4.
5.2 Results and discussion
5.2.1 Ethylene dissociation
For the dissociation experiments, the mole fraction of C2H4 in the C2H4/Ar was varied
from 1% to 100%. To validate the in situ laser diagnostics and measure other intermediate
species, GC-TCD and LAS measurements were collected concurrently in a continuously pulsed
plasma discharge of various frequencies, the results of which are presented in Fig. 5.2.1 for a
mixture of 6.25% C2H4 in Ar. At all conditions, more than 98.5% of the carbon is recovered by
the GC-TCD measurements. Any disagreement between the two measurement methods can be
attributed to stable species diffusion out of the plasma region introducing uncertainty in the path
length used for quantification. Due to the short time period of the transient experiments, species
diffusion is not expected to significantly affect those measurements. It is seen that, in the range
of 100-5000 Hz plasma frequency, all the intermediate species increase linearly with pulse
frequency. Acetylene concentration is greater than all other intermediates by at least a factor of
two, indicating that acetylene formation is the major pathway of ethylene dissociation. The next
most prominent intermediate species are the butene isomers, followed closely by propene. The
observation of butene and propene suggests that after plasma dissociation breaks down the fuel
molecules, a building up larger hydrocarbon species occurs, likely due to radical recombination
166
reactions. Therefore, inclusion of higher hydrocarbon pathways for radical recombination
reactions are needed in PAC kinetic models to adequately predict plasma activated fuel
dissociation, even for small hydrocarbons. Unfortunately, many of the current combustion
models for PAC do not take into account these large molecule formation pathways. Considerable
Figure 5.2.1. Major species concentrations during continuous plasma discharge measured
by GC-TCD sampling and laser absorption spectroscopy (open symbols) in a mixture of
6.25/93.75 C2H4/Ar at 60 Torr.
amounts of methane and hydrogen are also formed from radical recombination reactions. As
such, time dependent, in situ measurements of these intermediate species are critical to
improving the understanding of ethylene dissociation kinetics in plasma.
The time dependent measurements of C2H2 by mid-IR LAS during and after the 150 pulse
discharge (30 kHz, 5 ms burst duration) for all dissociation cases are plotted in Fig. 5.2.2. Time
zero represents the final plasma pulse in the burst. It is seen that C2H2 concentration generally
1
10
100
1000
10000
100000
0 1000 2000 3000 4000 5000
Mo
le F
rac
tio
n (
pp
m)
Plasma Frequency (Hz)
C2H4 C4H8C3H6 C2H6C2H2 C2H2 LASCH4 CH4 LASH2
167
increases with increasing fuel mole fraction of C2H4. However, as can be seen from the insert in
Fig. 5.2.2, the relative yield of C2H2 to initial C2H4 concentration decreases with the increase of
C2H4. Analysis of the reaction pathways reveals that the decrease of the relative acetylene yield
Figure 5.2.2. Acetylene mole fraction as a function of time for mixtures of C2H4/Ar after
150 pulses at 30 kHz repetition rate. Inset: Percent initial fuel conversion to C2H2 as a
function of initial fuel mole fraction.
is due to the decrease of excited Ar collisions with C2H4, which is one of the major pathways of
acetylene formation. This is also the cause of the reduction in measured C2H2 concentrations in
the 100% C2H4 experiment as compared to the 75% C2H4 experiment. The dip in C2H2
concentration at the end of the pulse burst in the high fuel loading experiments (>6.25% C2H4) is
due to residual radical reactions with C2H2 occurring once the plasma discharge is terminated.
Figure 5.2.3 shows the comparison of model predictions and the measured time histories of
C2H2 and CH4. HP-Mech considerably under-predicts the C2H2 concentration, particularly early
0
500
1000
1500
-5 -4 -3 -2 -1 0 1 2 3
Mo
le F
rac
tio
n C
2H
2(p
pm
)
Time from last pulse (ms)
1% C2H4 2.5% C2H46.25% C2H4 12.5% C2H425% C2H4 50% C2H475% C2H4 100% C2H4
0%
1%
2%
3%
4%
5%
0.01 0.1 1[C2H
2]/
[C2H
4]
Fuel Mole Fraction
168
in the burst of pulses. Moreover, the CH4 concentration is over-estimated by a factor of two.
Compared to HP-Mech, USC-Mech II displays the opposite trends, over-predicting the acetylene
concentration while greatly under-predicting the methane measurements by a factor of 10. Figure
5.2.4 shows the path fluxes as predicted by HP-Mech for three different fuel mole fractions,
integrated over the first 10 ms after the initial plasma pulse. It is seen that there are two major
pathways for C2H2 formation, one from electron impact dissociation of the fuel to form either
C2H3 followed by β-scission, or direct dissociation to C2H2, and the other from excited and
ionized argon collisional dissociation of fuel to C2H2, C2H3+, and C2H2
+. At high fuel
concentration, the former is dominant, but at a low fuel concentration the latter is dominant.
These two reaction pathways explain the results in Fig. 5.2.2 of decreased relative C2H2 yield
because the Ar collision pathway is terminated at high C2H4 concentration (low Ar
concentration). As such, the failure of the model to predict the acetylene measurements in Fig.
5.2.3 is largely due to the uncertainty of excited and ionized argon reactions with the fuel, as the
second reaction pathway accounts for up to 31% of fuel consumption for the 6.25% fuel loading
condition of Fig. 5.2.3.
Methane is formed by CH3 recombination with H, C2H3, and C2H5. The large discrepancy in
model prediction of CH4 also suggests the inaccuracy of fuel dissociation by excited and ionized
Ar, for these reactions have a large effect on radical formation. Moreover, these radicals are the
building blocks for larger hydrocarbons such as butene and propene, indicating that
recombination of these radicals must be included in future PAC models. Different from previous
reports of model validation using global properties such as ignition time, the present in situ
diagnostics of intermediate species clearly indicates that the existing plasma chemistry fails to
reproduce intermediate species.
169
Figure 5.2.3. Measurements and modeling of C2H2 and CH4 after 150 pulses at 30 kHz
repetition rate for a mixture of 6.25/93.75 C2H4/Ar.
Figure 5.2.4. Path flux analysis for fuel dissociation.
0.1
1
10
100
1000
10000
-5 -4 -3 -2 -1 0 1 2 3 4 5
Mo
le F
rac
tio
n (
pp
m)
Time from last pulse (ms)
C2H2, Experiment C2H2, HP Mech C2H2, USC Mech II
CH4, Experiment CH4, HP Mech CH4, USC Mech II
C2H4
C2H5
+ A
r* 2
3%
, 10
%, 0
%
C2H2
C2H3+
+ e 65%C2H
C2H2+
+ Ar+ 8%, 4%, 0%
+ X
10
0%
+ e-3
0%
C2 CH
C2H3+ H 43%, 6%, 0%
+ X
57
%, 9
5%
, 1
00
%
C2H4% = 1%, 6.25%, 100%X = RadicalM = Third body collider
Red = Combustion chemistry Blue = Plasma chemistry
+ H2 100%
+ e
<1%
, <1
%,
15
%
CH4
CH3 + H + MCH3 + C2H3
40
%, 9
5%
, 10
0%
CH3 + C2H5
170
5.2.2 Ethylene oxidation
Figure 5.2.5 shows the results of C2H4 oxidation by GC-TCD sampling and LAS in a steady
state plasma discharge at different repetition frequencies. The mixture is stoichiometric with a
75% argon dilution ratio. In these experiments, more than 92% of the carbon is recovered from
the measurements. Note that the oxidation products are very different from those of the
dissociation case. Propene and butene are no longer detectable, and the highest concentration
species is now formaldehyde. Considerable CO and acetaldehyde are also formed. The presence
of these species suggests the existence of low temperature oxidation pathways activated by the
plasma below 500 K.
Figure 5.2.6 shows the predicted and time-resolved measurements of temperature and
concentrations of C2H2, CH4, and H2O for C2H4 plasma-assisted oxidation. The comparison
between the model predictions and experimental results shows that both models greatly over-
predict the water concentration, especially USC-Mech II. Once again, C2H2 is under-predicted by
HP Mech and slightly over-predicted by USC-Mech II, while methane is over-predicted by HP
Mech and under-predicted by USC-Mech II. The large discrepancy between the model
predictions and the experimental data indicates that the PAC kinetic models fail to predict the
low temperature oxidation or are even missing important reaction pathways.
Both models over-predict the temperature after the discharge, which is due both to the over-
prediction of water concentration and the difficulty in modeling energy input channels for the
plasma discharge. Judging by the measured water concentration and the lack of carbon dioxide
in the steady state measurements, there is little heat release due to chemical reaction. Therefore,
the measured temperature increase is primarily a result of gas heating by the plasma. The total
energy added is approximately 225 mJ, which would result in a temperature rise of 102 K if the
171
Figure 5.2.5. Major species concentrations during continuous plasma discharge measured
by GC-TCD and laser absorption spectroscopy (open symbols) in a mixture of
6.25/18.75/75 C2H4/O2/Ar at 60 Torr.
Figure 5.2.6. Measurements and modeling of C2H2, CH4, H2O, and temperature after 150
pulses at 30 kHz repetition rate for a mixture of 6.25/18.75/75 C2H4/O2/Ar.
1
10
100
1000
10000
100000
1000000
0 1000 2000 3000 4000 5000
Mo
le F
rac
tio
n (
pp
m)
Plasma Frequency (Hz)
O2 C2H4CO CH2OC2H2 C2H2 LASCH4 CH4 LASC2H4O
0
100
200
300
400
500
1
10
100
1000
10000
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Te
mp
era
ture
(K
)
Mo
le F
rac
tio
n (
pp
m)
Time from last pulse (ms)
C2H2, Exp. C2H2, HP C2H2, USCCH4, Exp. CH4,HP CH4, USCH2O, Exp. H2O, HP H2O, USCT, Exp. T, HP T, USC
172
energy goes entirely to gas heating. Therefore, the measured temperature is in good agreement
with the measured energy input. It should be noted that after the plasma burst is terminated, the
mixture does not proceed to equilibrium concentrations due to the low temperature, resulting in
chemically frozen flow.
The path flux analysis of fuel oxidation integrated over the 10 ms after the initial plasma pulse
is shown in Fig. 5.2.6. There are three different major fuel consumption pathways. The first
pathway is direct collisional dissociation by excited and ionized Ar to form hydrocarbon
fragments. The second pathway is fuel oxidation by plasma generated O atoms. The last and
most interesting pathway is plasma activated low temperature oxidation involving O2 addition to
the fuel radicals, leading to C2H5O2 (ethyl peroxyl radical), C2H5O2H (ethyl hydroperoxide), or
O2C2H4OH and subsequent dissociation into formaldehyde and hydroxyl radicals. The ethyl
peroxide path is initiated by hydrogen addition to ethylene to form ethyl radical, followed by
oxygen addition to form ethyl peroxyl radical, about half of which is predicted to be collisionally
stabilized and the rest of which is converted to ethyl hydroperoxide by reaction with
hydroperoxy radicals (Miller et al. 2001, DeSain et al. 2003, Metcalfe et al. 2013).
Some of the ethyl peroxyl radicals are also predicted to react with ethylene to form
acetaldehyde. These pathways are chain terminating since none of these species are active
radicals at temperatures below 500 K. The other peroxy radical, O2C2H4OH, is formed via
hydroxyl addition to ethylene to form collisionally stabilized 2-hydroxyethyl radical (C2H4OH),
with subsequent O2 addition to arrive at O2C2H4OH. This species decomposes via a six-
membered ring formation to give two formaldehyde molecules and an OH so is thus chain
propagating. This is known as the Waddington sequence, originally proposed to describe propene
oxidation in the atmosphere, and is the dominant pathway for OH attack on the fuel below ≈800
173
K (Ray et al. 1973, Senosiain et al. 2006, Metcalfe et al. 2013). This low temperature oxidation
channel is mainly responsible for the formation of large amounts of formaldehyde observed in
Fig. 5.2.5. Note that the OH radical addition pathway is not included in most combustion
models built primarily for high temperatures. In USC-Mech II this reaction is modeled as an
abstraction reaction leading to C2H3 and H2O, explaining the drastic over-prediction of H2O by
this model.
Figure 5.2.7. Path flux analysis for fuel oxidation.
For larger hydrocarbons, the presence of oxygen addition pathways would lead to low
temperature chain branching and a source of radicals to reinitialize the reactions. The reason this
normally slow chemistry is possible for ethylene on the time scale of the plasma discharge is the
presence of H and O atoms. Atomic hydrogen is mainly formed by dissociation of the fuel by
electrons and excited Ar, the latter of which bears significant uncertainty. The branching ratio for
10
0%
C2H4
C2H5
C2H2
+ Ar+ 12%, 2%C2H3+
+ e
-3
0%
CH2CH2OH+ OH 33%, 22%
CH3+ HCOH + CH2CHO
+ e- 65%C2H
+ H 45%, 25%
C2H5O2
+ O2 + M96%
C2H5O2H
+ HO2
98%
HCO + CO
+ O
2 1
00
%
O2C2H4OH
+ O
21
00
%
2 CH2O + OH
R% = 4%, 25%Phi=1, 1X = RadicalM = Third body colliderRed = Combustion chemistryBlue = Plasma chemistry
O
e- + O2 Ar*+ O2O*+ O2
H
e- + C2H4
Ar*+ C2H4Ar+ + C2H4
O + C2H4O2 + C2H3
OH
CH3O2 + HHO2 + XCH2CHO+ O2
CH2 + O2O2C2H4OH
174
this reaction pathway was estimated to be similar to methane, ethane, and larger hydrocarbons
(Balamuta et al. 1983, Kosarev et al. 2009). Most of the atomic oxygen is produced by direct
impact dissociation of oxygen by electrons and excited Ar.
5.3 Conclusions
Quantitative in situ measurements of intermediate species in plasma-assisted C2H4/Ar
dissociation and C2H4/O2/Ar oxidation have been accomplished using mid-IR LAS coupled with
a mini-Herriot cell. Validation by GC-TCD sampling in a continuously pulsed plasma discharge
demonstrated the accuracy and robustness of the present mid-IR LAS diagnostic for the
quantification of H2O, C2H2 and CH4. The dissociation experiments showed that two acetylene
formation pathways exist, one by electron impact dissociation and the other by excited and
ionized argon reactions. The relative acetylene yield is governed by the former reaction at high
fuel loading, while the latter reaction dominates at low fuel loading conditions. It is also found
that a considerable amount of propene and butene are formed due to radical recombination,
suggesting that larger hydrocarbon kinetics need to be included to predict plasma activated fuel
dissociation of small hydrocarbon fuels. Plasma activated C2H4 oxidation experiments result in
completely different products from those of the dissociation experiment, producing much less
acetylene and negligible large hydrocarbons. Large amounts of formaldehyde and acetaldehyde
suggest that there exists a low temperature fuel oxidation pathway via oxygen addition reactions.
A new kinetic model (HP-Mech) for low temperature plasma activated fuel oxidation and
dissociation is assembled and assessed by comparison with experimental data and the results of
USC-Mech II. The new model shows that there exist three different fuel oxidation and
consumption pathways for low temperature plasma oxidation. The first pathway is the direct
175
collisional dissociation by excited and ionized Ar to form hydrocarbon fragments; the second
pathway is the direct oxidation by plasma generated radicals such as O; and the last is from
plasma activated low temperature oxidation via RO2 (C2H5O2 and O2C2H4OH) formation and
subsequent dissociation leading to formaldehyde and acetaldehyde. Both models failed to
quantitatively predict H2O, C2H2 and CH4 time histories, although HP-Mech has slightly better
performance. Large discrepancies in H2O and CH4 predictions suggest that there are large
uncertainties in excited and ionized argon dissociation reactions and the low temperature fuel
oxidation reactions. The present experiments provided in situ diagnostic data of intermediate
hydrocarbon species in low temperature plasma activated fuel dissociation and oxidation for the
first time. More quantitative in situ diagnostics of intermediate species including OH and CH2O
are needed to provide validation targets for kinetic models.
176
6 Kinetic study of methane oxidation in a NRP DBD flow
reactor
The results presented in this chapter were published in:
J. K. Lefkowitz, P. Guo, A. Rousso, Y. Ju (2015). "Species and temperature measurements of
methane oxidation in a nanosecond repetitively pulsed discharge." Phil. Trans. R. Soc. A
373(2048): 20140333.
I also presented the results in the following conferences:
J. K. Lefkowitz, P. Guo, A. Rousso, Y. Ju, “Low temperature oxidation of methane in a
nanosecond pulsed plasma discharge,” 53rd
AIAA Aerospace Sciences Meeting, Kissimmee,
Florida (2015) AIAA paper 2015-0665.
J. K. Lefkowitz, A. Rousso, P. Guo, Y. Ju, “A kinetic study of low temperature methane
oxidation in a nanosecond repetitively pulsed discharge,” 9th
U.S. National Combustion Meeting,
Cincinnati, OH (2015) paper 3B03.
J. K. Lefkowitz, “Kinetic study of low temperature methane oxidation in a nanosecond
repetitively pulsed dielectric barrier discharge”, 22nd
International Symposium on Plasma
Chemistry, Antwerp, Belguim (2015) presentation O-4-3.
I was responsible for data collection and processing, assembly of the kinetic model, analysis of
experimental and numerical results, and writing of the paper mentioned above. Peng Guo was
repsonsible for assembly of the numerical modelling software. Aric Rousso assisted in data
collection. The initial concept and guidance for the research was provided by Prof. Yiguang Ju.
177
6.1 Introduction
Our previous study of ethylene PAC in a room temperature flow reactor, discussed in Chapter
5, demonstrated the strong dependence of species concentrations to low temperature reaction
pathways, which proceeded primarily through O, OH, and O2 addition reactions. Recent
experiments of plasma activated cool flames of dimethyl ether and n-heptane further
demonstrated the importance of plasma activated low temperature chemistry in kinetic
combustion enhancement by plasma (Sun et al. 2014, Won et al. 2015a). It was found that the
reactions involving low temperature fuel oxidation are critical for predicting the experimental
measurements of plasma assisted combustion. Therefore, in order to extend our understanding of
fuel oxidation in low temperature PAC, and to resolve the issues observed in our previous
studies, we focus our study to the oxidation kinetics of the simplest hydrocarbon fuel, methane,
and to obtain time-resolved species time history in plasma discharge.
Previous studies of methane oxidation in plasma have been conducted in various experimental
and numerical platforms, including shock tubes (Kosarev et al. 2008a, Aleksandrov et al. 2009a,
Starikovskaya et al. 2009), counterflow flames (Sun et al. 2012, Sun et al. 2013), and flow
reactors (Lou et al. 2007, Uddi et al. 2009a, Uddi et al. 2009b, Yin et al. 2013b). In addition,
there have been extensive studies of plasma fuel reforming using methane (Suib et al. 1993,
Lesueur et al. 1994, Okumoto et al. 1998, Bromberga et al. 1999, Lee et al. 2007, Jasiński et al.
2008, Hwang et al. 2010, Zhang et al. 2013). Plasma reforming of methane has shown the
potential to efficiently produce hydrogen and/or syngas (Lesueur et al. 1994, Bromberga et al.
1999, Lee et al. 2007, Jasiński et al. 2008, Hwang et al. 2010), methanol (Okumoto et al. 1998),
and higher hydrocarbons (Suib et al. 1993). Recent studies of dry reforming and partial oxidation
of methane in a heated flow reactor with a DBD driven by an AC voltage waveform (Zhang et al.
178
2013, Zhang et al. 2015) revealed several interesting results. First, fuel reforming was
significantly more efficient for the oxidation case compared to the dry reforming case. Second,
while overall CH4 and O2 conversion is increased with initial temperature, reduced electric field
(E/N), and plasma power (i.e. frequency of the AC field), only the initial temperature could
affect the product selectivity. Modeling of the combustion chemistry (assuming some initial
radical concentrations from the plasma) indicated that increased temperatures increased the rate
of H-abstraction from CH4 by O, OH, and H radicals, which lead to subsequent reactions of fuel
intermediates with O2, increasing the overall fuel and oxidizer conversion. However, no
quantitative comparison with the measured species concentrations could be made due to the lack
of electron collision processes in the model, which has generally limited the analysis of plasma-
based fuel conversion systems. Some recent work in PAC has made steps toward more
quantitative modeling of the oxidation process.
In shock tubes, Kosarev et al. (Kosarev et al. 2008a) measured ignition delays after a fast
ionization wave (FIW) in CH4/O2/Ar mixtures at initial temperatures from 1230 – 1719 K and
pressures from 0.3 – 1.1 bar, finding that the ignition delay can be shortened by a factor of 30
using a nanosecond pulsed plasma, an enhancement that could not be replicated by assuming all
plasma energy went to gas heating. Numerical modeling of the ignition process including
separate consideration of the electron collision processes and combustion kinetics, revealed that
the plasma produced species were mostly quenched to form neutral radicals, particularly O atom,
which then reacted with CH4 and its intermediates, resulting in the initiation of chain branching
chemistry. However, no comparison between the actual species produced and the model
prediction were made in this study.
179
Lou et al. (Lou et al. 2007) studied oxidation of methane-air mixtures in a nanosecond
repetitively pulsed (NRP) discharge in a homogenous flow reactor at initial temperature 290 K
and pressure <100 Torr. For continuously pulsed plasma, it was found that up to 70% of the fuel
could be consumed, producing approximately 500 K temperature rise, without any flame. This
indicated that PAC could control fuel oxidation continuously, without any clear transition
between ignition and frozen flow. Further studies in a similar reactor (Uddi et al. 2009a, Uddi et
al. 2009b, Yin et al. 2013b) measured temperature, NO, O, and OH. In comparison to air
plasmas, the addition of methane did not affect the NO concentration profile or the peak O
concentration. There was, however, a decrease in the peak OH concentration, and there was a
significant increase in the O and OH consumption rates after the plasma, indicating reactions
were indeed occurring between methane and/or its intermediates with radicals formed in the
plasma. Modeling of the PAC processes could predict temperature and O atom concentration
simultaneously, but different combustion kinetic models had mixed results predicting the OH
profile, indicating the sensitivity of combustion kinetics at the relatively low temperatures in
these experiments (< 600 K).
In counterflow flames, Sun et al. (Sun et al. 2012) found that the extinction limits of a
partially premixed CH4/O2/Ar mixture could be extended by more than a factor of two when a
nanosecond pulsed discharge is applied to the mixture just before the exit of the fuel nozzle. The
main cause was found to be prompt reforming of the fuel stream to a mixture of heated H2,
CH2O, CO, CO2, and H2O, which was initiated by electron dissociation reactions in the plasma.
Modeling efforts found that O and CH2O concentrations were well predicted, but the
concentration of H2, CO, CO2, and H2O could not be matched by the model predictions. Sun et
al. (Sun et al. 2013) later integrated the plasma discharge directly into the counterflow burner
180
such that the plasma is stabilized between the burner nozzles. In this experiment, it was found
that the hysteresis between ignition and extinction could be smoothed into a continuous
transition. The reason for this change was again the activation of methane oxidation by radical-
produced O atoms, which initiated chain branching reactions at temperatures below the cross-
over temperature for hot ignition, allowing chemical heat extraction at methane concentrations
too small to otherwise sustain a flame.
To summarize: it has been found that a continuous transition from non-reacting to fully
reacted methane/oxygen/diluents mixtures can be achieved by controlling the plasma parameters
and mixture composition. The cause of this continuous transition is electron collision reactions
producing a mixture of electronically excited species, ions, and atoms, all of which quickly
quench to form mostly O atoms, but also H atoms and OH radicals. These radicals will react with
the fuel and fuel intermediates, and, depending on the temperature, can lead to ignition of the
mixture or to partial oxidation. While the power and E/N of the plasma affects the overall
amount of fuel conversion, the product species composition is largely controlled by the chemical
kinetics of combustion reactions at low temperature (300-700 K). However, a quantitative study
of simultaneous temperature and multi-species concentrations in low temperature methane PAC
is still lacking, and the model predictability in terms of speciation data is limited, preventing a
full assessment of the relevant reaction pathways.
The purpose of this chapter is to quantitatively measure time-resolved species production and
temperature histories in a homogeneous plasma discharge, and compare these results with kinetic
modeling, focusing on the low temperature initiation reactions. To accomplish this, the flow
reactor design, described in Chapter 3 is utilized. In situ and time-dependent measurements of
temperature and CH2O are performed using laser absorption spectroscopy (LAS) in a mini-
181
Herriot cell during a 300 pulse burst. CH2O was identified to be a major intermediate species
directly produced by plasma-generated O atoms (Kosarev et al. 2008a, Sun et al. 2012, Sun et al.
2013) so is thus an ideal candidate to determine the fidelity of the initial oxidation reactions. In
addition, temperature and species concentration were measured for a continuous discharge in
order to gain a more complete understanding of the kinetic processes involved. These results are
compared to modeling of the coupled plasma and chemical kinetic processes.
6.2 Results and discussion
6.2.1 Transient measurements
In order to model the discharge process, a reduced electric field (E/N) profile as a function of
time and an initial electron number density must be supplied as inputs to the ZDPlasKin part of
the solver. Because the 1-D solution for charge separation in an electric field is not solved, it is
not possible to calculate the reduced electric field a priori from the applied voltage waveform.
Thus, in order to provide the E/N, a square wave was assumed, and the peak E/N and discharge
duration were adjusted to bring the temperature prediction into approximate agreement with the
measurement, which resulted in an E/N of 180 Td and a discharge time of 3-4 ns, which needed
to be adjusted to keep total energy addition per pulse constant as gas temperature increased. In
the previous chapter, the concentration of a major intermediate species was used to fit the E/N
because this species was only sensitive to the reaction rate of electron dissociative ionization of
the fuel. In this case, temperature is used as the fitting parameter for the E/N due to the
sensitivity of heat addition on electron collision processes, and the importance of temperature on
the reaction rates of most combustion reactions. In order to confirm that the speciation results are
not sensitive to the choice of E/N, values of 160 – 200 Td were tested while keeping energy
182
deposition per pulse equal to the 180 Td condition by varying the discharge duration. This
resulted in <5% variation in the gas temperature, and fuel, oxidizer, and product species
concentrations. Therefore, while the E/N is not accurately known, this parameter is not found to
significantly change the results in a range of ±20 Td from the chosen value.
The measurement of temperature by two-line absorption in a 300 pulse burst discharge at 30
kHz repetition frequency is presented in Figure 6.2.1. The heat loss rate is fit using a modified
L/π value of 0.0064 m in Eq. 4.2.11 (instead of the calculated 0.0044 m) and a wall temperature
of 300 K. The temperature increases steadily throughout the pulse burst until it peaks on the last
pulse (at 10 ms), at which point heat losses take over and return the gas back to its original
temperature. In our previous study of stoichiometric C2H4/O2/Ar with 75% dilution, heat loss
could be neglected on the time scale of the experiment. However, there is a large effect of heat
loss when using He as the diluent. At 1 atm and 293.15 K, the thermal conductivity (λ in Eq.
