STUDY OF LIQUID FUEL FLAME TOPOLOGIES VIA NEW DIAGNOSTIC
CONSIDERATIONS

By

Amirreza Gandomkar

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

Mechanical Engineering—Doctor of Philosophy

2024

ABSTRACT

This thesis presents a comprehensive study of liquid fuel flame topologies through the

development and application of novel diagnostic techniques. The complexities associated with

liquid fuel combustion, particularly in the context of aviation and aerospace applications,

demand a deeper understanding of flame behavior and stability. Traditional diagnostic

methods often fall short due to the intricate interactions between liquid droplets, flame

surfaces, and multi-component fuel mixtures.

Our research focuses on addressing these challenges by introducing advanced diagnos-

tic approaches to investigate the structure, stability, and extinction characteristics of liquid

fuel flames. Key areas of exploration include the identification and analysis of reaction zones,

the impact of vaporization dynamics, and the effects of turbulent flow conditions on flame

stabilization. To achieve this, we employ Laser-Induced Fluorescence (LIF) and chemilu-

minescence imaging, alongside advanced numerical image processing algorithms to capture

high-resolution data on flame behavior. These methods enable us to discern fine details about

flame front interactions, droplet vaporization, and localized extinction events. By refining

these diagnostic tools, we aim to provide clearer insights into the parameters influencing

flame stability, such as equivalence ratio, mixing efficiency, and preheat temperature.

In addition, the study integrates computational simulations using CHEMKIN to vali-

date experimental results, allowing for a more comprehensive understanding of how liquid fuel

combustion behaves under varying conditions of turbulence and strain rates. The combined

experimental and computational approach ensures that the findings are both robust and

applicable to real-world aerospace scenarios.

The findings of this study contribute to the broader understanding of liquid fuel

combustion processes and offer valuable implications for the design and optimization of more

efficient and stable combustion systems in aerospace applications. This research not only

enhances our theoretical knowledge but also provides practical guidelines for improving flame

diagnostics and combustion performance.

Copyright by
AMIRREZA GANDOMKAR
2024

ACKNOWLEDGMENTS

First and foremost, I want to extend my heartfelt gratitude to my advisor, Dr. Pat-

ton Allison. His warm welcome into his lab, unwavering patience, and steadfast support

throughout every step of my journey have been invaluable. Dr. Allison’s exemplary passion

for knowledge and science, his commitment to constructive collaboration, and his insightful

guidance have profoundly shaped my thinking and were instrumental in the completion of

this manuscript. I am also deeply grateful to my committee members, Dr. Wichman, Dr.

Toulson, and Dr. Lira. Their constructive criticism, kind demeanor, and expert guidance

helped me navigate the path to selecting the correct research topic. Finally, I would like to

thank my family and friends, whose support and encouragement have been a constant source

of strength, even from miles away.

iv

TABLE OF CONTENTS

Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

Chapter 2 Experiments & Methodology . . . . . . . . . . . . . . . . . . . . . .

Chapter 3 Prevaporized Liquid Fuels . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 4 Spray Liquid Fuels . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 5 Chemiluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

55

68

80

Chapter 6 Summary and Discussion . . . . . . . . . . . . . . . . . . . . . . . .

97

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

v

Chapter 1

Introduction

1.1 Background and Motivation

Liquid fuel combustion is favorable for aviation applications in the aerospace field,

especially gas turbines. Little research has been conducted on liquid fuel reacting flow. The

interaction of liquid droplets with the flame surface, flammability limitation, and complex

fuel’s multi-component properties add up to the difficulty of the problem of liquid fuel combus-

tion. The interaction of liquid spray droplets with the flame surface and the slow vaporization

process compared to the dynamics of the flow raise questions about the flammability of the

mixture and the flame stabilization. The lean condition flammability and flame stabilization

depend on the complex conditions where turbulent chemistry of local mixing fuel-air and fluid

dynamics interact. This issue depends on several parameters: burner structure, fuels, strain

rate, equivalence ratio, and preheat temperature. The flame stabilization is determined by

lift-off and blow-off conditions. Lift-off is an intermediate stage before the blowout occurs,

where the mixture of air and fuel velocity is higher than the flame speed and the flame

detaches from the burner, stabilizing with an offset from the burner. As the mixture velocity

increases, the offset height increases until blow-off happens. Under blow-off conditions, the

velocity of the mixture is high enough that it does not allow the flame and chemistry to

1

stabilize around the burner. Flame stability and extinction result from the balance between

the fluid dynamics timescale and the mixing timescale, or simply put, the balance between

fluid velocity and flame speed. Flame speed depends on various parameters such as the equiv-

alence ratio, mixing efficiencies, and preheat temperature. Flame stability is determined by

whether the flow speed is higher or lower than the flame speed. For liquid flames, stabilization

and extinction are more complicated due to the additional factor of vaporization timescale.

In turbulent and high Reynolds number flow, the mixing and combustion timescales are

longer compared to the fluid dynamics timescale. This can cause severe quenching in the

flame, leading to its destabilization. The main goal of this research is to investigate local

extinctions by focusing on the coupling between the flame structure, the turbulent flow field,

and the interactions of droplets with the flame. Heavy hydrocarbon liquid combustion is

a complex phenomenon due to the cracking into multiple components and the interaction

between liquid droplets and the flame. Most diagnostic techniques are not applicable because

of the limitations imposed by combustion and liquid droplet interference. Consequently,

there is a lack of knowledge in this field, which intrigues researchers to investigate various

aspects of this subject. The motivations behind this research include studying turbulent

flame characteristics, flame stabilization, and the extinction of liquid fuel flames.

Also, the mixing process of liquid fuel is complicated compared to gaseous fuels.

Since the timescale for vaporization is relatively more significant than the fluid dynamics

timescale, the flammability of the mixture is unknown due to different local equivalence.

Additionally, liquid fuel studies with diagnostic techniques face several difficulties. Liquid

droplet interference with optical measurement is a severe problem in the diagnostic approach.

Hence, most studies involve gaseous fuels, while few studies have been done on liquid fuels.

Optical measurement methods in the combustion process are widely popular, and

monitoring species emissions and controlling parameters influencing the combustion phe-

nomenon is essential. The most common approach to studying the flame structure is Planar

Laser-Induced Fluorescence, PLIF. At the same time, chemiluminescence is the easiest way

2

to evaluate intermediate species, revealing vital information regarding flame characteristics.

These two approaches are implemented in our studies and will be introduced later.

Chemiluminescence is the light emitted from a chemical reaction. An excited inter-

mediate species that has high energy to reach a stable state releases its energy as photons

of light. The chemiluminescence light emitted by the flame is well suited for diagnostic

purposes. Chemiluminescence is used to obtain information in the reaction zone, which

is related to combustion performance, species formation, quenching, and emission. This

method has been widely used to measure equivalence ratio, heat release, flame front location,

and other potential information that can illustrate the condition of combustion. One of the

biggest advantages of chemiluminescence is that it is much cheaper than other measuring

methods in combustion and it can be set up much easier. Unlike laser diagnostic or other

methods, chemiluminescence does not need an external source for light since this process

emits light during species formation by heat release. This emitted light is used as a means

of diagnostic measurement since it has information about the local combustion processes.

The most important species to be investigated in the combustion process are OH1, CH2, and

possibly CO2

3. It has been noted that the relative intensities of different species depend

on the fuel equivalence ratio, pressure, strain, and performing conditions. Much research

has been done in numerical and experimental methods to find out the importance of species

intensities and connect them to the conditions of systems.

Laser Induced Fluorescence (LIF) is optical emission from molecules or atoms that

have been excited by an external source of energy. The flow is illuminated by a tunable

laser that overlaps an allowed electronic vibrational or rotational transition in an atom or

molecule. By absorbing the energy, the molecule reaches an excited state. The molecule

can release that energy in two ways: 1-colliding with another molecule, named collisional

quenching. 2-relaxing back to the ground state and emitting photons as fluorescence. LIF is

1Hydroxyl radical
2Methylidyne radical
3Carbon dioxide

3

a unique and useful method for measuring variables or visualization. It can be implemented

to measure various data such as density, concentration, velocity, temperature, pressure, etc.

It can be used at a single point, line, or even a sheet, which is called PLIF4. LIF (Laser-

Induced Fluorescence) is a resonant technique that requires the laser to be tuned to a specific

wavelength. Single-photon LIF is linear, but it often has a low signal, necessitating the use of

an Intensified CCD5 (ICCD) to amplify the signal. However, this amplification also increases

the noise level.

The LIF method differs for the liquid phase and gaseous phase. For the liquid phase,

fluorescence particles are seeded with the flow, and by absorbing energy and releasing it,

measurements of concentration, temperature, etc., are performed, while in the gaseous phase,

the seeded particle is not a proper way since the species have unique excitation wavelengths,

and the illuminating part is not just due to the particles. For the gas phase, a specific

wavelength for tuning the laser is needed to excite a specific species, and that excited species

will release photons after relaxation.

Examining flame structure gives insight into the influence of turbulent flow and is

beneficial for simulation and modeling. Many studies have investigated the effect of pressure

[Soika et al., 2003,Lachaux et al., 2005], turbulent intensity [Jenkins and Cant, 2002,Haq et al.,

2002], and Lewis number [Rutland and Trouv´e, 1993, Chakraborty and Cant, 2005, Renou

et al., 1998] on flame topology. Most of the previous studies focused on gaseous fuel flame. The

studies on the liquid-fuel flame topology are limited and restricted due to the interference

of liquid droplets in diagnostic techniques and the complexity of heavy hydrocarbon-fuel

combustion.

Walsh [Smooke et al., 1996] investigated CH and OH chemiluminescence in laminar

diffusion flame for methane-nitrogen fuel with experimental and simulation methods. They

used two different chemical mechanisms for the numerical procedure and then compared the

results with Laser Infrared Fluorescence diagnostic in their experiment. The first mechanism

4Planar Laser-Induced Fluorescence
5Charge-Coupled Device

4

was GRI Mech 2.11 and the second was an alternate hydrocarbon mechanism. The results

as shown in Figure 1.1 indicate that the GRI mechanism predicts CH and OH intensities

and their spatial resolution quite accurately while the alternate hydrocarbon mechanism has

a better lift-off match with the LIF method.

Figure 1.1: Comparison of OH PLIF and two simulations [Smooke et al., 1996]

One of the parameters which affect the emission and combustion are characteristics

of fuel. Armingol [Garc´ıa-Armingol and Ballester, 2014] researched different compositions

of fuels and their effect on chemiluminescence emission and they also tried to come up with

a correlation between chemiluminescence signals and equivalence ratio. They used about

12 different mixtures of CH4, CO, H2 and CO2. Numerical modeling with CHEMKIN was

also performed in their study with two different simulation models to understand which

one predicts the experimental results better. The models were performed by Chemkin for

turbulent flame situations, named as the ”free propagating laminar flame” and ”the perfectly

stirred reactor” (PSR). They observed that by adding methane to syngas (CO and H2) the

maximum concentration of OH increases because CH has an important role in formation of

OH.

An experiment focusing on the effect of diluent in a methane-air mixture with CO2 and

N2 was done by Guiberti [Guiberti et al., 2017]. Their experiments indicated that increasing

5

CO2 and N2 diluent would decrease the chemiluminescence of CH and OH. The decreasing

rate will magnify if the concentration of diluent increases. Also as shown in figure 1.2 ,

they expressed that the ratio of CH/OH does not change significantly with the change in N2

dilution, however increasing CO2 concentration led to decrease in the ratio of intensities.

Figure 1.2: Species chemiluminescence ratio vs percentage of diluted N2 and CO2 and
equivalence ratio [Guiberti et al., 2017]

Tripathi [Tripathi et al., 2012] investigated chemiluminescence of OH and CH in the

premixed flames of methane air for various equivalence ratios by observing the reaction zone.

The first impression in their results shown in Figure 1.3 is that background emission of CO2

decreases with increasing equivalence ratio, which is normal due to low production of CO2 at

rich flame combustion, but the impact of equivalence ratio on other species emission is not

certain due to the fact that they all are affected by the presence of CO2 in the background.

To eliminate background CO2 and obtain better results for CH and OH intensities, they used

a new partial least square regression (PLS-R) sensing method and measured OH and CH

chemiluminescence for a range of equivalence ratios. They showed that OH peak occurs at

an equivalence ratio of 0.8 and CH intensity peak happens around equivalence ratio (ϕ) of 1

and both intensities start to decrease after their peak with increasing equivalence ratio. They

also came up with a correlation for the ratio of OH/CH and equivalence ratio.

6

Figure 1.3: Species chemiluminescence ratio vs percentage of diluted N2 and CO2 and
equivalence ratio [Tripathi et al., 2012]

Higgins’s [Higgins et al., 2001] experiment focused on the influence of pressure, equiv-

alence ratio, and flow rate on OH chemiluminescence intensity in the lean combustion region,

laminar premixed methane-air flame. Their results seen in Figure 1.4 showed that in the lean

combustion region, increasing pressure decreases OH intensities due to increasing collisional

quenching at higher pressures. Also, OH chemiluminescence increased with increasing flow

rate since they have a proportional relationship with each other.

Figure 1.4: OH chemiluminescence intensity vs equivalence ratio and their correlation with
pressure [Higgins et al., 2001]

Panoutsos [Panoutsos et al., 2009] simulated 8 different mechanisms and sub-mechanisms

introduced by other researchers and compared them with experimental results. Their research

focused on the effect of equivalence ratio on CH and OH chemiluminescence and heat release

ratio. They expressed that OH could be helpful in tracking the heat release rate since it

has the lowest distance between the peak of OH and the peak of the heat release profile,

which can be seen in Figure 1.5. Also, OH and CH intensities were calculated and measured

7

for all mechanisms and experimental data, showing that these intensities depend greatly on

equivalence ratio and strain rate. However, it was mentioned that the ratio of OH to CH

depends only on the equivalence ratio and is independent of strain rate. All the mechanisms

in this research showed a good match with the experimental data. Finally, they mentioned

which mechanism works well for each species of interest.

Figure 1.5: Calculated peak-to-peak distances between heat release rate and HCO, OH ×
CH2O, OH and CH, as a function of equivalence ratio [Panoutsos et al., 2009]

Constandinos [Mitsingas et al., 2019] investigated PIV6 and LIF7 in turbulent pre-

mixed methane flame using a Bunsen flame burner stabilized with a pilot burner. They

obtained flame structures by analyzing the data of the CH and OH PLIF. Also, reactant

pockets residing in the product stream, product pockets residing in the reactant stream,

and interaction between out-of-plane and in-plane flame sheets were extracted from this

analysis. Furthermore, the PLIF and PIV experiments’ results measured the axial velocity

and fluctuations, and the reaction layer was tracked down.

Rosell [Rosell et al., 2017] studied multi-species (CH, OH, CH2O, and toluene) PLIF

to better understand the structure of turbulent flame and reaction zones. In order to track

6Particle Image Velocimetry
7Laser-Induced Fluorescence

8

the unburned fuel, they excited toluene, a product formed from the breakdown of methane.

The methane-air mixture was pumped through the McKenna burner with a pilot burner

as the stabilizer. A Laser Doppler Anemometer device was implemented in this work to

measure axial and radial velocities and their oscillations. They were able to calculate the

thickness of OH, CH, and CH2O layers and the thickness of overlapping layers between each

one by the results of PLIF. The preheat zone, which was studied by following an overlapping

layer of toluene and CH2O, depends on the Reynolds number. The thickness of the preheat

region increased by Reynolds number, but up until the flow is laminar. For turbulent flow,

the preheat zone is independent of the Reynolds number. They observed that the CH layer,

an indicator for fuel consumption rate, stayed thin through the experiments over variations

of Reynolds number. The same result was obtained for heat release, characterized by an

overlapping layer of OH and CH2O. Also, by having velocities distribution, Karlovitz number

and Damkohler number were obtained. Other features such as flame surface density, flame

wrinkling ratio, and fuel consumption velocity are also measured. For Ka < 40, the surface

wrinkle ratio depends on Reynolds number and flame height, and for Ka > 40, the surface

wrinkle ratio is independent of Reynolds number.

Tsushima [Tsushima et al., 1998] studied droplet cluster behaviors in a premixed spray

flame using the chemiluminescence method and Mie scattering technique. Their diagnostic

system, which allows for spatially resolved chemiluminescence measurements, is the only

viable option for monitoring the transient structure of flames, as laser-induced fluorescence

(LIF) cannot provide continuous time-series signals due to the intermittency of pulsed lasers

with droplet clusters. In the non-combusting case, there was no cluster seen in the images.

In contrast, the continuous spray image was eroded in the combusting case, or no Mie

scattering was detected, which is the sign of a droplet cluster. At the same time, the

OH chemiluminescence increased, which suggests that the reason behind these events is

flame propagation, not fluid motion. In the following period, increased OH intensity and

Mie scattering proved that there are combustion reactions inside the spray region and that

9

droplets burn with flame propagation. In another time step, they observed erosion in spray

in the downstream direction while there is no combustion, which means the upstream region

of spray prevents flame propagation. At last, observing the disappearance of droplet clusters

downstream, they came up with an estimation of disappearance speed.

Cessou [Stepowski et al., 1994] performed investigations of the OH concentration in the

stabilization zone of a two-phase jet flame. OH PLIF was used to observe flame structures in

the lift-off area of a spray jet flame. They observed that the flame propagates in two diverging

and opposed diffusion structures. The thickness of these two structures was measured and

reported. Lift-off height was also measured. They noticed that the chance of having a double-

branched structure is higher for low atomization velocity. Also, the droplets cross the inner

reaction zone to vaporize in the hot mixing structures in the double-branched structure. The

peak concentration level they observed was in agreement with the measurement reported by

others.

Lemaire [Lemaire et al., 2009] has evaluated liquid spray characteristics using a highly

efficient nebulizer in the McKenna piloted burner for three different liquid fuels. Estimation

of droplet size, evaporation time, and length of spray scale has been studied. The droplet

size and liquid spray measurements were measured using the Mie scattering technique at 532

nm and 1064 nm. It was reported that droplet size distribution for all three fuels was similar

to the spray structure exiting the nebulizer captured by the Mie scattering method. Liquid

completely vaporized over 15-16 cm above the burner for all fuels.

Meyer [Meyer et al., 2005] investigated the formation of soot in an unsteady turbu-

lent swirled stabilized liquid JP-8 spray combustor. They implemented several diagnostic

techniques such as PLIF, Mie scattering, and LII. With the LII method at the swirl section

exit area around 415 nm to 500 nm, soot formation and its volume fraction were measured.