4.2.11) of He is 0.1535 W m-1
K-1, while λ = 0.01738 W m
-1 K
-1 for Ar. Therefore, according to
Eq. 4.2.11, the heat loss will be an order of magnitude faster with He as the diluent so must be
considered for accurate model predictions.
Figure 6.2.2 presents the mole fraction of formaldehyde as a function of time at the same
conditions of Fig. 6.2.1. The model under-predicts the peak concentration of formaldehyde by a
factor of 5, indicating major limitation in the model’s predictive ability of this primary
intermediate. The primary formation pathway reported for formaldehyde in methane PAC
(Kosarev et al. 2008a, Sun et al. 2011) is:
CH3 + O → CH2O + H (M98)
183
therefore, formaldehyde is formed directly from two species produced in the plasma, since O
atom is produced in reactions M5-M7 and methyl radical is produced in M8. This may indicate
that the electron collision rates are not properly modeled, or that the rate of M98 is inaccurate.
To fully understand the reason for the disagreement between the model and the measurement,
additional species information is required.
Figure 6.2.1: Temperature measurements and model predictions during and after a 300
pulse burst at 30 kHz repetition rate and 8.76 kV peak voltage in a stoichiometric
CH4/O2/He mixture with 75% dilution.
290
300
310
320
330
340
350
360
0 5 10 15 20
Te
mp
era
ture
(K
)
Time (ms)
Model
Experiment
184
Figure 6.2.2: Formaldehyde measurements and model predictions during and after a 300
pulse burst at 30 kHz repetition rate and 8.76 kV peak voltage in a stoichiometric
CH4/O2/He mixture with 75% dilution.
6.2.2 Steady-state measurements
To collect a more complete set of product species, the plasma is run in continuous mode and
gas sampling is performed downstream for multiple species quantification using GC-TCD, as
well as temperature measurements by two-line absorption. Because of the fast thermal
conductivity rate of helium, the gas temperature will quickly reach a steady state in which the
heat loss to the walls balances the heat addition by the plasma. Figure 6.2.3 presents the gas
temperature during steady state NRP discharges at frequencies from 100 – 30,000 Hz. At lower
0
50
100
150
200
250
0 5 10 15 20
Mo
le F
rac
tio
n (
pp
m)
Time (ms)
CH2O Model
CH2O Experiment
185
Figure 6.2.3: Temperature measurements and model predictions in a continuous plasma at
30 kHz repetition rate and 8.76 kV peak voltage in a stoichiometric CH4/O2/He mixture
with 75% dilution.
frequencies (<10,000 Hz) the model and the measurement are in reasonable agreement.
However, at higher frequencies the model and the measurement diverge such that at 30,000 Hz
there is a 50 K differential between the measurement and the model. This is due to wall heating
in the continuous plasma, which decreases the heat loss rate and allows the steady state
temperature to rise higher than in the pulse burst case. In the model, the wall temperature is
assumed to be fixed at 300 K, and, in the absence of wall temperature measurements during the
discharge, the heat loss cannot be accurately predicted. Nevertheless, the overall trend is well
represented by the model.
275
300
325
350
375
400
425
450
475
500
100 1000 10000 100000
Te
mp
era
ture
(K
)
Frequency (Hz)
Temperature Model
Temperature Experiment
186
Figure 6.2.4 presents measured and predicted species concentration as a function of NRP
frequency. Figure 6.2.4a presents the concentrations of the reactants and water, the largest
product species. It is found that the fuel and oxygen consumption are predicted to within 5% of
the measured value, which is smaller than the experimental uncertainty. Since the consumption
of the reactants is almost entirely due to electron collision reactions and quenching of excited
species (as will be discussed in the following sections), we can assume that the total electron
collision rates are well modeled. The production of water is predicted to within 20% of the
measured value, which is in excellent agreement considering the order of magnitude
disagreement for water in our previous study of ethylene oxidation (Lefkowitz et al. 2015d).
Figure 6.2.4b presents the other major products: carbon monoxide, carbon dioxide, and
hydrogen. The model captures the correct trends and relative concentrations of the three species,
but under-predicts the absolute concentrations, particularly at the highest frequency conditions.
Nevertheless, carbon monoxide agreement is within 30% of the measured value, while carbon
dioxide and hydrogen are within 40%. The minor species are plotted in Figure 6.2.4c. Agreement
is comparatively poor between the model and measurements of formaldehyde, methanol, ethane,
ethylene, and acetylene. Similar to the time-dependent results, formaldehyde is under-predicted
by approximately a factor of five, while methanol is over-predicted by an order of magnitude.
The remaining species only appear in relatively small concentrations in the experiment, and are
all under-predicted by the model. In summary, the major trends of reactant consumption and
major product species production are well captured by the model, indicating that the electron
collision rates and dominant reaction pathways are well modeled, but the minor species modeling
results in significant disagreement, indicating that perhaps some secondary rates need further
attention.
187
Figure 6.2.4: Species measurements and model predictions in a continuous plasma at 30
kHz repetition rate and 8.76 kV peak voltage in a stoichiometric CH4/O2/He mixture with
75% dilution.
0
25
50
75
100
125
150
175
200
100 1000 10000 100000
Mo
le F
rac
tio
n (
pp
m)
x 1
00
0
Frequency (Hz)
O2 Experiment O2 Model
CH4 Experiment CH4 Model
H2O Experiment H2O Model
a
0
2000
4000
6000
8000
10000
12000
14000
100 1000 10000 100000
Mo
le F
rac
tio
n (
pp
m)
Frequency (Hz)
CO Experiment
CO Model
CO2 Experiment
CO2 Model
H2 Experiment
H2 Model
b
0
200
400
600
800
1000
100 1000 10000 100000
Mo
le F
rac
tio
n (
pp
m)
Frequency (Hz)
CH2O Experiment
CH2O Model
CH3OH Experiment
CH3OH Model
C2H2 Experiment
C2H2 Model
C2H4 Experiment
C2H4 Model
C2H6 Experiment
C2H6 Model
c
188
In addition to the major species measured in this study, the concentrations of plasma-
produced radical, excited, and ionized species are of central importance in PAC kinetics. The
model predictions for these species are presented in Fig. 6.2.5 for a single discharge pulse and
inter-pulse period for steady state 30,000 Hz operation (plot shown is for 3000th pulse). While
most excited species and ions are quenched during the inter-pulse time, some longer lifetime
species survive in appreciable concentrations, such as ,
, and
. The
reactions of these species with methane remains unclear (Kosarev et al. 2008a, Ombrello et al.
2010b) and may quench faster than predicted in the model. In addition, the reactions of
with helium is an estimate from (Stafford et al. 2004) as no measurements for this
rate could be found in the literature. The ground state radicals generated in the plasma react
slowly with methane, so these radicals build up to a steady state concentration. At higher gas
temperatures, the large radical concentration would lead to chain branching reactions and
ignition. However, methane at temperature below ≈ 800 K can only react in chain propagating
reactions, as will be discussed in the following section.
To understand which reactions are of importance, a path flux analysis is performed. Figure
6.2.6 presents the consumption pathways of methane for 30,000 Hz continuously pulsed plasma
at steady state temperature as predicted by the model (407 K). The major fuel consumption
pathways are through electron collision reactions, reactions with O(1D), and reaction with OH.
Electron collision dissociative excitation reactions (M8-M11) account for 16% of fuel
consumption, while ionization (M14) and dissociative ionization (M15) reactions account for
20%. The dissociative excitation reactions lead to methyl radical (CH3) and methylene radical
(CH2), while the dissociative ionization reaction leads to methyl cation (CH3+). Charge exchange
189
Figure 6.2.5: Model predictions of the concentration of radical, excited, and ionized species
in a continuous plasma at 30 kHz repetition rate and 8.76 kV peak voltage in a
stoichiometric CH4/O2/He mixture with 75% dilution.
1E+7
1E+8
1E+9
1E+10
1E+11
1E+12
1E+13
1E+14
1E+15
1E+16
1E-10 1E-9 1E-8 1E-7 1E-6 1E-5 1E-4
Co
nc
en
tra
tio
n (
cm
-3)
Time (s)
O
H
OH
O2(a1)
O2(b1)
O2(4.5eV)
O(1D)
O(1S)
He*
E
1E+7
1E+8
1E+9
1E+10
1E+11
1E+12
1E+13
1E+14
1E+15
1E+16
1E-10 1E-9 1E-8 1E-7 1E-6 1E-5 1E-4
Co
nc
en
tra
tio
n (
cm
-3)
Time (s)
CH4+
CH3+
CH2+
O2+
O+
He+
E
190
Figure 6.2.6: Path flux analysis of formaldehyde and methanol formation integrated over a
single pulse period during continuous discharge at 30 kHz repetition frequency and steady
state temperature conditions. Bold species represent those which are measured in Figure
6.2.4, red arrows refer to reactions from the combustion model, and blue arrows are from
the plasma model. R represents any radical species.
reactions for methyl cation found in the literature lead to more complex ions, the chemistry of
which was ignored in the current model. Therefore, the only consumption pathway for methyl
cation is dissociative recombination with electrons leading to methylene and hydrogen atom
(M75). In total, 10% of fuel consumption leads to methylene radical production. The methylene
radical is completely consumed by reaction with oxygen:
CH2 + O2 → CO + OH + H (M99)
→ CO2 + H + H (M100)
→ CO2 + H2 (M101)
CH4
CH3 + OH
+ e- 14%CH3 + H2O
+ OH 11%
CH3+ + H
CH2OH + H
CH3 + H
CH4+
+ O
2+
M 9
3%
CH3O2 + M
+ O 5%CH2O+ H
CH3O+ O2/OH CH3OH+ O2
+ e
-1
00
%
CH2 + H/H2
+ O
21
00
%
CH4 + O2+
+ O
21
00
%
CO + OH + H
CO2 + H + H
CH2O + O
CO2 + H2
+ O
21
00
%
CH2O+ H2/HO2
CH2O+ HO2 CH3OH+ CH2O/HCO+
O(1
D)
2%
+ O
27
%
191
→ CH2O + O (M102)
resulting in the formation of product species. Reaction M99 accounts for 47% of carbon
monoxide formation, and reactions M100 and M101 together account for 99% of carbon dioxide
formation. Reaction M101 accounts for 19% of hydrogen formation, while dissociative
excitation of methane into methylene radical (M9-M11) accounts for another 35% of hydrogen
production, accounting for a total of 54% of the hydrogen formation. Since all three species are
reasonably well predicted by the model, we can conclude that the reaction rates and branching
ratios in the CH4 + e → CH2 + H2 and CH2 + O2→ CO, CO2, H2 pathways are well modeled.
The other primary intermediate species is methyl radical, formed mainly by:
CH4 + e → CH3 + H + e (M8)
CH4 + O(1D) → CH3 + OH (M36)
CH4 + OH → CH3 + H2O (M103)
which together account for a total of 65% of fuel consumption. Considering that the model and
measurements of methane concentration are in good agreement, it is unlikely the total initial
consumption rates are significantly incorrect. In addition, M103 is responsible for 35% of water
formation, which is well predicted by the model. Methyl radical is consumed via:
CH3 + O → CH2O + H (M104)
CH3 + O2 +M → CH3O2 + M (M105)
The competition between these two rates determines the model prediction of methanol, as M104
leads to methyl peroxy radical (CH3O2), which reacts via:
192
CH3O2 + OH → CH3OH + O2 (M106)
CH3O2 + H → CH3O + OH (M107)
CH3O2 + O → CH3O + O2 (M108)
Reaction M106 is responsible for 82% of methanol formation, while further methoxy radical
(CH3O) reactions results in the remaining 18% of methanol formation, as presented in Figure
6.2.7. The methyl sub-model also plays a large role in formaldehyde production, as M104
accounts for 16% of formaldehyde formation, and methoxy radical reactions account for another
19%. An increase in the rate of M104 would increase formaldehyde formation while decreasing
methanol formation. However, both M104 and M105 have received considerable attention in the
literature (Baulch 2005, Fernandes et al. 2006), for they are major pathways in methane
combustion. Therefore, they are not expected to have significant uncertainty that could account
for the over-prediction of methanol and under-prediction of formaldehyde. Another possibility is
that there are other reactions participating in methyl radical oxidation that are not in the model,
for example:
CH3 + O(1D) → CH2O + H (M104´)
At 400 K, there is a difference of five orders of magnitude between the H-abstraction rate
from methane by O atom or by O(1D). If there is also such a difference between M104 and
M104´, it could change the amount of methyl radical going towards the formation of methanol as
compared to that going toward formaldehyde. However, no rates for this reaction could be found
in the literature. Reaction M104 has a rate of 1.12 × 10-10
cm3/s, so, in analogy with the five
order of magnitude difference in methane H-abstraction rates by O atom and O(1D), M104´ was
given a rate of 1.12 × 10-5
cm3/s. However, this only increased the CH2O concentration by 27%
193
and decreased the CH3OH concentration by 42%, which is not the order of magnitude change
needed to bring the model into quantitative agreement with the measurement. However, it did
improve model accuracy significantly, and deserves further exploration.
Figure 6.2.7: Path flux analysis of formaldehyde and methanol formation integrated over a
single pulse period during continuous discharge at 30 kHz repetition frequency and steady
state temperature conditions. Bold species represent those which are measured in Figure
6.2.4, red arrows refer to reactions from the combustion model, and blue arrows are from
the plasma model. R represents any radical species.
Another explanation for the disagreement may lie in the methyl peroxy submodel.
Consumption of methyl peroxy radical has not received nearly the same degree of attention as
the initial methyl radical reactions. Reaction M106 has only been estimated (Tsang et al. 1986),
while M107 and M108 were each measured once (Zellner et al. 1988). In the model, all three
rates are those suggested in (Tsang et al. 1986). If M106 is decreased, the reaction flux shifts
CH2OO(1D) + CH4
11%CH2 + O2
CH3 + OCH2OH + O2
8%
CH3O + R/O2
19
%
CH3OH
CH3O + R/RH CH3O2+ OH
194
towards the production of methoxy radical, which will increase the relative concentration of
formaldehyde as compared to methanol. The model was recomputed with a factor of 5 decrease
in M106, but the relative concentrations of formaldehyde to methanol was minimally affected
due to the fact that most of the methoxy is recycled to methyl radical, which loops back to the
production of methanol (see Fig. 6.2.6). Therefore, changing M106 has a non-monotonic effect
on the formaldehyde to methanol ratio. In the original paper measuring M107 by Šlemr and
Warneck (Slemr et al. 1977), the reaction is suggested to proceed through an alternate reaction
pathway that results in the direct formation of formaldehyde:
CH3O2 + H → CH2O + H2O (M107´)
which is dominant over the channel recommended by Tsang et al. (Tsang et al. 1986). Tsang et
al. determined that surface reactions in the Šlemr and Warneck measurement were responsible to
the observed formaldehyde concentration so dismissed any contribution from M107´. However,
changing M107 to M107´ also has a minimal effect on the formaldehyde to methanol ratio, since
M108 is a factor of three faster than M107 and dominates the H-abstraction reaction pathway.
Further investigation into the reaction pathways of methyl peroxy radicals is needed to resolve
this issue. The methoxy sub-model may also be responsible for the disagreement in methanol
concentration. However, similar to the methyl radical oxidation rates, methoxy has received
considerable attention (Tsang et al. 1986, Baulch 2005, Xu et al. 2010), for it is one of the
primary intermediates in methanol and other alcohol oxidation.
Dominating the formation pathways for formaldehyde are the reactions of O(1D) with
methane (M36 – M38), which are the most significant consumption pathways, accounting for
195
53% of total fuel consumption. The total rate and branching ratios of this reaction are taken from
Sander et al. (Sander et al. 2006), in which available rates were reviewed and a recommended
rate and branching ratios were prescribed. Even a small change in the branching ratio of this
reaction shifts the relative concentrations of methanol and formaldehyde significantly. As
discussed in the original reference (Sander et al. 2006), there is still significant disagreement
among measurements of the branching ratio for this reaction. Further studies of O(1D) + CH4
could lead to improved model predictions in PAC of methane. The disagreement for
formaldehyde and methanol formation in the low temperature region of this study thus remains
an open topic, and requires a more in depth understanding of the kinetic pathways involved,
particularly for the methyl peroxy reactions and the reaction of methane and O(1D).
All of the ethane, ethylene, and acetylene formation proceeds via radical-radical
recombination reactions. As the radical pool composition cannot be measured in this study, it is
difficult to identify which reactions are responsible for the disagreement. One possibility is that
more electron collision reactions need to be considered for electron collision dissociation of the
intermediate and product species. However, as each of the C2 species are only present in mole
fractions of under 100 ppm, the failure of the model to accurately predict these species does not
constitute a major shortcoming, so no major consideration was taken to improve the calculated
under-prediction of these species.
6.3 Conclusions
Time-dependent and steady-state measurements of temperature and species of a plasma
discharge in a stoichiometric methane-oxygen mixture with 75% helium dilution have been
performed using LAS and GC-TCD sampling techniques. In addition, a PAC modeling tool for
196
homogenous plasma assisted combustion systems as well as a high pressure plasma kinetic
mechanism (HP-Mech/plasma) for CH4/O2/He oxidation have been assembled and used to
simulate the experimental measurements. To match the E/N in the model to approximate the
experimental value, the E/N was varied to fit the time-dependent temperature profile. In situ
time-dependent measurements of CH2O in the plasma found a factor of five disagreement
between the measurement and the modeled results, prompting the need of further investigation
into low temperature plasma assisted combustion pathways. For this purpose, temperature
measurements in the continuous plasma and species sampling downstream of the plasma region
were performed. The temperature measurements indicated limitations in heat loss prediction due
to wall heating by the continuous plasma, under-predicting the temperature by up to 50 K.
However, overall prediction of the heating trend was captured. Species measurements found H2O
to be the largest product species, followed by CO, CO2, H2, CH2O, CH3OH, C2H4, C2H2, and
finally C2H6. The model accurately predicts fuel and oxygen consumption, as well as the
production of H2O, CO, CO2, and H2. However, the intermediate species CH2O, CH3OH, C2H2,
C2H4, and C2H6 are not well-predicted by the model. CH3OH is over-predicted by an order of
magnitude, and CH2O is again under-predicted by a factor of five. Reaction pathway analysis
indicates that the reactions primarily responsible for H2O, CO, CO2, and H2 formation proceed
via electron dissociative excitation and dissociative ionization reactions leading to methylene
radical. The routes leading to CH2O and CH3OH formation are largely through electron
dissociative excitation and H-abstraction reactions leading to methyl radicals, as well as directly
through the reactions of methane with O(1D). It is found that the methyl radical sub-mechanism
proceeds primarily through a molecular oxygen addition pathway leading to methyl peroxy
radicals. The low temperature consumption pathways of methyl peroxy radicals are not well
197
known and may be the reason for the significant failure of the model in terms of formaldehyde
and methanol prediction. Further studies into the reactions of methyl peroxy are necessary to
bring the model into agreement with the present measurements. In addition, the branching ratio
of methane and O(1D) can be altered to improve agreement. The sensitivity of PAC kinetics to
low temperature hydrocarbon chemistry indicates the need for further evaluation of reaction
processes at temperature below 700 K participating in the negative temperature coefficient range
of hydrocarbon oxidation, even for the smallest hydrocarbon species. In addition, a closer
examination of underlying assumptions of the model, such as the neglect of vibrationally excited
states of oxygen and methane, use of a single excited state to represent all excited states of argon,
neglect of complex and negative ions, and neglect of electron-electron collisions is necessary to
ascertain if these, in fact, were proper assumptions that should be carried over to future modeling
efforts.
198
7 Summary and future direction
7.1 Summary
The body of work contained in this thesis has covered topics ranging from applications of
PAC devices in engines through fundamental studies of reaction sequences in well defined
experimental platforms. To bridge this subject range, simple modeling techniques which add
insight into complex multi-dimensional phenomena have been developed and validated in the 0-
dimensional experimental platform. The main contributions of this work can be summarized as
follows: (1) measurement of low temperature dissociation and oxidation of methane and ethylene
fuels in an NRP discharge using a combination of in situ and ex situ techniques to determine the
kinetic pathways in PAC, (2) the development and initial validation of predictive models for
methane and ethylene PAC using coupled plasma-combustion chemical models, (3) analysis of
low temperature PAC oxidation pathways for small hydrocarbons and their effect on fuel
consumption and final product concentrations, (4) demonstration and explanation of inter-pulse
coupling for NRP plasma devices via both imaging and modeling techniques, and (5) proven
lean-limit and ignition time reduction in engines, including an IC engine and a PDE,
demonstrating the usefulness and also the limitations of two types of PAC devices for real world
combustion applications.
Chapter 1 covered the motivation for this thesis, basic concepts in PAC, a review of
significant previous work leading to the present state of the field, remaining questions of
scientific and engineering interest, and finally the objectives of the present body of work.
Emphasis was placed on the theory of ignition by localized plasmas and the debate centered on
the relative importance of thermal or kinetic effects during ignition. The motivation of this thesis
199
was to contribute to the understanding of this topic by studying the kinetic effects of PAC in
homogeneous discharges and in localized ignition phenomena. Previous studies ranging over 100
years of research have been unable to fully answer this question due to the complexity of the
physical phenomena involved and lack of methods for isolating the individual effects of plasma
on combustion. By observing localized discharges at high temperature and homogeneous
discharges at low temperature, using the same voltage waveform, the objective of improving
PAC understanding could be accomplished. In the process, a range of diagnostic and
computational tools were implemented in conjunction with specialized experimental platforms,
as described in Chapters 2-6.
In Chapter 2, two applications of plasma-assisted combustion were explored, one for
microwave-assisted sparks in IC engines, and one for nanosecond repetitively pulsed discharges
in PDE engines. In both experiments, it is demonstrated that the lean limit of engine operation
was extended and the time for flame development was reduced. The most significant finding in
Chapter 2 was that inter-pulse coupling plays a critical role for kinetic enhancement via plasma
produced radicals. Inter-pulse coupling is the process that occurs in NRP discharge ignition at
sufficiently high repetition frequencies, taking advantage of the efficient energy deposition
process during the breakdown phase of the discharge.
Chapters 3 and 4 described the development of the nanosecond repetitively pulsed dielectric
barrier discharge experimental platform and the ZDPlaskin-CHEMKIN modeling tool used to
carry out and analyze ethylene and methane PAC experiments in Chapters 5 and 6. The
necessary background concepts of laser absorption spectroscopy as well as a description and
confirmation of the operation of this diagnostic method were provided. A brief description of the
governing equations used in the modeling was also provided. In addition, Appendix A provides a
200
full derivation of the Boltzmann equation for finding the reaction rates of electron collision
processes. Using the experimental platform, a homogenous plasma was successfully initiated,
and measurements of methane, acetylene, water, formaldehyde, and temperature were all
accomplished in situ in a miniature Herriott cell within the discharge region. Modeling of the
discharge was successfully accomplished, the results of which are discussed in Chapters 5 and 6.
In Chapter 5, the dissociation and oxidation of ethylene in a NRP DBD was explored both
experimentally and numerically in the homogeneous NRP DBD reactor using IR LAS and GC
sampling techniques and fully coupled plasma-chemical models including low temperature
hydrocarbon oxidation kinetics. The major findings are that low temperature O2 and radical
addition pathways are a major part of the oxidation process of even small hydrocarbons in PAC
applications, and that these low temperature routes play a central role in determining the final
product distribution. Even at these low temperatures (< 500 K) ethylene could be continuously
oxidized and/or dissociated by the plasma. Another significant finding was the importance of
excited and ionized argon as a participant in the dissociation of ethylene. The yield of products
actually increased as the argon content of the gas is increased (and fuel fraction is reduced),
indicating that argon increases the efficiency of fuel dissociation and conversion into smaller
species, primarily acetylene.
In Chapter 6, the low temperature oxidation kinetics of methane in an NRP DBD was
explored. It was found that oxidation proceeded down two major pathways, one leading to
methyl radical formation and the other to methylene radical formation, resulting primarily in the
production of CO, CO2, H2O, H2, CH2O, and CH3OH. Similar to ethylene, the fuel was
consumed continuously but never reached an explosive mode. The kinetic model did a
significantly better job for methane prediction as compared to ethylene, correctly predicting all
201
major species and temperature. Significant disagreement was found for the prediction of minor
species, particularly formaldehyde and methanol. It was found that the main consumption
mechanism for methane oxidation was reactions with O(1D), which were not included in the
ethylene work. While species resulting from the methylene pathways were well predicted, the
species resulting from the methyl pathways were poorly predicted. As these pathways are
dominated by O2 addition reactions and the reactions of oxygenated radical species, particularly
methyl peroxy radical, the importance and uncertainty of low temperature oxygen addition
pathways for methane are again highlighted. Additional work is needed to determine the exact
branching ratio of O(1D) reaction with methane, which is already ongoing (Yang et al. 2015b),
and to determine the fate of methyl peroxy radical at low temperature.
To summarize the findings discussed above, it has been found that plasma discharges can
oxidize hydrocarbon fuels, even at temperatures far below their ignition threshold and at
equivalence ratios outside of the flammability/ignition limits. This is possible due to the
production of active species whose reactions with the fuel have significantly lower activation
energy than ground state species. It has been demonstrated that the effect of this chemical
process can be used to accelerate ignition in real engines, although the kinetic effects cannot be
decoupled from the thermal effects in these cases. The two effects go hand-in-hand, and the
relative importance of each is determined by the gas/plasma temperature. At low temperatures,
even fuels without NTC behavior can be continuously oxidized in a process similar to catalytic
fuel reforming. At higher temperatures, which result from high plasma density and/or long pulse
durations (or high frequencies in the case of AC fields), ignition is locally initiated by the plasma
and is sustained by chain branching combustion reactions, which can only proceed at
significantly high temperatures, the exact value of which is fuel dependent. Even in this case, it
202
can be demonstrated that super-equilibrium radical concentrations generated due to electron
collision reactions in the plasma can accelerate ignition as compared to purely thermal energy
addition. Therefore, to be able to understand and quantitatively model PAC and PAI, beginning
from the application of a given voltage waveform and electrode geometry, one must consider the
precise electron collision reactions and subsequent reactions of plasma-produced species, as well
as the relevant ground state reactions in the temperature regime of the gas and plasma.
This thesis has developed two kinetic models, for ethylene and methane, for the purpose of
these predictions. These have been partially validated in a 0-dimensional experimental apparatus
with both in situ TDLAS measurements as well as ex situ GC-TCD measurements. The model
has been expanded to explain observed enhancement for localized NRP discharge ignition of
both of these fuels, which has resulted in the discovery of inter-pulse coupling, a more efficient
ignition method for high-speed combustion environments. While some questions concerning the
importance of kinetic mechanisms in PAC have been answered herein, significant opportunities
for future research and applications have also been revealed, as will be discussed in the following
section.
In the Appendix, two additional chapters (B and C) have been included which discuss the
oxidation of t-butanol and acetone in counterflow flames. While these papers are not related to
PAC, they were a significant portion of the work performed in the course of this doctoral thesis,
and were included for reference.
203
7.2 Recommendations for future work
7.2.1 Further exploration of PAC-related reactions
7.2.1.1 The reaction of O(1D) with hydrocarbons
As discussed in Chapter 6, the reaction of O(1D) with fuel after the plasma discharge is a
significant contributor to overall fuel consumption. O(1D) is created in the plasma primarily due
to electron collisions resulting in dissociative excitation of O2, and a secondary source is the
dissociative recombination of O2+ with electrons. Since the rate of fuel reactions with O(
1D) can
be much faster than that of O(3P), as much as five orders of magnitude in the case of methane,
studies of this reaction even at low temperature can help shed light on the present uncertainties in
PAC modeling. For this purpose, a photolysis reactor coupled with TDLAS detection is under
development (Yang et al. 2015b) at Princeton University. When complete, this system will be
used to break down O3 into O2 and O(1D) via a photofragmentation reaction using laser light at
266 nm. The subsequent products will be monitored using appropriate TDLAS techniques,
including Faraday rotation spectroscopy (FRS) for detection of HO2 and OH (Brumfield et al.
2013, Brumfield et al. 2014, Kurimoto et al. 2015). In the case of methane, it will be possible to
monitor the direct products, including OH and CH2O, as well as the secondary products, such as
H2O and HO2. Therefore, both the total rate and the branching pathways of CH4 + O(1D) can be
measured and compared to existing values in the literature. Similar measurements can be made
with larger hydrocarbons, particularly ones displaying significant low temperature chemistry, to
determine the importance of this reaction for PAC and reduce one of the most uncertain
pathways for fuel oxidation at low temperature in plasmas.
204
7.2.1.2 Electron collision cross-sections for large hydrocarbons
The collision cross-sections and branching ratios for small hydrocarbons have been studied
and reliable values published over the years, i.e. (Nishimura et al. 1994, Janev et al. 2002, Janev
et al. 2003, Janev et al. 2004). However, cross-sections for larger molecules are not readily
available, including all n-alkanes larger than propane. Because there is significant interest in
these larger hydrocarbons in the combustion community, particularly due to their active
chemistry at low temperature (600 – 800 K) that mimics diesel and kerosene fuels, it would
greatly benefit the PAC community if these cross-sections could be measured. One technique is
to perform swarm studies to obtain the total and momentum transfer cross-section as a function
of electron energy. Another option is to compute the cross-sections using quantum chemistry
methods, for example, using the program Quantemol-N (Tennyson et al. 2007). However, any
computed rates need to be verified by experiment. In general, these type of collision cross-
sections are measured only for reactions of interest to plasma fusion devices or atmospheric
chemistry. In these applications, small hydrocarbon species such as ethylene and methane may
exist, but not many larger species. However, the increased attention on PAC should draw interest
to measuring these cross sections, which are necessary for modeling of real fuels or fuels with
NTC behavior. Without them, it is not possible to predictively model combustion events for
anything larger than propane at this time.
7.2.1.3 Energy transfer processes
One of the main arguments backing thermal processes as the dominant drivers of localized
plasma ignition is that, even though many electronically and vibrationally excited species are
formed in the plasma, the relaxation time of these species is much faster than the time scale of
combustion reactions. Therefore, the excited species do not matter, only the equilibrium
205
concentrations of neutral species and the temperature after the discharge need to be considered.
This view is certainly too simplistic if one wishes to model the discharge using an applied power
waveform as the input, in which the final temperature and neutral radical population is strongly
governed by the energy transfer processes involved. As discussed in Chapters 5 & 6 in this
thesis, as well as in much of the literature (Mintoussov et al. 2011, Popov 2011, Rusterholtz et al.
2013, Adamovich et al. 2015a, Adamovich et al. 2015b), excited species reactions are
responsible for transferring much of the electron energy to neutral species via dissociation and
quenching reactions. Without knowledge of the specific states occupied by the excited species
after the plasma, as well as their reaction rates, the radical population and the heat release rate
and final temperature (especially when considering heat loss rates vs. heat release rates) cannot
be predicted. Therefore, future research into the exact states occupied in and after the discharge
by all neutral species deserves closer attention in the PAC community.
Recently, the importance of vibrational temperature and population of vibrationally excited
states of N2 has been explored (Rusterholtz et al. 2013, Adamovich et al. 2015b). It was found
that, in NRP discharges, the super-equilibrium vibrational temperature and population of
vibrationally excited N2 can last up to the millisecond timescale, which is on the same order as
combustion processes. The vibrational-translational (V-T) relaxation rates then have a strong
effect on the rotational gas temperature, which is often not considered in modeling approaches.
In addition, the reaction rates of vibrationally excited species with neutrals, particularly
hydrocarbons, are not well known. Considering the importance of O2 reactions and their
temperature sensitivity in terms of attachment or H-abstraction from hydrocarbons and
hydrocarbon intermediates, the vibrational non-equilibrium of O2, as well as fuel molecules, may
be an important consideration in terms of the reaction rate of reaction pathways. This area has
206
many implications outside of PAC, including the chemistry behind shock waves and detonation
waves. Thus, it should be considered in future studies.
7.2.2 Further development of modeling approaches
There are many difficulties in properly modeling PAC, leading to some very fundamental
questions that remain open. One such question is if O-D modeling is actually appropriate for
modeling plane-to-plane or axisymmetic discharges considering the formation of sheath layers
and the transport of ions and electrons along the electric field direction creating non-uniform
distributions of species. Comparisons between O-D and 1-D modeling has shown some modest
benefits to the use of the much more complex 1-D model (Nagaraja et al. 2013, Nagaraja et al.
2014, Nagaraja et al. 2015a, Yang et al. 2015a, Yang et al. Submitted), the largest of which is the
accurate implementation of the applied voltage waveform for calculating the reduced electric
field. However, due to the slow speed of current calculations and the only modest effect on final
species concentrations of the 1-D approach, it is unclear if this level of accuracy is necessary.
Another major question is the similarity between different modeling approaches, and which
approximations are appropriate in different PAC conditions. For example, at what electron
density should electron-electron collisions be considered in solving the Bolzmann equation?
When does electron attachment become an important reaction process? What is the wall loss rate
of electrons, and what impact does that have on the concentration of ions after the discharge?
These questions have yet to be fully addressed by the community, but are of central importance.
Lastly, the availability of a unified modeling tool with accompanying kinetic models is a
major limitation, especially for researchers just entering the field. This problem is beginning to
be solved by using the plasma modeling tools in CHEMKIN PRO (Adamovich et al. 2015b).
207
However, this approach requires further validation, and is also limited to 0-D calculations.
Therefore, the availability and reliability of PAC modeling tools, let alone kinetic models,
remains an open question with major implications in terms of the ease and accessibility of
modeling PAC experiments.
7.2.3 Investigation of inter-pulse coupling
As discussed in Chapter 2, inter-pulse coupling may open the door for low energy ignition in
high-speed flows using NRP discharges. The mechanism for inter-pulse coupling has been
proposed, but additional measurements are necessary to confirm the hypothesis that super-
equilibrium radical build-up actually occurs in the discharge region. For this purpose, a wind
tunnel specifically designed to investigate inter-pulse coupling in a well defined laminar or
turbulent flow field is under development at the Air Force Research Laboratory. A schematic of
this experiment is presented in Fig. 7.2.1. The tunnel design allows the use of gaseous fuel/air
mixtures at various flow velocities and turbulent intensities. The electrodes can be controlled in
terms of tip shape, material, and separation distance. A pulse generator capable of repetition rates
up to 100 kHz will be used to extend the observed ignition time reduction with increasing pulse
frequency to higher frequencies. Optical access is available from three directions such that line
of sight and 90° observation are available.
Efforts will be made to determine the relationship between fluid conditions and pulse
repetition rate, building on the observation that multiple pulses can only be deposited into a
single kernel if the kernel residence time in the electrode region is sufficiently long. In addition,
measurement of radical concentration will be made using a combination of high speed PLIF (10s
of kHz) as well as TDLAS to determine the temporally resolved radical concentration in the
208
plasma and ignition kernel as it develops. The temperature will be measured using optical
emission spectroscopy as well as Rayleigh scattering. This combination of information can be
used in conjunction with the simple modeling described in Chapter 4 to ascertain the cause of
inter-pulse coupling.
Figure 7.2.1. Diagram of wind tunnel experiment for investigation of inter-pulse coupling
in NRP discharges.
In addition to the measurements above, careful attention will be made to measure the energy
deposition into the plasma. One central question that must be addressed is if NRP ignition is
indeed more efficient than a single, high energy pulse. In order to answer this question, careful
measurement of the input energy and reflected energy is an important part of this study.
The ultimate goal will be to understand the potential and limitations of the inter-pulse
coupling technique, particularly for high-speed flows. Is there a limit to the ability to sustain a
super-equilibrium radical pool? Is there a pulse frequency or pulse voltage beyond which no
209
additional benefit can be gained? How are these limits affected by flow conditions? Can the
limits of ignitability be extended to the flammability limit of the mixture? Can a kernel develop
into a self-propagating flame even for very high flow velocities? These are a number of the
practical questions which this future study will aim to answer in addition to the fundamental
question of how exactly inter-pulse coupling works.
7.2.4 Advanced diagnostics for species and temperature measurements
One of the central elements of the work presented in Chapters 5 and 6 has been the use of
TDLAS to investigate short timescale (ms) processes occurring in PAC. The true timescale of
chemical changes in the plasma extends down to the nanosecond level, and perhaps faster if
electron collision process are considered. While stable species can be measured using TDLAS,
even at these short timescales, many of the important species are either in too small of
concentration or have too much interference for absorption techniques. In addition, absorption
becomes more difficult for measuring multi-dimensional phenomena, necessitating the use of
other diagnostic techniques. As interest in PAC continues to develop, knowledge of these small
quantity and short lifetime species become critical to the understanding of PAC. Excellent
progress has been made along these lines, as discussed in the introductory chapter as well as the
introductions of Chapters 5 and 6. However, there are still many measurements necessary to be
made for further understanding of PAC kinetics. As mentioned in Section 7.2.1.1, measurements
of radicals such as HO2 will help to nail down the branching fractions of some of the most
important reactions in plasma chemistry. This can be achieved using FRS, which is also useful
for measurement of other small radical molecules, including OH, NO, CH3, etc. Use of this
technique can help clarify the role of each radical in the overall system, and can be made on a
time dependent basis. In addition, high speed PLIF can help determine the evolution of species
210
and temperature in complex flow environments during ignition events, measurements which will
be necessary to carry out the work mentioned in the previous section. The implementation of
these cutting edge techniques is necessary to understand the complex chemistry in PAC, and is a
vital direction for future research.
Despite the many species available to be measured using the current methods, many key
species in PAC still cannot be measured. For example, O(1D) measurements have not yet been
reported by any authors in the PAC community, and radicals resulting from electron collision
dissociation of hydrocarbons, such as methyl, ethyl, and vinyl radicals, would also improve the
understanding of the importance of electron collision processes. However, some methods, such
as resonance-enhanced multi-photon ionization (REMPI), are being developed for these types of
measurements (Wu et al. 2013a, Wu et al. 2013b). For low temperature chemistry, measurements
of oxygenated radicals is of central importance to the oxidation scheme, but are only now
becoming available via photoionization mass spectrometry (Zador et al. 2013). The continued
development of these advanced techniques for hard to measure species will help guide the
development of kinetic mechanisms in PAC, revealing new chemical pathways as well as
quantifying reaction rates for known reactions.
7.2.5 Potential applications for PAC
As of the writing of this thesis, no significant updates have been made to mainstream ignition
devices over the past few decades, despite the recent effort spent on investigating PAC. Without
any device that can fundamentally change or enable combustion devices in a cost-effective,
energy-efficient, and geometrically permissible way, it is difficult to justify the continued
research in this area. Therefore, the most pressing future for researchers in this area should be to
211
demonstrate the potential of PAC devices to enable combustion in regimes and devices that are
“game-changers.” The example most typically used is for ignition and/or flame holding in
scramjets. Several authors have been investigating plasma for this purpose, i.e. (Wagner et al.
1989, Takita et al. 2000, Takita et al. 2007, Jacobsen et al. 2008, Do et al. 2010a, Do et al.
2010b), and their results show significant promise. However, practical limitations prevent the use
of an arc or arcjet for use onboard a scramjet due to the high power consumption of such a
device. NRP pulsed ignition may sidestep this issue via the inter-pulse coupling effect,
potentially resulting in similar enhancement to an arc discharge but utilizing a small; fraction of
the power due to the low duty ratio of nanosecond pulsed devices. This application is currently
under consideration by the USAF.
Another potential application, as discussed in Chapter 2, is for pulsed detonation engines.
While it has been proven that these engines can be used to power a vehicle in flight, no practical
uses of PDE engines seem to be on the horizon. However, rotating detonation engines (RDEs)
are beginning to attract attention as potential replacements to traditional combustors in gas
turbines (Dyer et al. 2012). However, this type of engine is difficult to stabilize due to the
difficulty in initiating and sustaining a stable detonation wave in the annular channel. It is
possible that PAI can assist RDE initiation by the use of high frequency discharges, perhaps at
the revolution frequency of the detonation wave, providing a positive feedback to stabilize the
combustion event. Another possibility is, by the use of the inter-pulse coupling effect, create an
ignition event that can ignite the detonation wave at some desired location.
Another possible application is for use in spark-assisted homogeneous combustion
compression engines (SA-HCCI). These engines are meant to operate in both pure HCCI mode
and in a mode more resembling an SI engine depending on the load and engine speed (Glewen et
212
al. 2009). Excessive cyclic variation occurs in this engine type when transitioning between
modes, in which the probability of a spark ignition and a homogenous ignition event become
comparable. In some cases, a single cycle will exhibit characteristics of both of these ignition
types simultaneously. In order to control these instabilities, a means of ignition timing
manipulation must be applied on the fly. The ability of NRP ignition devices to control the
number of pulses and pulse frequency can provide this extra degree of ignition control. By
producing either a very weak or very strong ignition event, depending on the energy and power
deposited into the ignition kernel, it may be possible to produce a slowly or quickly expanding
flame. Therefore, by applying a variable degree of ignition timing on a cycle-to-cycle basis, it
should be possible to stabilize the transition regime between SI mode and HCCI mode.
This short list of applications does not include the more traditional uses of ignition devices in
SI and gas turbine engines, but it is certainly possible to improve combustion initiation in both of
these engine types (i.e. lean limit expansion and high altitude relight, respectively). However,
although there is literature on these applications, engine manufacturers have not yet adopted new
plasma devices into existing engines. Until the price is reduced and reliability proven, these
devices will continue to be only marginally attractive options for implementation production
engines. However, for new engine designs, plasma devices of some type or another may enable
combustion in regimes not possible without this technology. The road forward remains uncertain,
but the possibilities are real.
213
Appendix
A1. Solution of the Boltzmann Equation
The Boltzmann Equation derivation has been described previously (Hagelaar et al. 2005), but the
derivation is repeated here due to its central importance in this thesis. Note that bold symbols
represent vectors while all other symbols represent scalars. The Boltzmann Equation can be
simply expressed as:
( A1.1 )
Which can be expanded in six dimensional phase space to give:
( A1.2 )
Where D is the total derivative, f is the electron distribution function, is the velocity vector, e is
the elementary charge (1.602 × 10-19
coulombs), m is the mass of an electron (9.109 × 10-31
kg),
E is the electric field, is the velocity-gradient operator, and C is the rate of change of the
electron distribution due to collisions. In this chapter, we will follow the example of Hagelaar
and Pitchford (Hagelaar et al. 2005), to reduce the dimensionality of this general equation and
achieve a final form that can be solved numerically with relative ease.
The first assumption we can make is that the electric field and collision probabilities are
spatially uniform. If we imagine the electric field to be aligned in the z direction, then the field
points from –z to +z and it is uniform in the x and y direction. Therefore, in position space, f may
only vary in the z direction, so the second term of Eq. A1.2 becomes:
( A1.3 )
214
Where we have converted to spherical coordinates in velocity space such that v is the magnitude
of the velocity vector and θ is the angle between the electric field direction (z) and the velocity
vector. Because the electric field only varies in one direction, f only varies in velocity space in
the z direction, The third term of thus Eq A1.2 becomes:
( A1.4 )
Where ¸ ¸ and ¸are the unit vectors in the r, θ, and φ directions, and φ is the angle
between the projection of r onto the x-y plane and the x axis. The unit vectors can be rewritten as:
( A1.5 )
( A1.6 )
( A1.7 )
Equation A1.4 can then be simplified:
( A1.8 )
Therefore, equation A1.2 is written:
( A1.9 )
To solve this differential equation, we can expand f in terms of Legendre polynomials of cosθ.
While it is possible to perform this expansion for as many terms as is desired, in general it has
215
been found that two terms is enough to accurately represent plasmas with moderate values of
E/N (Holstein 1946, Hagelaar et al. 2005). First, we will rewrite Eq. A1.9 in terms of cosθ :
( A1.10 )
The series expansion for f is then:
( A1.11 )
Where Pn are the Legrende polynomials, with each one representing an orthogonal nth
degree
polynomial function described by:
( A1.12 )
The two term approximation is then written as:
( A1.13 )
Where f0 is not dependent on θ and is thus called the isotropic term (i.e, only varies with v, z, t),
and f1 is the coefficient of the angular term, which determines the anisotropic perturbation (i.e.,
does vary with θ). Note that f1 is defined as a negative term, which will be important later. Also,
f0 is normalized such that if we integrate over velocity space, we simply end up with the electron
number density ne:
216
( A1.14 )
Plugging Eq. A1.13 into A1.10, we have:
( A1.15 )
We now multiply by the Legrende polynomials (only P0 and P1) and integrate over cosθ to get:
( A1.16 )
( A1.17 )
The right hand side of Eq. A1.16 represents the change in f0 due to collisions, which represents
changes in the overall distribution function from all types of collisional processes. The right hand
side of Eq. A1.17 represents the change in the anisotropic term due to collisions, both elastic and
inelastic. Elastic collisions will change the trajectory of the electron by some angle θ, while
inelastic collisions are considered to consume the initial momentum of the electron and then
release the electron isotropically. Therefore, the right hand side of Eq. A1.17 is simply some
proportionality factor f1 times the frequency of collisions:
( A1.18 )
217
Where N is the number density of the gas, and is the momentum transfer cross section,
defined as:
( A1.19 )
Where xk is the mole fraction of species k and σk is the effective momentum transfer cross section
of species k. We can rewrite equations A1.16 and A1.17 in terms of the electron energy:
( A1.20 )
( A1.21 )
Where the constant γ replaces the constants in the equation:
( A1.22 )
And the electron energy, in eV, is defined as:
( A1.23 )
We can now make some further assumptions about the energy, space, and time dependencies
of the electron distribution function. Considering that the electron number density cannot be
constant in space and time since there are processes that consume and produce electrons (i.e,
ionization and attachment), we cannot assume that the distribution function is constant in space
or time. However, for uniform discharges, as the one described thus far in the simplification of
the Boltzmann Equation, we can use the fact that energy of the electrons is independent of space
218
and time. Therefore, we can separate the space and time dependent part of the distribution
function from the energy dependent part:
( A1.24 )
Here, the F0,1 term describes the energy distribution, and ne(z,t) describes the distribution of the
electron number density. In order for this new formulation to match the earlier normalization of
the distribution function, we have normalized F0 as:
( A1.25 )
From here, we can consider two cases: a spatially uniform plasma with time dependence, or a
steady state plasma with spatial dependence. For the experimental work presented herein, the
spatially uniform and time dependent solution best approximates the nanosecond pulsed
dielectric barrier discharge. This is because there is not significant time for the plasma to reach a
steady state; however, the plasma is approximately uniform in the low pressures (60 Torr) used
for all experiments. We can then rewrite Eq. A1.21 as:
( A1.26 )
The time rate of change of the electron number density in spatially uniform can only be from
ionization and attachment processes. The frequency of reactions between a particle “A” and
collider “B” is defined as (Fridman 2008):
( A1.27 )
219
Where is the reaction frequency, is the interaction cross section, and is the relative
velocity of species A and B. To be more accurate, we can consider the cross section to be a
function of velocity, and for the velocity to have some distribution f( ). Then, averaging over all
velocities, we have:
( A1.28 )
The change in electron number density will be the electron number density times the reaction
frequency. If we sum over all ionization and attachment processes, convert to energy units, and
replace the distribution function using Eq. A1.24, we have:
( A1.29 )
Plugging into Eq. A1.26, we arrive at:
( A1.30 )
Now that we have solved for F0 we can plug into Eq. A1.21 and after some algebra, arrive at:
( A1.31 )
220
The two terms on the right hand side account for contributions from all collisional processes in
the term, while the term ensures that F0 is normalized to unity. The collision term includes
all collisions with neutral gas particles of type k, as well as collisions with other electrons:
( A1.32 )
Each type of process takes a different form. Here, we will consider elastic collisions,
ineleastic collisions, ionization, and attachment. Electron-electron collisions are only typically
important at very high electron densities (ne/N > 10-5
), which are not typical of the experiments
presented in the preceding chapters. Therefore, for simplicity, we will not discuss the effects of
electron-electron collisions.
Elastic collisions can be described as (Hagelaar et al. 2005):
( A1.33 )
Where Mk is the mass of the collision partner, kB is the Boltzmann constant, and T is the bulk gas
temperature. The first term accounts for electrons in collisions where energy is lost to the
collision partner, while the second term accounts for collisions in which energy is gained from
the collision partner. The second term assumes a Maxwellian distribution of the neutral gas
particles. For nanosecond pulsed discharges, the electron energy is not in equilibrium with the
bulk gas, i.e, the energy of neutral gas particles is much lower than the electron energy.
Therefore, the contribution of neutral particles scattering electrons into higher energy states is
not as significant as collisions which reduce the electron energy.
221
Inelastic collisions refer either to collisions in which the electron imparts some energy to the
collision partner inducing a transition to a higher energy state, or to collisions with excited
species in which the collision partner imparts some discrete amount of energy to the electron.
The effect this has on the electron energy distribution is to remove electrons from one part of the
energy distribution and insert them into another part. These terms can be described by:
( A1.34 )
Where uk is the threshold energy of the collision. The first term on the right hand side is for
electrons initially at energy ε that get scattered to another area of the distribution due to energy
gain or loss in a collision, while the second term is for electrons with energy ε + uk (uk can be
positive or negative depending on if the collision is an excitation or de-excitation process) losing
or gaining energy and ending up in the part of the distribution centered at ε.
The form of the ionization term is similar to the inelastic scattering term, except that two
electrons are emitted. For our purposes, we will make the assumption that the two scattered
electrons after ionization share the remaining energy after the ionization energy has been
removed:
( A1.35 )
Finally, we consider attachment processes. These can only remove electrons from the
distribution, so we have:
( A1.36 )
We can now combine all of these terms and put them into Eq. A1.31:
222
( A1.37 )
This can be reformulated as:
( A1.38 )
Which is the same form as a convection-diffusion continuity equation. Here, we define
( A1.39 )
( A1.40 )
( A1.41 )
( A1.42 )
( A1.43 )
Eq. A1.38 can now be solved numerically, as described in (Hagelaar et al. 2005).
223
A2. Chemical kinetic study of tertiary-butanol in a flow reactor and a
counterflow diffusion flame
The results presented in this chapter were published in:
J. K. Lefkowitz, J. S. Heyne, S. H. Won, S. Dooley, H. H. Kim, F. M. Haas, S. Jahangirian, F.L.
Dryer, Y. Ju, “A Chemical Kinetic Study of tertiary-Butanol in a Flow Reactor and a
Counterflow Diffusion Flame,” Combust. Flame 159 (2012) 968-978.
Prof. Joshua Heyne presented some of these results in the following conference:
J. Heyne, J. K. Lefkowitz, F. M. Haas, S. H. Won, S. Dooley, H. H. Kim, S. Jahangirian, F. L.
Dryer, Y. Ju, “Combustion Kinetics Study of t-Butanol,” 7th
U.S. National Combustion Meeting,
Atlanta, Georgia (2011).
Also, Prof. Stephen Dooley presented the results in the following conference:
J. K. Lefkowitz, J. S. Heyne, S. H. Won, S. Dooley, H. H. Kim, F. M. Haas, S. Jahangirian, F. L.
Dryer, and Y. Ju. “A Chemical Kinetic Study of the Alternative Transportation Fuel, tertiary-
Butanol,” 49th
AIAA Aerospace Sciences Meeting, Orlando, Florida (2011) AIAA paper 2009-
698.
I was responsible for data collection, calculation of the numerical modeling, analysis of results,
and writing of the paper mentioned above for all section pertaining to the counterflow burner. I
was assisted in all of these tasks by Dr. Sang Hee Won, and Mr. Hwanho Kim assisted in data
collection. Sections of this work pertaining to flow reactor results were performed by Prof.
Joshua S. Heyne, with assistance from the.remaining coauthors. The initial concept and guidance
for the research was provided by Prof. Yiguang Ju and Prof. Frederick L. Dryer.
224
A2.1 Introduction
Oxygenate fuels have been investigated recently as an alternative fuel or fuel additive for
ground transportation. In particular, increased attention has been given to biofuels as a means to
generate sustainable supplies of environmentally friendly, renewable alternatives to petroleum
based transportation fuels. The most prominent alternative today is ethanol, which has been
commercialized as a gasoline additive, though use in diesel and advanced engine cycles is also
under investigation. The benefits from the use of alcohol fuels in spark ignition applications
include lower NOX and soot particle production with relatively minor decreases in thermal
efficiency . However, there are some challenges that must be overcome to enable the sustainable
use of alcohols more generally in transportation applications. Ethanol, methanol, and other
alcohol fuels have negative attributes such as incompatibility with materials, low energy density
in comparison to petroleum products, and hydrophilic properties. Considerable efforts have been
made to identify suitable alcohol fuels with higher energy density. Therefore, the butanol
isomers have recently attracted significant attention from industry and research communities. For
the implementation of the butanol isomers as transportation fuels, understanding their
combustion characteristics would aid significantly in evaluating potential problems and
optimizing efficiency and environmental benefits through fuel blending with or complete
substitution for petroleum derived fuels.
Studies to evaluate the combustion characteristics of butanol isomers have grown in number
recently. The oxidation characteristics of the butanol isomers have been studied by experiments
in shock tubes (Tsang 1964, Dorko et al. 1971, Gonzalez et al. 1971, Lewis et al. 1974, Newman
et al. 1979, Choudhury et al. 1990, Moss et al. 2008, Noorani et al. 2010), jet-stirred reactors
(Dagaut et al. 2009a, Dagaut et al. 2009b), and briefly in an atmospheric pressure flow reactor
225
(Norton et al. 1991) to produce ignition delay measurements and speciation profiles for the
homogenous mixtures. In addition, the butanol isomers have been studied in both non-premixed
and premixed flame systems. Extinction limits in diffusion flames have been measured
experimentally with pure oxygen as an oxidizer in the counterflow configuration (Veloo et al.