OH PLIF at 283.922 nm wavelength gave them the distribution of OH and tracked down the

local equivalence ratio. Mie scattering was also performed to mimic spray patterns in the

combustion flow. They observed that decreasing temperature and lack of oxidizer increase

10

the soot formation in lean combustion.

Flame surface density is one of the essential details of flame structure, which expresses

the flame’s wrinkling scale, and it can lead to the flame burning velocity or consumption speed.

Zhang [Zhang et al., 2019] experimented with OH PLIF and CH chemiluminescence for a

methane-air mixture in bluff-body and swirl stabilizer flame. The behavior of flame root and

burning velocity are studied using flame surface density data. Weijie and Meng Zhang [Zhang

et al., 2014] studied turbulent methane flame enriched with hydrogen using instantaneous OH

PLIF. Their result indicated that the addition of hydrogen enhanced flame surface density,

especially in higher turbulent cases. The Flame Surface Density’s (FSD) profile for the mean

progress variable was symmetric with the maximum location at < c >= 0.5. For liquid fuel,

as it appears, there is no background of flame surface measurement with CH PLIF. Usually,

OH PLIF is conducted for capturing FSD, and due to interference of Mie scattering in the

PLIF experiment, there are limited PLIF studies for liquid fuels.

Rodrigues [Rodrigues et al., 2015] investigated the addition of air and hot diluted co-

flow to the ethanol swirl spray flames. They implemented Laser Doppler Anemometry (LDA)

and Phase Doppler Anemometry (PDA) to measure velocities of gas and spray, respectively.

Alongside high-speed visualization, they measured temperature in the spray region using

anti-Stokes Raman spectroscopy. Their experiment’s result can be used for further simulation

models since it contains various information about the different regions of spray flame. They

observed that hot diluted co-flow has a significant effect on the liquid breakup mechanics.

In hot diluted where the droplet is surrounded by hot gas, the maximum temperature shifts

toward the lean section, and vaporization of droplets accelerates.

Sadiki [Sadiki et al., 2005] experimentally investigated the vaporization rate and com-

bustion of kerosene in a turbulent spray flame. They utilized Phase Doppler Particle Analyzer

(PDPA) to capture velocity and sizes of droplets. In their burner facilities, a methane-air

mixture piloted burner was used to stabilize the spray flame. OH chemiluminescence was

performed at the wavelength of 308 nm as well. Gas chromatography was performed to

11

measure the concentration of various gases such as O2, H2, CH4. Due to resolution and sec-

ondary light scattering from droplets, Mie scattering images were only used to show droplet

dispersion. They observed that faster evaporation, which can be achieved by decreasing the

centrifugal Stokes number, does not imply faster combustion. They expressed that dispersion

of fuel in the gaseous phase is slower than in the liquid phase. Therefore, the represented

flame length is significantly longer compared to the other one.

The DNS results as shown in Figure 1.6 indicate the effect of Lewis number of liquid

fuels on heat release and fuel consumption in the region of high curvature [Aspden et al.,

2017, Savard and Blanquart, 2015, Savard et al., 2015]. Furthermore, the instantaneous CH

PLIF imaging visualizes a substantial variation of CH signals in high-curvature areas for

high-Le-number fuels. While for fuels with Le number around unity, there is weak variation in

the CH fluorescence signal [McManus et al., 2021]. This behavior intrigues the idea that there

may be a relation between Le number, the CH signal, and heat release. The quantification

between the CH signal and curvature for different fuels is required to prove this idea.

Figure 1.6: Strong variation of temperature, fuel consumption, and heat release around
regions of high curvature captured by DNS results [Aspden et al., 2017]

12

1.2 Gaps and holes in the Research

In this section, I am going to talk about gaps and holes in diagnostic methods in flame

studies. First, in the chemiluminescence method, it should be noted that the light emitted by

species in combustion produces a relatively low signal compared to other methods, making

it more challenging to detect and extract detailed information about flame characteristics.

To address the low signal in chemiluminescence, an intensifier can be applied in front of the

camera or detector. However, intensifiers typically cause image blurriness. This blur can be

partially reduced using image processing techniques.

A primary challenge in this method is detecting CO2 chemiluminescence, as it spans

a wide range of wavelengths and affects measurements of other species. While some studies

have attempted to calculate CO2 chemiluminescence, they are limited to light hydrocarbon

fuels. The models for CO2 are generally confined to a specific range of equivalence ratios and

do not fully predict all chemiluminescence trends. For heavier hydrocarbons like kerosene,

which exhibit more complex behavior, there is no detailed study. To achieve more accurate

measurements for kerosene, a detailed mechanism tailored to this fuel is necessary to improve

our understanding of the problem.

In Laser-Induced Fluorescence (LIF), interference from species bands in other LIF

signals remains an unresolved issue. For example, in CO LIF, there are reports of interference

from the C2 Swan band in the LIF signals. Similarly, in OH LIF (0,0) around 308 nm,

interference occurs due to the proximity of the CH C-X8 band at 314 nm. Additionally,

for visualizing spray flames, single-wavelength LIF is often unsuitable due to interference

from droplets that scatter the fluorescence signal. Some researchers suggest using LIF at an

alternative wavelength to mitigate droplet scattering.

Another challenge is the effect of kerosene fuel on diagnostics and flame behavior.

Currently, there is no comprehensive mechanism for kerosene combustion, though some

surrogates have been proposed to simulate it. The intermediate species produced during

8Transition from the third excited electronic state to the ground electronic state

13

kerosene combustion and their impact on flame front characteristics, chemiluminescence, and

LIF remain largely unknown. The interaction between fuel composition and flame structure,

which can lead to flame extinction or extended flammability, requires further investigation.

A detailed mechanism to accurately predict the reactions involved in kerosene combustion is

needed.

Regarding gaps in flame studies, further research is suggested on the interaction

between flames and sprays. Studying the effects of product recirculation, unsteady mixing,

or preheating could yield valuable insights. Flame extinction caused by droplet clusters or

changes in flame length due to such interactions has been explored by only a few researchers.

Overall, research on liquid spray flames is limited, indicating a need for the application or

combination of various diagnostic techniques in this area.

Conventional laser diagnostics are not recommended for liquid fuel combustion due

to interference from liquid fuel and intermediate combustion species. To study the reac-

tion layers of flames and gather information on the extinction process, a new diagnostic

method is necessary to enable imaging of liquid fuel combustion despite interference from

fuel components. In collaboration with the Air Force Research Laboratory (AFRL), a new

diagnostic technique has been developed, using CH PLIF imaging on prevaporized heavy

hydrocarbons to study flame structure and extinction. This technique allows for the analysis

of liquid fuel flame structure without interference from fuel components. Initial observations

suggest that a Lewis number effect may influence the fluorescence signal in regions of high

curvature, motivating further investigation into the correlation between Lewis number and

signal/curvature.

Additionally, simultaneous CH/OH PLIF observations indicate that OH PLIF may

be unable to detect local extinction within the flame structure. Since OH species have a

slower quenching rate compared to CH, they may remain present even after flame extinction,

potentially leading to an overestimation of flame characteristics in OH PLIF. The OH PLIF

signal typically shows a simpler, two-sided structure without product stream orientation,

14

whereas CH PLIF provides a more detailed view of breaks in the flame structure. As shown

in Figure 1.7, simultaneous CH/OH PLIF suggests that OH PLIF fails to detect certain local

extinction regions that CH PLIF captures. This study proposes that CH PLIF may offer a

more reliable diagnostic technique for turbulent flames where local extinctions are prevalent.

Figure 1.7: Local extinctions, circled, observed in the flame surface via CH PLIF imaging
in turbulent liquid fuel jet flames [McManus et al., 2021]

Furthermore, I aim to expand our research on spray liquid flames during the extinction

process using our advanced diagnostic technique. The primary motivations for this thesis

include:

1. How are finite-rate chemistry effects in low-temperature processes represented in changes

to the local flame structure and heat release?

2. How do global extinction behaviors in spray flames differ from those in prevaporized or

gaseous fuel flames? Are there experimental indicators of extinction in turbulent sprays?

3. Is our standard diagnostic toolbox for gaseous flames sufficient for studying liquid fuels?

If not, what adjustments are necessary?

a. Interference-free fluorescence and chemiluminescence from fuel, droplets, or polycyclic

aromatic hydrocarbons (PAH)

b. Development of a numerical algorithm to address the complexities of turbulent structures

This research aims to examine the turbulent flame structure in liquid fuels, investi-

gating the effects of fuel type and equivalence ratio using CH PLIF and CH*/OH* chemilu-

15

minescence. A new PLIF diagnostic technique compatible with liquid fuels is utilized in this

study. Turbulent flame properties such as flame curvature, flame surface density, turbulent

flame speed, and turbulent flame area are analyzed using an innovative image processing

approach applied to CH PLIF images. Later in this work, the image processing technique

for identifying reactant and product regions will be introduced. Curvature and flame surface

density measurements for prevaporized and spray flames of n-dodecane and n-heptane are

plotted and compared to premixed methane under similar conditions, as defined in the Borghi

diagram [Borghi, 1985]. Finally, CH and OH chemiluminescence data are compared with

simulations from various sub-mechanisms to evaluate their accuracy.

This thesis proposes an investigation into turbulent liquid flame characteristics by ana-

lyzing the topological features of liquid fuel combustion near blow-off, linking local extinction

characteristics to global phenomena.

1.3 Research Objectives and Contributions

In this research, the primary objective is to investigate the turbulent flame structure of

heavy hydrocarbon liquid fuels using advanced diagnostic techniques. By implementing CH

PLIF (Planar Laser-Induced Fluorescence) to capture flame structure and devising a novel

numerical image processing method, this study seeks to measure the turbulent characteristics

of flames and lkocal extinction process in liquid fuels. These types of topological statistics

have not been experimentally performed before in liquid fuels.

The main complication with UV diagnostics for liquid fuels is interference from fluores-

cence of the fuel itself, Mie scattering from droplets or soot, and Raman scattering from the

fuel. The C-X Q branch CH method performs well in liquid fuels because it is on-resonance, as

same input laser wavelength to imaged wavelength, thus avoiding Raman and fuel PLIF, and

because the method requires low pulse energies, there is minimal interference. The R-branch

excitation works for droplet laden flows because it is off-resonance and the Mie scattering

from droplets can be effectively filtered out. Hence CH can be performed in spray flames to

16

see local extinctions. The capability to conduct CH PLIF in both spray and prevaporized

liquid fuels is innovative and valuable for obtaining extinction data, which is not accurately

captured with OH PLIF

Our primary objective is to develop an algorithm that accurately captures the flame

structure and distinguishes between the reactant and product side of the flame. This algorithm

enhances the understanding of flame topology by improving the accuracy and reliability of

identifying combustion zones.

It addresses the challenge of capturing the dynamic and

complex structure of turbulent flames, providing a more detailed and nuanced understanding

of the combustion process. Furthermore, the algorithm can highlight the shortcomings

of common approaches used in the combustion industry. This contribution is crucial for

advancing diagnostic techniques in combustion research.

Additionally, we aim to analyze the turbulent flame features of prevaporized liquid

fuels, focusing on flame structure statistics, extinction events, and the interaction of turbulent

flow with the flame. The detailed measurement and analysis of these features offer valuable

insights into the combustion behavior of prevaporized fuels. By understanding how turbulence

affects flame stability and extinction, this research contributes to the optimization of fuel

combustion processes, leading to more efficient and stable combustion in practical applications.

This is particularly relevant for improving the performance of gas turbines and other aerospace

propulsion systems.

Also, we want to investigate the turbulent flame features of liquid spray fuels, examin-

ing droplet interactions with the flame, fragmentation, vaporization, and flame stabilization.

This objective highlights the differences and similarities between spray and prevaporized

fuel, offering a comprehensive understanding of how liquid droplets influence flame behav-

ior. The insights gained from this investigation are critical for optimizing fuel injection and

atomization techniques, thereby enhancing combustion efficiency and reducing emissions in

liquid-fueled engines. This research also contributes to the development of more accurate

predictive models for spray combustion.

17

Chemiluminescence is an easy diagnostic for industrial applications to observed the

flame and derive information about heat release and rough mixture ratio. For CFD efforts,

the combustion process is calculated from a chemical mechanism optimized for particular

fuels. For comparisons between CFD and industrial measurements, often the only available

data on the flame is chemiluminescence. Thus, the CFD calculation also attempts to predict

chemiluminescence and require a chemiluminescence chemical sub-mechanism in addition to

the primary fuel mechanism. Often, these sub mechanisms are included as part of the main

mechanism.

However, issues arise because the primary fuel mechanism is not optimized for the

prediction of the primary chemiluminescence species and also because the chemiluminescence

sub-mechanisms have been optimized for methane, not liquid fuels. CFD modelers are often

unaware of the assumptions under which the mechanism and sub-mechanisms have been

derived and optimized. This drives motivation to check if such mechanisms should be trusted

for CFD-experimental comparison of liquid fuels.

The analysis of chemiluminescence measurements enhances the accuracy of predictive

models for combustion behavior. These improved models support the design of more reliable

and efficient combustion systems, benefiting a wide range of applications in aerospace and

beyond. The findings of this research have practical implications for improving combustion

efficiency and stability in aerospace applications. The optimized combustion processes and

advanced diagnostic techniques developed through this study contribute to the design of

more efficient and environmentally friendly propulsion systems.

The further objective is to conduct chemiluminescence measurements and simulations

of CH*/OH* radicals for liquid fuel, validating combustion mechanisms and integrating

findings with other diagnostic methods. The application of chemiluminescence techniques

provides a powerful tool for probing flame chemistry and validating theoretical combustion

models. By integrating these measurements with other diagnostic methods, the research

enhances the overall understanding of flame chemistry, leading to more accurate simulations

18

and predictions of combustion behavior. This contribution is essential for advancing both

experimental and computational studies in combustion science.

Furthermore, future implementations of this research include exploring the application

of the developed algorithm and diagnostic techniques to other types of liquid fuels and

combustion systems. This includes extending the methodologies to biofuels and alternative

energy sources, thereby broadening the impact of the research on sustainable energy solutions.

Additionally, the integration of advanced computational fluid dynamics (CFD) models with

experimental data can further refine predictive capabilities, leading to more precise control

over combustion processes.

Another future direction is the application of the developed techniques in industrial

combustion systems, such as power plants and internal combustion engines, to enhance

efficiency and reduce emissions on a larger scale. Collaborations with industry partners could

facilitate the transfer of these technologies from the laboratory to real-world applications,

driving innovation in combustion engineering.

This research advances the understanding of liquid fuel combustion processes and

offers practical implications for improving combustion efficiency and stability in aerospace

applications. The research advances diagnostic methods in combustion science, providing

more accurate tools for studying flame structures and dynamics. The findings pave the

way for future studies and innovations in diagnostic techniques and combustion system

design, contributing to the broader field of fluid dynamics and combustion science. The

insights gained from this research support the optimization of combustion processes for

both prevaporized and spray liquid fuels. The continued exploration and development of

these techniques promise significant advancements in both academic research and practical

applications in the energy and aerospace sectors.

19

1.4 Outline of Dissertation

The chapters of this thesis delve into the study of liquid fuel flame topologies through

the application of novel diagnostic techniques, addressing the complexities of liquid fuel

combustion in various contexts.

Chapter 1 introduces the motivation behind studying liquid fuel flame topologies,

outlining the research objectives and significance. This chapter provides an overview of

the challenges in liquid fuel combustion and the need for advanced diagnostic techniques to

enhance our understanding.

Chapter 2 details the experiments and methodologies employed in this study.

It

introduces the experimental setup, validation procedures and explains the data collection

and analytical methods used. The chapter also covers the diagnostic tools such as Laser-

Induced Fluorescence (LIF), high-speed imaging, and chemiluminescence, along with the

experimental protocols for both prevaporized and spray liquid fuels.

Chapter 3 focuses on prevaporized liquid fuels, examining their characteristics and

flame topology. It presents the diagnostic findings related to flame stability and extinction

events and discusses the results in the context of theoretical models and combustion efficiency

implications.

Chapter 4 explores spray liquid fuels, analyzing the interaction of droplets with the

flame and the resulting fragmentation and vaporization. This chapter highlights the diagnostic

findings on flame stabilization and quenching and provides a comparative analysis with

prevaporized fuels, emphasizing the impact of turbulent flow conditions.

Chapter 5 investigates the application of chemiluminescence in studying liquid fuel

flames.

It elucidates the fundamentals of chemiluminescence in combustion and its role

in validating combustion mechanisms proposed by other researchers through quantitative

analysis. The chapter discusses the advantages and limitations of this technique, as well as

its integration with other diagnostic methods.

Finally, Chapter 6 summarizes the key findings of the research, discussing the implica-

20

tions for future studies and practical applications in the aerospace industry and combustion

system design.

It concludes with reflections on the research contributions and potential

directions for further investigation.

1.5 Summary

The introduction of this dissertation emphasizes the significance of understanding

turbulent flame structures in heavy hydrocarbon liquid fuels, particularly for aerospace

applications like gas turbines. The research aims to fill the knowledge gaps in liquid fuel

combustion by developing and applying advanced diagnostic techniques to study the complex

interactions between liquid droplets and flames.

The importance of liquid fuel combustion is highlighted, focusing on the challenges

posed by the interactions between liquid droplets and flames and the limitations of existing

diagnostic methods. This research is motivated by the need to better understand turbulent

flames and improve diagnostic accuracy, seeking to capture flame structures and interactions

more precisely.

The significance of this research lies in its contributions to developing more efficient

and environmentally friendly combustion technologies by advancing diagnostic techniques and

improving the understanding of flame dynamics. The findings support optimizing combustion

processes for both prevaporized and spray liquid fuels, contributing to designing more efficient

and stable combustion systems.

Furthermore, this research suggests potential future directions for further investiga-

tions, including applications to other fuel types and integration with computational models

for better control over combustion processes. This study’s advancements in diagnostic meth-

ods and understanding of liquid fuel combustion processes offer practical implications for

improving combustion efficiency and stability in aerospace applications, paving the way for

future studies and innovations in the broader fluid dynamics and combustion science field.

21

Chapter 2

Experiments & Methodology

2.1

Introduction

In this chapter, I will detail the experimental procedures chosen to study the turbulent

flame structures of pre vaporized and spray liquid fuels. Each experiment’s conditions and

the diagnostic techniques implemented are explained in detail to provide a comprehensive

understanding of the methodologies used. The major objective of this chapter is to intro-

duce and devise a novel numerical algorithm designed to decipher intricate turbulent flame

structures. This algorithm focuses on capturing and calculating flame edges, and relevant

statistics, thereby providing valuable insights into the dynamics of liquid fuel combustion.