2010), and the detailed speciation of the intermediates formed has been measured in butanol
doped methane/air coflowing flames (McEnally et al. 2005) and premixed low pressure (40
mbar) flames (Yang et al. 2007, Oßwald et al. 2011). The flame speeds have also been measured
in both premixed counterflow flames and spherically propagating flames to provide validation
data for kinetic modeling (Gu et al. 2010, Veloo et al. 2011).
However, the majority of previous studies have focused on normal-butanol (n-butanol),
whereas the iso-, secondary- and tertiary-butanol isomers have received considerably less
attention. Unlike the other butanol isomers, tertiary-butanol (t-butanol) has been utilized for
decades as a fuel additive (Mak 1986, Hamid et al. 2004). Therefore, a significant production
infrastructure already exists for t-butanol. Interestingly, there are still comparatively few
computational investigations of kinetic model development on the tertiary isomer. Only the
works of Moss et al. (Moss et al. 2008) and, very recently, Van Geem et al. (Van Geem et al.
2010) and Grana et al. (Grana et al. 2010) have provided kinetic models. All three models
suggest different rate constants for the initial fuel consumption reactions. Particularly, the
unimolecular elimination reaction to form isobutene and water varies by at least a factor of two
across the temperature range of 900-1500 K. A comprehensive validated model for t-butanol
combustion is not yet available. In order for such a model to be constructed successfully, it is
important to develop detailed validation data across the temperature and pressure domains
relevant to applications.
226
Motivated by the above, the objective of this study is to characterize the global reactivity of t-
butanol in different temperature and pressure conditions and provide detailed speciation profiles
to aid in the validation of kinetic modeling efforts. Both the low temperature (< 1000 K) and
high temperature (> 1200 K) oxidation of t-butanol are studied in the Princeton Variable Pressure
Flow Reactor (VPFR) and in a counterflow diffusion flame environment, respectively. These
venues are utilized to obtain speciation measurements that can be used to evaluate the accuracy
of t-butanol combustion models. The kinetic roles of the major intermediate species formed in t-
butanol oxidation (acetone, isobutene, and methane) are further investigated by measurement of
their extinction limits in diffusion flames. Numerical computations using available kinetic
models for t-butanol as well as acetone, isobutene, and methane are employed and compared to
experiments in order to understand the chemical kinetic details of the oxidation of this fuel at
vastly different temperature conditions and illuminate more specific means to improve current t-
butanol models.
A2.2 Experimental methodology
A2.2.1 Variable Pressure Flow Reactor
The experiments reported herein are conducted in the VPFR facility. The design,
instrumentation, and experimental methodology of this apparatus have been discussed in detail
previously (Held et al. 1998, Fischer et al. 2000, Li et al. 2001, Li et al. 2004b, Haas et al. 2009)
and are therefore only briefly discussed here.
Nitrogen carrier gas is heated by electric resistance heaters and homogenously mixed with
oxygen as it enters a 10.2 cm diameter quartz test section. The test section is surrounded by
thermostated electrical resistance heaters, which maintain the reactor wall temperatures at near-
227
adiabatic conditions. Liquefied t-butanol (99.7+ %; Aldrich Chemical Co., Ltd.) is supplied by
syringe pump to a liquid evaporator system, where it is gas-blast vaporized and diluted by
nitrogen gas at 423 K. The liquefied t-butanol (melting point ~298 K) is held within the syringe
pump at a constant temperature (308 K) by a thermostated heating tape wrapped around the
pump and piping leading to the evaporator to ensure that the fuel does not solidify during the
experiment. The diluted fuel vapor flows radially through a central injector tube and is rapidly
mixed with a much larger flow of N2/O2 carrier mixture. The fuel vapor and carrier flows meet at
the entrance to a conical mixer/diffuser where they mix prior to entering the test section. The
fuel/carrier gas mixture exits the mixer/diffuser section into the constant area test section at
Reynolds numbers where the reactor flow field is well characterized (> 6000).
During operation of the VPFR, flow rates of carrier, diluent, and reactants are metered and
held constant for the duration of each set of experiments reported. The initial gas temperature in
the test section is held constant under conditions of non-reacting flow. The pressure inside the
reactor is also held constant. Under these conditions, the reactor operates as a near-adiabatic,
isobaric, steady flow system at an initially constant temperature. Subsequent introduction of
small amounts of fuel (1 % carbon or less) into this system from cycling on fuel flow to the
evaporator negligibly perturbs the test section flow field and pressure. Reactant addition
establishes a reacting flow, which is sampled at a discrete location along the test section axis.
Reactant continues to flow through the evaporator until all on-line analytical readings, including
the temperature rise, achieve steady state (a few minutes). After steady state conditions are
achieved, these readings are digitally recorded, a portion of the gas is stored in a thermostated
vessel for offline analysis, and fuel flow is turned off. New sampling configurations are then
228
established under the same or different initial test section conditions to generate a spatial
sample/temperature rise profile along the test section.
The reacting mixture is sampled at incremental axial positions downstream of the
mixer/diffuser using a water-cooled, stainless steel sampling probe. The sampled stream is
continuously extracted and convectively quenched by wall heat transfer. The axial temperature
of the reacting flow is measured at the same location as the sampling probe tip using a silica-
coated R-type thermocouple. With all flows established except the reactant flow, the temperature
at the sampling location varies by less than ±1.5 K and is known to have a relative uncertainty of
±1.5 K and an absolute accuracy better than ±6 K (Held 1993). Estimated measurement
uncertainty in reported pressure is ±0.2 atm.
The sampling configuration inside the test section is altered by translating the mixer/diffuser
relative to the fixed sampling location. This procedure is employed at various mixer/diffuser
positions to obtain stable species and sample temperature profiles as a function of sampling
configuration for a given set of initial test section conditions. Residence times for specific
sampling configurations along the test section are calculated using experimental axial velocity
profile information determined for the test section under non-reactive flow conditions. These are
then corrected for reaction conditions using a Reynolds number correlation technique (Zhao et al.
2008). Conservative uncertainty in residence time is approximately ±1.6 % of the reported value.
Oxidation of t-butanol is presently studied using 0.25 % (2500 ppm) t-butanol at an
equivalence ratio (φ) of 1.0 corresponding to 1.50 % O2 (15000 ppm). Global reactivity of t-
butanol is determined by varying the reaction temperature over the range 675-950 K and at fixed
pressure (12.5 atm) and residence time (1.8 s).
229
A2.2.2 Sample analysis
Sampled gas flows through heated Teflon lines (373 K) to online analytical equipment: a
Fourier transform infrared (FTIR) spectrometer, an electrochemical oxygen analyzer, and non-
dispersive infrared (NDIR) analyzers for carbon monoxide and carbon dioxide. Uncertainties in
the measurements reported here are O2 ≤ 4 %; CO ≤ 3 %; CO2 ≤ 3 % of the reported reading.
Quantification of water was performed via online Fourier transform infrared (FTIR)
spectroscopy, with estimated uncertainties of ≤ 5 % of the reported value.
For offline analysis, a sample line in parallel to the FTIR sample line transfers gaseous
samples to fifteen stainless steel sample storage loops (VICI Valco Inc., 10 ml volume) attached
to a multi-position valve (MPV) stored in a chamber heated to 400 K. A gas chromatograph (GC,
Agilent 7890A) equipped with a (J&W HP-PLOT Q, 30 m length, 535 µm diameter, 40 µm film
thickness) column is utilized for offline chemical analysis. This column is followed by a flame
ionization detector (FID). This column is used for separation and quantification of hydrocarbons
and oxygenates. The detection limits of the GC-FID setup are better than 0.5 part-per-million
molar (ppm). The carrier gas for both columns is nitrogen and an oven temperature program of
223–533 K is utilized to achieve chromatographic separation of species.
Identification of species is performed by retention time comparisons against those of pure
substances for the same chromatographic conditions. Oxygenate (acetone and t-butanol)
calibration was performed by flowing liquid sample in the flow reactor at fixed nitrogen flow
rates and varying the liquid sample flow rate to produce gas phase calibration mixtures. Samples
of the calibration mixtures were stored in MPVs and subsequently analyzed by the GC-FID in
the same manner as other storage loop samples. Standard calibration gases (Airgas and Air
230
Liquide) were used for calibration of all other species (permanent gases, and hydrocarbons ≤
C6).
Quantified concentrations for select light species including CH4 and C2H4 are also compared
against those recorded (online) from FTIR spectra. Unless specified differently in the description
of an experiment, reported species concentrations are based on quantification by NDIR analyzers
for CO and CO2; the electrochemical oxygen analyzer for O2; the FTIR for H2O; and the GC-FID
for other hydrocarbon and oxygenate species. The repeatability of the GC-FID analysis method
is ensured by the regular analysis of standard gas mixtures of known composition (Airgas and
Air Liquide). The maximum estimated uncertainties on the species quantifications using the FID
are ±5 %.
A2.2.3 Counterflow diffusion flame extinction and speciation
Diffusion flame extinction limits for t-butanol (99.7+ %; Aldrich Chemical Co., Ltd.) and its
major intermediates, acetone (99.9+ %; Aldrich Chemical Co., Ltd.), isobutene (99+ %; Aldrich
Chemical Co., Ltd.), and methane (99+ % Airgas, Inc.) are measured in an atmospheric pressure
counterflow burner integrated with a fuel vaporization system, previously described in (Won et
al. 2010, Won et al. 2011). This system is also used to measure spatially-resolved species
profiles from a t-butanol diffusion flame. A schematic of the experimental apparatus is shown in
Fig. A2.1, and the system is briefly discussed here.
Liquid fuel and heated nitrogen are injected into the vaporization chamber (300 cc) through
coaxial inner (0.2 mm) and outer (1 mm) nozzles, respectively, to obtain steady state atomization
and vaporization. The flow rates of liquid fuels are controlled by a syringe pump system
(Harvard Apparatus, PHD 22/2000) heated to a constant temperature (308 K) by thermostated
231
heating tape wrapped around the syringe and tubing leading to the vaporization system. The flow
rate of nitrogen to the vaporization system is controlled using a calibrated sonic nozzle, and the
temperature of the nitrogen flow is elevated by using an in-line electric heater (Omega) and
monitored by a K-type thermocouple at the outlet of the heater. To reduce the heat loss from the
chamber and avoid fuel condensation, the outer surface of the chamber is surrounded by two
electrical knuckle heaters. Full, steady vaporization is maintained through PID control of the
vaporization chamber exit temperature.
Figure A2.1. Counterflow burner integrated with flame sampling system.
The temperature at the exit of the vaporization chamber is set slightly above the vaporization
temperature of the liquid fuel. It has been observed that at too high of temperatures unsteady
vaporization occurs within the vaporization chamber, causing temperature and pressure
fluctuations at the chamber exit. Therefore, the temperature at the exit of the fuel injection nozzle
in the vaporization system is maintained at 400 K for t-butanol, isobutene, and methane
experiments, and 365 K for acetone experiments. The temperature was maintained to ±10 K
using a PID controller. Vaporized fuel/nitrogen mixture is transported to the upper burner nozzle
Heated N2
N2
Air
Fuel
P
G
Positioning
Stage
TC TC
Temperature
Controller
PGTC
Mixing Chamber Heater
PG Pressure Gauge
TC Thermocouple
Needle Valve
On/Off Valve
Three-way Valve
VacuumMultiple
Port
Valve
PG
N2 Bypass
N2 Source
Stagnation
Plane
232
through a thermally insulated tube. The upper nozzle exit temperature is kept at 400 ± 5 K using
PID control. Oxidizer flow rate to the lower burner nozzle is controlled using a calibrated sonic
nozzle, and the temperature of this flow is maintained at 298 ± 5 K. Upper and lower nozzle
nitrogen coflows are controlled by calibrated sonic nozzles as well, and their temperatures are
maintained at 400 ± 5 K on the upper (fuel) side and 298 ± 5 K at the lower (oxidizer) side.
The two hydrodynamically converging nozzles are designed to prevent boundary layer
separation, and both have 13 mm i.d. at the exit. Honeycombs of 50 mm length are installed 20
mm upstream of the nozzle exits to obtain a uniform velocity profile. Particle image velocimetry
measurements confirm flow uniformity at the nozzle exit. The distance between the nozzles was
varied from 6 to 18 mm for a variety of strain rates to test for consistent and repeatable results
for extinction experiments, and 9 mm was selected as an acceptable separation distance for all
measurements. Extinction strain rates were measured by increasing the flow velocities of both
the fuel and oxidizer streams while matching the respective flow momentums until the flame was
no longer visible. Global strain rate, a, was utilized to represent the flow conditions at the
extinction limits, and was calculated via Equation (A2.1) (Seshadri et al. 1978):
(A2.1)
Where U represents the velocity at the nozzle exit, L represents the distance between the fuel
and oxidizer nozzles, ρ represents the density at the nozzle exit, and the subscripts f and o
represent “fuel” and “oxidizer”, respectively.
To obtain the spatial profile of stable species in the diffusion flame, a micro-probe sampling
system is incorporated into the counterflow burner. This system is comprised of a 363 µm o.d.
=2𝑈
1 +
𝑈
𝑈
233
and 220 µm i.d. fused silica microprobe (SGE Analytical Science, P/N: 0624469) which draws
sample gas from the flame into heated sample storage loops. All tubing involved in the sampling
is stainless steel and heated to 373 ± 10 K to prevent condensation of the sample gas.
The probe position is controlled by a translation stage, actuated in the burner’s axial and
radial dimensions by stepper motors with an accuracy of 5 µm per step. The actual position of
the probe is monitored via digital photography utilizing a camera with an optical accuracy of
approximately 20 µm per pixel in the burner plane. All measurements are taken with the probe
fixed 1 mm radially away from the burner axis. Experiments show that, at the stagnation plane,
similar sampling results are obtained between 0 and 2 mm from the axis of symmetry. Therefore,
the 1 mm offset from the burner axis reduces perturbations to the flame structure while safely
remaining in the region of consistent measurements.
Dilution of the sample gas is necessary in order to raise the pressure of the sample in the
storage loops to the value required by the current GC configuration. High pressure gas (roughly
275 kPa) in the storage loops ensures that the sample gas can fully clear the exiting gas in the
tubing connecting the MPV to the GC before being injected for analysis. Dilution of the sample
gas is achieved by mixing N2 with the sample in a heated stainless steel chamber, keeping note of
the ratio of sample pressure to dilution pressure utilizing pressure gauges (Omega) with an
accuracy of ±0.7 kPa. This contributes ±1 % uncertainty to the species quantification. The
diluted mixture is then transferred into the storage loops for offline analysis with the same gas
chromatograph used in the flow reactor experiments and described in the previous section. The
uncertainty associated with species quantification in the gas chromatograph is ±5 %, which
brings the total uncertainty in species measurements to ±6 % of the reported value. The error
234
associated with intrusive measurements of the flame is uncertain, and requires further study to
quantify.
Numerical simulations of the extinction limit and sampling experiments are performed using
the OPPDIF module of the CHEMKIN package (Kee et al. 2003) with a modified arc-length
continuation method for both plug flow and potential flow conditions. Further details on
calculations were described previously (Won et al. 2010, Won et al. 2011).
A2.3 Results and discussion
A2.3.1 t-Butanol oxidation in the Variable Pressure Flow Reactor
The oxidation of a 2500/15000/982500 ppm (φ =1.0) mixture of t-butanol/O2/N2 was studied
at 12.5 atm and temperatures of 675-950 K at a constant residence time of 1.8 seconds. The
experimental observations are shown in Fig. A2.2. It is observed that t-butanol decomposes
rapidly at early residence times corresponding to positions inside the mixer/diffuser section of
the reactor, in part as a result of higher concentrations during the mixing of fuel vapor/diluent
with carrier and heterogeneous reactions on surfaces. There is no further significant
heterogeneous conversion noted outside the diffuser section as illustrated in Fig. A2.3, which
displays the pyrolytic conversion of t-butanol to water and isobutene in the mixer/diffuser
section and downstream. Similar observations have been noted previously in the study of other
oxygenated fuels in this reactor, for example, ethanol (Li et al. 2004a, Haas et al. 2009) and
methyl formate (Dooley et al. 2010). Consistent with the previous studies, the heterogeneous
decomposition rate is a very weak function of temperature, characteristic of a catalytic process.
In the oxidative experiments shown in Fig. A2.2, approximately 707 ppm fuel is observed to be
converted heterogeneously to isobutene and water at all reaction temperatures lower than 780 K.
235
Homogeneous gas phase oxidation is comparable at 780 K, and is principally responsible for the
consumption of t-butanol at higher temperatures (780-950 K).
Figure A2.2. Flow reactor oxidation of t-butanol/O2/N2 2500/15000/982500 ppm at 12.5 atm
and a residence time of 1.8 seconds. Experimental data are symbols and lines are Grana et
al. (Grana et al. 2010) kinetic model computations.
Despite the complicating effects noted above, Fig. A2.2 reveals that t-butanol oxidation does
not exhibit low temperature reactivity and the negative temperature coefficient behavior
characteristic of alkylperoxy radical (RO2) chemistry. The gas phase consumption of fuel is
coincident with the production of large amounts of water, methane, acetone, and carbon
monoxide. It is likely that nearly all the isobutene observed in the experiment can be attributed to
236
Figure A2.3. Species concentrations as a function of residence time in the flow reactor at
6.1 atm, 658 K, and t-butanol/N2 2500/997500 ppm. Symbols are experimental data, dashed
line represents the end of the mixer/diffuser region, and the solid horizontal line illustrates
no further destruction via heterogeneous conversion of t-butanol in the test section.
heterogeneous conversion within the mixer/diffuser section of the reactor. At temperatures lower
than the onset of gas phase oxidation (~780 K), the prescribed mass of carbon to the reactor may
be accounted for by the summation of the measured t-butanol and isobutene concentrations. This
strongly indicates that no homogeneous processes are occurring below 780 K. At temperatures
greater than 780 K, isobutene is not measured at concentrations higher than those attributable to
237
heterogeneous chemistry, indicating that the rate of gas phase consumption is always greater that
the rate of gas phase production of isobutene.
Approximately 84-97 % of the carbon is recovered in the experimental quantification of the
measured species. Methane and acetone are produced in significant quantities peaking at ~395
ppm and 343 ppm respectively. Other measured species include acetaldehyde, isopropanol,
propylene, ethane, and ethylene. Of these species, ethylene is produced in the highest
concentrations peaking at ~168 ppm at 841 K. Propylene measurements peak at a lower
temperature (823 K) and concentration (~89 ppm). Isopropanol is observed in low concentrations
(< 50 ppm). Five small concentration (< 100 ppm) species were detected but were not able to be
identified; these account for the carbon that is not recovered in the experiment. Based on their
retention times, it is likely that the largest concentration species are C3 or C4 oxygenates as they
do not have the retention times of calibrated C3-C5 hydrocarbons.
As a result of the symmetry of the t-butanol molecular structure, only isobutene, acetone, and
propen-2-ol may be formed as major intermediates of mechanistic relevance. The chemical
pathways responsible for these intermediates are shown in Fig. A2.4. There are two hydrogen
abstraction − radical beta-scission processes possible for t-butanol consumption, represented as
Paths 1 and 3 in Fig. A2.4. An alkoxy radical may be produced by hydrogen abstraction from the
alcohol position; this radical may then beta-scission to produce acetone and a methyl (CH3)
radical. Alternately, an alkylhydroxy radical may be produced by hydrogen abstraction from the
methyl position; this radical may subsequently beta-scission to form propene-2-ol and methyl or
isobutene and a hydroxyl (OH) radical. It is far more likely that the latter scenario, Path 3 in Fig.
A2.4, will dominate over Path 1 due to a factor of nine increase in available abstraction sites and
the weaker bond dissociation energy of the C-H bond of the methyl group (101.11 kcal/mol) as
238
compared to the O-H bond in the alcohol group (104.07 kcal/mol), Fig. A2.5 (Grana et al. 2010).
The observed quantity of methane is strongly indicative of the production of large amounts of
methyl. Methyl production from alkylhydroxy radical beta-scission will be accompanied by the
formation of propen-2-ol. Alkyl radical decomposition via the cleavage of C-C bonds is
thermochemically favored over the alternative route involving cleavage of the C-O bond. In
summary, it appears that t-butanol is predominantly oxidized by the hydrogen abstraction
pathway depicted in Path 3 of Fig. A2.4 followed by alkyl radical beta-scission to form propen-
2-ol and methyl.
Figure A2.4. Primary consumption paths of t-butanol. All pathways result in isobutene,
water, acetone, or propen-2-ol as stable intermediate products.
239
Figure A2.5. Bond dissociation energies of t-butanol (Grana et al. 2010).
In apparent contrast to this mechanistic discussion, significant quantities of propen-2-ol were
not definitively observed in the flow reactor experiments. One of the unidentified species from
the GC-FID analysis could be propen-2-ol on the basis of the relative retention times to known
species. Unfortunately, this cannot be confirmed as propen-2-ol cannot be obtained and
calibrated. However, it is well known that propen-2-ol has a propensity to tautomerize to form
acetone (Turecek et al. 1988). Consequently, the absence of major concentrations of propen-2-ol
in the experimental analysis is to be expected and, moreover, is consistent with the mechanistic
discussion. The observed concentration of acetone suggests a significant amount of t-butanol
reacts through Paths 1 and/or 3 in Fig. A2.4.
On the basis of our experimental observations, there is no evidence to suggest that the gas
phase molecular elimination reaction of t-butanol to form isobutene and water is significantly
active in the oxidation of t-butanol at these flow reactor conditions (600-950 K). It is noted
however that Norton and Dryer (Norton et al. 1991) did observe significant quantities of
isobutene to be formed in their atmospheric pressure t-butanol flow reactor oxidation study at
significantly higher temperatures and shorter residence times of 1027 K and ~100 ms. Finally, on
the basis of the experimental observations presented here and as per the above discussion, one
should expect methyl chemistry to play a significant role in the combustion of t-butanol as it is
likely to be present in abundance.
240
A2.3.2 Extinction limits and species profiles in the counterflow diffusion flame
A2.3.2.1 Species profile measurements
To determine the t-butanol consumption pathways at high temperature, species profiles have
been measured in diffusion flames using the sampling technique for Xf = 0.28 and a = 100 s-1
.
These results are presented in Fig. A2.6. Measured species include t-butanol, isobutene, acetone,
methane, acetylene, ethylene, ethane, and propylene. Propane and 1,3-butadiene were also
measured but are not reported due to their small concentrations in the flame. Note that the GC
analytical technique employed precludes detection of small oxygenated species such as CH2O,
CO2, CO, and H2O. The results show that isobutene is the most abundant intermediate species in
the flame. This indicates that molecular elimination, Path 4 in Fig. A2.4, is strongly active at
high temperature conditions, as opposed to its negligible contribution in the lower temperature
flow reactor experiments. The result is consistent with the flow reactor results of Norton and
Dryer (Norton et al. 1991) at 1027 K, discussed above.
Figure A2.6. Speciation profile of the t-butanol diffusion flame, Xf = 0.28, a = 100s-1
,
compared with the Grana et al. model (Grana et al. 2010).
1 2 3 4 5 6
0
2000
4000
6000
8000
10000
12000
14000
16000 t-butanol/20
isobutene
methane
acetone
Sp
ec
ies
co
nc
en
tra
tio
n,
pp
m
Distance from the burner, mm
1 2 3 4 5 6
0
2000
4000
6000
8000
10000
12000
14000
16000 acetylene
ethylene
ethane
propylene
Distance from the burner, mm
241
Consistent with the lower temperature VPFR results, significant quantities of methane and
acetone are also observed in the flame. Acetone and methane may be produced via Paths 1, 2, or
3 in Fig. A2.4. The high temperature and rich equivalence ratio environment of diffusion flames
provide conditions where thermal decomposition can be a significant pathway to fuel
consumption. Therefore, it is expected that Path 2, C-C bond homolysis of the weakest bond in
the molecule at ~87.5 kcal mol-1
, will be active in the diffusion flame. This path may result in
two separate product sets, 1) acetone, methyl, and H atom or 2) propen-2-ol, methyl, and H atom.
It is not possible to definitively detect propen-2-ol using the GC-FID analytical technique
employed here. However, in the premixed burner stabilized t-butanol flame studied by Oβwald et
al. (Oßwald et al. 2011) using photo-ionization molecular beam mass spectrometry, both acetone
and propen-2-ol are measured. Oβwald et al. note that propen-2-ol exists in quantities of at least
an order of magnitude less than acetone. As this study and that of Oβwald et al. are concerned
with similar temperature regimes, it is expected that the quantity of acetone present in the
diffusion flame will also far outweigh that of propen-2-ol (Turecek et al. 1988). The absence of
any unidentified large concentration species in the GC analysis corroborates this postulation.
Similar to the flow reactor observations at much lower temperatures, methane is observed as
an abundant intermediate, and the fact that it does not increase quantitatively with acetone is
evidence that it is a significant product in the further oxidation of isobutene, acetone, and
propen-2-ol. As in the present flow reactor study, the high concentration of methane is indicative
of a radical pool containing a large quantity of methyl. There are also significant concentrations
of ethane, which is indicative of methyl recombination reactions as ethane is not a product of
beta bond scission processes.
242
A2.3.2.2 Extinction limit measurements
The extinction limit of diffusion flames is affected simultaneously by both chemical kinetics
and mass transport, in a highly coupled manner. Previous work on diffusive flame extinction
(Diévart et al. 2011, Won et al. 2011) suggests that, by multiplying the fuel mole fraction by the
heat of combustion for a specific fuel, the kinetic effect on extinction can be better displayed. For
fuels of different molecular weight, a transport-weighted enthalpy term can be defined to
normalize for the effects of fuel quantity, internal energy, and the rate of diffusive transport,
thereby leaving only the kinetic effect. Won et al. define the Transport-weighted enthalpy (TWE)
as the product of fuel concentration (Xf ), the standard state enthalpy of combustion (ΔHc°), and
the inverse square root of the ratio of fuel molecular weight to diluent molecular weight
((MWF/MWN2)-1/2
) (Won et al. 2012). By comparing different fuels on a TWE vs. aE plot one may
determine the difference in reactivity as if all fuels have the same potential for heat release and
the same rate of mass diffusion.
To provide a measure of the high temperature chemical kinetic/mass diffusive coupling, the
extinction limit of t-butanol is measured and shown in Fig. A2.7. Based on the diffusion flame
sampling experiment, it is proposed that t-butanol initially decomposes into: 1) isobutene and
water, 2) acetone, methyl, and H atom, or 3) propene-2-ol, methyl and H atom, Fig. A2.4.
Therefore, in order to provide for the systematic validation of the significant intermediate species
particular to t-butanol oxidation, the extinction strain rates of acetone, isobutene, and methane
are also determined and presented in Fig. A2.7. The extinction limit of each fuel on a mole
fraction basis is observed to increase in the order of methane < t-butanol < acetone < isobutene.
As Won et al discuss (Won et al. 2012), on a mole fraction comparison, this ordering is dictated
by a coupled interaction of chemical kinetics, mass diffusion, and fuel potential energy. The
243
standard state heats of combustion of acetone, isobutene, methane, and t-butanol are 1790
kJ/mol, 2700 kJ/mol, 890.3 kJ/mol, and 2644 kJ/mol respectively (NIST). Following Won et al,
analysis of these measurements by a TWE metric reveals that the extinction limit increases in the
order of t-butanol < methane isobutene < acetone, exposing the relative chemical kinetic
potential of each fuel and the relative effect of the formation of each intermediate species formed
in the oxidation of t-butanol flames. Though not measured in our flame experiments, water is
expected to be a major intermediate species formed early in the flame structure as a co-product
with isobutene from the t-butanol molecular elimination reaction. For a variety of reasons, it is
expected that formation of water will inhibit flame reactivity.
Figure A2.7. Extinction strain rate measurements of t-butanol, isobutene, acetone, and
methane compared with computations of the Grana et al. model (solid lines) (Grana et al.
2010) and the Wang et al. model (dashed lines) (Wang et al. 2007).
0.1 0.2 0.3 0.4
0
200
400
600 acetone
isobutene
methane
t-butanol
Grana model
Wang model
Fuel Mole Fraction, XF
Ex
tin
cti
on
str
ain
ra
te,
aE
/ s
-1
5 10 15 20
0
200
400
600
acetone
isobutene
methane
t-butanol
Grana model
Wang model
Transport-weighted Enthalpy, J cm-3
244
A2.3.3 Model comparison
Computations using the kinetic models of Moss et al. (Moss et al. 2008), Van Geem et al.
(Van Geem et al. 2010), and Grana et al. (Grana et al. 2010) were compared to the flow reactor
observations. Due to the heterogeneous conversion in the mixer/diffuser section, the model
computations are performed using the initialization technique described by Zhao et al. (Zhao et
al. 2008) with the use of the measured species concentrations at the lowest temperature. For the
computations, it is assumed that the heterogeneous conversion is constant for each temperature
and that the only species formed heterogeneously are water and isobutene. Of the above models,
the Grana et al. model is observed to best reproduce the experimental results, so the following
discussions focus on this model.
The Grana et al. model captures the general shape and character of all species profiles
measured in the flow reactor study, Fig A2.2. The rate of production analysis at 775 K, 12.5
atm, 707/707/1793/15000/981793 isobutene/water/t-butanol/oxygen/nitrogen and 1.8 seconds for
the flow reactor conditions appears in Fig. A2.8 as bold numbers. The Grana et al. model does
not treat the formation of acetone from the t-butyl-ol radical in the same manner as the above
discussion. Instead the t-butyl-ol radical is described to directly react to form acetone and
methyl with no consideration given to the propen-2-ol intermediate (Grana et al. 2010).
Nonetheless, the model lags in the computation of acetone at flow reactor temperatures. Grana et
al. describe the formation of acetone as being mainly due to the beta scission of a t-butyl-ol
radical to form acetone and methyl. This reaction contributes to approximately 92 % of the t-
butyl-ol destruction at 830 K (the peak concentrations of acetone). The formation of acetone and
methane in the experiment are out-of-phase with one another, implying alternative methane
245
formation routes. The model does not predict as much low temperature acetone production
suggesting radical reactions with t-butanol that lead to acetone are perhaps too slow.
It is also observed that this model predicts a slower overall rate of oxidation than observed in
the flow reactor experiment. Regrettably, due to the complicating effects of the heterogeneous
conversion of fuel in the mixer/diffuser section of the reactor it is not possible to conclusively
state the absolute fidelity of model performance at these conditions.
Figure A2.6 shows the fuel and major intermediate species measured for a Xfuel = 0.28, a =
100 s-1
diffusion flame compared to computations with the Grana et al. model. At the higher
temperature diffusion flame conditions, the model predicts a later consumption of fuel and
narrower profiles for all intermediate species compared to the experimental measurements. It is
presently unclear whether this disagreement is due to perturbations introduced by the intrusive
sampling technique or to kinetic/transport model fidelity. It is also observed that isobutene is
greatly over-predicted by the model, with the peak concentration computed to be nearly a factor
of two greater than measurements, indicating that that the relative rates between paths 1-4 may
be inaccurate, and/or that isobutene consumption at flame conditions is prescribed too slowly.
There is good agreement for acetone and methane, and the location of the peak concentration is
well predicted for almost all of the species measured.
The consumption pathways of t-butanol have been analyzed by integrating the rates of all
reactions over the entire counterflow diffusion flame computational domain at conditions of Xf =
0.3 and a = 100 s-1
. The chemical pathways responsible for t-butanol consumption into the
primary intermediates according to the Grana et al. model are presented in Fig. A2.8 in italic
numbers. It is shown that approximately 75 % of t-butanol decomposes into isobutene and water
246
Figure A2.8. Flux analysis for t-butanol using the Grana et al. model (Grana et al. 2010).
Bold numbers indicate flow reactor simulations at 775 K, 12.5 atm, and φ =1.0; italicized
numbers indicate diffusion flame simulations at Xf = 0.3, a = 100s-1
, normal numbers
indicate paths with the same flux for either condition.
via unimolecular elimination, Path 4 in Fig. A2.8, a reaction that is of negligible importance at
the cooler flow reactor conditions. The rate constant of this reaction is rather uncertain, as
mentioned in Section 1, and is of principal importance to predicting isobutene concentrations
under flame conditions. While the hydrogen abstraction − beta-scission pathway to form acetone
247
is dominant in the low temperature flow reactor study, here it represents only ~3 % of total t-
butanol consumption, while the thermal decomposition path to form acetone accounts for 19 %
of total t-butanol consumption.
Predicted extinction limits of t-butanol, acetone, isobutene, and methane diffusion flames are
also compared with measurements using the Grana et al. model (Fig. A2.7). The model over-
predicts the extinction limit of the t-butanol and isobutene diffusion flames, while performing
relatively well for methane and accurately reproducing the measured acetone extinction limits.
Therefore, it appears that the discrepancy between the experimental and modeling results for the
extinction limit of t-butanol is due in some part to the fidelity of the isobutene sub-model. It now
appears that the model of Grana et al. predicts faster isobutene chemistry than the measurements
for both the flow reactor and extinction strain rate experiments, and therefore it is very likely that
the branching ratio must be improved for accurate computation of t-butanol combustion.
The extinction limits of isobutene diffusion flames using the Wang et al. isobutene model
(Wang et al. 2007) are also computed. As shown in Fig. A2.7, Wang et al.’s model under-
predicts the extinction limit compared to the measurements. It is interesting to point out that
between the two models there is almost a factor-of-two discrepancy in the extinction limit
prediction. Consideration has been given to the effect of the estimated mass transport parameters
on the computed extinction limit, since the collision diameter for isobutene in the Grana et al.
model is ~25 % larger than that in the Wang et al. model. The transport parameter of isobutene is
re-estimated based on a correlation to molecular weight suggested in previous work (Diévart et
al. 2011). It appears that the differences in transport parameters between the two models
accounts for less than 10 % of the difference in the computed extinction strain rate. Thus, the
248
discrepancy between the two models is primarily caused by differences in the chemical reactions
and kinetic parameters of the two isobutene oxidation sub-models.
As stated above, it is expected that the early formation of water will have an inhibitive effect
on t-butanol reactivity. The Grana et al. model predicts that 75 % of t-butanol decomposes via
the water elimination reaction in the diffusion flame, releasing 18.6 % of the total water
produced in the flame, which is the second largest of any reaction. To get a qualitative
understanding of the effect of the water elimination reaction, the reaction was removed from the
Grana et al. model, and the extinction limit of t-butanol was recalculated. The new extinction
limit is approximately 100 s-1
greater than the previous prediction for the case of Xf = 0.3 (TWE
= 14.86). Therefore, it is clear that the reaction pathway through the water elimination reaction is
a large factor in limiting the reactivity of t-butanol. It has also been shown that isobutene is more
reactive than t-butanol, so it seems logical that the early formation of water is significant in
limiting the extinction strain rate.
A2.3.4 Sub-model analysis
A2.3.4.1 Acetone
Figure A2.9. Rate of production analysis of the acetone diffusion flame with Xf = 0.2, a =
100s-1
using the Grana et al. model (Grana et al. 2010).
249
Further analysis has been conducted to understand the performance of the sub-models of the
t-butanol model of Grana et al. for each of the major intermediate species. A path flux analysis
for acetone was conducted for the acetone diffusion flame with Xf = 0.2 and a = 100 s-1
and is
summarized in Fig. A2.9. Acetone consumption follows only two paths at these conditions, with
50 % of the fuel being consumed via hydrogen abstraction and another 50 % via thermal
decomposition. The hydrogen abstraction route forms a stable species (mostly molecular
hydrogen), ketene, and a methyl, while the primary thermal decomposition route leads to two
methyls and carbon monoxide. Regardless of the specific path, as in t-butanol oxidation, methyl
is the predominant chain carrier in the acetone fuel sub-model.
A2.3.4.2 Isobutene
The sub-model for isobutene is not as simple as that for acetone. The primary consumption
pathway in the diffusion flame is shown in Fig. A2.10 for the Grana et al. model with Xf = 0.2
and a = 100 s-1
. The computations predict that 83 % of isobutene initially breaks down into the
resonantly stable isobutenyl radical, while another 17 % reacts to form primary and tertiary
isobutyl radicals. Both isobutyl radicals decompose directly into methyl and propylene via
thermal decomposition, thus adding to the methyl dominated radical pool. The isobutene
consumption route through the isobutenyl radical is split in two: isobutenyl either decomposes
into methyl and allene via thermal decomposition or vinylacetylene and propylene via hydrogen
abstraction. Propylene is measured as an intermediate in the diffusion flame, and the observed
concentrations are far less than the model computations. The model predicts 86 % of propylene
to be formed directly from isobutyl and isobutenyl radical reactions. Therefore, the failure of the
Grana et al model to predict this species is further indicative of errors in the isobutene sub-
mechanism.
250
Figure A2.10. Rate of production analysis of isobutene diffusion flame, Xf = 0.2, a = 100s-1
.
Top: Grana et al. model (Grana et al. 2010). Bottom: Wang et al. model (Wang et al. 2007).
To understand further the apparent uncertainties in isobutene oxidation, Wang et al.’s
isobutene model (Wang et al. 2007) is also analyzed in terms of fuel consumption pathways at
the same conditions (Fig. A2.10). Significant differences are observed between the two models.
In the Grana et al. model, 60 % of isobutene is decomposed thermally, forming 7 % of total H
atom production in the isobutene flame, whereas in the Wang et al. model only 17 % of the fuel
251
decomposes by this pathway, accounting for just 2 % of total H atom production. In the flame,
the production of H atoms is extremely significant for predicting extinction limits, as radical
chain branching will result from the reaction with molecular oxygen, producing O atoms and OH
radicals, which will further engage in radical chain branching or heat releasing reactions. This
evaluation of isobutene flame chemistry, and of the uncertainties in its description, highlights the
need for improvement in the understanding of this molecule’s oxidation kinetics in order to
fundamentally model t-butanol oxidation.
A2.4 Concluding remarks
The VPFR study shows that the gas-phase oxidation of t-butanol becomes dominant at ~780
K and 12.5 atm pressure and a residence time of 1.8 seconds. Low temperature oxidation and
negative temperature coefficient behavior does not occur at the studied conditions. Large
quantities of acetone and methane are observed upon the oxidation of t-butanol. Methane is a
result of methyl radical presence, so the observed methane and acetone concentrations are
consistent with a hydrogen abstraction – radical beta-scission mechanism to produce propen-2-ol
or acetone and methyl. The lack of isobutene production in the gas phase chemistry indicates
that, at the low temperature conditions of the flow reactor study, t-butanol is consumed by a
bimolecular radical-oriented reaction rather than by molecular elimination to form water and
isobutene.
In contrast, a sampling study of t-butanol counterflow diffusion flames shows a different
mechanism of oxidation at these high temperature conditions as indicated by the abundant
quantities of isobutene in addition to acetone and methane that were observed. Under flame
conditions, the isobutene and acetone measurements provide mechanistic evidence of a
252
significant competition for fuel consumption between molecular elimination, thermal
decomposition, and hydrogen abstraction – radical beta-scission mechanisms. When speciation
of the t-butanol counterflow diffusion flame is compared to results of the lower temperature flow
reactor study, it is apparent that at high temperature conditions the consumption routes resulting
in the formation of isobutene are dominant over the routes that form acetone.
The extinction limit of t-butanol and the primary intermediate species formed by its high
temperature oxidation have been determined. By a metric of transport-weighted enthalpy (Won
et al. 2012) the extinction limit is found to increase in the order of t-butanol < methane
isobutene < acetone. Therefore, the relative chemical kinetic potential of each species is exposed.
An available kinetic model for t-butanol oxidation has been exercised against the
experimental observations reported here. Under flow reactor conditions, the model is found to
perform reasonably well in reproducing the overall gas phase oxidation at temperatures of 675-
950 K. However, it was observed that at flame temperatures the production of isobutene by this
model is enhanced when compared to experiment. The model is also found to over-predict the
extinction strain rate of isobutene in addition to that of t-butanol.
It is concluded that a better understanding of isobutene chemical kinetics is presently limiting
the development of high fidelity kinetic models for t-butanol combustion. In addition, there is
also uncertainty in the mechanism of initial fuel consumption affecting the intermediate species
population, especially so at high temperatures where the molecular elimination reaction is active.
These specific areas are recommended for further study in order to produce kinetic models for t-
butanol of high fidelity.
253
A3. Uncertainty assessment of species measurements in acetone counterflow
diffusion flames
The results discussed in this chapter were published and presented in:
J. K. Lefkowitz, S. H. Won, Y. Fenard, Y. and Y. Ju (2013). "Uncertainty assessment of species
measurements in acetone counterflow diffusion flames." Proc. Combust. Inst. 34(1): 813-820.
I was responsible for data collection, computation of numerical modeling, analysis of results, and
writing the paper. Sang Hee Won assisted in data collection, analysis of results, and writing of
the paper. Yann Fenard assisted with data collection. The initial concept and guidance for the
research was provided by Prof. Yiguang Ju.
A3.1 Introduction
Kinetic model validation in flames not only requires accurate measurements of global
properties such as flame speeds and extinction limits, but also needs species measurements of
flame structures. Extinction limits of diffusion flames in the counterflow configuration have been
rigorously measured to test the fidelity of the proposed kinetic models (Honnet et al. 2009, Won
et al. 2010, Won et al. 2011, Won et al. 2012). Since the extinction limit of diffusion flames is
not only governed by the kinetics of fuel chemistry, but also by the convective and diffusive
transport in a coupled manner, its measurement provides a unique advantage to evaluate the
overall performance of a kinetic model (Won et al. 2012).
However, model agreement with the extinction limit of diffusion flames is not sufficient to
assess the fidelity of a kinetic model in a broad parametric range. A good example is the recently
proposed skeletal kinetic model for n-propylbenzene oxidation (Won et al. 2011). Although the
254
model has shown good agreement with global parameters, such as the extinction limit of
diffusion flames (Won et al. 2011) and flame speeds (Ji et al. 2012), it completely fails to predict
the formation of styrene, which has been observed as a key intermediate species of n-
propylbenzene oxidation (Gudiyella et al. 2012).
In order to obtain a detailed evaluation of the performance of a chemical kinetic model, the
speciation profiles of flames must be measured. To quantify fuel oxidation in diffusion flames, a
sampling probe technique coupled with gas chromatograph (GC) analysis has been widely used
(Tsuji 1982, Smooke et al. 1988, Sinha et al. 2004, Gaïl et al. 2007, Saxena et al. 2007,
Bufferand et al. 2009, Jahangirian et al. 2009, Sarathy et al. 2009, Frassoldati et al. 2010,
Lefkowitz et al. 2012a). Recently, in a study of tertiary-butanol oxidation (Lefkowitz et al.
2012a), the speciation profiles of a diffusion flame have been measured by the sampling
technique. The detailed species results present valuable information for the understanding of
high temperature tertiary-butanol chemistry. Unfortunately, many measurements using this
technique in diffusion flames have displayed large and consistent discrepancies with model
predictions, namely an earlier fuel consumption profile and broader intermediate species profiles
(Gaïl et al. 2007, Frassoldati et al. 2010, Lefkowitz et al. 2012a). Few researchers have
emphasized these large and consistent uncertainties using the probe technique. The large
differences between measurements and model predictions have simply been attributed to the
inaccuracy of the kinetic models without careful assessment of the uncertainty associated with
the intrusive sampling technique. A recent study of uncertainty in extinction limit measurements
using particle image velocimetry (PIV) showed that a sufficiently large nozzle separation
distance in the counterflow configuration is needed to obtain consistent extinction limit
measurements (Sarnacki et al. 2012). However, the uncertainty associated with the probe
255
sampling technique for species measurement in a counterflow diffusion flame remains
unresolved, and, despite its importance, no appropriate guidance for quantitative measurements
is available.
Motivated by the above discussions, the objective of the present study is to investigate and
quantify the uncertainty of species measurements in acetone diffusion flames by using non-
intrusive diagnostic approaches such as PIV and planar laser induced fluorescence (PLIF).
Acetone, which provides a strong LIF signal, is chosen as the fuel and also serves to non-
intrusively quantify the fuel profile. Possible reductions of the flow perturbation by the sampling
probe are investigated by changing the flame location from the oxidizer side to the fuel side of
the stagnation plane. Lastly, the speciation profile of an acetone diffusion flame is compared to
published models in order to assess the current understanding of this important intermediate
species.
A3.2 Experiments and numerical simulations
Speciation profiles in the diffusion flame have been measured with acetone (99.9+% in
purity) as the fuel in an atmospheric pressure counterflow burner integrated with a fuel
vaporization system. A schematic of the experimental apparatus is shown in Fig. A3.1. Details of
experiments were reported previously (Won et al. 2010, Won et al. 2011, Lefkowitz et al. 2012a,
Won et al. 2012).
The liquid fuel and heated nitrogen are injected into the vaporization chamber through
coaxial nozzles. The vaporized fuel/diluents mixture is transported directly to the upper burner.
The upper nozzle exit temperature is kept at 400 ± 5 K using a PID controller. The oxidizer flow
rate to the lower burner nozzle is maintained at 298 ± 5 K. The two hydrodynamically
256
converging nozzles of 13 mm i.d. at the exits are employed to prevent boundary layer separation.
The global strain rate, 1 / 2
2 1 2 1 22 / 1 / ( / )a U L U U
, is utilized (Seshadri et al. 1978).
Here, U is the flow velocity at the nozzle exit, L the separation distance between the nozzles,
the density, and subscripts 1 and 2 represent the upper and lower nozzles, respectively. To
incorporate the change of flame position between air and oxygen oxidizers, a is determined at the
oxidizer and fuel sides for air and oxygen oxidizer, respectively.
Figure A3.1. Schematic of experiments.
To measure the spatial distribution of stable species, a micro-probe sampling system is
employed. This system is comprised of a 363 µm o.d. and 220 µm i.d. fused silica microprobe
(SGE Analytical Science, P/N: 0624469) that draws sample gas from the flame into a mixing
chamber. The sample gas is diluted by nitrogen and stored in heated sample storage loops (373 ±
10 K) for GC analysis. To confirm choked flow in the sample probe, the pressure in the mixing
chamber is measured as a function of time during sampling. A linear correlation of pressure with
time is observed, confirming the existence of aerodynamic quenching in the sample probe . The
probe position is controlled by a translation stage actuated in the burner’s axial and radial
Heated N2
N2La
se
r s
he
et
for
LIF
an
d P
IV
Air/O2
Fuel/N2(He)
P
G
Positioning
Stage
TC TC
Temperature
Controller
PG TC
Mixing Chamber Heater
PG Pressure Gauge
TC Thermocouple
Needle Valve
On/Off Valve
Three-way Valve
VacuumMultiple
Port
Valve
PG
N2 Bypass
N2 Source
257
dimensions by stepper motors with an accuracy of 5 µm per step. All measurements are taken
with the probe fixed 1 mm radially away from the centerline of the nozzles. The uniformity of
species profiles has been preliminarily tested in the radial direction by sampling measurements,
and uniform species profiles are confirmed between 0 and 2 mm from the axis of symmetry. The
sample gas is analyzed by the GC-FID system (Agilent 7890A). Details of the sample gas
preparation and the uncertainty have been reported previously (Lefkowitz et al. 2012a).
For velocity profile measurements at the nozzle exit, a PIV system has been used with
alumina particles (0.3 m diameter) seeded into the oxidizer flow. The PIV system is composed
of a dual head Nd:YAG laser (New Wave Research, Solo PIV), a synchronizer (TSI, Model
610034), and a CCD camera (TSI, Model 630059). The PLIF technique has been used to
measure the fuel (acetone) and OH profiles in the diffusion flame. Details of PLIF measurements
have been reported previously (Won et al. 2010, Won et al. 2011). The OH and acetone PLIF
signals are captured through UG-11 and WG-305 optical filters for OH, and a WG-305 optical
filter for acetone. For the simultaneous OH and acetone PLIF measurements, the laser
wavelength of OH excitation is used and only one optical filter, the WG-305, is placed in front of
the ICCD camera.
Numerical simulations of the sampling experiments are performed using the OPPDIF module
of the CHEMKIN package (Kee et al. 2003). The structures of acetone diffusion flames are
computed with four kinetic models: Pichon et al. (Pichon et al. 2009), Fischer et al. (Fischer et al.
2000), Grana et al. (Grana et al. 2010), and Wang et al. (Wang et al. 2007), which will be
hereafter referred to as models 1 through 4, respectively. Further details on calculations were
described previously (Won et al. 2010, Won et al. 2011, Won et al. 2012).
258
A3.3 Results and discussion
A3.3.1 Discrepancy in species distribution measurements
Figure A3.2 shows the fuel (acetone) and ethylene (C2H4) profiles measured by sampling in
the acetone diffusion flame at a fuel mole fraction of Xf = 0.20, strain rate of a = 100 s-1
, and a
separation distance of L = 9 mm for air as the oxidizer. Compared to the numerical results, the
sampling results exhibit two noticeable disparities. One is much earlier fuel consumption in the
measurements, and the other is the broadened and irregular spatial profile of C2H4. The spatial
Figure A3.2. Comparison of speciation measurements of acetone and ethylene (C2H4) in an
acetone diffusion flame to numerical results with models 1 (solid line) and 2 (dashed line);
Xf = 0.20, a = 100 s-1
, L = 9 mm, and the oxidizer is air.
0
2000
4000
6000
8000
0
0.05
0.1
0.15
0.2
0.25
0 1 2 3 4 5 6
C2 H
4m
ole
frac
tion
[pp
m]A
ce
ton
e m
ole
fra
cti
on
Distance from fuel nozzle exit [mm]
Sampling
Pichon et al. model
Kaiser et al. model
acetone
ethylene, C2H4
Xf = 0.20, a = 100 1/s, and L = 9 mm
air for oxidizer
Stagnation
plane
259
profile of C2H4 indicates a slight movement of flame location at 4 mm from the nozzle exit due
to the flow perturbation of the probe. The noticeable mismatch in fuel profiles has been observed
frequently not only in our present and previous measurements (Lefkowitz et al. 2012a), but also
in other similar studies (Gaïl et al. 2007, Frassoldati et al. 2010, Grana et al. 2010). The reason
for the mismatches has been attributed either to possible fuel loss in the sampling tube by
catalytic wall reactions or to significant flow perturbations by the intrusive sampling probe. It
was also argued previously that the measurement of the temperature profile is important to
account for the flow perturbation (Bufferand et al. 2009). However, a recent study (Sarnacki et
al. 2012) has revealed that the initial velocity profile at the nozzle exits could be significantly
perturbed by thermal expansion in the reaction zone, depending on the separation distance
between the two nozzles. Consequently, sampling measurements in the counterflow diffusion
flame requires more quantitative analysis not only on the uncertainty associated with the
intrusive measurement but also on the validity of the one-dimensional approximation of the
experimental system. In order to address this issue, the velocity and fuel consumption profiles
are measured by PIV and PLIF, respectively.
A3.3.2 Effects of separation distance
Figure A3.3(a) shows the axial velocity profiles in the radial direction, taken at the oxidizer
nozzle exit, for different nozzle separation distances. The velocity was measured by PIV in the
presence of an acetone diffusion flame (reacting flow) at Xf = 0.20, a = 100 s-1
with air as the
oxidizer. Although not shown in this figure, the velocity profiles without the presence of the
flame (non-reacting flow) have been taken to confirm the validity of the converging counterflow
nozzles. The axial velocity profiles shown in Fig. A3.3 are normalized by the mean velocity
measured from the non-reacting flow. The results show significant deviation from uniformity at
260
small nozzle separation distances. This finding is consistent with the recent report (Sarnacki et al.
2012). The flow non-uniformity is defined as the difference between the maximum and
minimum velocities (excluding the outer boundary layer) divided by the maximum velocity, and
the results as a function of separation distance are embedded, Fig. A3.3(b). It is clear that by
increasing the separation distance from 7 to 21 mm, the axial velocity profile becomes more
uniform, thus allowing the one-dimensional plug flow approximation to be employed in the
calculations.
Figure A3.3. Radial velocity profiles from PIV measurements at the nozzle exit for three
separation distances (a) and the evaluated non-uniformity with separation distance L (b)
for an acetone diffusion flame at Xf = 0.20, a = 100 s-1
, and the oxidizer is air.
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
0 0.2 0.4 0.6 0.8 1 1.2
No
rma
lize
d a
xia
l ve
loc
ity U
= u
/uo
Normalized radial distance R = r/ro
L = 9 mm
L = 15 mm
L = 21 mm (a)
0
10
20
30
40
50
5 10 15 20 25
No
n-U
nif
orm
ity
(um
ax-u
min
)/u
ma
x [%
]
Separation distance L [mm]
(b)
261
The fuel (acetone) profiles in the axial direction are measured by the PLIF technique near the
centerline (40 pixels, corresponding to 2 mm) at three different separation distances and for both
air and pure oxygen as the oxidizer (to vary the relative location of the flame to the stagnation
plane). The PLIF results (closed symbols) are plotted in Fig. A3.4 and compared with the
numerical predictions (lines). The results in Fig. A3.4 clearly show significant disparities of fuel
profiles for the cases of L = 9 mm and 15 mm, as expected from the velocity measurements (Fig.
A3.3). For the case of L = 25 mm, where the velocity profiles are found to be uniform, the fuel
profiles measured by PLIF exhibit excellent agreement with those of the numerical computations
for both air and oxygen as oxidizers.