This foundational work sets the stage for the subsequent analysis and interpretation of data,

which are crucial for advancing our understanding of flame behavior in turbulent conditions.

2.2 Diagnostic techniques

Diagnostic techniques in combustion are essential tools for understanding and ana-

lyzing the complex processes that occur during the combustion of fuels. These techniques

provide valuable insights into the chemical and physical phenomena that govern flame be-

havior, including species concentration, temperature distribution, pressure dynamics, and

22

overall thermal efficiency. The ability to accurately monitor and measure these parameters

is crucial for optimizing combustion systems, improving fuel efficiency, reducing emissions,

and advancing our fundamental knowledge of combustion science.

One of the most widely used diagnostic techniques in combustion research is Pla-

nar Laser-Induced Fluorescence (PLIF). PLIF is renowned for its ability to provide high-

resolution, two-dimensional images of chemical species within a flame. By exciting specific

molecules with a laser and capturing the resultant fluorescence, researchers can visualize the

spatial distribution of key intermediates and reaction products. This technique is particu-

larly useful for studying flame structure, turbulence-chemistry interactions, and pollutant

formation processes.

Another powerful diagnostic method is chemiluminescence, which involves detecting

the light emitted by excited species as they return to their ground state. Chemiluminescence

is a relatively straightforward technique that can be used to evaluate the presence and

concentration of intermediate species such as CH*, OH*, and C2*. These species are often

indicators of important combustion reactions, and their detection provides critical information

about flame stability, ignition processes, and reaction kinetics.

In addition to PLIF and chemiluminescence, several other diagnostic techniques are

commonly employed in combustion studies. These include laser-induced incandescence (LII)

for soot measurement, tunable diode laser absorption spectroscopy (TDLAS) for species

concentration and temperature measurement, and particle image velocimetry (PIV) for flow

field visualization. Each of these methods offers unique advantages and can be tailored to

specific research objectives.

The combination of these diagnostic techniques in combustion research allows for a

comprehensive analysis of flame behavior and combustion dynamics. By combining multiple

methods, researchers can obtain a more complete picture of the complex interactions occurring

within a flame, leading to more accurate models and better-engineered combustion systems.

In this chapter, I will go through the principles and applications of PLIF and chemi-

23

luminescence, the two primary diagnostic techniques utilized in our studies. I will discuss

their implementation in our experimental setups, the advantages they offer, and the specific

insights they provide into the turbulent flame structures of prevaporized and spray liquid

fuels. Through this exploration, I aim to highlight the critical role of advanced diagnostic

techniques in advancing combustion science and technology.

2.2.1 Chemiluminescence

There are several different methods in optical procedures, and chemiluminescence is

one of them. Chemiluminescence is the light emitted from a chemical reaction. An excited

intermediate species with high energy to reach a stable state releases its energy as photons of

light. The chemiluminescence light emitted by the flame is well suited for diagnostic purposes.

Chemiluminescence is used to obtain information in the reaction zone related to

combustion performance, species formation, quenching, and emission. This method has

been widely used to measure equivalence ratio, heat release, flame front location, and other

potential information that can illustrate combustion conditions. One of the most significant

advantages of chemiluminescence is that it is much cheaper than other measuring methods

in combustion, and it is way more accessible.

Unlike laser diagnostic or other methods, chemiluminescence does not need an external

source for light since this process emits light during species formation by heat release. This

emitted light is used as a diagnostic measurement since it has information about the local

combustion processes. Flame chemiluminescence has been employed in various combustion

applications, from simple Bunsen flames to full-scale gas turbine engines, as a diagnostic tool.

Flame chemiluminescence is a well-established diagnostic in gaseous flames, and its behavior

has been represented in several simulations [Nori and Seitzman, 2009, Panoutsos et al., 2009].

Due to adverse circumstances for laser diagnostics or restricted optical access in complicated

turbulent flows, chemiluminescence is sometimes the only experimental diagnostic available.

Chemiluminescence imaging is frequently utilized as a validation and comparison

metric for mathematical models [Hossain and Nakamura, 2014] .The most important species

24

to be investigated in the combustion process are OH, CH and possibly CO2. By measuring

these species, this method is capable of evaluating equivalence ratio sensing [Hardalupas and

Orain, 2004], heat release [Guyot et al., 2010], and marking reaction zone [Guethe et al.,

2012]. There is a different approach to this investigation, such as numerical modeling or

experimental research. It has been noted that the relative intensities of different species

depend on the fuel equivalence ratios, pressure, strain, and performing conditions.

Several studies have been done in numerical and experimental methods to find the

importance of species intensities and connect them to the conditions of systems. Knowing the

mechanism of different species is the most crucial part of every chemiluminescence modeling.

However, only methane oxidation processes have been used to analyze these sub-mechanisms

in the past. Furthermore, chemiluminescence as a diagnostic tool for higher alkane species and

kerosene fuels has not been adequately investigated. As shown in Figure 2.1, the real problem

is the spectroscopic balance between the genuine chemiluminescence signal from the species

of interest and a wideband background pedestal attributable to CO2 [Lauer and Sattelmayer,

2010]. In methane flames, the ratio between these two species has fluctuated as a function

of the equivalence ratio [Nori and Seitzman, 2008]. When most of the chemiluminescence

signal is due to the background elevation rather than the CH, C2, or OH species, which

have been linked to flame performance, the issue arises of what is being photographed. This

study, which is the second part of the study, includes the experiment results to evaluate

the mechanisms and modeling. These models will investigate kerosene fuels’ performance,

allowing for correlations between experimental results and combustion simulations. Nori and

Seitzman [Nori and Seitzman, 2009] studied many aircraft kerosene fuels, including Jet-A,

surrogates, and JP-8.

Several main mechanisms and sub-mechanisms are used to calculate main species and

intermediate ones. Most chemiluminescence physics and kinetics are validated for methane

with GRI Mech 3 [Smith et al., 2011] mechanism. For n-Heptane and n-Dodecane, the

JetSurF2.0 [Wang et al., 2010] mechanism has provided detailed formation and quenching

25

Figure 2.1: Chemiluminescence spectrum of a methane-air flame. The radical emissions
from OH ⋆, CH ⋆, and C ⋆
2 [Lauer and
Sattelmayer, 2010]

2 are superimposed by the broadband pedestal from CO⋆

reactions with a built-in sub-mechanism to calculate CH and OH. JetSurF2.0 consists of 348

species and 2163 reactions.

The development effort centers on n-Dodecane and n-butyl-cyclohexane. However,

the model also includes the high-temperature chemistry of all n-alkanes up to n-Dodecane.

HyChem [Wang et al., 2018, Xu et al., 2018] is a detailed high-temperature mechanism that

includes seven reaction steps to describe JP-8 POSF10264 thermal decomposition and USC

Mech IIa as the foundational fuel chemistry model. The cracked products considered are

C2H4, H2, CH4, C3H6, 1 − C4H8, i − C4H8, benzene, toluene, CH3. The chemiluminescence

sub-mechanism requires formation reaction, quenching, and relaxation reactions and their

constant rates to simulate CH and OH emissions. The main formation reactions for CH and

OH are as follows:

C2H + O ⇐⇒ CH ⋆ + CO

C2H + O2 ⇐⇒ CH ⋆ + CO2

CH + O2 ⇐⇒ CO + OH ⋆

(2.1)

(2.2)

(2.3)

Several reactions in the quenching and relaxing processes should be addressed and

26

considered. These mechanisms and sub-mechanisms will be addressed later in the Chemilu-

minescence chapter. Since radicals have a slow formation rate and a quick rate of elimination

and quenching, they have a low concentration in flames. The concentrations were measured

as a post-process to the flame modeling and were considered in a quasi-steady state with no

impact on the overall kinetics. The formation and removal quenching reaction rates can be

used to determine the concentration of species. The concentrations of CH(moles/cm3) and

OH can be calculated using the equations below:

[CH ⋆] =

k1[C2H][O] + k2[C2H][O2]
ΣjkQ,j[Mj] + Ac

[OH ⋆] =

k3[CH][O2]
ΣkQ,j[Mj] + Ac

(2.4)

(2.5)

Where K1 and K2 are the reactions constant rates for two formation reactions (R1

and R2), kQ,jis the quenching rate constant by the species j, and Ac is the Einstein coefficient

for spontaneous emission. K3 is the constant reaction rate for formation reaction R3. The

photon emission rate, i (mole photon,cms) is calculated by:

iCH ⋆ = Ac[CH ⋆]

iOH ⋆ = Ac[OH ⋆]

(2.6)

(2.7)

The profiles through the 1D simulation are integrated to calculate the full spatial

formation of each species, ICH (Photons, cms), where L is the integration path length across

the 1D domain:

(cid:90)

ICH ⋆ =

iCH ⋆dx

(2.8)

The total emission rate P (Photons/sec), is given by the flame area and can be derived

by:

27

PCH ⋆ = ICH ⋆ × Af = ICH ⋆

m
ρuSL

(2.9)

Chemiluminescence studies on the characteristics of spectral emission of kerosene

have been done by Lee [Lee and Seo, 2015]. Intermediate species of CH, OH, and C2 were

investigated for the flame of a swirl combustor. They expressed that the high luminosity of

kerosene is due to a high attribution of background CO2 chemiluminescence. Additionally,

OH and CH showed similar behavior as a function of equivalence ratio: as the equivalence

ratio increased, both CH and OH chemiluminescence intensities decreased. However, the

C2 chemiluminescence was affected by changes in the equivalence ratio. In rich combustion

conditions or with higher inlet temperature, the C2 intensity increased. Finally, they found

a correlation between the ratio of CH/OH and the equivalence ratio.

Chemiluminescence and simulation of CH/OH for liquid fuel require detailed reaction

mechanisms which depend on the fuels, temperature, and pressure. Due to the complexity

of liquid fuel simulation, it is common to use gaseous fuel simulation results to estimate the

CH/OH species in liquid fuel combustion.

Comparisons with experimental chemiluminescence measurements will determine the

optimal scale for model validation. Experimental measurements are required to determine

which sub-mechanisms provide correct predictions. Experiments validate modeling attempts

for scaling CH and OH species for varied operating circumstances.

2.2.2 Laser Induced Fluorescence

2.2.2.1 LIF Concept

Laser-Induced Fluorescence (LIF) is the optical emission from molecules or atoms

that have been excited by an external source of electromagnetic radiation. To understand

fluorescence thoroughly, it is crucial to study the energy states and electronic transitions in

molecules. Typically, molecules are in their ground state, the lowest energy state. When a

tunable laser illuminates the flow, it overlaps an allowed electronic, vibrational, or rotational

28

transition in an atom or molecule. The molecule undergoes electronic excitation upon

absorbing light, moving to a higher energy state. This state, known as an excited electronic

state, depends on the number of absorbed photons and the resulting energy level.

After excitation, the molecule may undergo vibrational relaxation within the excited

state, releasing excess energy through non-radiative processes. This process brings the

molecule to the lowest vibrational level of the excited state. The excited molecule then

undergoes a radiative transition, or fluorescence emission, returning to the ground state.

During this process, the excess energy acquired during excitation is emitted as a photon with

lower energy than the absorbed photon. This emission of photons from a higher energy state

is termed fluorescence.

To advance deeper into molecular energy transitions and their relation to fluorescence

emission, it is important to understand the various types of transitions in molecules, each

associated with changes in their energy levels. The three main types of transitions are

electronic, vibrational, and rotational. These transitions determine how molecules interact

with each other and influence their bond lengths.

Electronic transitions involve changes in the electronic energy levels of a molecule,

where electrons move between different electronic states, such as from the ground state (e.g.,

labeled X) to excited states (e.g., labeled A, B, C). The energy associated with electronic

transitions is often in the ultraviolet (UV) or visible range.

Vibrational transitions involve changes in the vibrational energy levels of a molecule.

As atoms within the molecule move relative to each other, the molecule vibrates. Transi-

tions between vibrational states are associated with changes in these vibrational motions.

The energy associated with vibrational transitions is generally lower than that of electronic

transitions and falls in the infrared (IR) range. Vibrational levels are usually denoted by ”v”.

Rotational transitions involve changes in the rotational energy levels of a molecule.

Molecules can rotate around their axes, and transitions between rotational states are associ-

ated with changes in rotational motion. The energy associated with rotational transitions is

29

lower than both electronic and vibrational transitions and typically falls in the microwave

range. Rotational levels are usually denoted by ”J”.

Transition selection rules are principles that govern which transitions are allowed or

forbidden in a molecule. These rules help predict and understand the spectra observed in

various spectroscopic techniques. At room temperature, typically only the lowest energy

vibrational state (v = 0) is populated, so v0 = 0 and ∆v = 1 are common transitions. States

with J ̸= 0 can be populated since they represent the fine structure of vibrational states

and have smaller energy differences than successive vibrational levels. Additionally, ∆J = 1

because a photon contains one quantum of angular momentum, adhering to the principle of

conservation of angular momentum.

Figure 2.2: Diagram of P,Q, and R branches transitions

For diatomic molecules, the transition ∆J = 0 (i.e., J” = 0 and J ′ = 0) with v0 = 0

and ∆v = +1 is forbidden, so pure vibrational transitions are not typically observed. The

rotational selection rule gives rise to an R-branch (when ∆J = +1) and a P-branch (when

∆J = −1). Each line of the branch is labeled R(J) or P(J), where J represents the value

of the lower state. The R-branch transition occurs when ∆J = +1, where the rotational

30

quantum number in the excited state is one more than in the ground state. The P-branch

transition occurs when ∆J = −1, where the rotational quantum number in the excited state

is one less than in the ground state. The Q-branch transition occurs when ∆J = 0, where the

rotational quantum number remains the same in both states. These branches are presented

in Figure 2.2.

2.2.2.2 LIF application in combustion

LIF is a unique and valuable method for measuring variables or visualization. It can

be implemented to measure various data such as density, concentration, velocity, temperature,

and pressure. It can be used at a single point, line, or even a sheet called PLIF. LIF is a

resonant technique requiring the laser to be tuned to a specific wavelength. Single-photon

LIF is linear, but it can have a low signal that requires ICCD to intensify the signal, but

on the other hand, it increases the noise level. LIF method differs for the liquid phase and

gaseous phase. For the liquid phase, fluorescent particles are seeded with the flow, and by

absorbing energy and releasing it, measurement of concentration and temperature. While

performing in the gaseous phase, the seeded particle is not proper since the species have a

unique excitation wavelength. The illuminating part is not just due to the particles. For gas,

a phase-specific wavelength for tuning laser is needed to excite a specific species, and that

excited species will release photons after relaxation.

In the combustion study, OH is excited in the UV near 282nm for (1,0) or 306nm for

(0,0). OH PLIF produces a high signal due to the abundance of OH, and it can be used as a

marker for product and turbulent transport. On the other hand, CH is excited in the UV/vis

near 390 nm for B-X1, 413 nm for A-X2, and 314nm for C-X. CH PLIF can identify the

flame front in premixed flame, and it can identify the stoichiometric surface in non-premixed

flame.

The study of turbulent flame topology has grown drastically recently, and quantifying

1Transition from second excited electronic state to ground electronic state
2Transition from first excited electronic state to ground electronic state

31

flame features is of great use to associate with direct numerical simulation (DNS). One of

the favored diagnostic techniques in combustion is CH/OH PLIF. Recent advancements in

CH PLIF using C-X band transition have been applied in prevaporized and spray flames.

Several features of flame structure can be measured with this method, such as flame front,

flame curvature, flame surface density, and consumption speed.

Wang [Wang et al., 2019] studied turbulent liquid spray flame supported by pilot

flame using OH LIF and CO PTLIF for the first time in shooting flames. A laser beam

focused at 230 nm wavelength at two-photon LIF was used to excite the CO, and for OH

PLIF, a laser beam at 282.75 nm wavelength was used for Q1 (5) rotational transition of

OH radical excitation. OH PLIF results indicate that the reaction zone is located near the

nozzle exit primarily, and it propagates uniformly by moving away from the burner since the

droplets vaporized thoroughly. Analyzed data of CO TPLIF shows that near nozzle exit, a

solid spectral line around 460 nm appears, which is due to C2 Swan bands.

Heat release rate or flame burning rate is one of the characteristics that cannot be

measured directly, and there are indirect measurements and correlations to predict it. HCO

is the direct method and tracker of heat release rate, but it is hard to capture because of

quenching and removal reactions. Since they have low concentrations in flames, it is hard to

detect them by single-pulse measurements. However, Paul [Paul and Najm, 1998] utilized a

new method for single-shot measurements to follow up on the heat release rate in flames. It

is suggested that OH and CH2O can be the species that can mark the heat release rate. They

applied OH PLIF at 283 nm and CH2O PLIF at 338.1 nm on a turbulent premixed dimethyl

ether air V-flame vortex interaction and compared the results to a simulation outcome. It was

observed that the OH surface was in contact with the CH2O surface through the experiments,

and the flame structure was hardly affected by the vortex interaction. It has been expressed

that OH and CH2O PLIF can single pulse measure the flame front and heat release rate.

Stabilization and flame structure in turbulent spray flame has been investigated by

Stepowski [Stepowski et al., 1994]. Flame of atomized methanol liquid mixed with flowing air

32

through a duct has been studied by OH LIF measurement, and velocity distribution has been

captured with the PDV3 method. The laser was pumped at the wavelength of 283.93 nm,

and detection occurred at 315 nm. Although an elastic scattering of laser has been neglected

by using a high pass filter, it cannot avoid big methanol droplets of Raman scattering around

309 nm which can be detected as saturated signal regions of white circles. Figure 2.3 shows

flame structure at different velocities of fuel air mixtures to measure the lift off location and

evaluate the effect of methanol droplets on the flame structure. Methanol droplets have

caused discrete regions in the LIF result. Also, they stated that by increasing flow velocity,

the lift off location increases which may affect the flame stabilization.

Figure 2.3: OH LIF in spray liquid flame for two different velocity and appearance of
discrete region in spray axis due to large droplet scattering [Stepowski et al., 1994]

CH radical is formed within the reaction zone; therefore, it is the flame front marker.

The extinction and penetration of pockets within reactants are possible by studying CH.

However, due to the low concentration of CH and interference of other species, fluorescence

signal detection of CH is low.