The comparison between acetone PLIF and the sampling results (open symbols) provides
insightful information concerning flow perturbations due to intrusive measurements by a
microprobe. In the case of air as the oxidizer, the sampling measurements with L = 9 and 15 mm
match with acetone PLIF measurements in the region of early fuel decomposition, but deviate in
terms of the fuel consumption rate closer to the stagnation plane. From this comparison, it can be
concluded that the mismatch of fuel profiles between the sampling measurements and model
results is predominantly caused by inappropriate boundary conditions in the counterflow burner
and only partially due to flow perturbations from the probing.
Interestingly, the sampling measurements at L = 25 mm show completely different results
between using air and oxygen as the oxidizer. In the case of air, where the flame is located on the
oxidizer side of the stagnation plane, the sampling measurements mismatch completely with
those from PLIF and numerical results, indicating a significant flow perturbation by the probe. In
contrast, the sampling results in the case of pure oxygen as the oxidizer, where the flame is
located on the fuel side of the stagnation plane, match reasonably well with both the LIF and
262
Figure A3.4. Comparison of acetone profiles between PLIF measurements (closed
symbols), sampling measurements (open symbols), and numerical results with model 1
(lines) for three separation distances for acetone diffusion flames at a = 100 s-1
. With air
oxidizer at Xf = 0.20 (circular symbols and solid lines) and with oxygen oxidizer at Xf = 0.05
(triangle symbols and dashed lines); red (L = 9 mm), blue (L = 15 mm), and black (L = 25
mm).
numerical results. This finding is very important for species measurements. In the case of air, the
flame locates on the oxidizer side close to the stagnation plane, where the diffusive terms are
dominant over the convective terms. Consequently, when the sampling probe approaches the
stagnation plane, which is also the region of the reaction zone, any slight perturbation of the flow
field by the existence of the probe breaks the balance between local convection and diffusion,
blocking the transport of fuel molecules to the flame zone and leading to a large flow
0.00
0.05
0.10
0.15
0.20
0.25
0 2 4 6 8 10 12 14 16
Ac
eto
ne
mo
le f
rac
tio
n
Distance from fuel nozzle [mm]
L = 9 mm
L = 15 mmL = 25 mm
air oxidizer
O2 oxidizer
263
perturbation. In the case of oxygen, the diffusion flame locates far upstream on the fuel side.
Therfore, throughout the fuel consumption region, the convective term outweighs the diffusive
term in the fuel transport equation, resulting in only a minimal perturbation of the fuel profile,
and it can be concluded that the composition of the fuel and oxidizer streams plays an important
role in the flow perturbation by the probe, since it determines the location of the diffusion flame.
A3.3.3 Quantification of the flow perturbation due to the sampling probe
To quantify the flow perturbation by the probe, both the fuel profiles and the OH
distributions are measured in terms of spatial location by simultaneous acetone and OH PLIF
measurements. For both the air and oxygen oxidizer cases, the diffusion flames are established at
L = 25 mm, a = 100 s-1
, and Xf = 0.20 and 0.05, respectively. Direct PLIF images are obtained by
varying the position of the sampling probe from the fuel nozzle exit (0 mm) to 15 mm in 1 mm
intervals. Some of the representative images are shown in Fig. A3.5 for air oxidizer cases (a-d)
and for oxygen oxidizer cases (e-h). The fuel profiles for only oxygen as the oxidizer are shown
in Fig. A3.5(i), since the overall behavior of flow perturbations are found to be qualitatively
identical for both oxidizer cases.
Without the existence of the probe in Fig. A3.5(a, e), the fuel profiles from acetone PLIF
match well with the model results as shown in Fig. A3.4. When the probe is inserted at the fuel
nozzle exit (Fig. A3.5(b, f)), although there exists a visible radial deformation of the flame
structure just below the probe, the flow perturbation is found to be minimal at the centerline. As
the probe is moved downstream to 12 mm, just above the stagnation plane, the flow perturbation
264
Figure A3.5. Direct images of simultaneous acetone and OH PLIF measurements for the
separation distance L = 25 mm at a = 100 s-1
; (a-d) at Xf = 0.20 with air oxidizer, (e-h) at Xf
= 0.05 with oxygen oxidizer; (a and e) without probe, (b and f) probe at 0 mm from the fuel
nozzle, (c and g) at 12 mm, and (d and h) at 15 mm (white dotted line indicates the peak
OH position for the case without the probe). Bottom plot (i) shows acetone and OH profiles
at the centerline as a function of distance for oxygen oxidizer cases, for a number of probe
positions, along with numerical results (model 1).
becomes noticeable, but remains within 0.3 mm in terms of the fuel profile shift, Fig. A3.5(c,
g). However, when the probe is positioned beyond the stagnation plane, the flow perturbation
becomes significant, resulting in a shift of the entire fuel profile and the reaction zone to the
oxidizer side by more than 1 mm, Fig. A3.5(d, h), which is comparable to the entire flame
0
0.01
0.02
0.03
0.04
0.05
0.06
0.00
0.20
0.40
0.60
0.80
1.00
1.20
8 9 10 11 12 13 14 15
Ac
eto
ne
mo
le
frac
tion
LIF
sin
ga
ls [
a. u
]
Distance from the fuel nozzle [mm]
probe at 0 mmprobe at 8 mmprobe at 12 mmprobe at 15 mmPichon et al. model
Xf = 0.05, a = 100 s-1
oxygen for oxidizer
Acetone LIF
OH LIF
(e) (f) (g) (h)OHacetone
(a) (b) (c) (d)O
2,o
xid
ize
ra
ir,
ox
idiz
er
265
thickness. These results are shown quantitatively in Fig. A3.5(i). It is concluded that in order to
achieve reliable speciation measurements using the probe sampling technique, the flame should
be located far on the fuel side of the stagnation plane, which can be accomplished by varying the
fuel/oxygen ratio.
A simple scaling analysis is performed based on the Reynolds number around the cylindrical
probe, ReP, with L = 25 mm and a = 100 s-1
. The ReP with a 363 µm o.d. probe varies from 0.5 to
2.5, depending on the location of the probe. This result indicates that the flow field is near the
Stokes flow regime, so the flow field is able to fully recover with little or no wake. This
rationalizes why the flow uncertainty quantified by PLIF ( 0.3 mm) is comparable to the outer
diameter of the probe. Since the Stokes regime is defined as ReP << 1, helium dilution is tested
here to reduce the ReP by a factor of 8.
To examine the impact of binary mass diffusivity on sampling, the direct PLIF image and
corresponding numerical results with helium dilution in the fuel stream are presented in Fig.
A3.6. Unfortunately, the results show a complete mismatch between the model and the PLIF
measurements. This mismatch can be explained based on the Reynolds number at the nozzle exit,
Ren. In the case of nitrogen dilution, the Ren is about 300, whereas with helium dilution the Ren
is reduced to about 60. The numerical model has been developed based on the interaction of two
near-field (developing regime) solutions of opposing jets at high Ren (Seshadri et al. 1978). The
length of the developing regime in a single jet can be estimated from boundary layer theory (Lee
et al. 1997). With Re of 60, the length is estimated to be 12 mm, which is comparable to half of L
in this measurement. As a result, the flow condition tested with helium dilution in Fig. A3.6 fails
to meet the approximation made in the theory, thus resulting in the significant mismatch
observed.
266
Figure A3.6. (a) Image of acetone and OH PLIF measurements for an acetone diffusion
flame; fuel side is acetone/helium, oxidizer is oxygen; L = 25 mm, Xf = 0.04, and a = 100 s-1
,
(b) Acetone and OH profiles from numerical results (model 1).
It is noteworthy that the OH PLIF signal, which is indicative of the reaction zone position, is
remarkably curved even at the centerline as compared to the nitrogen dilution cases shown in
Fig. A3.5(e). Although the Ren can be increased at higher strain rates, this flame curvature effect
remains significant, thus resulting in local extinction at the centerline as shown in Fig. A3.7 for a
= 175 s-1
. Both acetone and OH PLIF signals and the Abel-transformed chemiluminescence
signal in Fig. A3.7 clearly show local extinction at the centerline and also the existence of an
edge flame (Chung 2007). The curved fuel profile and flame can be attributed to the excessively
fast diffusion of helium in the radial direction, causing strong multi-dimensional effects.
0 10
00
20
00
30
00
0
0.0
1
0.0
2
0.0
3
0.0
4
0.0
5
05
10
15
20
25
OH mole fraction [ppm]
Acetone mole fraction
Dis
tan
ce
from
fue
l no
zzle
[mm
]
acetone
OH
Stagnation Plane
OH LIF
Acetone LIF
(a) (b)
L=
25 m
m
267
Figure A3.7. Simultaneous OH and acetone PLIF images (a) and abel-transformed cross
section image of flame chemiluminesence (b) for acetone/helium diffusion flame with
oxygen oxidizer at Xf = 0.04 and a = 175 s-1
.
A3.3.4 Speciation profiles of intermediate species
Figure A3.8 summarizes the acetone oxidation pathways in the diffusion flame. The general
oxidation pathways of acetone are found to be qualitatively similar in all the models, but
quantitatively different. All models reveal that the methyl radical, CH3, plays a significant role as
a major intermediate either from uni-molecular decomposition reactions or from H abstraction
reactions from acetone. The ratio of the two fuel decomposition channels significantly affects the
radical pool population, controlling the global reactivity (Won et al. 2012). Since all primary fuel
consumption pathways lead to the production of CH3, it is important to measure ketene to
confirm the flux by the H abstraction channel in acetone consumption suggested in the model.
Unfortunately, since the GC system used in the present study is not able to quantify the ketene
concentration, the models are evaluated based only on the formation of C2H6 and its
consumption pathways through C2H4 and C2H2.
OH LIFAcetone LIF
(a)(b)
268
Figure A3.8. Schematic of acetone oxidation pathways in the diffusion flame, UD: uni-
molecular decomposition reaction, H-abs: H abstraction reaction.
Figure A3.9 shows the comparisons between the numerical results with four different kinetic
models and the sampling measurements of (a) acetone, (b) C2H6, (c) C2H4, and (d) C2H2 with the
quantified uncertainty from the previous section. The results in Fig. A3.9(a) show that all models
predict the rate of fuel consumption reasonably well and within the uncertainty of measurements.
In Fig. A3.9(b), the measured C2H6 mole fractions are remarkably higher than those from the
model predictions. Note that the radical pool concentration in the diffusion flame is an order of
magnitude greater than those in flow reactor (Lefkowitz et al. 2012a) and shock tube
experiments (Vasu et al. 2010). For example, the maximum concentrations of H, O, and OH are
in the range of 1000~4000 ppm in the flame, whereas 1~10 ppm in the other experiments. Thus,
model accuracy in radical recombination reactions can be better examined by species
measurements in diffusion flames (Pichon et al. 2009, Dooley et al. 2011).
AcetoneCH3COCH3
CH3COCH2
CH3CO
KeteneCH2CO
CH3
C2H6CH2O
UD
UD
H-abs
UD
UDUD
HCCOCH4
269
Figure A3.9. Comparison of speciation profiles for intermediate species of an acetone
diffusion flame; oxygen oxidizer, L = 25 mm, Xf = 0.20, a = 100 s-1
, with numerical results.
In this regard, the accuracy of the CH3 recombination reaction, CH3 + CH3 (+M) = C2H6 (+M),
becomes questionable for all tested models, since C2H6 is predominantly generated by this
recombination reaction. A similar disparity is also found for the profile of CH4, which is
produced from the recombination reaction of CH3 + H (+M) = CH4 (+M). Similar sensitive
behavior of acetone kinetics to these two recombination reactions has been previously shown
(Pichon et al. 2009). The models show significantly better agreement for C2H4 and C2H2 profiles.
The main production pathway of C2H4 is via hydrogen abstraction from C2H6 and subsequent
beta-scission of C2H5 (ethyl radical). Production of C2H2 is via hydrogen abstraction from C2H4
and subsequent beta-scission of C2H3 (vinyl radical). The comparison for both C2H4 and C2H2
reveals that only one model, model 3, is able to emulate both species concentrations
0
10000
20000
30000
40000
50000
60000
9 9.5 10 10.5 11 11.5 12
Ac
eto
ne
mo
le f
rac
tio
n
[pp
m]
Distance from fuel nozzle [mm]
measurementmodel 1model 2model 3model 4
(a) acetone0
1000
2000
3000
4000
5000
6000
9 9.5 10 10.5 11 11.5 12
Sp
ec
ies
mo
le f
rac
tio
n
[pp
m]
Distance from fuel nozzle [mm]
measurementmodel 1model 2model 3model 4
(b) C2H6
0
1000
2000
3000
4000
9 9.5 10 10.5 11 11.5 12
Sp
ec
ies
mo
le f
rac
tio
n
[pp
m]
Distance from fuel nozzle [mm]
measurementmodel 1model 2model 3model 4
(c) C2H40
300
600
900
1200
1500
9 9.5 10 10.5 11 11.5 12
Sp
ec
ies
mo
le f
rac
tio
n
[pp
m]
Distance from fuel nozzle [mm]
measurementmodel 1model 2model 3model 4
(d) C2H2
270
simultaneously. It is noteworthy that large disagreement among the models for the well studied
C2 chemical kinetics still exists.
A3.4 Concluding remarks
In an effort to quantify the effect of flow perturbation on the uncertainty of species sampling
in counterflow diffusion flames, non-intrusive PIV and PLIF measurements have been employed
in acetone diffusion flames to measure the velocity, fuel, and OH profiles. The PIV and PLIF
results have revealed that the separation distance between the two nozzles needs to be
sufficiently large to achieve uniform radial velocity profiles at the nozzle exit for quantitative
species measurement. It is found that the composition of the fuel and oxidizer streams needs to
be carefully selected, since it determines the diffusion flame location and has a strong effect on
the uncertainty of species sampling. PLIF measurements indicate that the diffusion flame needs
to be located at the fuel side far from the stagnation plane in order to obtain reliable speciation
measurements. In convection dominated environments, the uncertainty of the flow perturbation
by the probe has been quantified and found to be comparable to the outer diameter of the probe,
0.3 mm. A simple Reynolds number analysis shows that the flow near the probe is just on the
outskirts of the Stokes regime. In order to suppress the Reynolds number further, helium dilution
of the fuel stream has been tested, but exhibits significant deviation against the model prediction
due to excessive diffusion process. The structure of an acetone diffusion flame has been analyzed
through a comparison between the models and speciation measurements. From the measurements
of C2H6, C2H4, and C2H2, it has been revealed that further detailed understanding of C2 chemistry
is required in order to improve the fidelity of acetone kinetic models.
271
References
Adamovich, I. V., et al. (2009a). "Plasma assisted ignition and high-speed flow control: non-
thermal and thermal effects." Plasma Sources Sci. Technol. 18(3): 034018.
Adamovich, I. V., Nishihara, M., Choi, I., Uddi, M. and Lempert, W. R. (2009b). "Energy
coupling to the plasma in repetitive nanosecond pulse discharges." Phys. Plasmas 16(11):
113505.
Adamovich, I. V. and Lempert, W. R. (2015a). "Challenges in understanding and predictive
model development of plasma-assisted combustion." Plasma Phys Control. Fusion 57(1):
014001.
Adamovich, I. V., Li, T. and Lempert, W. R. (2015b). "Kinetic mechanism of molecular energy
transfer and chemical reactions in low-temperature air-fuel plasmas." Philos. Trans. R. Soc.
London, Ser. A 373(2048).
Aleksandrov, N. L., Kindysheva, S. V., Kosarev, I. N., Starikovskaia, S. M. and Starikovskii, A.
Y. (2009a). "Mechanism of ignition by non-equilibrium plasma." Proc. Combust. Inst. 32(1):
205-212.
Aleksandrov, N. L., Kindysheva, S. V., Kukaev, E. N., Starikovskaya, S. M. and Starikovskii, A.
Y. (2009b). "Simulation of the ignition of a methane-air mixture by a high-voltage nanosecond
discharge." Plama Phys. Rep. 35(10): 867-882.
Alves, L. L., et al. (2013). "Comparisons of sets of electron–neutral scattering cross sections and
swarm parameters in noble gases: II. Helium and neon." J. Phys. D: Appl. Phys. 46(33): 334002.
Anikin, N. B., Starikovskaia, S. M. and Starikovskii, A. Y. (2006). "Oxidation of saturated
hydrocarbons under the effect of nanosecond pulsed space discharge." J. Phys. D: Appl. Phys.
39(15): 3244-3252.
Armstrong, B. H. (1967). "Spectrum Lines Profiles: The Voigt Function." J Quant Spectrosc Ra
7: 61-88.
Balamuta, J., Golde, M. F. and Ho, Y.-S. (1983). "Product distributions in the reactions of
excited noble-gas atoms with hydrogen-containing compounds." J. Chem. Phys. 79(6): 2822.
272
Ballal, D. R. and Lefebvre, A. H. (1975). "The influence of flow parameters on minimum
ignition energy and quenching distance." Symposium (International) on Combustion 15(1):
1473-1481.
Bao, A., Utkin, Y. G., Lou, G. and Adamovich, I. V. (2007). "Ignition of Ethylene–Air and
Methane–Air Flows by Low-Temperature Repetitively Pulsed Nanosecond Discharge Plasma."
IEEE Trans. Plasma Sci. 35(6): 1628-1638.
Baulch, D., et al. (1992). "Evaluated kinetic data for combustion modelling." J. Phys. Chem. Ref.
Data 21(3): 411-734.
Baulch, D. L. (2005). "Evaluated Kinetic Data for Combustion Modeling: Supplement II." J.
Phys. Chem. Ref. Data 34(3): 757.
Becker, K. H., Groth, W. and Schurath, U. (1971). "The quenching of metastable O2(1Δg) and
O2(1 g+) molecules." Chem. Phys. Lett. 8: 259-262.
Biagi, S. F., Biagi-v8.9 database.
Biagi, S. F., Biagi-v7.1 database.
Blanc, M., Guest, P., von Elbe, G. and Lewis, B. (1947a). "Ignition of explosive gas mixtures by
electric sparks. I. Minimum ignition energies and quenching distances of mixtures of methane,
oxygen, and inert gases." J. Chem. Phys. 15(11): 798-802.
Blanc, M., Guest, P., von Elbe, G. and Lewis, B. (1948). "Ignition of explosive gas mixtures by
electric sparks: III. Minimum ignition energies and quenching distances ofmixtures of
hydrocarbons and ether with oxygen and inert gases". Symposium on Combustion and Flame,
and Explosion Phenomena 3(1), Elsevier.
Blanc, M. V., Guest, P. G., von Elbe, G. and Lewis, B. (1947b). "Ignition of Explosive Gas
Mixtures by Electric Sparks. I. Minimum Ignition Energies and Quenching Distances of
Mixtures of Methane, Oxygen, and Inert Gases." J. Chem. Phys. 15(11): 798.
Boyle, A. and Llewellyn, F. (1947). "The electrostatic ignitibility of various solvent vapour‐air
mixtures." Journal of the Society of Chemical Industry 66(3): 99-102.
273
Bozhenkov, S. A., Starikovskaia, S. M. and Starikovskii, A. Y. (2003). "Nanosecond gas
discharge ignition of H2− and CH4− containing mixtures." Combust. Flame 133(1-2): 133-146.
Bradford, B. W. and Finch, G. I. (1948). "The mechanism of ignition by electric discharges."
Proceedings of the Symposium on Combustion 1: 112-126.
Bromberga, L., Cohna, D. R., Rabinovicha, A. and Alexeev, N. (1999). "Plasma catalytic
reforming of methane." Int. J. Hydrogen Energy 24: 1131-1137.
Brown, P. N., Byrne, G. D. and Hindmarsh, A. C. (1989). "VODE: A variable-coefficient ODE
solver." SIAM journal on scientific and statistical computing 10(5): 1038-1051.
Brumfield, B., Sun, W., Ju, Y. and Wysocki, G. (2013). "Direct In Situ Quantification of
HO2from a Flow Reactor." J. Phys. Chem. Lett. 4(6): 872-876.
Brumfield, B., Sun, W., Wang, Y., Ju, Y. and Wysocki, G. (2014). "Dual modulation Faraday
rotation spectroscopy of HO2 in a flow reactor." Opt. Lett. 39(7): 1783-1786.
Bufferand, H., Tosatto, L., La Mantia, B., Smooke, M. D. and Gomez, A. (2009). "Experimental
and computational study of methane counterflow diffusion flames perturbed by trace amounts of
either jet fuel or a 6-component surrogate under non-sooting conditions." Combust. Flame
156(8): 1594-1603.
Burke, M. P., Chaos, M., Ju, Y., Dryer, F. L. and Klippenstein, S. J. (2012). "Comprehensive
H2/O2 kinetic model for high‐pressure combustion." Int. J. Chem. Kinet. 44(7): 444-474.
Busby, K., Corrigan, J., Yu, S.-T., Williams, S., Carter, C., Schauer, F., Hoke, J., Cathey, C. and
Gundersen, M. (2007). "Effects of Corona, Spark and Surface Discharges on Ignition Delay and
Deflagration-to-Detonation Times in Pulsed Detonation Engines". 45th AIAA Aerospace
Sciences Meeting, Reno, NV.
Canny, J. (1986). "A Computational Approach to Edge Detection." IEEE Trans. Pattern Anal.(6):
679-698.
Capitelli, M., Ferreira, C. M., Osipov, A. I. and Gordiets, B. F. (2000). Plasma kinetics in
atmospheric gases, Springer.
274
Cathey, C., Cain, J., Wang, H., Gundersen, M. A., Carter, C. and Ryan, M. (2008). "OH
production by transient plasma and mechanism of flame ignition and propagation in quiescent
methane–air mixtures." Combust. Flame 154(4): 715-727.
Cathey, C. D., Tang, T., Shiraishi, T., Urushihara, T., Kuthi, A. and Gunderen, M. A. (2007).
"Nanosecond Plasma Ignition for Improved Performance of an Internal Combustion Engine."
IEEE Trans. Plasma Sci. 35(6): 1664-1668.
Chen, Z. and Ju, Y. (2007). "Theoretical analysis of the evolution from ignition kernel to flame
ball and planar flame." Combust. Theor. Model. 11(3): 427-453.
Chen, Z. (2009). Studies on the Initiation, Propagation, and Extinction of Premixed Flames.
Mechanical and Aerospace Engineering. Princeton, NJ, Princeton University. Ph.D.
Chen, Z., Burke, M. P. and Ju, Y. (2009a). "Effects of compression and stretch on the
determination of laminar flame speeds using propagating spherical flames." Combust. Theor.
Model. 13(2): 343-364.
Chen, Z., Burke, M. P. and Ju, Y. (2009b). "Effects of Lewis number and ignition energy on the
determination of laminar flame speed using propagating spherical flames." Proc. Combust. Inst.
32(1): 1253-1260.
Chen, Z., Burke, M. P. and Ju, Y. (2011). "On the critical flame radius and minimum ignition
energy for spherical flame initiation." Proc. Combust. Inst. 33(1): 1219-1226.
Choudhury, T., Lna, M., Lin, C.-Y. and Sanders, W. (1990). "Thermal decomposition of t-butyl
alcohol in shock waves." Combust. Sci. Technol. 71(4-6): 219-232.
Chung, S. H. (2007). "Stabilization, propagation and instability of tribrachial triple flames."
Proc. Combust. Inst. 31(1): 877-892.
Cohen, N. and Westberg, K. (1991). "Chemical Kinetic Data Sheets for High‐Temperature
Reactions. Part II." J. Phys. Chem. Ref. Data 20(6): 1211-1311.
Dagaut, P., Sarathy, S. M. and Thomson, M. J. (2009a). "A chemical kinetic study of n-butanol
oxidation at elevated pressure in a jet stirred reactor." Proc. Combust. Inst. 32(1): 229-237.
275
Dagaut, P. and Togb , C. (2009b). "Experimental and modeling study of the kinetics of oxidation
of butanol− n-heptane mixtures in a jet-stirred reactor." Energy & Fuels 23(7): 3527-3535.
DeMore, W. B., Sander, S., Golden, D., Hampson, R., Kurylo, M., Howard, C., Ravishankara,
A., Kolb, C. and Molina, M. (1992). "Chemical kinetics and photochemical data for use in
stratospheric modeling." JPL Publication 97-4.
Demtröder, W. (2013). Laser spectroscopy: basic concepts and instrumentation, Springer Science
& Business Media.
DeSain, J. D., Klippenstein, S. J., Miller, J. A. and Taatjes, C. A. (2003). "Measurements, theory,
and modeling of OH formation in ethyl+ O2 and propyl+ O2 reactions." J. Phys. Chem. A
107(22): 4415-4427.
Dievart, P., et al. (2014). "On the pyrolysis and oxidation chemistry of of methyl formate and
methanol." Combust. Flame.
Diévart, P., Won, S. H., Uddi, M., Dooley, S., Dryer, F. L. and Ju, Y. (2011). "Measurements of
Extinction Limits of Methyl Decanoate Diffusion Flames and A Kinetic Study". 49th AIAA
Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition.
Do, H., Cappelli, M. A. and Mungal, M. G. (2010a). "Plasma assisted cavity flame ignition in
supersonic flows." Combust. Flame 157(9): 1783-1794.
Do, H., Im, S.-k., Cappelli, M. A. and Mungal, M. G. (2010b). "Plasma assisted flame ignition of
supersonic flows over a flat wall." Combust. Flame 157(12): 2298-2305.
Dooley, S., Burke, M., Chaos, M., Stein, Y., Dryer, F., Zhukov, V. P., Finch, O., Simmie, J. and
Curran, H. (2010). "Methyl formate oxidation: Speciation data, laminar burning velocities,
ignition delay times, and a validated chemical kinetic model." Int. J. Chem. Kinet. 42(9): 527-
549.
Dooley, S., Dryer, F., Yang, B., Wang, J., Cool, T., Kasper, T. and Hansen, N. (2011). "An
experimental and kinetic modeling study of methyl formate low-pressure flames." Combust.
Flame 158(4): 732-741.
Dorko, E., McGhee, D., Painter, C., Caponecchi, A. and Crossley, R. (1971). "Shock tube
isomerization of cyclopropane." J. Phys. Chem.-U.S. 75(16): 2526-2530.
276
Dunlea, E. J., Talukdar, R. K. and Ravishankara, A. R. (2005). J. Phys. Chem. A. 109(3912-
3920).
Dyer, R., Naples, A., Kaemming, T., Hoke, J. and Schauer, F. (2012). "Parametric testing of a
unique rotating detonation engine design". 50th AIAA Aerospace Sciences Meeting including
the New Horizons Forum and Aerospace Exposition.
ElSabbagh, M., Kado, S., Ikeda, Y. and Sasaki, K. (2011). "Measurements of Rotational
Temperature and Density of Molecular Nitrogen in Spark-Plug Assisted Atmospheric-Pressure
Microwave Discharges by Rotational Raman Scattering." Japanese Journal of Applied Physics
50(7): 076101.
Farooq, A., Jeffries, J. B. and Hanson, R. K. (2008). "In situcombustion measurements of H2O
and temperature near 2.5 µm using tunable diode laser absorption." Meas Sci Technol 19(7):
075604.
Farooq, Z., Chestakov, D. A., Yan, B., Groenenboom, G. C., van der Zande, W. J. and Parker, D.
H. (2014). "Photodissociation of singlet oxygen in the UV region." PCCP 16(7): 3305-3316.
Fenn, J. B. (1951). "Lean Flammability Limit and Minimum Spark Ignition Energy. Commercial
Fluids and Pure Hydrocarbons." Ind. Eng. Chem. 43(12): 2865-2869.
Fernandes, R. X., Luther, K. and Troe, J. (2006). "Falloff Curves for the Reaction CH3 + O2 (+
M) → CH3O2 (+ M) in the Pressure Range 2−1000 Bar and the Temperature Range 300−700
K." J. Phys. Chem. A 110(13): 4442-4449.
Fischer, S., Dryer, F. and Curran, H. (2000). "The reaction kinetics of dimethyl ether. I:
High‐temperature pyrolysis and oxidation in flow reactors." Int. J. Chem. Kinet. 32(12): 713-
740.
Flitti, A. and Pancheshnyi, S. (2009). "Gas heating in fast pulsed discharges in N2–O2 mixtures."
Eur. Phys. J.-Appl. Phys. 45(2): 21001.
Florescu-Mitchell, A. I. and Mitchell, J. B. A. (2006). "Dissociative recombination." Physics
Reports 430(5-6): 277-374.
Frank-Kamenetskii, D. A. (1955). Diffusion and heat transfer in chemical kinetics, Princeton
University Press.
277
Frassoldati, A., Cuoci, A., Faravelli, T., Niemann, U., Ranzi, E., Seiser, R. and Seshadri, K.
(2010). "An experimental and kinetic modeling study of n-propanol and iso-propanol
combustion." Combust. Flame 157(1): 2-16.
Fridman, A. (2008). Plasma Chemistry. New York, Combridge University Press.
Gaïl, S., Thomson, M. J., Sarathy, S. M., Syed, S. A., Dagaut, P., Diévart, P., Marchese, A. J.
and Dryer, F. L. (2007). "A wide-ranging kinetic modeling study of methyl butanoate
combustion." Proc. Combust. Inst. 31(1): 305-311.
Ganguly, B. N. and Parish, J. W. (2004). "Absolute H atom density measurement in pure
methane pulsed discharge." Appl. Phys. Lett. 84(24): 4953.
Glewen, W. J., Wagner, R. M., Edwards, K. D. and Daw, C. S. (2009). "Analysis of cyclic
variability in spark-assisted HCCI combustion using a double Wiebe function." Proc. Combust.
Inst. 32(2): 2885-2892.
Gonzalez, M., Lew, L. and Cunningham, R. (1971). "Determinacion de la Cinetica de
Descomposicion Termica de Alcoholes e Hidrocarburos Mediante un Reactor Pulso." Lab.
Ensayo Mater. Invest. Tecnol. Prov. Buenos Aires Publ. Ser. 2: 103-123.
Grana, R., Frassoldati, A., Faravelli, T., Niemann, U., Ranzi, E., Seiser, R., Cattolica, R. and
Seshadri, K. (2010). "An experimental and kinetic modeling study of combustion of isomers of
butanol." Combust. Flame 157(11): 2137-2154.
Gu, X., Huang, Z., Wu, S. and Li, Q. (2010). "Laminar burning velocities and flame instabilities
of butanol isomers–air mixtures." Combust. Flame 157(12): 2318-2325.
Gudiyella, S. and Brezinsky, K. (2012). "High pressure study of n-propylbenzene oxidation."
Combust. Flame 159(3): 940-958.
Haas, F. M., Chaos, M. and Dryer, F. L. (2009). "Low and intermediate temperature oxidation of
ethanol and ethanol–PRF blends: An experimental and modeling study." Combust. Flame
156(12): 2346-2350.
Hagelaar, G. J. M. and Pitchford, L. C. (2005). "Solving the Boltzmann equation to obtain
electron transport coefficients and rate coefficients for fluid models." Plasma Sources Sci.
Technol. 14(4): 722-733.
278
Hamid, H. and Ali, M. A. (2004). Handbook of MTBE and other gasoline oxygenates, CRC
Press.
Hanson, R. K. (2011). "Applications of quantitative laser sensors to kinetics, propulsion and
practical energy systems." Proc. Combust. Inst. 33(1): 1-40.
Hanson, R. K. (2013). Quantitative Laser Diagnostics for Combustion Chemistry and Propulsion.
Princeton University.
Harrison, A. and Weinberg, F. (1971). "Flame stabilization by plasma jets". Proceedings of the
Royal Society of London A: Mathematical, Physical and Engineering Sciences 321(1544), The
Royal Society.
Hayashi, M., Hayashi database.
Hayashi, M., Capitelli, M. and Bardsley, J. (1990). "Nonequilibrium processes in partially
ionized gases." NATO ASI series, 220b.
Held, T. J. (1993). The oxidation of methanol, isobutene and methyl tertiary-butyl ether. Ann
Arbor, Princeton University. 9407111: 294-294 p.
Held, T. J. and Dryer, F. L. (1998). "A comprehensive mechanism for methanol oxidation." Int.
J. Chem. Kinet. 30(11): 805-830.
Herzberg, G. (1950). Molecular spectra and molecular structure. New York, NY, Van Nostrand
Reinhold Company.
Heywood, J. B. (1988). Internal Combustion Engine Fundamentals, McGraw-Hill.
Hicks, A., Norberg, S., Shawcross, P., Lempert, W. R., Rich, J. W. and Adamovich, I. V. (2005).
"Singlet oxygen generation in a high pressure non-self-sustained electric discharge." J. Phys. D:
Appl. Phys. 38(20): 3812-3824.
Holstein, T. (1946). "Energy Distribution of Electrons in High Frequency Gas Discharges."
Phys. Rev. 70(5-6): 367-384.
279
Honnet, S., Seshadri, K., Niemann, U. and Peters, N. (2009). "A surrogate fuel for kerosene."
Proc. Combust. Inst. 32(1): 485-492.
http://www.fidtechnology.com/ http://www.fidtechnology.com/.
http://www.knite.com/, http://www.knite.com/.
Hwang, N., Song, Y.-H. and Cha, M. S. (2010). "Efficient Use of CO2 Reforming of Methane
With an Arc-Jet Plasma." IEEE Trans. Plasma Sci. 38(12): 3291-3299.
Ikeda, Y., Nishiyama, A., Katano, H., Kaneko, M. and Jeong, H. (2009a). "Research and
development of microwave plasma combustion engine (Part II: engine performance of plasma
combustion engine)." SAE Technical Paper.
Ikeda, Y., Nishiyama, A., Wachi, Y. and Kaneko, M. (2009b). "Research and development of
microwave plasma combustion engine (Part I: concept of plasma combustion and plasma
generation technique)." SAE Technical Paper.
Itikawa, Y. (2005). "Cross Sections for Electron Collisions with Water Molecules." J. Phys.
Chem. Ref. Data 34(1): 1.
Jacobsen, L. S., Carter, C. D., Baurle, R. A., Jackson, T. A., Williams, S., Bivolaru, D., Kuo, S.,
Barnett, J. and Tam, C.-J. (2008). "Plasma-Assisted Ignition in Scramjets." J. Propul. Power
24(4): 641-654.
Jahangirian, S., McEnally, C. S. and Gomez, A. (2009). "Experimental study of ethylene
counterflow diffusion flames perturbed by trace amounts of jet fuel and jet fuel surrogates under
incipiently sooting conditions." Combust. Flame 156(9): 1799-1809.
Janev, R. K. and Reiter, D. (2002). "Collision processes of CHy and CHy+ hydrocarbons with
plasma electrons and protons." Phys. Plasmas 9: 4071.
Janev, R. K. and Reiter, D. (2003). "Unified analytic representation of hydrocarbon impurity
collision cross-sections." J. Nucl. Mater. 313-316: 1202-1205.
Janev, R. K. and Reiter, D. (2004). "Collision processes of C[sub 2,3]H[sub y] and C[sub
2,3]H[sub y][sup +] hydrocarbons with electrons and protons." Phys. Plasmas 11(2): 780.
280
Jasiński, M., Dors, M. and Mizeraczyk, J. (2008). "Production of hydrogen via methane
reforming using atmospheric pressure microwave plasma." J. Power Sources 181(1): 41-45.
Ji, C., Dames, E., Wang, H. and Egolfopoulos, F. N. (2012). "Propagation and extinction of
benzene and alkylated benzene flames." Combust. Flame 159(3): 1070-1081.
Ju, Y. and Maruta, K. (2011). "Microscale combustion: Technology development and
fundamental research." Prog. Energy Combust. Sci. 37(6): 669-715.
Ju, Y., Lefkowitz, J. K., Wada, T., Yang, X., Won, S. H. and Sun, W. (2015a). "Plasma assisted
combustion: kinetic studies and new combustion technology". 53rd AIAA Aerospace Sciences
Meeting, American Institute of Aeronautics and Astronautics.
Ju, Y. and Sun, W. (2015b). "Plasma assisted combustion: Dynamics and chemistry." Prog.
Energy Combust. Sci. 48: 21-83.
Kee, R. J., et al. (2003). Chemkin Collection, Release 3.7.1. I. Reaction Design. San Diego, CA.
Kim, H. H., Won, S. H., Santner, J., Chen, Z. and Ju, Y. (2013). "Measurements of the critical
initiation radius and unsteady propagation of n-decane/air premixed flames." Proc. Combust.
Inst. 34(1): 929-936.
Kim, M. K., Ryu, S. K., Won, S. H. and Chung, S. H. (2010a). "Electric fields effect on liftoff
and blowoff of nonpremixed laminar jet flames in a coflow." Combust. Flame 157(1): 17-24.
Kim, W., Mungal, M. G. and Cappelli, M. A. (2008). "Formation and role of cool flames in
plasma-assisted premixed combustion." Appl. Phys. Lett. 92(5): 051503.
Kim, W., Godfrey Mungal, M. and Cappelli, M. A. (2010b). "The role of in situ reforming in
plasma enhanced ultra lean premixed methane/air flames." Combust. Flame 157(2): 374-383.
Kogelschatz, U. (2003). "Dielectric-barrier Discharges: Their History, Discharge Physics, and
Industrial Applications." Plasma Chem. Plasma Process. 23(1): 1-46.
Konnov, A. (2000). "Detailed reaction mechanism for small hydrocarbons combustion." Release
0.5.
281
Kosarev, I. N., Aleksandrov, N. L., Kindysheva, S. V., Starikovskaia, S. M. and Starikovskii, A.
Y. (2008a). "Kinetics of ignition of saturated hydrocarbons by nonequilibrium plasma: CH4-
containing mixtures." Combust. Flame 154(3): 569-586.
Kosarev, I. N., Aleksandrov, N. L., Kindysheva, S. V., Starikovskaia, S. M. and Starikovskii, A.
Y. (2008b). "Kinetic mechanism of plasma-assisted ignition of hydrocarbons." J. Phys. D: Appl.
Phys. 41(3): 032002.
Kosarev, I. N., Aleksandrov, N. L., Kindysheva, S. V., Starikovskaia, S. M. and Starikovskii, A.
Y. (2009). "Kinetics of ignition of saturated hydrocarbons by nonequilibrium plasma: C2H6- to
C5H12-containing mixtures." Combust. Flame 156(1): 221-233.
Kossyi, I., Kostinsky, A. Y., Matveyev, A. and Silakov, V. (1992). "Kinetic scheme of the non-
equilibrium discharge in nitrogen-oxygen mixtures." Plasma Sources Sci. Technol. 1(3): 207.
Kramida, A., Ralchenko, Y., Reader, J. and Team, N. A. (2014). NIST Atomic Spectra Database
(version 5.2). http://physics.nist.gov/asd National Institute of Standards and Technology,
Gaithersburg, MD.
Kurimoto, N., Brumfield, B., Yang, X., Wada, T., Diévart, P., Wysocki, G. and Ju, Y. (2015).
"Quantitative measurements of HO2/H2O2 and intermediate species in low and intermediate
temperature oxidation of dimethyl ether." Proc. Combust. Inst. 35(1): 457-464.
Lacoste, D., Xu, D., Moeck, J. and Laux, C. (2013a). "Dynamic response of a weakly turbulent
lean-premixed flame to nanosecond repetitively pulsed discharges." Proc. Combust. Inst. 34(2):
3259-3266.
Lacoste, D. A., Xu, D. A., Moeck, J. P. and Laux, C. O. (2013b). "Dynamic response of a weakly
turbulent lean-premixed flame to nanosecond repetitively pulsed discharges." Proc. Combust.
Inst. 34(2): 3259-3266.
Lawton, J. and Weinberg, F. J. (1969). Electrical aspects of combustion, Clarendon Press.
Lee, D., Kihm, K. and Chung, S. (1997). "J. Fluids Engineering." Trans. ASME 119: 716-718.
Lee, D. H., Kim, K. T., Cha, M. S. and Song, Y. H. (2007). "Optimization scheme of a rotating
gliding arc reactor for partial oxidation of methane." Proc. Combust. Inst. 31(2): 3343-3351.
282
Lefkowitz, J. K., Heyne, J. S., Won, S. H., Dooley, S., Kim, H. H., Haas, F. M., Jahangirian, S.,
Dryer, F. L. and Ju, Y. (2012a). "A chemical kinetic study of tertiary-butanol in a flow reactor
and a counterflow diffusion flame." Combust. Flame 159(3): 968-978.
Lefkowitz, J. K., Ju, Y., Tsuruoka, R. and Ikeda, Y. (2012b). "A Study of Plasma-Assisted
Ignition in a Small Internal Combustion Engine". 50th AIAA Aerospace Sciences Meeting
including the New Horizons Forum and Aerospace Exposition, American Institute of
Aeronautics and Astronautics.
Lefkowitz, J. K., Ju, Y., Stevens, C. A., Ombrello, T., Schauer, F. and Hoke, J. (2013). "The
Effects of Repetitively Pulsed Nanosecond Discharges on Ignition Time in a Pulsed Detonation
Engine". 49th AIAA/ASME/SAE/ASEE Joint Propulsion Conference, American Institute of
Aeronautics and Astronautics.
Lefkowitz, J. K., Windom, B. C., MacDonald, W., Adams, S., Chen, T., Uddi, M. and Ju, Y.
(2014). "Time Dependent Measurements of Species Formation in Nanosecond-Pulsed Plasma
Discharges in C2H4/O2/Ar Mixtures". 52nd Aerospace Sciences Meeting, National Harbor, MD.
Lefkowitz, J. K., Guo, P., Ombrello, T., Won, S. H., Stevens, C. A., Hoke, J. L., Schauer, F. and
Ju, Y. (2015a). "Schlieren imaging and pulsed detonation engine testing of ignition by a
nanosecond repetitively pulsed discharge." Combust. Flame.
Lefkowitz, J. K., Guo, P., Rousso, A. and Ju, Y. (2015b). "Species and temperature
measurements of methane oxidation in a nanosecond repetitively pulsed discharge." Phil. Trans.
R. Soc. A 373(2048): 20140333.
Lefkowitz, J. K., Guo, P., Rousso, A. and Ju, Y. (2015c). "Low temperature oxidation of
methane in a nanosecond pulsed plasma discharge". 53rd AIAA Aerospace Sciences Meeting,
American Institute of Aeronautics and Astronautics.
Lefkowitz, J. K., Uddi, M., Windom, B. C., Lou, G. and Ju, Y. (2015d). "In situ species
diagnostics and kinetic study of plasma activated ethylene dissociation and oxidation in a low
temperature flow reactor." Proc. Combust. Inst. 35(3): 3505-3512.
Lesueur, H., Czernichowski, A. and Chapelle, J. (1994). "Electrically Assisted Partial Oxidation
of Methane." Int. J. Hydrogen Energy 19(2): 139-144.
Lewis, B. and von Elbe, G. (1947). "Ignition of Explosive Gas Mixtures by Electric Sparks. II.
Theory of the Propagation of Flame from an Instantaneous Point Source of Ignition. 1." J. Chem.
Phys. 15(11): 803-808.
283
Lewis, D., Keil, M. and Sarr, M. (1974). "Gas phase thermal decomposition of tert-butyl
alcohol." J. Am. Chem. Soc. 96(14): 4398-4404.
Li, J., Kazakov, A. and Dryer, F. L. (2001). "Ethanol pyrolysis experiments in a variable
pressure flow reactor." Int. J. Chem. Kinet. 33(12): 859-867.
Li, J., Kazakov, A. and Dryer, F. L. (2004a). "Experimental and numerical studies of ethanol
decomposition reactions." J. Phys. Chem. A 108(38): 7671-7680.
Li, J., Zhao, Z., Kazakov, A. and Dryer, F. L. (2004b). "An updated comprehensive kinetic
model of hydrogen combustion." Int. J. Chem. Kinet. 36(10): 566-575.
Li, T., Adamovich, I. V. and Sutton, J. A. (2013). "A Burner Platform for Examining the Effects
of Non-Equilibrium Plasmas on Oxidation and Combustion Chemistry." Combust. Sci. Technol.
185(6): 990-998.
Lofthus, A. and Krupenie, P. H. (1977). "The spectrum of molecular nitrogen." J. Phys. Chem.
Ref. Data 6(1): 113-307.
Lou, G., Bao, A., Nishihara, M., Keshav, S., Utkin, Y. G., Rich, J. W., Lempert, W. R. and
Adamovich, I. V. (2007). "Ignition of premixed hydrocarbon–air flows by repetitively pulsed,
nanosecond pulse duration plasma." Proc. Combust. Inst. 31(2): 3327-3334.
Lutz, A. E., Kee, R. J. and Miller, J. A. (1988). "SENKIN: A FORTRAN program for predicting
homogeneous gas phase chemical kinetics with sensitivity analysis." Sandia National
Laboratories Report SAND87-8248.
Macheret, S., Shneider, M. and Murray, R. (2006). "Ionization in Strong Electric Fields and
Dynamics of Nanosecond-Pulse Plasmas." Phys. Plasmas 13(2): 023502.
Mak, F. (1986). "Economics of producing unleaded gasoline blends with Oxinol and MTBE."
Energy 11(3): 253-260.
Maly, R. and Vogel, M. (1979). "Initiation and propagation of flame fronts in lean CH 4-air
mixtures by the three modes of the ignition spark". Symposium (International) on Combustion
17(1), Elsevier.
284
Maly, R. (1981). "Ignition model for spark discharges and the early phase of flame front
growth". Symposium (International) on Combustion 18(1), Elsevier.
Maly, R., Saggau, B., Wagner, E. and Ziegler, G. (1983). "Prospects of ignition enhancement."
SAE Technical Paper.
Maly, R. (1984). Spark Ignition: Its Physics and Effect on the Internal Combustion Engine. Fuel
Economy in Road Vehicles Powered by Spark Ignition Engines. J. C. Hilliard and G. S.
Springer. New York, NY, Plenum Press: 91-148.
Mayer, S. W. and Schieler, L. (1968). "Activation Energies and Rate Constants Computed for
Reactions of Oxygen with Hydrocarbons." J. Phys. Chem. 72: 2628-2631.
McEnally, C. S. and Pfefferle, L. D. (2005). "Fuel decomposition and hydrocarbon growth
processes for oxygenated hydrocarbons: butyl alcohols." Proc. Combust. Inst. 30(1): 1363-1370.
Metcalfe, W. K., Burke, S. M., Ahmed, S. S. and Curran, H. J. (2013). "A Hierarchical and
Comparative Kinetic Modeling Study of C1− C2Hydrocarbon and Oxygenated Fuels." Int. J.
Chem. Kinet. 45(10): 638-675.
Michael, J. B., Dogariu, A., Shneider, M. N. and Miles, R. B. (2010). "Subcritical microwave
coupling to femtosecond and picosecond laser ionization for localized, multipoint ignition of
methane/air mixtures." J. Appl. Phys. 108(9): 093308.
Michael, J. B., Chng, T. L. and Miles, R. B. (2013). "Sustained propagation of ultra-lean
methane/air flames with pulsed microwave energy deposition." Combust. Flame 160(4): 796-
807.
Miles, R. B. (2010). MAE 521: Optics and Lasers. Princeton University.
Miller, J. A. and Klippenstein, S. J. (2001). "The reaction between ethyl and molecular oxygen
II: Further analysis." Int. J. Chem. Kinet. 33(11): 654-668.
Mintoussov, E. I., Pendleton, S. J., Gerbault, F. G., Popov, N. A. and Starikovskaia, S. M.
(2011). "Fast gas heating in nitrogen–oxygen discharge plasma: II. Energy exchange in the
afterglow of a volume nanosecond discharge at moderate pressures." J. Phys. D: Appl. Phys.
44(28): 285202.
285
Mintusov, E., Serdyuchenko, A., Choi, I., Lempert, W. R. and Adamovich, I. V. (2009).
"Mechanism of plasma assisted oxidation and ignition of ethylene–air flows by a repetitively
pulsed nanosecond discharge." Proc. Combust. Inst. 32(2): 3181-3188.
Mitchell, J. B. A. (1990). "The dissociative recombination of molecular ions." Physics Reports
186(5): 215-248.
Montello, A., Yin, Z., Burnette, D., Adamovich, I. V. and Lempert, W. R. (2013). "Picosecond
CARS measurements of nitrogen vibrational loading and rotational/translational temperature in
non-equilibrium discharges." J. Phys. D: Appl. Phys. 46(46): 464002.
Morsy, M. H. (2012). "Review and recent developments of laser ignition for internal combustion
engines applications." Renew. Sust. Energ. Rev. 16(7): 4849-4875.
Moss, J. T., Berkowitz, A. M., Oehlschlaeger, M. A., Biet, J., Warth, V., Glaude, P.-A. and
Battin-Leclerc, F. (2008). "An experimental and kinetic modeling study of the oxidation of the
four isomers of butanol." J. Phys. Chem. A 112(43): 10843-10855.
Nagaraja, S., Yang, V. and Adamovich, I. (2013). "Multi-scale modelling of pulsed nanosecond
dielectric barrier plasma discharges in plane-to-plane geometry." J. Phys. D: Appl. Phys. 46(15):
155205.
Nagaraja, S. and Yang, V. (2014). "A ‘frozen electric-field’ approach to simulate repetitively
pulsed nanosecond plasma discharges and ignition of hydrogen–air mixtures." J. Phys. D: Appl.
Phys. 47(38): 385201.
Nagaraja, S., Li, T., Sutton, J. A., Adamovich, I. V. and Yang, V. (2015a). "Nanosecond plasma
enhanced H 2/O 2/N 2 premixed flat flames." Proc. Combust. Inst. 35(3): 3471-3478.
Nagaraja, S., Sun, W. and Yang, V. (2015b). "Effect of non-equilibrium plasma on two-stage
ignition of n-heptane." Proc. Combust. Inst. 35(3): 3497-3504.
Napartovich, A. P., Dyatko, N. A., Kochetov, I. V. and Sukharev, A. G., TRINITI database.
Newman, C., O'neal, H., Ring, M., Leska, F. and Shipley, N. (1979). "Kinetics and mechanism
of the silane decomposition." Int. J. Chem. Kinet. 11(11): 1167-1182.
286
Nishimura, H. and Tawara, H. (1994). "Total electron impact ionization cross sections for simple
hydrocarbon molecules." Journal of Physics B: Atomic, Molecular and Optical Physics 27(10):
2063.
Nishiyama, A. and Ikeda, Y. (2012). "Improvement of lean limit and fuel consumption using
microwave plasma ignition technology." SAE Technical Paper.
NIST, http://webbook.nist.gov/chemistry/.
Noorani, K. E., Akih-Kumgeh, B. and Bergthorson, J. M. (2010). "Comparative High
Temperature Shock Tube Ignition of C1−C4 Primary Alcohols." Energy & Fuels 24(11): 5834-
5843.
Norton, T. S. and Dryer, F. L. (1991). "The flow reactor oxidation of C 1− C 4 alcohols and
MTBE". Symposium (International) on Combustion 23(1), Elsevier.
Okumoto, M., Rajanikanth, B. S., Katsura, S. and Mizuno, A. (1998). "Nonthermal Plasma
Approach in Direct Methanol Synthesis from CH4." IEEE Transactions on Industry Applications
34(5): 940-944.
Ombrello, T., Qin, X., Ju, Y. and Carter, C. D. (2006). "Combustion enhancement via stabilized
piecewise nonequilibrium gliding arc plasma discharge." AIAA J. 44: 142.
Ombrello, T. and Ju, Y. (2008a). "Kinetic Ignition Enhancement of H2 Versus Fuel-Blended Air
Diffusion Flames Using Nonequilibrium Plasma." IEEE Trans. Plasma Sci. 36(6): 2924-2932.
Ombrello, T., Ju, Y. and Fridman, A. (2008b). "Kinetic Ignition Enhancement of Diffusion
Flames by Nonequilibrium Magnetic Gliding Arc Plasma." AIAA J. 46(10): 2424-2433.
Ombrello, T., Won, S. H., Ju, Y. and Williams, S. (2010a). "Flame propagation enhancement by
plasma excitation of oxygen. Part I: Effects of O3." Combust. Flame 157(10): 1906-1915.
Ombrello, T., Won, S. H., Ju, Y. and Williams, S. (2010b). "Flame propagation enhancement by
plasma excitation of oxygen. Part II: Effects of O2(a1Δg)." Combust. Flame 157(10): 1916-
1928.
Orrin, J., Vince, I. and Weinberg, F. (1981). "A study of plasma jet ignition mechanisms".
Symposium (International) on Combustion 18(1), Elsevier.
287
Oßwald, P., Güldenberg, H., Kohse-Höinghaus, K., Yang, B., Yuan, T. and Qi, F. (2011).
"Combustion of butanol isomers – A detailed molecular beam mass spectrometry investigation
of their flame chemistry." Combust. Flame 158(1): 2-15.
Pai, D. Z., Lacoste, D. A. and Laux, C. O. (2010a). "Nanosecond repetitively pulsed discharges
in air at atmospheric pressure—the spark regime." Plasma Sources Sci. Technol. 19(6): 065015.
Pai, D. Z., Lacoste, D. A. and Laux, C. O. (2010b). "Transitions between corona, glow, and
spark regimes of nanosecond repetitively pulsed discharges in air at atmospheric pressure." J.
Appl. Phys. 107(9): 093303.
Pancheshnyi, S., Eismann, B., Hagelaar, G. J. M. and Pitchford, L. C. (2008). ZDPlasKin.
http://www.zdplaskin.laplace.univ-tlse.fr University of Toulouse, LAPLACE, CNRS-UPS-INP,
Toulouse, France.