The simple approach to capturing CH radicals is to excite them via A-X and B-X

bands near 340 nm by higher energy. However, by increasing the energy of the pulse of

the beam, other species, such as PAH and OH, will excite as well. Since their absorbance

coefficients at CH excitation wavelengths are not zero, and the concentration of these species

3Photonic Doppler Velocimetry

33

is much higher than CH, they will interfere with the CH signal. On the other hand, liquid

droplets in spray or prevaporized cases can absorb the energy and fluorescence at a specific

wavelength. This is known as the Mie scattering effect, which interferes with the CH signal.

Carter [Carter et al., 2014, Carter et al., 2016] introduced a new approach in the CH

PLIF experiment to overcome this problem. Instead of exciting CH at the A-X (1-0) band

at a 430 nm wavelength, they excite CH in the range of 310-320 nm wavelengths via the

C-X (0-0) band, where CH has higher signal detection. There are several OH transition

wavelengths in this region, which can create some interference in the CH signal, as shown in

Figure 2.4.

Figure 2.4: Detailed spectrum with inset images for specific wavelengths [Carter et al.,
2016]

This interference can be avoided if only CH PLIF are desired. However, it poses the

possibility of capturing simultaneous CH/OH PLIF with one camera and a single wavelength,

as shown in Figure 2.5. The most prominent feature of this approach is that the excitation

can be done with a small amount of laser energy, which makes high-speed imaging possible.

Since the laser beam has low energy, the interference of droplet or PAH Mie scattering is

minimized and neglected. As can be seen, there are three different regions: first, the reactant,

which has minimal signal; second, the flame surface, which has the maximum signal; and

last, the product, which has intermediate values of signals.

34

Figure 2.5: Combined CH–OH PLIF image from premixed turbulent Bunsen flame [Carter
et al., 2016]

For liquid spray cases, the excitation wavelength differs compared to prevaporized

studies. This idea was tested before by Hammack [Hammack et al., 2018] in turbulent liquid

spray jet flame using Hi-Pilot burner. In this study, instead of exciting CH fluorescence at Q

branches around 314 nm, the R-branches were chosen to excite the CH radicals. The main

reason for this alteration is to avoid the interference of droplet Mie scattering by using a

custom Semrock (AFRL 0002) and UG-5 filters. Figure 2.6 shows the transition wavelength

spectroscopy in the R-branches region.

Figure 2.6: Excitation wavelength of CH in R-branches region [Carter et al., 2016]

35

The excitation occurs at 311 nm, where a UG5 filter around 311 nm is used to block

the Mie scattering interference of liquid fuel droplets and components. However, since the

filter does not completely block the fluorescence at 311 nm, two custom-made Semrock filters

around 312 nm are used to improve rejection of droplet fluorescence signals. Another goal

of this research is to compare CH PLIF and OH PLIF by studying flame surface structure

like flame surface density. Compared to OH PLIF, where the product and reactant sides are

oriented and distinguished, CH PLIF requires more detailed binarization and side detection

processing due to flame fragmentation and local extinctions. Observations of simultaneous

OH-CH PLIF in these flames often show layers of extinguished CH. At the same time,

localized OH signal remains, which suggests the imaging of OH alone may not capture in-

plane extinction and may lead to an over-prediction of flame surface density. In this study,

to simulate OH PLIF and evaluate the overprediction, the edges of extinction areas in CH

images are connected artificially. The flame surface is calculated separately for this pseudo-

OH PLIF. Hence, the flame surface density, conditioned on the local progress variable, is

calculated for all cases, comparing fueling effects and equivalence ratio. Lastly, to determine

potential differences in imaging methodologies between the CH PLIF and OH PLIF, the

flame surface density is measured for pseudo-OH PLIF and compared to the CH PLIF-derived

measurements. CH and OH PLIF have their benefits and limits. CH PLIF depicts flame

location [Smolke et al., 2018] and provides a two-sided flame structure, while OH PLIF is a

marker of heat release location, and the flame structure is one-sided. The stream’s orientation

is known when OH PLIF is used, and the flame is divided into reactant and product. However,

in CH PLIF, since the flame edges are captured, there is no distinguishable sign between

reactant and product. CH PLIF can detect breaks and extinction in the flame, while OH

PLIF is unable to detect it and produces a false flame edge. Overall, each method has its

application based on these advantages and disadvantages. To study flame structures like

flame surface density and curvature, OH PLIF is preferred. To research flame extinction and

blow-off cases, CH PLIF is suitable.

36

2.2.3 Diagnostic Procedure

To capture the turbulent prevaporized flame structure, a CH PLIF experiment was

conducted using the (v’=0, v”=0) band to excite CH species with a 10 Hz Nd: YAG laser

at 314.4 nm. The CH PLIF method used in this experiment was previously introduced by

Carter. The measurements were acquired at laser pulse energies of 1.3 mJ, which is in the

saturated regime for this method. A cylindrical and spherical lens with -50 mm and 1000 mm

focal lengths were employed to create the desired laser sheet. A PIMAX III intensified camera

was used with a UV-Cerco 100 mm lens to visualize and capture the CH signal. In order

to minimize the Rayleigh scattering, horizontal beam polarization is used. To obtain the

ideal output, the settings of gain and gate were set to 51 and 250ns, respectively. Lastly, the

observed field of view in this experiment was 35 x 35 mm, and the resolution was measured to

be 34 microns per pixel. In order to capture liquid spray flame structure, a CH PLIF similar

to prevaporized experiment was conducted. But instead of exciting the CH at CX Q-branch,

the R-branch excitation was implemented. The excitation is around 311nm wavelength and to

prevent the interference of Mie scattering fluorescence, two filters of UG5 and Semrock 0002

were installed in imaging equipment. The field of view in this experiment was determined

to be 50 50 2 . By applying a binning process, the resolution of 49 /pixel was achieved. For

CH and OH excitations, the amount of laser energy was recorded as 1-2 mJ/pulse and 3-4

mJ/pulse, respectively.

2.3

Image processing

CH radical, which is known as a flame surface marker and is formed within the flame

reaction zone, has been used to identify turbulent topological features such as the formation

of pockets and holes [Rosell et al., 2017, Li et al., 2010]. Tracing the CH radical can lead to

identifying local extinction, and it is not transported from the reaction zone [Smolke et al.,

2018]. CH and OH PLIF images have been used to measure turbulent flame features such as

curvature, surface density, and consumption speed [Driscoll, 2008, Zhang et al., 2020]. But

37

when using only CH PLIF, the information of product zones extracted from the OH PLIF is

missing. So, problems such as curvature sign, flame breaking, or extinction, etc., appear in

these cases where only CH PLIF is provided.

Figure 2.7: Flow chart of image processing in this method

As shown in Figure 2.7, the flow chart of the presented method is briefly explained.

First, CH PLIF images (Figure 2.8a) are processed through background subtraction and

noise elimination to start the logic behind this method. A 3x3 pixel square median filter is

also applied to reduce the noise. The images are binarized using Otsu’s method [Otsu, 1979].

Using Otsu’s method, the CH PLIF image becomes a matrix of 0 and 1, where the number

1 indicates that a pixel is within the reaction zone and number 0 means that the pixel is

either in the product or reactant region (Figure 2.8b). Then, by applying a gradient to the

binarized images, it can identify the inner and outer surfaces of the flame. Using the gradient

over the binarized image, the outcome shows either 0, +1, or -1 1/pixel.

38

(a)

(b)

Figure 2.8:
(a) Original raw and (b) binarized images for turbulent Re = 10,000, at
equivalence ratio ϕ = 1.2 of n-Heptane flame at 35 mm height above the burner with 60 m/s
pilot flame

Now, by setting a reference point at the centerline of the image close to the burner

exit, which indicates the reactant region, and moving through the gradient of the binarized

image, it can identify all of the regions, such as reactants, products, and flame. For example,

if the start point is in the reactant region where the value of the binarized signal is zero and

the direction of movement is known, as long as the gradient value stays zero, it means the

region is still reactant. And by seeing a difference in gradient, the reaction zone is identified.

All image processing and coding were developed and done in MATLAB R2018a and its image

processing toolbox.

The next step is to assign reaction zones, or what is used to call from now on, families.

For this purpose, it has to detect the edges of the flame and then define families. The edges

of the flame are easily known by searching through the binarized image. After that, to assign

families, it first sets those points and searches for their neighbors with a value of 1 (since the

image is binarized). If none of the neighbors have the value of 1, then this family is done,

39

and it looks for the next points and assigns another family. By repeating this procedure,

it divides the whole flame into various reaction zones and applies the image processing to

these zones or families. However, to eliminate noise and error, a threshold value for several

points to consider them a family is set. The threshold value is arbitrary as long as it can be

assumed that the number of points forms a reaction zone. Based on our observation through

various images, the value of 40 points is considered the threshold value. Figure 2.9 shows a

sample of families detected using this method.

Figure 2.9: Families identified for turbulent Re = 10,000, at equivalence ratio ϕ = 1.2 of
n-Heptane flame at 35 mm height above the burner with 60 m/s pilot flame

The main problem in the current image processing and curvature calculation is the

correct sign of the curvature since the information regarding the orientation relative to the

reactant or product stream, which determines the sign of the curvature, is not known. As

shown in Figure 2.10, normal vectors to the family’s surface are used as a guideline for

knowing the direction of movement in the flame surface, and in further steps it is used to

find pockets of reactant and products in the reaction zones. Also, this is a good approach

to avoid any miscalculation and it is a starting point to identify the sign of curvature, and

this technique is similar to other research in this area [Haq et al., 2002, Kim and Pitsch,

40

2007, Falkenstein et al., 2020]. To make sure that all of the vectors are towards the outside

of the surface, the method uses the normal vector to make a surface parallel to our original

surface. If the values of the binarized image in that new surface were zero, it means it is

outside of the flame. Vice versa, if the value of the new surface is 1, it means it is inside of

the flame surface and it has to correct the normal vector.

Figure 2.10: Normal vectors defined for a small segment of the binarized image for turbulent
Re = 10,000, at equivalence ratio ϕ = 1.2 of n-Heptane flame at 35 mm height above the
burner with 60 m/s pilot flame

To identify holes or product/reactant pockets and reaction zones, based on the ob-

servation, using the normal vectors seems to be a good idea. First, it’s necessary to locate

the center of gravity of the family by taking the mean values of the family’s points x and

y coordinates. Then, using that center of gravity as a reference point, calculate the angle

between two vectors. The first vector is the line going from the reference point to a point on

the family surface, and the second vector is the normal vector at that point. To measure the

angle between two vectors for each point, the following equation can be used:

θi = cos−1

(cid:19)

(cid:18) rg,i · ηi
|rg,i||ηi|

41

(2.10)

Where ηi is the normal vector on the ithe point of the family and rg,i is the vector

from the gravity point.

If the average of these angles for a family member is between 90 and 270, then that

family has a product/reactant pocket or hole in itself; otherwise, that family is a reaction zone

of the flame. For now, the holes are filled and excluded from our calculation, and curvature

is not calculated on the hole surface. This method lacks in detecting whether the hole or

pocket is reactant or product, and so the sign of curvature is not determined on these regions

based on it.

The next step is to find the edges of each family and use them as initial points for

each one. These edges are used as initial points to start the calculation of curvature and

move clockwise with the help of normal vectors until reaching the other edge point or the last

one, then start over from the last point to the initial point. To find the edges of the families,

the skeleton of the families is extracted using the longest distance between the edges and

branches to eliminate small and unnecessary ones. Another alternative idea is to measure

the curvature on the skeleton branches; however, it may not be representative of the whole

flame.

The final step is to derive the curvature value from the flame surfaces and edges. The

following equation is used to measure the curvature, which is an improvised version to have

better results. In this equation, the s parameter is the length of the flame alongside the

initial point that has been chosen.

(cid:115)(cid:18) d2x
ds2

(cid:19)2

(cid:19)2

+

(cid:18) d2y
ds2

H =

(2.11)

It’s important to mention that this approach is specifically devised for the prevaporized

liquid fuel flame structure near the base of the burner, where there is no sign of fragmentation

or global extinction. However, the behavior of the flame changes significantly at higher axial

locations. In the middle section of the flame, fragmentation occurs due to increased turbulence

and interaction with surrounding air, requiring a different algorithm to accurately model and

42

address these phenomena. Similarly, at the tip of the flame, global extinction happens because

of the limited fuel supply and the mixing dynamics, necessitating yet another specialized

algorithm to handle these challenges. The next section delves into these additional algorithms,

providing detailed solutions for managing the distinct behaviors observed in various parts of

the flame.

Since the flame edge can be affected by high-frequency digital imaging, some sort

of smoothing is required [Haq et al., 2002, Chen, 2007]. The Savitzky-Golay [Savitzky and

Golay, 1964] algorithm, which is a low-pass filter, is used on the flame edges to smooth them

and improve the procedure of calculation. This equation is a positive analytical value which

needs to have a sign due to being convex or concave toward the reactant or products. Positive

curvature flames are convex toward the reactant, or they are concave toward the product.

These flame surfaces are curved outward, toward the reactant. The negative curvature flames

are concave toward the reactant, or convex toward the product. These flame surfaces are

curved inward, toward the products. The method sets the initial points on the outer surface

of each family, and also every region’s identity is known. So, a correction factor is assigned

to the final results to identify the sign of positive or negative of the curvature.

It should be noted that, to have improvements in our calculations, an interpolation

is applied to the points of the families. The rate factor of interpolation is 2, which means

the number of points is doubled and between each two points, a new point is introduced and

gets the average value of the two points. Also, a wavelet noise eliminator was applied to the

calculation to smooth the flame edges even more and decrease the effects of digital imaging.

With a greater number of points and smoother paths, higher accuracy and lower error in the

calculation of curvature is achieved.

According to the study, the highest detectable curvature is 5000 1/m, which corre-

sponds to a circle with a radius of 0.2 mm. The CH layer thickness limit predicted by the

CHEMKIN simulation is likewise approximately 200 microns [McManus et al., 2021].

43

2.3.1 Spacial Case Algorithms

Here I try to categorize different flame structures and explain why each one of them

requires a special algorithm to overcome the complexities and difficulties of their nature.

Based on the field of view and how turbulent flame is captured, I can divide them into three

different categories. First, the base of the burner which has been explained in the previous

section. The second part is the middle section where the initial fragmentation happens and

the flame divides into two parts. And lastly, the tip of the flame where the flame begins to

break into small fragments and global extinction occurs. I try to come up with a unique

algorithm for each case. Also, a separate individual algorithm is provided for spray liquid

flame structures.

2.3.1.1 The Middle Section of The Flame

This section addresses the part of the flame where the first major fragmentation occurs,

causing the main jet flame to break due to local extinction when it comes into contact with

the surrounding air. To resolve the flame edge detection for this section, it is crucial to locate

the breaking or fragmentation point. Most of the steps are similar to the main algorithm up

to the point of connecting the reaction layers’ endpoints. Our goal is to identify the breaking

point and categorize the flame patches or reaction layers into two groups: one before and

one after the breaking point. I then apply the previous algorithm, used for the base of the

burner, to these two sections. Based on our observations, almost all fragmentation starts in

the upper half of the flame, at a higher axial location compared to the center point of all

patches. Additionally, most of these fragmentations can be identified by the angles they form

with the center point. By restricting the patches to those whose endpoints form an angle

between 60 degrees and 120 degrees, I can isolate the reaction layers that are at a higher

axial location than the breaking point.

Next, I sort all the patches below the breaking point and designate them as Group A.

I then identify the point with the maximum axial coordinate within Group A and use this

44

maximum point to determine which of the remaining patches belong to Group A.

For the other patches that are not yet sorted, I compare their endpoints to the

maximum point individually. If the distances from both endpoints to the maximum point

are greater than 50 pixels, these patches are labeled as Group B, which are considered the

flame patches above the breaking point. Otherwise, they are included in Group A. This

iterative process continues until all patches are designated as either Group A or Group B.

Figure 2.11: Rprevaporized n-dodecane at the middle section with a Reynolds number of
10000

Figure 2.11 displays a sample image of prevaporized n-dodecane in the middle section

with a Reynolds number of 10,000. In this image, the reaction zones above the breaking point

have been identified and isolated. Figure 2.12 illustrates the differences between application

of the original base of the burner algorithm and the current algorithm to the images from the

middle section. This approach prevents the omission of flame edges detected by the previous

algorithm and ensures accurate assignment of regions to products or reactants.

45

(a)

(b)

Figure 2.12: Product and reactant regions detected by (a) special algorithm for the middle
section of the flame, (b)base of the burner algorithm

2.3.1.2 The Tip of The Flame Case

The final part of the flame that requires a specific algorithm due to its complex

structure is the tip of the flame, where multiple small fragmentations are present. This field

of view usually consists of two parts: the remaining portion of the middle section flame and

multiple isolated small patches. To organize them, I must identify and exclude all small

reaction layers that are no longer part of the main flame. By excluding these parts, only

the main flame patches remain, which can then be processed by the previous algorithm to

effectively connect the edges.

To determine if a patch is considered isolated, the distances of its endpoints from the

center point and other patches are measured. Generally, a patch is considered isolated if

both endpoints are far away (around 30 to 50 pixels) from other patches or if the nearest

patches to its endpoints are the same. After identifying all isolated patches, I recheck them

to see if any two patches are paired together. This involves examining the distances between

their endpoints and those of other patches. If the minimum distances of both endpoints from

46

other endpoints are larger than 50 pixels, they are labeled as solo isolated patches. If two

or more patches are close to each other within their neighborhood, they will connect to each

other by artificial edges, as they are part of a larger flame fragmentation that has broken into

smaller patches. Figure 2.13 shows a sample processed image of the flame’s tip of methane

at a Reynolds number of 10,000.

Figure 2.13: Processed image of methane flame at the tip section with a Reynolds number
of 10000

For liquid spray with prevaporized liquid fuel flames, similar algorithm to the flame’s

tip is used. This is due to the presence of various small flame fragmentations around the

main jet flame. The algorithm exclude isolated patches based on their size and their distance

from other patches. After excluding the isolated patches, the main prevaporized base of the

flame algorithm is applied to detect the flame edges. However, the criteria to identify isolated

patches in the spray case may differ slightly. Since the small patches are much closer to the

main flame, a smaller threshold distance is required compared to the prevaporized case.

47

2.3.2 Flame Surface Density Procedure

The reactant and product sides should be identified following binarizing the images

locally utilizing a moving box. It is impossible to distinguish between distinct flame areas

due to extinction. The method solves the problem by artificially connecting the ends of

each reaction zone. The binarization method applies again to fill the reaction zone. The

binarization procedure was utilized to derive the progress value contour of 500 photos. After

that, a map of progress variables between 0 and 1 is created by combining them and taking

an average. The density of the flame surface is computed using the following equation (Eq.