Pancheshnyi, S., et al. (2010). LXcat. http://nl.lxcat.net/home/.
Pancheshnyi, S. V., Lacoste, D. A., Bourdon, A. and Laux, C. O. (2006). "Ignition of Propane–
Air Mixtures by a Repetitively Pulsed Nanosecond Discharge." IEEE Trans. Plasma Sci. 34(6):
2478-2487.
Paterson, C. C. and Campbell, N. (1918). "Some characteristics of the spark discharge and its
effect in igniting explosive mixtures." Proceedings of the Physical Society of London 31(1): 168.
Pearse, R. W. B. and Gaydon, A. G. (1965). The Identification of Molecular Spectra 3rd ed.
London, U.K., Chapman & Hall.
Peng, M.-W., Shy, S., Shiu, Y.-W. and Liu, C.-C. (2013). "High pressure ignition kernel
development and minimum ignition energy measurements in different regimes of premixed
turbulent combustion." Combust. Flame 160(9): 1755-1766.
Phelps, A., Phelps database.
Pichon, S., Black, G., Chaumeix, N., Yahyaoui, M., Simmie, J., Curran, H. and Donohue, R.
(2009). "The combustion chemistry of a fuel tracer: Measured flame speeds and ignition delays
and a detailed chemical kinetic model for the oxidation of acetone." Combust. Flame 156(2):
494-504.
288
Pilla, G., Galley, D., Lacoste, D. A., Lacas, F., Veynante, D. and Laux, C. O. (2006).
"Stabilization of a turbulent premixed flame using a nanosecond repetitively pulsed plasma."
Plasma Science, IEEE Transactions on 34(6): 2471-2477.
Pilla, G. L., Lacoste, D. A., Veynante, D. and Laux, C. O. (2008). "Stabilization of a swirled
propane–air flame using a nanosecond repetitively pulsed plasma." Plasma Science, IEEE
Transactions on 36(4): 940-941.
Popov, N. A. (2011). "Fast gas heating in a nitrogen–oxygen discharge plasma: I. Kinetic
mechanism." J. Phys. D: Appl. Phys. 44(28): 285201.
Raizer, Y. P. (1991). Gas Discharge Physics. Berlin, Springer-Verlag.
Rakitin, A. E. and Starikovskii, A. Y. (2008). "Mechanisms of deflagration-to-detonation
transition under initiation by high-voltage nanosecond discharges." Combust. Flame 155(1-2):
343-355.
Rapp, V. H., DeFilippo, A., Saxena, S., Chen, J.-Y., Dibble, R. W., Nishiyama, A., Moon, A.
and Ikeda, Y. (2012). "Extending lean operating limit and reducing emissions of methane spark-
ignited engines using a microwave-assisted spark plug." Journal of Combustion 2012.
Ray, D. J. M., Ruiz Diaz, R. and Waddington, D. J. (1973). "Gas-phase oxidation of butene-2:
The role of acetaldehyde in the reaction." Symposium (International) on Combustion 14(1): 259-
266.
Reif, F. (2009). Fundamentals of statistical and thermal physics, Waveland Press.
Rothman, L. S., et al. (2009). "The HITRAN 2008 molecular spectroscopic database." J Quant
Spectrosc Ra 110(9-10): 533-572.
Rusterholtz, D. L., Lacoste, D. A., Stancu, G. D., Pai, D. Z. and Laux, C. O. (2013). "Ultrafast
heating and oxygen dissociation in atmospheric pressure air by nanosecond repetitively pulsed
discharges." J. Phys. D: Appl. Phys. 46(46): 464010.
Samukawa, S., et al. (2012). "The 2012 Plasma Roadmap." J. Phys. D: Appl. Phys. 45(25):
253001.
289
Sander, S. P., et al. (2006). "JPL Publication 06-2." Jet Propulsion Laboratory, California
Institute of Technology: Pasadena, CA,.
Santner, J., Dryer, F. L. and Ju, Y. (2013). "The effects of water dilution on hydrogen, syngas,
and ethylene flames at elevated pressure." Proc. Combust. Inst. 34(1): 719-726.
Sarathy, S. M., Thomson, M. J., Togbé, C., Dagaut, P., Halter, F. and Mounaim-Rousselle, C.
(2009). "An experimental and kinetic modeling study of n-butanol combustion." Combust. Flame
156(4): 852-864.
Sarnacki, B. G., Esposito, G., Krauss, R. H. and Chelliah, H. K. (2012). "Extinction limits and
associated uncertainties of nonpremixed counterflow flames of methane, ethylene, propylene and
n-butane in air." Combust. Flame 159(3): 1026-1043.
Saxena, P. and Williams, F. A. (2007). "Numerical and experimental studies of ethanol flames."
Proc. Combust. Inst. 31(1): 1149-1156.
Schauer, F. R., Stutrud, J. S. and Bradley, R. P. (2001). "Detonation Initiation Studies and
Performance Results for Pulsed Detonation Engine Applications". 39th AIAA Aerospace
Sciences Meeting, Reno, NV.
Senosiain, J. P., Klippenstein, S. J. and Miller, J. A. (2006). "Reaction of ethylene with hydroxyl
radicals: a theoretical study." J. Phys. Chem. A 110(21): 6960-6970.
Seshadri, K. and Williams, F. (1978). "Laminar flow between parallel plates with injection of a
reactant at high Reynolds number." Int. J. Heat Mass Transfer 21(2): 251-253.
Shiraishi, T., Urushihara, T. and Gundersen, M. (2009). "A trial of ignition innovation of
gasoline engine by nanosecond pulsed low temperature plasma ignition." J. Phys. D: Appl. Phys.
42(13): 135208.
Singleton, D., Pendleton, S. J. and Gundersen, M. A. (2011). "The role of non-thermal transient
plasma for enhanced flame ignition in C2H4–air." J. Phys. D: Appl. Phys. 44(2): 022001.
Sinha, A. and Thomson, M. J. (2004). "The chemical structures of opposed flow diffusion flames
of C3 oxygenated hydrocarbons (isopropanol, dimethoxy methane, and dimethyl carbonate) and
their mixtures." Combust. Flame 136(4): 548-556.
290
Slemr, F. and Warneck, P. (1977). "Kinetics of the Reaction of Atomic Hydrogen with Methyl
Hydroperoxide." Int. J. Chem. Kinet. 9: 267-282.
Smooke, M. D., Puri, I. K. and Seshadri, K. (1988). "A comparison between numerical
calculations and experimental measurements of the structure of a counterflow diffusion flame
burning diluted methane in diluted air." Symposium (International) on Combustion 21(1): 1783-
1792.
Sohn, C. H., Son, J. W., Won, S. H. and Ju, Y. (2015). "Computational studies of diffusion cool
flame structures of n-heptane with/without ozone sensitization with a reduced chemistry."
Journal of Mechanical Science and Technology 29(3): 1297-1305.
Stafford, D. S. and Kushner, M. J. (2004). "O[sub 2]([sup 1]Δ) production in He∕O[sub 2]
mixtures in flowing low pressure plasmas." J. Appl. Phys. 96(5): 2451.
Stancu, G. D., Kaddouri, F., Lacoste, D. A. and Laux, C. O. (2010). "Atmospheric pressure
plasma diagnostics by OES, CRDS and TALIF." J. Phys. D: Appl. Phys. 43(12): 124002.
Starik, A. M. and Titova, N. S. (2006). "Intensification of the oxidation of rich methane/air
mixtures by O2 molecules excited to the a 1Δg state." Kinet. Catal. 47(4): 487-496.
Starik, A. M., Titova, N. S., Bezgin, L. V. and Kopchenov, V. I. (2009). "The promotion of
ignition in a supersonic H2–air mixing layer by laser-induced excitation of O2 molecules:
Numerical study." Combust. Flame 156(8): 1641-1652.
Starik, A. M., Kozlov, V. E. and Titova, N. S. (2010). "On the influence of singlet oxygen
molecules on the speed of flame propagation in methane–air mixture." Combust. Flame 157(2):
313-327.
Starikovskaia, S., Kukaev, E., Kuksin, A., Nudnova, M. and Starikovskii, A. (2004). "Analysis
of the spatial uniformity of the combustion of a gaseous mixture initiated by a nanosecond
discharge." Combust. Flame 139(3): 177-187.
Starikovskaia, S. M., Starikovskii, A. Y. and Zatsepin, D. V. (2001). "Hydrogen oxidation in a
stoichiometric hydrogen-air mixture in the fast ionization wave." Combust. Theor. Model. 5(1):
97-129.
Starikovskaia, S. M. (2006). "Plasma assisted ignition and combustion." J. Phys. D: Appl. Phys.
39(16): R265-R299.
291
Starikovskaia, S. M. (2014). "Plasma-assisted ignition and combustion: nanosecond discharges
and development of kinetic mechanisms." J. Phys. D: Appl. Phys. 47(35): 353001.
Starikovskaya, S. M., Aleksandrov, N. L., Kosarev, I. N., Kindysheva, S. V. and Starikovskii, A.
Y. (2009). "Ignition with low-temperature plasma: Kinetic mechanism and experimental
verification." High Energy Chemistry 43(3): 213-218.
Starikovskii, A. Y., Anikin, N. B., Kosarev, I. N., Mintoussov, E. I., Starikovskaia, S. M. and
Zhukov, V. P. (2006). "Plasma-assisted combustion." Pure Appl. Chem. 78(6): 1265-1298.
Starikovskiy, A., Aleksandrov, N. and Rakitin, A. (2012). "Plasma-assisted ignition and
deflagration-to-detonation transition." Philos. Trans. R. Soc. London, Ser. A 370(1960): 740-
773.
Starikovskiy, A. and Aleksandrov, N. (2013). "Plasma-assisted ignition and combustion." Prog.
Energy Combust. Sci. 39(1): 61-110.
Stevens, C., Gamezo, V., King, P., Schauer, F. and Hoke, J. (2011). "Interactions of Detonations
with Ramps". 49th AIAA Aerospace Sciences Meeting including the New Horizons Forum and
Aerospace Exposition, Orlanda, FL.
Stockman, E. S., Zaidi, S. H., Miles, R. B., Carter, C. D. and Ryan, M. D. (2009).
"Measurements of combustion properties in a microwave enhanced flame." Combust. Flame
156(7): 1453-1461.
Suib, S. L. and Zerger, R. P. (1993). "A Direct, Continuous, Low-Power Catalytic Conversion of
Methane to Higher Hydrocarbons via Microwave Plasmas." Journal of Catalysis 139: 381-391.
Sun, W., Uddi, M., Ombrello, T., Won, S. H., Carter, C. and Ju, Y. (2011). "Effects of non-
equilibrium plasma discharge on counterflow diffusion flame extinction." Proc. Combust. Inst.
33(2): 3211-3218.
Sun, W., Uddi, M., Won, S. H., Ombrello, T., Carter, C. and Ju, Y. (2012). "Kinetic effects of
non-equilibrium plasma-assisted methane oxidation on diffusion flame extinction limits."
Combust. Flame 159(1): 221-229.
292
Sun, W., Won, S. H., Ombrello, T., Carter, C. and Ju, Y. (2013). "Direct ignition and S-curve
transition by in situ nano-second pulsed discharge in methane/oxygen/helium counterflow
flame." Proc. Combust. Inst. 34(1): 847-855.
Sun, W., Won, S. H. and Ju, Y. (2014). "In situ plasma activated low temperature chemistry and
the S-curve transition in DME/oxygen/helium mixture." Combust. Flame 161(8): 2054-2063.
Takashima, K., Yin, Z. and Adamovich, I. V. (2013). "Measurements and kinetic modeling of
energy coupling in volume and surface nanosecond pulse discharges." Plasma Sources Sci.
Technol. 22(1): 015013.
Takita, K., Uemoto, T., Sato, T., Ju, Y., Masuya, G. and Ohwaki, K. (2000). "Ignition
characteristics of plasma torch for hydrogen jet in an airstream." J. Propul. Power 16(2): 227-
233.
Takita, K., Abe, N., Masuya, G. and Ju, Y. (2007). "Ignition enhancement by addition of NO and
NO 2 from a N 2/O 2 plasma torch in a supersonic flow." Proc. Combust. Inst. 31(2): 2489-2496.
Tennyson, J., Brown, D. B., Munro, J. J., Rozum, I., Varambhia, H. N. and Vinci, N. (2007).
"Quantemol-N: an expert system for performing electron molecule collision calculations using
the R-matrix method". Journal of Physics: Conference Series 86(1), IOP Publishing.
Tsang, W. (1964). "Thermal Decomposition of Some Tert‐Butyl Compounds at Elevated
Temperatures." J. Chem. Phys. 40(6): 1498-1505.
Tsang, W. and Hampson, R. F. (1986). "Chemical Kinetic Data Base for Combustion Chemistry.
Part I. Methane and Related Compounds." J. Phys. Chem. Ref. Data 15(3): 1087.
Tsuji, H. (1982). "Counterflow diffusion flames." Prog. Energy Combust. Sci. 8(2): 93-119.
Tsuji, M., Kobarai, K., Obase, H., Kouno, H. and Nishimura, Y. (1991). "Dissociative excitation
of CH4 by collisions with helium active species." J. Chem. Phys. 94(1): 277.
Tsuji, M., Kouno, H., Matsumura, K.-i., Funatsu, T., Nishimura, Y., Obase, H., Kugishima, H.
and Yoshida, K. (1993). "Dissociative charge-transfer reactions of Ar+ with simple aliphatic
hydrocarbons at thermal energy." J. Chem. Phys. 98(3): 2011.
293
Tucker, K. C., King, P. I. and Schauer, F. R. (2008). "Hydrocarbon Fuel Flash Vaporization for
Pulsed Detonation Combustion." J. Propul. Power 24(4): 788-796.
Turecek, F., Brabec, L. and Korvola, J. (1988). "Unstable enols in the gas phase. Preparation
ionization, energies, and heats of formation of (E)-and (Z)-2-buten-2-ol, 2-methyl-1-propen-1-ol,
and 3-methyl-2-buten-2-ol." J. Am. Chem. Soc. 110(24): 7984-7990.
Uddi, M., Jiang, N., Adamovich, I. V. and Lempert, W. R. (2009a). "Nitric oxide density
measurements in air and air/fuel nanosecond pulse discharges by laser induced fluorescence." J.
Phys. D: Appl. Phys. 42(7): 075205.
Uddi, M., Jiang, N., Mintusov, E., Adamovich, I. V. and Lempert, W. R. (2009b). "Atomic
oxygen measurements in air and air/fuel nanosecond pulse discharges by two photon laser
induced fluorescence." Proc. Combust. Inst. 32(1): 929-936.
Uddi, M., Das, A. K. and Sung, C.-J. (2012). "Temperature measurements in a rapid
compression machine using mid-infrared H2O absorption spectroscopy near 7.6 μm." Appl.
Optics 51(22): 5464-5476.
Uddi, M., Lefkowitz, J. K. and Ju, Y. (2013). "Species Measurements of Ethylene Oxidation in a
Nanosecond-Pulsed Plasma Discharge Using QCL Absorption Spectroscopy Near 7.6µm". 51st
AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace
Exposition, American Institute of Aeronautics and Astronautics.
Van Geem, K. M., Pyl, S. P., Marin, G. B., Harper, M. R. and Green, W. H. (2010). "Accurate
high-temperature reaction networks for alternative fuels: butanol isomers." Industrial &
Engineering Chemistry Research 49(21): 10399-10420.
Vasu, S. S., Davidson, D. F., Hanson, R. K. and Golden, D. M. (2010). "Measurements of the
reaction of OH with n-butanol at high-temperatures." Chem. Phys. Lett. 497(1): 26-29.
Velazco, J. E., Kolts, J. H. and Setser, D. W. (1978). "Rate constants and quenching mechanisms
for the metastable states of argon, krypton, and xenon." J. Chem. Phys. 69(10): 4357.
Veloo, P. S., Wang, Y. L. and Egolfopoulos, F. N. (2010). 2010 Spring Technical Meeting of the
Western States Sections of the Combustion Institute, University of Colorado, Boulder, CO.
Veloo, P. S. and Egolfopoulos, F. N. (2011). "Flame propagation of butanol isomers/air
mixtures." Proc. Combust. Inst. 33(1): 987-993.
294
Vu, T. M., Won, S. H., Ombrello, T. and Cha, M. S. (2014). "Stability enhancement of ozone-
assisted laminar premixed Bunsen flames in nitrogen co-flow." Combust. Flame 161(4): 917-
926.
Wagner, T. C., O'Brien, W. F., Northam, G. B. and Eggers, J. M. (1989). "Plasma torch igniter
for scramjets." J. Propul. Power 5(5): 548-554.
Wang, F., Liu, J. B., Sinibaldi, J., Brophy, C., Kuthi, A., Jiang, C., Ronney, P. and Gundersen,
M. A. (2005). "Transient Plasma Ignition of Quiescent and Flowing Air/Fuel Mixtures." IEEE
Trans. Plasma Sci. 33(2): 844-849.
Wang, H., You, X., Joshi, A. V., Davis, S. G., Laskin, A., Egolfopoulos, F. and Law, C. K.
http://ignis.usc.edu/USC_Mech_II.htm.
Wang, H., You, X., Joshi, A. V., Davis, S. G., Laskin, A., Egolfopoulos, F. and Law, C. K.
(2007). "USC Mech Version II. High-Temperature Combustion Reaction Model of H2/CO/C1-
C4 Compounds." University of Southern California, Los Angeles, CA, accessed Jan 4: 2011.
Wang, Q., Zhang, G., Liu, Y., Hou, L., Liu, C., Wang, Z. and Huang, J. (2013a). "Visual features
of microwave ignition of methane-air mixture in a constant volume cylinder." Appl. Phys. Lett.
103(20): 204104.
Wang, S., Davidson, D. F. and Hanson, R. K. (2013b). "High-temperature laser absorption
diagnostics for CH2O and CH3CHO and their application to shock tube kinetic studies."
Combust. Flame 160(10): 1930-1938.
Wang, Z. H., Yang, L., Li, B., Li, Z. S., Sun, Z. W., Aldén, M., Cen, K. F. and Konnov, A. A.
(2012). "Investigation of combustion enhancement by ozone additive in CH4/air flames using
direct laminar burning velocity measurements and kinetic simulations." Combust. Flame 159(1):
120-129.
Weideman, J. A. C. (1994). "Computation of the Complex Error Function." SIAM J. Numerical
Analysis 31(5): 1497-1518.
Weinberg, F., Hom, K., Oppenheim, A. and Teichman, K. (1978). "Ignition by plasma jet."
Nature 272: 341-343.
295
Wolk, B., DeFilippo, A., Chen, J.-Y., Dibble, R., Nishiyama, A. and Ikeda, Y. (2013).
"Enhancement of flame development by microwave-assisted spark ignition in constant volume
combustion chamber." Combust. Flame 160(7): 1225-1234.
Won, S. H., Cha, M. S., Park, C. S. and Chung, S. H. (2007). "Effect of electric fields on
reattachment and propagation speed of tribrachial flames in laminar coflow jets." Proc. Combust.
Inst. 31(1): 963-970.
Won, S. H., Ryu, S. K., Kim, M. K., Cha, M. S. and Chung, S. H. (2008). "Effect of electric
fields on the propagation speed of tribrachial flames in coflow jets." Combust. Flame 152(4):
496-506.
Won, S. H., Sun, W. and Ju, Y. (2010). "Kinetic effects of toluene blending on the extinction
limit of n-decane diffusion flames." Combust. Flame 157(3): 411-420.
Won, S. H., Dooley, S., Dryer, F. L. and Ju, Y. (2011). "Kinetic effects of aromatic molecular
structures on diffusion flame extinction." Proc. Combust. Inst. 33(1): 1163-1170.
Won, S. H., Dooley, S., Dryer, F. L. and Ju, Y. (2012). "A radical index for the determination of
the chemical kinetic contribution to diffusion flame extinction of large hydrocarbon fuels."
Combust. Flame 159(2): 541-551.
Won, S. H., Jiang, B., Dievart, P., Sohn, C. H. and Ju, Y. (2015a). "Self-sustaining n-heptane
cool diffusion flames activated by ozone." Proc. Combust. Inst. 35.
Won, S. H., Jiang, B., Diévart, P., Sohn, C. H. and Ju, Y. (2015b). "Self-sustaining n-heptane
cool diffusion flames activated by ozone." Proc. Combust. Inst. 35(1): 881-888.
Wu, Y., Zhang, Z. and Ombrello, T. M. (2013a). "Spatially resolved measurement of singlet
delta oxygen by radar resonance-enhanced multiphoton ionization." Opt. Lett. 38(13): 2286-
2288.
Wu, Y., Zhang, Z., Ombrello, T. M. and Katta, V. R. (2013b). "Quantitative Radar REMPI
measurements of methyl radicals in flames at atmospheric pressure." Appl. Phys. B-Lasers O.
111(3): 391-397.
Xu, K., Xu, Z. F. and Lin, M. C. (2010). "Ab initiokinetic prediction of branching rate constants
for reactions of H atoms with CH3O and CH2OH." Molecular Physics 105(19-22): 2763-2776.
296
Yalin, A. P., Laux, C. O., Kruger, C. H. and Zare, R. N. (2002a). "Spatial profiles of N 2 +
concentration in an atmospheric pressure nitrogen glow discharge." Plasma Sources Sci.
Technol. 11(3): 248.
Yalin, A. P., Zare, R. N., Laux, C. O. and Kruger, C. H. (2002b). "Temporally resolved cavity
ring-down spectroscopy in a pulsed nitrogen plasma." Appl. Phys. Lett. 81(8): 1408.
Yang, B., Oßwald, P., Li, Y., Wang, J., Wei, L., Tian, Z., Qi, F. and Kohse-Höinghaus, K.
(2007). "Identification of combustion intermediates in isomeric fuel-rich premixed butanol–
oxygen flames at low pressure." Combust. Flame 148(4): 198-209.
Yang, S., Nagaraja, S., Yang, V., Sun, W., Lefkowitz, J. K. and Ju, Y. (2015a). "Numerical and
Experimental Investigation of Nanosecond-Pulsed Plasma Activated C2H4/O2/Ar Mixtures in a
Low Temperature Flow Reactor". 53rd AIAA Aerospace Sciences Meeting, American Institute
of Aeronautics and Astronautics.
Yang, S., Gao, X., Yang, V., Sun, W., Nagaraja, S., Lefkowitz, J. K. and Ju, Y. (Submitted).
Numerical and Experimental Investigation of Nanosecond Pulsed Plasma Activated C2H4/O2/Ar
Mixtures in a Flow Reactor. J. Propul. Power.
Yang, X., Lefkowitz, J. K., Brumfield, B. E., Chen, Q., Wyscoki, G. and Ju, Y. (2015b).
"Kinetics studies of O3/O2/CH3OH/Ar mixtures in a photolysis flow reactor". 9th U.S. National
Combustion MeetingCincinnati, OH
Yang, X. L., Shen, X. B., Santner, J. and Ju, Y. (2013). "Experimental and Kinetic Studies of
Acetylene Flames at Elevated Pressures." Proc. Combust. Inst. 35.
Yetter, R. A. and Glassman, I. (2008). Combustion, Elsevier, Inc.
Yin, Z. and Adamovich, I. (2011). "Ignition Delay and Time-Resolved Temperature
Measurements in Nanosecond Pulse Hydrogen-Air and Ethylene-Air Plasmas at Elevated Initial
Temperatures."
Yin, Z., Adamovich, I. V. and Lempert, W. R. (2013a). "OH radical and temperature
measurements during ignition of H2-air mixtures excited by a repetitively pulsed nanosecond
discharge." Proc. Combust. Inst. 34(2): 3249-3258.
297
Yin, Z., Montello, A., Carter, C. D., Lempert, W. R. and Adamovich, I. V. (2013b).
"Measurements of temperature and hydroxyl radical generation/decay in lean fuel–air mixtures
excited by a repetitively pulsed nanosecond discharge." Combust. Flame 160(9): 1594-1608.
Yin, Z., Eckert, Z., Adamovich, I. V. and Lempert, W. R. (2015). "Time-resolved radical species
and temperature distributions in an Ar–O2–H2 mixture excited by a nanosecond pulse
discharge." Proc. Combust. Inst. 35(3): 3455-3462.
Zador, J., Huang, H., Welz, O., Zetterberg, J., Osborn, D. L. and Taatjes, C. A. (2013). "Directly
measuring reaction kinetics of QOOH--a crucial but elusive intermediate in hydrocarbon
autoignition." Physical chemistry chemical physics : PCCP 15(26): 10753-10760.
Zellner, R., Hartmann, D., Karthauser, J., Rhasat, D. and Weibring, G. (1988). "A Laser
Photolysis/LIF Study of the Reactions of O(3P) Atoms with CH3 and CH302 Radicals." J. Chem.
Soc., Faraday Trans. 2 84(5): 549-568.
Zhang, X. and Cha, M. S. (2013). "Electron-induced dry reforming of methane in a temperature-
controlled dielectric barrier discharge reactor." J. Phys. D: Appl. Phys. 46(41): 415205.
Zhang, X. and Cha, M. S. (2015). "Partial oxidation of methane in a temperature-controlled
dielectric barrier discharge reactor." Proc. Combust. Inst. 35(3): 3447-3454.
Zhao, Z., Chaos, M., Kazakov, A. and Dryer, F. L. (2008). "Thermal decomposition reaction and
a comprehensive kinetic model of dimethyl ether." Int. J. Chem. Kinet. 40(1): 1-18.
Zhukov, V. P., Rakitin, A. E. and Starikovskii, A. Y. (2008). "Effect of High-Voltage Pulsed
Discharges on Deflagration to Detonation Transition." J. Propul. Power 24(1): 88-93.
Ziegler, G. F., Wagner, E. P., Saggau, B., Maly, R. and Herden, W. (1984). "Influence of a
breakdown ignition system on performance and emission characteristics." SAE Technical Paper.
Ziegler, G. F., Wagner, E. P. and Maly, R. R. (1985a). "Ignition of lean methane-air mixtures by
high pressure glow and arc discharges". Symposium (International) on Combustion 20(1),
Elsevier.
Ziegler, G. F. W., Wagner, E. P. and Maly, R. R. (1985b). "Ignition of lean methane-air mixtures
by high pressure glow and ARC discharges." Symposium (International) on Combustion 20(1):
1817-1824.
298
Zuzeek, Y., Choi, I., Uddi, M., Adamovich, I. V. and Lempert, W. R. (2010). "Pure rotational
CARS thermometry studies of low-temperature oxidation kinetics in air and ethene–air
nanosecond pulse discharge plasmas." J. Phys. D: Appl. Phys. 43(12): 124001.
Zuzeek, Y., Bowman, S., Choi, I., Adamovich, I. V. and Lempert, W. R. (2011). "Pure rotational
CARS studies of thermal energy release and ignition in nanosecond repetitively pulsed
hydrogen-air plasmas." Proc. Combust. Inst. 33(2): 3225-3232.