2):

In this equation, Af and Lf are time average surface area and length of flame in

the control volume of ∆x2. For calculating the flame surface density, the flame front edge

is considered. Damkohler suggested that turbulence affects flamelet regimes to wrinkle

and disturb the reaction layer. This idea implies that the laminar local properties remain

constant. Furthermore, following his assumption, the ratio of turbulent to laminar burning

velocity ST /SL equals the ratio of turbulent to the laminar burning area (AT /AL). AT and

AL represent a wrinkled turbulent and laminar flame’s burning area, respectively. Discroll

[Driscoll, 2008] showed that the ratio of turbulent to the laminar area could be calculated

with the following equation:

ST /SL = AT /AL =

(cid:90)

Σdx

(2.12)

In this equation, η is the normal axis to < c >= 0.5 from the binarized and averaged

CH PLIF. The flame surface density is explained and calculated in this section.

2.3.3 OH Pseudo method

As shown in Figure 2.14, the OH PLIF pictures are straightforward, dividing the

reactant and product areas. It cannot, however, detect extinction events or flame edge breaks.

Figure 2.15 depicts the reaction zone’s outer surface (product side) for the CH and pseudo-

48

OH techniques. Because only the edge of the reaction zone is visible with CH PLIF (Figure

2.15a), the extinctions and breaks may hinder the distinction between the reactant and

product sides. The locations on the flame breaks are artificially linked to imitate OH PLIF

compared with the CH PLIF technique. The OH PLIF, which differentiates the reactant

and product parts, is similarly represented by this artificially linked surface, which I call

OH-Pseudo. The extinction endpoints on the CH PLIF are depicted in Figure 2.15b within

a red rectangular box containing the fabricated lines. The flame surface density is estimated

using two distinct approaches to determine the potential differences.

Figure 2.14: OH PLIF image sample of prevaporized JP8 at equivalence ratio ϕ = 0.9

The flame surface area per unit is assumed to be the same as the flame surface

perimeter. An 11 by 11 interrogation box is used to quantify the flame edge in each pixel

for the whole image. In the inquiry box, the length of the flame for 500 CH PLIF photos is

estimated by counting the number of pixels representing the flame front. The flame surface

density in the center of the interrogation box is calculated by dividing the entire length by

the area of the interrogation box. The average flame surface density is measured using a

2 threshold for each progress variable. Then, variation of measured Σ based on progress

49

variable c is presented.

(a)

(b)

Figure 2.15: Product and reactant regions detected by (a) special algorithm for the middle
section of the flame, (b)base of the burner algorithm

2.4 Apparatus

This section outlines the experimental setup and procedures used to capture flame

statistics for prevaporized liquid fuels, spray liquid fuels, and chemiluminescence experiments.

The study of prevaporized liquid fuel flames was conducted using a McKenna burner, while

a HiPilot burner was utilized for spray liquid fuel flames. The details of each experiment are

explained in the following sections.

2.4.1 McKenna Burner

The experimental setup used in this study is schematically depicted in Figure 2.16a.

The setup is based on a McKenna burner with a central tube diameter of 7.6 mm and

a flat flame plate with a 60 mm diameter utilized as a pilot. A methane-air mixture at

stoichiometric conditions was used for the pilot burner with a mean flow velocity of 60 cm/s.

50

Figure 2.16: (a) Piloted McKenna burner and imaging setup. (b) Prevaporizer schematic

In this experiment, three different fuels, n-Dodecane, n-Heptane, and methane, were

operated at a jet Reynolds number of 10,000.

In order to study the liquid fuels in the

premixed regime, a temperature controllable pre-vaporizer system, shown in Figure 2.16b,

was implemented. Liquid fuels pass through a hot water bath to be preheated before mixing

with air. Inline heaters were used to increase the air temperature to the desired point. Then

the heated air and heated fuels enter the insulated mixing chamber for mixing. With control

over water bath temperature and inlet heater, the degree of prevaporization can be tuned.

51

Table 2.1: Operating conditions for prevaporized, premixed jet flames

Re

Fuel

ϕ

Exit Velocity (m/s)

Laminar Flame

Speed (cm/s)

10000

20000

Methane

0.8, 1, 1.2

n-Heptane

0.8, 1, 1.2

JP8

0.8, 1, 1.2

n-Dodecane

0.8, 1, 1.2

Methane

0.8, 1, 1.2

n-Heptane

0.8, 1, 1.2

JP8

0.8, 1, 1.2

n-Dodecane

0.8, 1, 1.2

46

42

43

42

92

84

86

84

75.5

85.7

68.6

77.2

75.5

85.7

68.6

77.2

Turbulent, prevaporized premixed combustion of different fuels are presented in Table

2.1. A mixture of methane-air with a 60 cm/s flow rate runs through the pilot flow for all

the cases. Equivalence ratios of 0.8, 1, and 1.2 are presented for this work. The preheating

temperature is set to 470 K such that all liquid fuels are vaporized properly.

2.4.2 HI-Pilot Burner

For spray liquid flame, the schematic of the experiment setup can be seen in Figure

2.17. The setup is based on the Hi-Pilot burner for both prevaporized and spray liquid fuels

up to 99,000 Reynolds numbers. The burner consists of a nebulizer prevaporizing system to

produce a fully vaporized and premixed mixture of fuel and air. Additionally, a 260-micron

diameter needle is implemented in the burner to inject liquid fuel into the mixture. The

jet Bunsen burner is surrounded by a flat plate pilot burner to provide extra heating and

stabilization in the main jet flame.

52

Figure 2.17: Hi-Pilot burner schematic

The conditions and fuel selection in the spray liquid experiment were based on two

different categories. The first category is based on the turbulent intensity of the flow. Three

different flows with Reynolds numbers of 10K, 25K, and 50K were tested while the turbulence

intensity increased for each case. The second category in the experiment condition is fuel

preparation. This includes premixed methane alongside liquid spray fuels, pure premixed

prevaporized liquid fuels, and lastly, prevaporized liquid fuel alongside spray liquid fuels.

These two categories, combined together, provided wide and diverse options toward the

investigation of turbulent heavy hydrocarbon liquid fuel flames.

2.4.3 Chemiluminescence Experiment Setup

For the chemiluminescence experiment, the following setup as shown in Figure 2.18

has been used to capture CH and OH signals. Prevaporized methane, n-Heptane, n-Dodecane,

and JP8 at 470K temperature and Re=10K run through the piloted McKenna jet burner.

A specific CH filter centered at 430 nm wavelength with a ±10 nm margin was utilized to

capture the CH chemiluminescence signal. The field of view captures the entire flame length.

The spectroscopy of the flame has been done via Ocean Optics Flame model with 0.3 nm

resolution and 600-micron diameter fiber. The exposure time varied from 1 to 10 seconds,

and the chemiluminescence signal is the average signal of 3 similar measurements.

53

Figure 2.18: Schematic setup of chemiluminescence experiment

54

Chapter 3

Prevaporized Liquid Fuels

3.1

Introduction

In this chapter, several pre vaporized liquid fuels are tested to analyze their turbulent

flame structures. The flame structures and statistics of these fuels are measured and compared

to those of pure methane flame at two different turbulence intensities. The aim is to find

a relationship between CH fluorescence signals in regions with high curvature values and

interpret any influence of the Lewis number, as suggested by the DNS results. Additionally,

this chapter compares the use of CH PLIF with the commonly used OH PLIF to determine

which method better captures flame details for a more comprehensive understanding of

turbulent combustion.

3.2

Imaging

Study of CH PLIF using the (v’=0, v”=0) band enables capturing the turbulent

flame structures induced by the 10 Hz Nd:YAG laser at 314.4 nm wavelength. By emitting

1.3 mJ energy per pulse of the laser, this method is designed to measure CH PLIF in the

saturated regime of the laser. It was initially developed and presented by Carter [Carter

et al., 2014, Carter et al., 2016]. An optical system consisting of a cylindrical lens of 50 mm

55

and a spherical lens of 1000 mm produces the desired laser sheet. In order to visualize the

results, a PIMAX III intensified camera is used together with a UV-Cerco 100-mm focal

length lens. Polarization of horizontal beams minimizes the Rayleigh scattering effect. In

order to provide optimal visual impressions, the gate factor and gain are set to 51% and

250 ns, respectively. During these experiments, the field of view was 35 x 35 mm, and the

resolution was measured to be 34 microns per pixel [Gandomkar et al., 2021b].

3.3 Fuel Differences

In this study, CH PLIF was conducted for combustion of premixed, pre-vaporized

liquid fuels at two different turbulence intensities of Re=10,000 and 20,000, and different

equivalence ratios of ϕ=0.8, 1, and 1.2. Liquid fuels of n-Heptane, JP8, and n-Dodecane

have been selected and prevaporized for these experiments to compare with gaseous methane

flame. After imaging and processing them, which were explained in the previous chapter,

curvature is measured by using a novel method to identify reactant/product sides of the flame

edges and assign the correct sign of the curvature [Gandomkar et al., 2021a]. The following

sections express the turbulent characteristics of flame structures.

3.4 Flame Statistics

Let’s start this section with the curvature Probability Density Function (PDF) of

liquid fuels and methane flames at different Reynolds numbers and equivalence ratios. Figure

3.1a expresses the curvature PDF of liquid fuels and methane at two different Reynolds

numbers at equivalence ratio ϕ = 1.2, 5 mm above the burner (the base of the burner). It is

observed that for both turbulence intensities, liquid fuels have a wider curvature range than

methane cases. The PDF of low curvature values for methane is larger, which indicates the

flame structure for liquid fuels is more wrinkled compared to methane. Although the negative

curvature patterns for all liquid fuels are similar to each other and have a significant difference

with methane, the positive curvature is totally different. Based on the experimental and

56

DNS observations, it can be assumed that the flame extinctions and breaks occur around the

regions with positive curvatures. It can be interpreted that liquid fuel flames contain more

breaks than methane since the gap between the extreme negative and positive curvature

PDFs is higher. Also, it seems that with higher turbulence intensity the flame structure gets

more wrinkled and distribution of extreme and higher curvatures increases. This behavior

has been observed for all fuels, even methane, where distribution of smooth structures with

small curvature decreases with higher intensity of turbulence.

(a)

(b)

Figure 3.1: Curvature PDF of (a) all fuels with equivalence ratio ϕ = 1.2, (b)dodecane with
different equivalence ratios, 5 mm above the burner with 60 cm/s pilot flame at turbulent
Re = 10000 and 20000

Figure 3.1b depicts the curvature PDF of n-Dodecane at Re = 10, 000 and 20, 000,

5mm above the burner for different equivalence ratios of ϕ. It presents that the flame PDF

profile gets wider by decreasing the equivalence ratio. This behavior could be due to flame

extinction or more importantly, low signal detection. As the equivalence ratio decreases,

the amount of CH fluorescence decreases as well. This causes ambiguity in detecting flame

extinctions, which could lead to more extinction detection in lower equivalence ratios. It can

be observed that lower equivalence ratio and lower CH fluorescence signal conclude more

57

wrinkling distribution. Although it should be mentioned that this distribution is not only

due to the extinctions of flame, but not capturing flame edges due to lower signal of CH.

The statistics of curvature PDFs are presented below in Table 3.1. The variance of

PDFs indicates that a wide range of curvature is occupied by liquid fuels rather than methane.

It may suggest that the flame structure for liquid fuels is more wrinkled and complicated,

and higher curvature values are achieved.

It is a similar trend among all fuels that the

mean of curvature decreases by increasing the equivalence ratio. It can be interpreted that

a lower equivalence ratio leads to lower signal detection, which can cause extinction and

breaks in flame structure. The skewness of the PDF distribution for liquid fuels decreases by

increasing the equivalence ratio, while for methane it increases. This behavior expresses that

for liquid fuels, the distribution of curvature PDF merges toward a symmetric distribution.

In other words, the probability of positive curvature increases by increasing the equivalence

ratio. All fuels show positive skewness for all equivalence ratios. A positive skewness means

the distribution is right skewed and the mean of curvature is higher than the median value.

While the flame goes from a lean condition to a rich one, the mean increases. It seems the

rich liquid fuel flame has a higher probability of positive curvature compared to lean flames.

Increasing the equivalence ratio increases the kurtosis of curvature PDF for all fuels. It means

by increasing the equivalence ratio, the distribution diverges from normal distribution and

its tail gets thicker. So the probability of higher curvature values increases.

58

Table 3.1: Statistical measurement of curvature PDFs

ϕ Mean (1/m) Variance∗(10e6) Skewness Kurtosis

0.8

1

1.2

0.8

1

1.2

0.8

1

1.2

0.8

1

1.2

Methane

Heptane

JP8

Dodecane

-47.1

-42.5

-29.3

-194.4

-230.4

-183.0

-303.9

-256.2

-156.6

-295.3

-266.1

-271.7

2.9

2.2

1.9

3.6

2.9

2.7

3.1

2.8

2.6

3.6

3.0

2.7

0.0751

0.1022

0.1227

0.1052

0.0805

0.0697

0.1171

0.0842

0.0733

0.1393

0.1043

0.0758

3.7

4.5

4.9

3.1

3.6

3.8

3.5

3.7

4.0

3.2

3.6

3.8

I showed the effect of turbulence intensity briefly before, but let’s analyze its effect

in more detail in Figure 3.2. Figure 3.2a shows negative curvature percentage of total

distribution at different Reynolds numbers for all liquid fuels and methane. It appears for all

fuels, the distribution of negative curvatures decreases at higher Reynolds numbers. Moreover,

Figure 3.2b depicts the width of PDF corresponding to 10% of the maximum value for all

fuels. The PDF distribution appears to broaden for all fuels by increasing the intensity level.

59

(a)

(b)

Figure 3.2: Turbulence intensity effect on flame statistics. a) Probability density functions
of negative curvatures. b) width of 10Max Probability density functions , of all fuels at
equivalence ratio ϕ = 1.2 and two different Re numbers, 5 mm above the burner with 60
cm/s pilot flame

Additionally, frequency of local extinctions or breaks within the flame and their

average length have been calculated and presented in Figure 3.3. Figure 3.3a displays the

average quantities of extinction frequency for all fuels at different Reynolds numbers. It

can be concluded that for most fuels the probability of local extinction increases at higher

levels of turbulence. Number of breaks and extinctions for liquid fuels are significantly higher

compared to methane regardless of Reynolds number, which suggests that the local extinction

is not solely dependent on intensity level and fuel’s properties may have an important role in

it. Also, Figure 3.3b expresses the average length of extinctions at different Reynolds numbers

for all fuels. As expected, the length of breaks is higher for liquid fuel flames compared to

methane flame, and average length of break enlarges at higher turbulence levels.

60

(a)

(b)

Figure 3.3: Turbulence intensity effect on flame statistics. (a)Local extinction frequency
(b)Averaged size of extinctions, for all fuels at equivalence ratio ϕ = 1.2 and two different Re
numbers, 5 mm above the burner with 60 cm/s pilot flame

The next feature of flame topology, flame surface density, is determined and presented

for all cases in Figure 3.4. Figure 3.4a shows the mean of calculated flame surface density

(Σ) for each progress variable (c) for all fuels with CH PLIF and OH Pseudo methods at

equivalence ratios of 1.2. The sizing step of the progress variable is 0.1, including a threshold

of 2% to extract the flame surface density. The OH Pseudo method shows a higher flame

surface density than CH PLIF for all cases. The reason is that the OH Pseudo method

connects all edges of the flame, so the extinction or flame breaks are discarded. In this case,

liquid fuels all have higher FSD1 than methane, which suggests the flame length distribution

increases compared to the flame area. However, in CH PLIF, methane flame presents the

highest FSD among the fuels. The reason is mainly the flame extinction that occurs in

liquid fuels. The difference in CH FSD between CH PLIF and OH Pseudo methods suggests

that the extinction rate in methane seems to be the lowest compared to liquid fuels, and

n-Dodecane shows the highest extinction rate due to the difference of CH PLIF and Pseudo

1Flame Surface Density

61

OH PLIF.

(a)

(b)

Figure 3.4: Flame surface density of CH PLIF & Pseudo OH PLIF for turbulent flame
Re=10,000 for (a) all fuels at equivalence ratio ϕ = 1.2, (b)dodecane at different equivalence
ratios, 5 mm above the burner with 60 cm/s pilot flame

The effect of equivalence ratio on flame surface density is shown in Figure 3.4b where

FSD of n-Dodecane for two CH and OH Pseudo methods has been expressed at various

equivalence ratios. The flame surface density of n-dodecane increases by increasing the

equivalence ratio, which means the flame distribution area decreases compared to flame

length. The differences between CH and OH Pseudo methods increase significantly for lower

equivalence ratios. This decrease is not just due to the extinction rate, where a lower CH

signal due to the lean condition of combustion intensifies the effect of extinction.

The next turbulent feature which can be studied based on the CH PLIF is the total

actual burning area relative to the exit velocity of the burner. Figure 3.5 shows the averaged

ratio of length to area of the flame for each fuel at different exit velocities of fuel with a

specific equivalence ratio. The laminar flame speed normalizes the exit velocity of the fuel

regarding equivalence ratio.

62

Figure 3.5: Average ratio of length to area of turbulent flame at Re=10,000 and all fuels
flame 5 mm above the burner with 60 cm/s pilot flame

The ratio of turbulent speed to laminar flame speed has been depicted in Figure 3.6a

for fuels at equivalence ratio ϕ = 1.2 across the axial location. The liquid fuels’ turbulent

velocities increase rapidly along the axial location in CH PLIF and OH Pseudo methods.

At the same time, methane seems to have constant or slow progress across the flame height.

The equivalence ratio’s effect on the turbulent flame speed of n-Dodecane has been displayed

in Figure 3.6b where turbulent velocity increases alongside the axial location for the rich

flame. At the same time, it remains constant halfway through stoichiometric conditions and

decreases for the lean flame.

63

(a)

(b)

Figure 3.6: Turbulent velocity of turbulent flame Re=10,000 for (a) all fuels at equivalence
ratio ϕ = 1.2, (b)dodecane at different equivalence ratios, 5 mm above the burner with 60
cm/s pilot flame

3.5 Lewis Number Effect

The algorithm captures turbulent flame structure using a technique to distinguish

reactant and product sides of the flame and accounts for curvature sign correction. Also,

this method enables the investigation of non-unity Lewis number (Le ̸= 1) effects on flame

structure and curvature distribution in prevaporized, premixed turbulent liquid fuel jet

flames. The Figure 3.7 presents curvature PDF of n-dodecane, n-heptane, JP8, and methane

at Re = 10, 000 for 5 mm above the burner using CH PLIF measurement. It is observed that

liquid fuels have a wider curvature range than methane cases. The PDF of low curvature

values for methane is larger, which indicates the flame structure for liquid fuels is more

wrinkled than methane. Although the negative curvature patterns for all fuels are similar,

the positive curvature is different. The flame extinction and break occur around the regions

with positive curvatures. This is may be due to the following reasons:

1- The flame surface of positive curvature increases toward the reactant which is cooler and

64

may lead to the losing heat and cooling flame temperature and quenching flame

2- High positive curvature zones experience dilution of heat and reactive species because the

flame is forced to spread over a larger area, which weakens the concentration of fuel and

oxidizer in the reaction zone, reducing the chemical reaction rates

3- When a flame front is convex (positive curvature) toward the reactants, it experiences

higher strain rates due to the turbulent flow and increased flame stretching. The flame speed

must increase to keep up with the expanding flame surface. This stretching effect can thin

the reaction zone, making it harder for the flame to maintain the high temperatures and

reaction rates.

It can be interpreted that liquid fuel flames contain more breaks than methane since

their positive curvature PDF is higher than methane.

(a)

(b)

Figure 3.7: Signal-curvature correlation of turbulent flame Re=10,000 for (a) all fuels at
equivalence ratio ϕ = 1.2, (b)dodecane at different equivalence ratios, 5 mm above the burner
with 60 cm/s pilot flame

Figure 3.7a depicts the relation between CH signal and curvature for n-Dodecane, JP8,

n-Heptane, and methane at the equivalence ratio of 1.2. This result shows the effect of the

Le number on curvature/signal behavior. As discussed earlier, the DNS result relates to heat

65

release and high curvature regions. Also, experimental observation expresses a relationship

between CH signal and high curvature area. This behavior is more decisive for fuels with high

Le numbers like n-Dodecane. For liquid fuels with a higher Le number, the signal increases

as the curvature decreases, while methane does not follow a similar trend.

The positive and negative curvatures seem to have a similar range of signal values

for methane. Based on Figure 3.7b, for n-Dodecane, it appears that the negative curvatures

have a higher signal compared to positive curvatures. The average signal value reported for

high negative curvature (-5000) is higher than the positive curvature (5000). A linear fitting

line is represented for each curvature/signal to quantify the correlation while R-2 square

coefficients are measured.

CHEMKIN simulations of flame chemiluminescence in experimental conditions are

conducted to estimate the CH concentration’s behavior regarding the turbulence effect and

equivalence ratio. This study aims to determine the effect of strain rate on CH concentration

and provide evidence of the Le number effects shown in the previous section. Two counter-

flow flame modules utilized with the stoichiometric methane-air mixture and preferred fuel

with a specific equivalence ratio measures the CH concentration in different strain rates. For

these calculations, GRI Mech 3, HyChem, and JetSurf 2.0 have been utilized for methane,

JP8, and n-Dodecane, respectively. The total value of CH concentration and heat release are

calculated and presented below in Figure 3.8.

66

(a)

(b)

Figure 3.8: Measured quantities of (a) Total CH concentration, and (b)Total Heat Release
over strain rate for heavy liquid fuels and methane at different equivalence ratios

As shown in Figure 3.8a, by increasing strain rate, for most fuels, the CH concentration

decreases until global extinction occurs. However, for the lean methane case, a peculiar

behavior is observed. It seems the CH production accumulated over the course of strain rate.

Additionally, Figure 3.8b shows the effect of strain rate on the amount of heat release

for all fuels and equivalence ratios. It appears the value of heat release decreases by increasing

the strain rate. This behavior is expected, as increasing strain rate prolongs the chemical

time scale and shortens the dynamic time scale. These alterations lead to lower reaction

rates and CH formation overall in higher Reynolds numbers.

67

Chapter 4

Spray Liquid Fuels

4.1

Introduction

In this chapter, various combinations of spray liquid fuels alongside prevaporized liquid

fuels are used to analyze their turbulent flame structures. The flame structures and statistics

of these fuels are measured at different turbulence intensities. The goal is to investigate

droplet interactions with the flame, fragmentation, vaporization, and flame stabilization by

analyzing the turbulent flame statistics of liquid spray fuels. The measurement procedures

and calculations are similar to those used in the prevaporized case. Most of the spray liquid

experiments are conducted using OH PLIF, which is not the primary focus of this research.

However, a few selected cases are measured with CH PLIF for detailed analysis in our studies.

4.2

Imaging

In order to capture the liquid spray flame structure, a CH PLIF similar to the pre-

vaporized experiment was conducted. But instead of exciting the CH at CX Q-branch, the

R-branch excitation was implemented. The excitation is around 311nm wavelength, and to

prevent interference from Mie scattering, two filters of UG5 and Semrock 0002 were installed

in imaging equipment. The field of view in this experiment was determined to be 50*50. By

68

applying a binning process, the resolution of 49 micron/pixel was achieved. For CH and

OH excitations, the amount of laser energy was recorded as 1-2 mJ/pulse and 3-4 mJ/pulse,

respectively.

In the next sections, I try to evaluate statistics of spray flames to highlight their

impact on prevaporized flames. Additionally, the Lewis number effect of spray flame has

been investigated and curvature-signal relation has been depicted to demonstrate liquid fuel

spray’s effect on turbulent flames.

4.3 Flame Statistics

Figure 4.1 shows sample CH PLIF images of prevaporized methane flame with different

additions of liquid n-Heptane spray. As can be seen, the effect of the Le number is similar to

what we saw earlier in prevaporized cases. For cases with higher Le number fuels, the region

of higher signal is the region with high negative curvature. The signal value for premixed

methane with a liquid spray of n-Heptane is much higher compared to pure premixed methane.

Figure 4.1: Pure Premixed methane at equivalence ratio ϕ = 0.9 (Left), Premixed methane
at equivalence ratio ϕ = 0.75 alongside spray liquid n-Heptane with 9.2 mL/min (Middle),
and 20 mL/min (Right) flow rate

69

The challenges in topology analysis of liquid spray flames are concerned with the

algorithm to detect flame structure and calculate turbulent features. In this case, extinction

and breaks occur more frequently compared to the prevaporized case, because the interaction

of liquid droplets with the flame surface may lead to severe quenching. This may affect

the performance of the existing algorithm, which was developed based on local moderate

extinction. Also, strong Raman scattering from droplet pockets appears in the imaging due

to the high density of fuel components in liquid droplets. Additionally, the low concentration

of CH in lean or extremely rich conditions leads to low CH fluorescence signal, which makes

topology measurement difficult.

Figure 4.2: 2D map of averaged FSD for pure methane, and methane with different flow
rates of n-Heptane

Figure 4.2 shows the 2-D FSD map of three different pure methane, methane, and

spray n-Heptane cases with two different flow rates. Adding liquid sprays generally broadens

and extends the flame compared to the pure prevaporized case. For methane, adding 9.2 ml/s

of n-Heptane improves FSD and wrinkling of the flame, while adding 20 ml/s of n-Heptane

decreases FSD across the axial location.

It can be interpreted that the liquid spray can

enhance turbulent flame features up to a point. After that, it seems that interaction of larger

70

and more frequent liquid droplets with the flame surface causes extinction and breaks out of

the flame.

I tried to analyze the effect of liquid fuel sprays on flame structure and statistics for

different turbulent intensities. The first objective is the distribution of signal values captured

from the flame edges. For this observation, we focus on methane and n-Heptane jet flames

with addition of n-dodecane and n-Heptane for two different turbulent intensities. The effect

of n-Dodecane addition to methane flame’s signal distribution is shown in Figure 4.3. For both

turbulent cases, the addition of liquid spray increases the distribution of higher signal values,

indicating that regions with higher CH concentration increase compared to lean methane

flame and the distribution of signal is much similar to the rich methane flame. This can be

explained by the fact that the interaction of liquid droplets can have other effects besides

flame quenching. It appears that for some regions, the flame can vaporize liquid droplets

to increase the equivalence ratio in these regions, which leads to a richer flame and higher

CH formation and concentration. As shown in Figure 4.3b, for higher turbulent intensity,

the distribution of signal for lower liquid spray rate leans toward the lean flame, while the

higher spray case behaves similarly to the rich flame. This behavior is probably due to the

effect of Reynolds number and turbulence of flow on the mixing process. As regions with

higher CH formation are created, since the mixing process is much faster, some of the high

CH regions may break up and scatter into other regions, so their local equivalence ratios

decrease. However, as the liquid spray flow rate increases, the dispersion rate is not enough

compared to the increase in the number of high CH regions.

71

(a)

(b)

Figure 4.3: Flame signal PDF of pure methane with addition of n-Dodecane sprays (a)
Case 2, (b) case 3

For the prevaporized n-Heptane with addition of n-Heptane liquid spray, for all cases

of spray, the distribution of high signal values increases as predicted, shown in Figure 4.4.

As shown in Figure 4.4a, for the lower turbulent intensity, it appears that none of the spray

addition cases have a distribution profile similar to the rich flame of prevaporized n-Heptane.

This is probably due to the incapability of the flame and turbulent intensity to vaporize

liquid droplets completely, mix, and disperse the local rich region throughout the entire flame.

However, for the higher turbulent intensity, the distribution of signal for different flow rates

of sprays is intriguing, as presented in Figure 4.4b. The signal PDF1 for the lowest spray

flow rate follows the lean prevaporized n-Heptane flame PDF, with a slight increase in the

high signal distribution, as if the quenching rate of the flame is canceled out with additional

vaporized droplets to the flame. However, by increasing the flow rate of the spray to 30 mL,

the signal PDF of the flame closely resembles the PDF of the rich prevaporized n-Heptane

flame. It seems the dispersion rate of the vaporized droplets is higher than the extinction

1Probability Density Function

72

rate caused by the droplets. Lastly, for the highest spray flow rate of 40 mL, the distribution

profile reverts to the lean condition.

It may be due to the fact that the extinction and

quenching rate of the flame by the droplets is much higher, such that even vaporizing some

portions of the droplets cannot compensate for it.

(a)

(b)

Figure 4.4: Flame signal PDF of prevaporized n-Heptane with addition of n-Heptane sprays
(a) Case 2, (b) case 3

The next feature of the turbulent flame to study is the flame curvature and the effect

of liquid spray on its distribution. In the following, Figure 4.5 shows the effect of spray

addition on the distribution of signal values in the PDF of methane and n-Heptane. The

effect of n-Dodecane spray addition on the flame curvature of methane is shown in Figure

A.2a. The distribution of curvature PDF with addition of n-Dodecane is much similar to

the rich methane flame. The spray addition cases have a wider distribution compared to

the rich methane flame, indicating more wrinkling and sharp edges compared to the smooth

structure of the rich methane flame. While increasing the spray flow rate after a point, it

has the opposite effect. For the highest flow rate of spray, the curvature maximum PDF

73

is smaller than the maximum of the lean methane flame. This means the flame structure

is not as smooth as the lean methane flame, and its edges have been demolished by the

impact of liquid droplets on the flame. Figure A.2c shows the curvature PDF of prevaporized

n-Heptane with the addition of liquid spray of n-Heptane with different flow rates. It appears

that n-Heptane with equivalence ratio of ϕ = 0.75 plus different flow rates of sprays show

similar distribution of smooth curvature (λ ≈ 0) compared to rich prevaporized n-Heptane.

Another interesting outcome is the higher probability distribution of high negative curvatures

for flames with liquid sprays compared to the rich flame, which suggests that spray cases

have complex and wrinkled flame structures due to the interaction of liquid droplets with

flames, although the overall distribution of their curvatures look similar.

(a)

(b)

Figure 4.5: Signal-curvature correlation of turbulent flame Re=10,000 for (a) all fuels at
equivalence ratio ϕ = 1.2, (b)dodecane at different equivalence ratios, 5 mm above the burner
with 60 cm/s pilot flame

Figure 4.6 shows the turbulent-to-laminar area ratio for pure methane and methane

with n-Heptane sprays. The ratio of turbulent-to-laminar area, which indicates the ratio of

turbulent to laminar flame speed, is higher for the lean case of methane. Additionally, adding

74

Figure 4.6: Ratio of turbulent flame aria to laminar flame area for pure methane and
methane with spray n-Heptane liquid

a spray of n-Heptane to methane spreads the flame longer, though the turbulent-to-laminar

flame speed decreases along the flame. This may be due to the interaction of droplets with

the flame.

Figure 4.7 shows different behavior for pure prevaporized n-Heptane and n-Heptane

spray added to prevaporized n-Heptane. It appears that prevaporized n-Heptane with equiv-

alence ratio ϕ = 0.75 with 9.2ml/s spray n-Heptane shows a similar pattern to prevaporized

n-Heptane with equivalence ratio ϕ = 0.9. A similar trend is observed for prevaporized

n-Heptane with equivalence ratio ϕ = 1.1, and prevaporized n-Heptane with equivalence

ratio ϕ = 0.75 added with 14.5 ml/s spray n-Heptane. The interesting point is that the

global equivalence ratio of these cases is similar to the pure prevaporized cases. The global

equivalence ratio of ϕ = 0.75 plus 9.2 ml/s spray n-Heptane is 0.9, and the global equivalence

ratio of ϕ = 0.75 plus 14.5 ml/s spray n-Heptane is 1.1. However, the difference in wrinkling

between the pure and spray cases is significant. It suggests that although they have similar

global equivalence, in the spray case, the length of flame per area increased, and it is more

wrinkled compared to the pure case.

75

Figure 4.7: Ratio of turbulent flame aria to laminar flame area for prevaporized n-Heptane
with spray n-Heptane liquid

4.4 Lewis number effect

Similar to prevaporized research, after a comprehensive flame statistics analysis, fuel

effect regarding Lewis number is investigated for liquid spray flames. In order to study Lewis

number effect on liquid spray flames, measuring CH signal on the flame edges is essential.

After detecting flame edges and interpreting the CH signal, the probability of regions with

CH signal was presented in the previous section. It was shown that more regions with higher

CH signals are received in spray cases than pure cases. This behavior is observed for all liquid

fuels regarding the type of it. It presents that liquid fuel increases CH formation rates in the

reaction zone, which leads to higher CH concentration and higher received signal. In order

to depict the effect of the Le number, the correlation of CH signal values with corresponding

curvature is required. Figure 4.8 shows the correlation of mean CH signal and curvature for

turbulent methane flame alongside different concentrations of liquid n-Dodecane sprays at

two different turbulent intensities.

It is observed that pure methane with lean and rich conditions shows somehow similar

behavior at different curvature regions regarding the CH signal. However, adding liquid spray

76

changes this behavior significantly. It appears that adding liquid sprays increases the signal

values at high negative curvature and lowers it at positive curvature, just what we expect

for prevaporized liquid fuels. It can be assumed that as long as the methane flame is strong

enough to evaporate liquid sprays, the vapor liquid fuel burns as well, and the characteristics

of the methane flame change accordingly. Although adding liquid spray can enhance and

improve the methane flame, it can have opposite effects as well when the amount of liquid

spray increases beyond a certain point. As shown in Figure 4.8a, when we add 24 ml of

dodecane to Case 2 of the methane flame, the signal-curvature correlation tends toward the

pure cases. These cases may be a combination of evaporation and quenching in regions where

the liquid spray makes contact with methane flames. The overall balance between these two

factors seems to be in favor of quenching, so the total CH signal drops. Interestingly, when

we add 30 ml of n-Dodecane to the methane flame in Case 3, which has higher turbulence

intensity, the flame is capable of overcoming the quenching effect caused by liquid sprays as

shown in Figure 4.8b. It can be assumed that the turbulence intensity can improve the flame

when making contact with liquid droplets and enhance the evaporation process.

77

(a)

(b)

Figure 4.8: Signal-curvature correlation of methane turbulent flame with different liquid
n-Dodecane sprays at (a) turbulence intensity case b, and (b) turbulence intensity case c

Figure 4.9 shows the signal-curvature correlation for pure n-Heptane lean and rich

flames and two cases with added liquid spray of n-Heptane. The idea with these cases was

to recreate locally lean and rich n-Heptane flame with a combination of prevaporized and

liquid spray of n-Heptane and to see if the characteristics of spray flames were similar to

prevaporized ones. For n-Heptane cases with liquid spray, they seem to follow neither the rich

nor lean cases. Adding liquid spray to the prevaporized n-Heptane shows similar behavior

to pure prevaporized lean cases in regions of negative curvature, while the positive curvature

regions are completely different. It appears that due to turbulence or other unknown reasons,

the vaporized droplets increase the CH production and thus CH signal in negative curvature

regions. Furthermore, the addition of liquid spray to n-Heptane flame seems to behave like

a methane flame where all regions of curvature correlate to a similar CH signal.

78

Figure 4.9: Signal-curvature correlation of prevaporized n-Heptane turbulent flame with
different liquid n-Heptane sprays at turbulence intensity case b

79

Chapter 5

Chemiluminescence

5.1

Introduction

In this chapter, modeling simulations and experimental investigations are employed

to identify the most accurate CH sub-mechanism for heavy hydrocarbon liquid fuels. Four

chemiluminescence mechanisms for CH and OH are analyzed, each composed of CH* and

OH* sub-mechanisms. These combinations affect the predicted CH*/OH* ratio. Photon

emission rates for heavy fuels such as n-dodecane, kerosene, and JP-8 are post-processed

from 1-D unstrained, laminar flame simulations at 300K and 470K. The simulation results

are compared with experimental CH measurements to evaluate which sub-mechanism best

predicts CH*.

The chapter also focuses on the use of chemiluminescence as a practical tool for study-

ing combustion phenomena, including heat release, flame motion, and ignition/extinction

events. Particularly useful in industrial settings with limited optical access, chemilumines-

cence is often applied in internal combustion engine research. However, since most data

are based on methane, this chapter questions whether similar conclusions apply to heavier

liquid fuels and addresses whether chemiluminescence is a suitable technique for studying

such fuels.

80

Motivated by the need to compare experimental and computational approaches, the

chapter evaluates chemical mechanisms that include chemiluminescence sub-mechanisms.

Additionally, it assesses the spectroscopic validity of chemiluminescence data, ensuring con-

sistency between experimental and theoretical models. These comparisons are crucial for

validating both experimental methods and computational simulations.

The goals of this chapter are to test both full chemical mechanisms and chemilumines-

cence sub-mechanisms, ensuring that they predict the correct trends with varying equivalence

ratios and allow reliable comparisons across different fuels. By doing so, the chapter enhances

the understanding of chemiluminescence’s role in combustion processes, particularly for heavy

liquid fuels.

However, the chapter acknowledges several concerns and limitations. Certain quench-

ing species are not included in the models, and the prediction of base formation species may

be incomplete. Additionally, turbulence and flame area effects, as well as factors like Lewis

number, stretch, and curvature, are not fully addressed. These factors could influence the ac-

curacy of chemiluminescence predictions, especially in more complex flame conditions. These

concerns highlight the need for further refinement in both experimental and computational

approaches when studying heavy fuels.

5.2 CH Spectroscopy

5.3 Mechanisms and sub-mechanisms for CH and OH

In order to model CH* and OH* emission, the chemiluminescence sub-mechanism

needs formation reactions, quenching and relaxation reactions, as well as their constant rates.

The general mechanism for measuring CH and OH is provided for each fuel specifically. GRI

Mech3 is the general mechanism for methane, while for n-heptane and n-dodecane, the Jetsurf

2 is the main mechanism that provides the reactions. JetSurF2.0 consists of 348 species and

2163 reactions. The development effort centers on n-dodecane and n-butyl-cyclohexane, but

81

the model includes the high-temperature chemistry of all n-alkanes up to n-dodecane and

mono-alkylated cyclohexanes, including n-propylcyclohexane, ethylcyclohexane, methylcyclo-

hexane, and cyclohexane, as well. Highly detailed HyChem mechanism is usually used for

JP8, where it uses seven reaction steps describing JP-8 POSF10264 thermal decomposition

and USC Mech IIa as the foundational fuel chemistry model. The cracked products con-

sidered are C2H4, H4, H2, CH4, C3H6, 1-C4H8, i-C4H8, benzene, toluene, CH3, and H. The

oxidation kinetics of these cracking products are described by USC Mech IIa. After using

CHEMKIN and main mechanisms for each fuel, the final products of CH, OH, and other

species are used to calculate CH* and OH* with detailed sub-mechanisms. As mentioned in

previous chapters, the main formation reactions for CH* and OH* are as:

C2H + O ⇐⇒ CH ⋆ + CO

C2H + O2 ⇐⇒ CH ⋆ + CO2

CH + O2 ⇐⇒ CO + OH ⋆

(5.1)

(5.2)

(5.3)

There are several reactions in quenching and relaxing processes that should be men-

tioned and considered. The mechanisms used to evaluate CH* and OH* are presented in the

following tables, which contain formation and quenching reactions with their reaction rate

coefficients.

82

Table 5.1: Mechanism 1: CH∗ and OH∗ chemiluminescence mechanism from Peeters [De-
vriendt et al., 1996], Tamura et al. [Tamura et al., 1998], and Walsh et al. [Walsh et al.,
1998].[k = AT B exp (cid:0) −Ea

(cid:1)]

RT

No.

Reactions

A

B

Ea

Ref.

1

2

3

4

5

6

7

8

9

10

11

12

C2H + O ⇌ CH ∗ + CO

1.08 × 1013

C2H + O2 ⇌ CH ∗ + CO2

2.17 × 1010

CH ∗ → CH + hv

1.85 × 106

CH ∗ + H2O ⇌ CH + H2O 5.3 × 1013

CH ∗ + CO2 ⇌ CH + CO2

2.4 × 10−1

CH ∗ + CO ⇌ CH + CO

2.44 × 1012

0

0

0

0

4.3

0.5

0

0

0

0

Peeters

Peeters

Tamura

Tamura

-1694 Tamura

0

Tamura

CH ∗ + H2 ⇌ CH + H2

1.47 × 1014

0

1361 Tamura

CH ∗ + O2 ⇌ CH + O2

2.48 × 106

2.14

-1720 Tamura

CH ∗ + N2 ⇌ CH + N2

3.03 × 102

3.4

-381 Tamura

0

0

0

0.5

0.5

0.5

0.5

0.5

0.5

0.5

167

Tamura

0

0

Walsh

Tamura

-861 Tamura

-968 Tamura

-787 Tamura

-444 Tamura

-482 Tamura

-1238 Tamura

-635 Tamura

CH ∗ + CH4 ⇌ CH + CH4

1.73 × 1013

CH + O2 ⇌ CO + OH ∗

3.25 × 1013

OH ∗ → OH + hv

1.45 × 106

13 OH ∗ + H2O ⇌ OH + H2O 5.92 × 1012

14 OH ∗ + CO2 ⇌ OH + CO2

2.75 × 1012

15

16

17

18

OH ∗ + CO ⇌ OH + CO 3.23 × 1012

OH ∗ + H2 ⇌ OH + H2

2.95 × 1012

OH ∗ + O2 ⇌ OH + O2

2.1 × 1012

OH ∗ + N2 ⇌ OH + N2

1.08 × 1011

19 OH ∗ + CH4 ⇌ OH + CH4

3.36 × 1012

83

Table 5.2: Mechanism 2: CH∗ and OH∗ chemiluminescence mechanism from Peeters [De-
vriendt et al., 1996], Tamura et al. [Tamura et al., 1998], and Porter et al. [Porter et al.,
1967].[k = AT B exp (cid:0) −Ea

(cid:1)]

RT

No.

Reactions

A

B

Ea

Ref.

1

2

3

4

5

6

7

8

9

10

11

12

C2H + O ⇌ CH ∗ + CO

1.08 × 1013

C2H + O2 ⇌ CH ∗ + CO2

2.17 × 1010

CH ∗ → CH + hv

1.85 × 106

CH ∗ + H2O ⇌ CH + H2O 5.3 × 1013

CH ∗ + CO2 ⇌ CH + CO2

2.4 × 10−1

CH ∗ + CO ⇌ CH + CO

2.44 × 1012

0

0

0

0

4.3

0.5

0

0

0

0

Peeters

Peeters

Tamura

Tamura

-1694 Tamura

0

Tamura

CH ∗ + H2 ⇌ CH + H2

1.47 × 1014

0

1361 Tamura

CH ∗ + O2 ⇌ CH + O2

2.48 × 106

2.14

-1720 Tamura

CH ∗ + N2 ⇌ CH + N2

3.03 × 102

3.4

-381 Tamura

0

0

0

0.5

0.5

0.5

0.5

0.5

0.5

0.5

167

Tamura

0

0

Porter

Tamura

-861 Tamura

-968 Tamura

-787 Tamura

-444 Tamura

-482 Tamura

-1238 Tamura

-635 Tamura

CH ∗ + CH4 ⇌ CH + CH4

1.73 × 1013

CH + O2 ⇌ CO + OH ∗

6.0 × 1010

OH ∗ → OH + hv

1.45 × 106

13 OH ∗ + H2O ⇌ OH + H2O 5.92 × 1012

14 OH ∗ + CO2 ⇌ OH + CO2

2.75 × 1012

15

16

17

18

OH ∗ + CO ⇌ OH + CO 3.23 × 1012

OH ∗ + H2 ⇌ OH + H2

2.95 × 1012

OH ∗ + O2 ⇌ OH + O2

2.1 × 1012

OH ∗ + N2 ⇌ OH + N2

1.08 × 1011

19 OH ∗ + CH4 ⇌ OH + CH4

3.36 × 1012

84

Table 5.3: Mechanism 3: CH∗ and OH∗ chemiluminescence mechanism from Peeters [El-
samra et al., 2005], Tamura et al. [Tamura et al., 1998], and Porter et al. [Porter et al.,
1967].[k = AT B exp (cid:0) −Ea

(cid:1)]

RT

No.

Reactions

A

C2H + O ⇌ CH ∗ + CO

6.02 × 1012

B

0

Ea

Ref.

457

Elsamara

C2H + O2 ⇌ CH ∗ + CO2

6.02 × 10−4

4.4

-2285 Elsamara

CH ∗ → CH + hv

1.85 × 106

CH ∗ + H2O ⇌ CH + H2O 5.3 × 1013

CH ∗ + CO2 ⇌ CH + CO2

2.4 × 10−1

CH ∗ + CO ⇌ CH + CO

2.44 × 1012

0

0

4.3

0.5

0

0

Tamura

Tamura

-1694 Tamura

0

Tamura

CH ∗ + H2 ⇌ CH + H2

1.47 × 1014

0

1361

Tamura

CH ∗ + O2 ⇌ CH + O2

2.48 × 106

2.14

-1720 Tamura

CH ∗ + N2 ⇌ CH + N2

3.03 × 102

3.4

-381

Tamura

1

2

3

4

5

6

7

8

9

10

11

12

0

0

0

0.5

0.5

0.5

0.5

0.5

0.5

0.5

167

Tamura

0

0

Porter

Tamura

-861

Tamura

-968

Tamura

-787

Tamura

-444

Tamura

-482

Tamura

-1238 Tamura

-635

Tamura

CH ∗ + CH4 ⇌ CH + CH4

1.73 × 1013

CH + O2 ⇌ CO + OH ∗

6.0 × 1010

OH ∗ → OH + hv

1.45 × 106

13 OH ∗ + H2O ⇌ OH + H2O 5.92 × 1012

14 OH ∗ + CO2 ⇌ OH + CO2

2.75 × 1012

15

16

17

18

OH ∗ + CO ⇌ OH + CO

3.23 × 1012

OH ∗ + H2 ⇌ OH + H2

2.95 × 1012

OH ∗ + O2 ⇌ OH + O2

2.1 × 1012

OH ∗ + N2 ⇌ OH + N2

1.08 × 1011

19 OH ∗ + CH4 ⇌ OH + CH4

3.36 × 1012

85

Table 5.4: Mechanism 4: CH∗ and OH∗ chemiluminescence mechanism from Peeters [El-
samra et al., 2005], Tamura et al. [Tamura et al., 1998], and Carl et al. [Carl et al.,
2003].[k = AT B exp (cid:0) −Ea

(cid:1)]

RT

1

2

3

4

5

6

7

8

9

10

11

12

No.

Reactions

A

C2H + O ⇌ CH ∗ + CO

6.02 × 1012

B

0

Ea

Ref.

457

Elsamara

C2H + O2 ⇌ CH ∗ + CO2

6.02 × 10−4

4.4

-2285 Elsamara

CH ∗ → CH + hv

1.85 × 106

CH ∗ + H2O ⇌ CH + H2O 5.3 × 1013

CH ∗ + CO2 ⇌ CH + CO2

2.4 × 10−1

CH ∗ + CO ⇌ CH + CO

2.44 × 1012

0

0

4.3

0.5

0

0

Tamura

Tamura

-1694 Tamura

0

Tamura

CH ∗ + H2 ⇌ CH + H2

1.47 × 1014

0

1361

Tamura

CH ∗ + O2 ⇌ CH + O2

2.48 × 106

2.14

-1720 Tamura

CH ∗ + N2 ⇌ CH + N2

3.03 × 102

3.4

-381

Tamura

CH ∗ + CH4 ⇌ CH + CH4

1.73 × 1013

CH + O2 ⇌ CO + OH ∗

6.0 × 1010

OH ∗ → OH + hv

1.45 × 106

13 OH ∗ + H2O ⇌ OH + H2O 5.92 × 1012

14 OH ∗ + CO2 ⇌ OH + CO2

2.75 × 1012

15

16

17

18

OH ∗ + CO ⇌ OH + CO

3.23 × 1012

OH ∗ + H2 ⇌ OH + H2

2.95 × 1012

OH ∗ + O2 ⇌ OH + O2

2.1 × 1012

OH ∗ + N2 ⇌ OH + N2

1.08 × 1011

19 OH ∗ + CH4 ⇌ OH + CH4

3.36 × 1012

0

0

0

0.5

0.5

0.5

0.5

0.5

0.5

0.5

167

Tamura

0

0

Carl

Tamura

-861

Tamura

-968

Tamura

-787

Tamura

-444

Tamura

-482

Tamura

-1238 Tamura

-635

Tamura

Generally, radicals have low concentrations in flames, because they have a low rate

of formation and a rapid rate of removal and quenching. The concentrations were evaluated

as a post-process to the flame simulation, and assumed to be in a quasi-steady state, having

little overall effect on the total kinetics. Concentrations of species can be calculated by

86

using the formation and removal reaction rates. The following equation is for measuring the

concentration of CH* (moles/cm3) and OH*:

[CH ⋆] =

k1[C2H][O] + k2[C2H][O2]
ΣjkQ,j[Mj] + Ac

[OH ⋆] =

k3[CH][O2]
ΣkQ,j[Mj] + Ac

(5.4)

(5.5)

Where K1 and K2 are the reaction constant rates for two formation reactions (R1

and R2), kQ,j is the quenching rate constant by species j, and Ac is the Einstein coefficient

for spontaneous emission. K3 is the constant reaction rate for formation reaction R3. The

photon emission rate, i (mole photon,cms) is calculated by:

iCH ⋆ = Ac[CH ⋆]

iOH ⋆ = Ac[OH ⋆]

(5.6)

(5.7)

The profiles through the 1D simulation are integrated to calculate the full spatial

formation of each species, ICH (Photons cm-2 s-1), where L is the integration path length

across the 1D domain:

(cid:90)

ICH ⋆ =

iCH ⋆dx

(5.8)

The total emission rate, P (Photons/s), is given by the flame area, and can be derived

by:

PCH ⋆ = ICH ⋆ × Af = ICH ⋆

m
ρuSL

(5.9)

87

Normalized by the fuel mass flow rate:

P(CH ∗)
˙mf

= I(CH ∗)

(cid:33)

(cid:32)1 + ˙ma
˙mf
ρuSL

P(OH ∗)
˙mf

= I(OH ∗)

(cid:33)

(cid:32)1 + ˙ma
˙mf
ρuSL

(5.10)

(5.11)

These equations are presented to calculate the concentration of radical species. As

shown in the equations, these radical species depend on the concentration of species, which

can be measured through detailed mechanisms. Each fuel has its own specific mechanisms

that predict intermediate and final species.

Detailed chemical and kinetic calculations were done for a one-dimensional, unstrained,

adiabatic premixed flame in CHEMKIN for methane, JP-8 kerosene, and dodecane. For

methane, the GRI-mech 3 mechanism [Smith et al., 2011] with 325 reactions and 53 species

was used. The HyChem Fuel A1 (JP-8) detailed mechanism [Wang et al., 2018, Xu et al.,

2018] was utilized to simulate the combustion of JP-8. Finally, for dodecane, the JetSurf

2 [Wang et al., 2010] detailed high-temperature chemistry mechanism of all n-alkanes up to

n-dodecane, which contains 2163 reactions and 348 species, was employed. The simulations

have been done for temperatures of 300K and 470K and an equivalence ratio of 0.6 to 1.2.

The results from CHEMKIN are post-processed with different CH* and OH* sub-mechanisms

to evaluate each radical concentration.

5.4 Comparison to Mechanisms

Initially, I aim to depict the simulated distribution of different species for the combus-

tion of each fuel. Then, by utilizing various sub-mechanisms for CH* and OH*, it is possible

to determine which sub-mechanism can predict these species.

Preliminary CHEMKIN simulations using JetSurf2.0 [Wang et al., 2010] are given

below in Figure 2-4 for a freely propagating dodecane flame at ambient temperature and

88

pressure. The JetSurf mechanism inherently calculates the species concentrations of CH*

and OH* using the built-in model reactions by Renlund [Renlund et al., 1982] and Hall and

Petersen [Hall and Petersen, 2006]. Figure 5.1 depicts the spatial structure of combustion

intermediates of formaldehyde, HCO, CH, and OH. The structure of these profiles provides

information on the low-temperature preheat zones, reacting layers, and post-flame heat

release. Using the intermediate species profiles and the modeling approaches, profiles for CH*

and OH* can also be computed as given in Figures 5.2 and 5.3. The integration of these

profiles of chemiluminescent species can then be used for comparison against the measured

spectroscopic quantities from experiments.

Figure 5.1: Spatial structure of combustion intermediates of formaldehyde, HCO, CH, and
OH in a freely propagating n-Dodecane flame

Figure 5.2 shows the distribution of radical CH within a freely propagating rich n-

Dodecane flame, obtained from equilibrium condition, as described before. The concentration

of radical CH is two orders smaller than the CH. Figure 5.3 expresses the distribution of OH

and radical OH, which was measured similarly to radical CH. OH radical concentration is 8

89

degrees smaller than OH species concentration.

Figure 5.2: Spatial structure of combustion intermediate CH and CH* in a freely propa-
gating n-Dodecane flame

Figure 5.3: Spatial structure of combustion intermediate OH and OH* in a freely propa-
gating n-Dodecane flame

The four mechanisms presented in Tables 1 to 4 are suggested sub-mechanisms to

predict OH * and CH *. In addition, the JetSurf chemical mechanism includes a built-in

sub-mechanism for CH* and OH* excited species. The results for the ratios of CH*/OH* are

representative of the combination of sub-mechanisms included in the above table for each

pairing of CH* or OH*.

90

Figure 5.4 depicts the normalized signal values of CH* measured during experiments

for different fuels and equivalence ratios. As can be seen, the CH* signal level increases

with the equivalence ratio for all fuels, however, the rate of increase is higher for liquid

fuels compared to methane. The heavy hydrocarbon liquid fuels seem to produce higher

CH* compared to methane. This can be due to the multi-component effect and more CH

formation elements within the fuels.

Figure 5.4: Experimental normalized values of measured CH* signal for fuels at different
equivalence ratio

The measurement of CH* with simulation tools such as CHEMKIN is tested to

evaluate whether the sub-mechanisms can predict the CH* production for each fuel. Figure

5.5 shows the CH* measurements with two different sub-mechanisms introduced by Peeters

and Elsamra. For CH, as Figure 5.5 shows, the sub-mechanism simulations do not predict the

experimental trend. The experiments follow the expected trend where heavy hydrocarbon

fuels have a higher CH* signal than lighter fuels like methane. However, the simulation is

not capable of predicting the correct trend.

91

(a)

(b)

Figure 5.5: simulation measurement of CH* signal based on (a) Peeters and (b) Elsamra
sub-mechanisms for fuels at different equivalence ratios

The other important intermediate species of interest is OH. Figure 5.6 shows the

experimental signals of OH* captured for different fuels at different equivalence ratios. The

trend of OH is pretty similar to CH* distribution. As the equivalence ratio increases, the

OH* signal increases as well. The level of OH* signal is higher for liquid fuels than methane,

which can be due to several reasons. The multi-component fuels break apart into smaller

species and produce many more reactions compared to methane. These reactions can lead

to excessive production of CH* compared to methane at the same equivalence ratios.

92

Figure 5.6: Experimental normalized values of measured OH* signal for fuels at different
equivalence ratio

For the evaluation of sub-mechanisms, the same procedure is used for the calculation of

the OH signal. Figure 5.7 shows the different sub-mechanisms’ capability to predict the OH*

values for different fuels. Similar to CH*, the trend of sub-mechanism OH* does not follow

the experimental measurements. It seems all sub-mechanisms can follow the experimental

trend with equivalence ratios. However, the sub-mechanisms are not capable of distinguishing

the effect of fuels. The prediction of all of the sub-mechanisms for liquid fuels is similar to

each other.

93

(a)

(b)

Figure 5.7: simulation measurement of OH* signal based on (a) Walsh, (b) Porter, and (c)
Carl sub-mechanisms for fuels at different equivalence ratio

(c)

The ratio of CH*/OH* in different equivalence ratios is shown in Figure 5.8. This

ratio is significantly higher for liquid fuels at any given equivalence ratio. It shows that by

increasing the equivalence ratio, the ratio of CH*/OH* increases as well, and the rate of

increase is higher for liquid fuels compared to methane.

94

Figure 5.8: Experimental values of CH*/OH* for fuels at different equivalence ratio

Figure 5.9 shows the simulated ratio of CH*/OH* for different sub-mechanisms for

fuels at different equivalence ratios. As it is shown, the ratio of CH*/OH* does not follow

any specific trend with the equivalence ratio, compared to experimental results, where the

ratio of CH*/OH* increases with the equivalence ratio.

(a)

(b)

Figure 5.9: simulation measurement of CH*/OH* based on (a) Peeters-Carl, (b) Elsamra-
Carl sub-mechanisms for fuels at different equivalence ratio

95

The sub-mechanism built into JETSURF 2 is not capable of predicting either CH*

or OH* like other sub-mechanisms. The predicted CH and OH measurements do not match

experimental values. These results show the failure of chemical mechanisms to predict CH

and OH and indicate the necessity of further investigation into intermediate species and

finding a better mechanism to represent CH/OH chemiluminescence more accurately.

96

Chapter 6

Summary and Discussion

In this thesis, I investigated the complexities of liquid fuel combustion in aerospace

applications, particularly focusing on the challenges posed by gas turbines. Liquid fuel

combustion differs significantly from gaseous fuel combustion due to factors such as the

slower vaporization rate, complex fuel composition, and the turbulent nature of combustion.

These factors present unique challenges in achieving flame stabilization, understanding droplet

interactions, and defining the flammability limits of fuel mixtures, especially under varying

turbulent conditions. Traditional diagnostic methods like chemiluminescence and Planar

Laser-Induced Fluorescence (PLIF) have faced limitations in studying liquid fuels, particularly

due to issues like signal interference from droplets and the low intensity of emissions from

the combustion process.

To address these challenges, our research aimed to develop advanced diagnostic tech-

niques and computational models to improve the accuracy of flame diagnostics and gain a

deeper understanding of turbulent liquid fuel combustion. A key innovation introduced in this

work is the use of CH PLIF for effectively capturing flame structures and extinction events,

without being affected by the interference of liquid droplets. This was achieved through

collaboration with the Air Force Research Laboratory (AFRL), which allowed us to develop

and implement a new diagnostic methodology. In addition to CH PLIF, I introduced a nu-

97

merical image processing algorithm to enhance the accuracy of flame structure measurements,

particularly focusing on turbulence, flame curvature, and local extinction events.

Throughout the study, several experimental approaches were applied to measure

turbulent flame characteristics and understand the influence of parameters such as equivalence

ratio, fuel composition, and turbulence intensity on combustion dynamics.

I found that

liquid fuels exhibit higher local extinction frequencies and larger flame breaks compared to

methane flames, indicating that flame behavior in liquid fuels is affected by factors beyond

turbulence intensity alone. The flame surface density (FSD) of liquid fuels was found to

be significantly different from that of methane, especially under lean conditions, which

highlighted the importance of accurate diagnostic techniques that account for flame breaks

and extinction events.

Another major aspect of the research was to understand the influence of liquid fuel

sprays on flame structure. Our study showed that adding liquid sprays to methane and

n-Heptane flames increases CH concentrations and alters the signal distribution, resulting in

higher CH emissions, especially in highly turbulent conditions. However, I also found that the

addition of liquid sprays leads to more complex flame structures, characterized by increased

flame wrinkling and surface instability. These effects varied with the turbulence intensity

and spray flow rate, which ultimately determined the balance between flame enhancement

and quenching. The results highlighted the complexity of droplet interactions with the flame

and the impact of turbulence on the combustion process.

In addition, I explored the role of chemiluminescence as a diagnostic tool for studying

combustion phenomena in heavy hydrocarbon fuels such as kerosene, JP-8, and n-Dodecane.

Chemiluminescence is widely used in industrial combustion studies due to its simplicity and

cost-effectiveness. However, I found discrepancies between the experimental measurements

of CH*/OH* ratios and those predicted by the chemical mechanisms for heavier fuels. This

indicated that the existing chemical models might not fully capture the intermediate species

formation and reaction pathways for heavy fuels. Therefore, the study emphasized the need

98

for further refinement of chemical mechanisms to improve the accuracy of combustion models

for heavy liquid fuels.

The research also identified several limitations in existing diagnostic methods. For

example, background emissions from CO2 and C2 significantly affect the accuracy of CH

chemiluminescence measurements. To mitigate this, I developed techniques to subtract the

contributions of these interfering species, and I emphasized the use of narrower optical filters to

improve measurement accuracy. These improvements allowed for more precise determination

of CH concentrations, which is critical for understanding the combustion characteristics of

heavy hydrocarbon fuels.

Overall, the research presented in this thesis provides significant advancements in the

study of liquid fuel combustion. By developing advanced diagnostic tools, refining compu-

tational models, and addressing the challenges of measuring turbulent flame characteristics,

this work contributes to the optimization of combustion processes in aerospace applications.

The findings from this study can lead to improved combustion efficiency, stability, and re-

duced emissions in liquid-fuel-based systems, which are essential for the development of more

efficient and environmentally friendly propulsion technologies.

6.1 Key Results and Conclusions:

• This is the first application of C-X CH PLIF for any liquid fuel and represents the

highest-signal to noise imaging of CH for these fuels. This is the first time that local

extinctions have been visualized in liquid fuels.

• Developed a novel CH PLIF diagnostic technique to effectively visualize flame structures

and detect extinction events without droplet interference.

• The ability to perform CH PLIF in liquid fuels, both spray and prevaporized, is novel

and useful in the access to extinction data which is not accurately captured on OH

PLIF. The OH species is longer lived even after a local extinction has occurred and is

therefore not necessarily an accurate marker of when extinctions occur. However, CH

99

is a short-lived molecule exclusively produced in the heat release region of the flame

thickness, whose presence directly indicate reaction or quenched processes.

• Created a numerical algorithm and the OH Pseudo method to enhance the accuracy of

flame structure measurements, addressing limitations in traditional PLIF techniques.

• Demonstrated that liquid fuels exhibit higher local flame extinction frequencies and

longer flame breaks compared to methane flames, highlighting the impact of fuel prop-

erties on flame behavior.

• Higher and lower CH PLIF signals are observed in regions of high positive versus high

negative curvature. The value CH PLIF signal is controlled with the concentration

of the CH radical which correlated with the local heat release value. Thus, varying

CH fluorescence is indicative or vary heat release. This curvature driven variance is

a Lewis number effect and not observed strongly in methane flames because the Le is

approximately equal to 1. This is the first experimental observation of Lewis number

curvature effects in liquid fuels, which is corroborated by DNS calculations of other

liquid fuels.

• Where the fluorescence signal of CH is low is correlated with a lower heat release and

even extinction of the flame surface. This is a new extinction mechanism for liquid

fuels, with larger Le number, that has only been previously seen in hydrogen which has

a low Le number.

• Found that flame surface density (FSD) is significantly affected by fuel type and flame

conditions, with methane showing higher FSD under lean conditions compared to liquid

fuels.

• Showed that adding liquid sprays leads to higher CH concentrations and increased

flame wrinkling, indicating more complex and unstable flame structures, depending on

turbulence intensity and spray flow rate.

• Identified discrepancies between experimental and simulated CH*/OH* ratios for heavy

100

hydrocarbon fuels, indicating the need for further refinement of chemical models.

• The primary conclusion of the chemiluminescence effort suggest that the chemical

mechanism for the heavy fuels should not be trusted to accurately account for chemi-

luminescence differences in fuels. The general experimental trend shows that heavier

fuels generate more CH* and OH*, which is not predicted by the calculations.

• In fact, particularly for OH*, chemical mechanisms predict the same production of

radical signal for three different fuels since the same concentration of chemiluminescent

reactants are predicted for each fuel. Thus, the mechanisms are unable to distinguish

even between different liquid fuels.

• Highlighted the role of turbulence intensity and fuel composition in determining the

balance between flame enhancement and quenching in liquid fuel sprays.

These results provide critical insights into the combustion dynamics of liquid fuels,

offering guidelines for optimizing combustion processes in aerospace propulsion systems and

contributing to the development of more efficient and stable combustion technologies.

6.2 Next Steps

For future work, several avenues could be pursued to advance the understanding of

chemiluminescence in heavy hydrocarbon fuels. One approach would be to refine the existing

chemical mechanisms further by incorporating a broader range of intermediate species and

quenching reactions. This could involve developing more sophisticated models that account

for the detailed interactions between fuel components, radicals, and other intermediates during

combustion. Integrating these mechanisms into higher-dimensional simulations, such as 2D

and 3D turbulent flame models, could also provide a more comprehensive understanding of

how these fuels behave under realistic flow conditions.

Additionally, future work should aim to improve the accuracy of experimental mea-

surements by developing new diagnostic tools or refining current ones. For example, using

101

more advanced spectrometers with narrower optical filters could help reduce interference from

species like CO2 and C2, providing a clearer signal for CH and OH chemiluminescence. Fur-

thermore, implementing more advanced image-processing techniques to separate overlapping

emissions could enhance the resolution and reliability of the collected data.

Another potential direction would be to expand the scope of the research to include

other heavy fuels, such as alternative and sustainable biofuels. Investigating the chemilumi-

nescence characteristics of these fuels could provide insights into their combustion properties,

helping optimize their use in propulsion systems and industrial applications. Comparing these

findings with traditional fuels like JP-8 could also support the development of generalized

models applicable across a wider range of fuel types.

Incorporating computational fluid dynamics (CFD) models with refined chemilumi-

nescence mechanisms could also help predict flame behavior more accurately in practical

combustion systems. Integrating these models with real-world data would enable the de-

velopment of robust predictive capabilities that can guide the design and optimization of

industrial combustion systems, such as power plants and aerospace engines, to enhance their

efficiency and reduce emissions. Collaborating with industry partners could facilitate the

transfer of these refined diagnostic and computational techniques from the laboratory to

industrial applications, broadening the impact of this research.

102

Attributions and Other Contributions

Chapters of this work are extended text based on several manuscripts that were

authored by Amirreza Gandomkar: [Gandomkar et al., 2020, Gandomkar et al., 2021b, Gan-

domkar et al., 2021a, Gandomkar et al., 2021c]. Two additional articles on prevaporized and

spray liquid fuels are currently prepared for submission and are expected to be published in

the near future.

Chapter 2

This chapter is based on work done in collaboration with Savvas Gkantonas [Gan-

domkar et al., 2021a]. Both authors contributed equally to the initial development of the

methodology of flame edge detection. Gandomkar resumed the initial draft and finalized the

algorithms to measure flame statistics.

Chapter 3

This chapter is based on unpublished work done in collaboration with Dr. Campbell

Carter and Thomas A McManus from Air Force Research Laboratory. This research was

inspired by the original study on liquid fuels [McManus et al., 2021]. Most of the prevaporized

liquid fuels experiments were conducted by Dr. Allison and Dr. Carter in the AFRL.

Gandomkar contributed to the original text, image processing, data analysis, and experimental

design.

Chapter 4

This chapter is based on work done in collaboration with Dr. Campbell Carter from

Air Force Research Laboratory. Most of the spray liquid fuels experiments were conducted

by Dr. Allison and Dr. Carter in the AFRL1. Gandomkar contributed to the original text,

image processing, data analysis, and experimental design.

1Air Force Research Laboratory

103

Chapter 5

This chapter is based on work done in collaboration with John Schihl [Gandomkar

et al., 2020] and another unpublished research. Both authors contributed equally to the final

text of the original manuscript, data collection and analysis, and experimental design. Schihl

contributed to the experiment data acquisition. Gandomkar provided CHEMKIN simulation

results and the final figures.

104

BIBLIOGRAPHY

[Aspden et al., 2017] Aspden, A., Bell, J., Day, M., and Egolfopoulos, F. (2017). Turbulence–
flame interactions in lean premixed dodecane flames. Proceedings of the Combustion
Institute, 36(2):2005–2016.

[Borghi, 1985] Borghi, R. (1985). On the structure and morphology of turbulent premixed
flames. In Recent Advances in the Aerospace Sciences: In Honor of Luigi Crocco on His
Seventy-fifth Birthday, pages 117–138. Springer.

[Carl et al., 2003] Carl, S., Van Poppel, M., and Peeters, J. (2003). Identification of the ch+
o2→ oh (a)+ co reaction as the source of oh (a- x) chemiluminescence in c2h2/o/h/o2
atomic flames and determination of its absolute rate constant over the range t= 296 to
511 k. The Journal of Physical Chemistry A, 107(50):11001–11007.

[Carter et al., 2014] Carter, C. D., Hammack, S., and Lee, T. (2014). High-speed planar
laser-induced fluorescence of the ch radical using the cˆ 2\varsigmaˆ+-xˆ 2\varpi\left (0,
0\right) c 2 σ+-x 2 π 0, 0 band. Applied Physics B, 116:515–519.

[Carter et al., 2016] Carter, C. D., Hammack, S., and Lee, T. (2016). High-speed flamefront
imaging in premixed turbulent flames using planar laser-induced fluorescence of the ch c-
x band. Combustion and Flame, 168:66–74.

[Chakraborty and Cant, 2005] Chakraborty, N. and Cant, R. (2005).

Influence of lewis
number on curvature effects in turbulent premixed flame propagation in the thin reaction
zones regime. Physics of Fluids, 17(10).

[Chen, 2007] Chen, Y.-C. (2007). Measurements of flame-front curvature based on fourier

transformation. Combustion Theory and Modelling, 11(3):333–349.

[Devriendt et al., 1996] Devriendt, K., Van Look, H., Ceursters, B., and Peeters, J. (1996).
Kinetics of formation of chemiluminescent ch (a¡/sup¿ 2¡//sup¿ δ) by the elementary
reactions of c¡/inf¿ 2¡//inf¿ h (x¡/sup¿ 2¡//sup¿ σ¡/sup¿+¡//sup¿) with o (¡/sup¿ 3¡//sup¿
p) and o¡/inf¿ 2¡//inf¿(x¡/sup¿ 3¡//sup¿ σ¡/inf¿ g¡//inf¿¡/sup¿-¡//sup¿): A pulse laser
photolysis study. Chemical Physics Letters, 261(4-5):450–456.

[Driscoll, 2008] Driscoll, J. F. (2008). Turbulent premixed combustion: Flamelet structure
and its effect on turbulent burning velocities. Progress in Energy and Combustion Science,
34(1):91–134.

[Elsamra et al., 2005] Elsamra, R. M., Vranckx, S., and Carl, S. A.

(2005). Ch (a2δ)
formation in hydrocarbon combustion: The temperature dependence of the rate constant of
the reaction c2h+ o2→ ch (a2δ)+ co2. The Journal of Physical Chemistry A, 109(45):10287–
10293.

105

[Falkenstein et al., 2020] Falkenstein, T., Kang, S., and Pitsch, H.

(2020). Analysis of
premixed flame kernel/turbulence interactions under engine conditions based on direct
numerical simulation data. Journal of Fluid Mechanics, 885:A32.

[Gandomkar et al., 2021a] Gandomkar, A., Gkantonas, S., Schihl, J., McManus, T. A.,
Carter, C., and Allison, P. M. (2021a). Non-unity lewis number effects on curvature
statistics in prevaporized, turbulent liquid fuel jet flames: Curvature calculation method-
ology.

[Gandomkar et al., 2021b] Gandomkar, A., Gkantonas, S., Schihl, J., McManus, T. A.,
Carter, C. D., and Allison, P. M. (2021b). Fuel effects on turbulent premixed jet flame
topology as resolved by ch plif imaging. In AIAA Scitech 2021 Forum, page 0079.

[Gandomkar et al., 2021c] Gandomkar, A., Gkantonas, S., Svoboda, A., McManus, T. A.,
Carter, C., and Allison, P. M. (2021c). New perspectives of turbulent flame structure via
ch plif in prevaporized liquid fuels.

[Gandomkar et al., 2020] Gandomkar, A., Schihl, J., and Allison, P. M. (2020). Applicability
of flame chemiluminescence from liquid, heavy-hydrocarbon fuels. In AIAA Scitech 2020
Forum, page 0523.

[Garc´ıa-Armingol and Ballester, 2014] Garc´ıa-Armingol, T.

(2014).
Influence of fuel composition on chemiluminescence emission in premixed flames of
ch4/co2/h2/co blends. International journal of hydrogen energy, 39(35):20255–20265.

and Ballester, J.

[Guethe et al., 2012] Guethe, F., Guyot, D., Singla, G., Noiray, N., and Schuermans, B.
(2012). Chemiluminescence as diagnostic tool in the development of gas turbines. Applied
Physics B, 107:619–636.

[Guiberti et al., 2017] Guiberti, T., Durox, D., and Schuller, T. (2017). Flame chemilu-
minescence from co2-and n2-diluted laminar ch4/air premixed flames. Combustion and
Flame, 181:110–122.

[Guyot et al., 2010] Guyot, D., Guethe, F., Schuermans, B., Lacarelle, A., and Paschereit,
C. O. (2010). Ch/oh chemiluminescence response of an atmospheric premixed flame under
varying operating conditions. In Turbo Expo: Power for Land, Sea, and Air, volume 43970,
pages 933–944.

[Hall and Petersen, 2006] Hall, J. M. and Petersen, E. L. (2006). An optimized kinetics
model for oh chemiluminescence at high temperatures and atmospheric pressures. Inter-
national journal of chemical kinetics, 38(12):714–724.

[Hammack et al., 2018] Hammack, S. D., Skiba, A. W., Lee, T., and Carter, C. D. (2018).
Ch plif and piv implementation using cx (0, 0) and intra-vibrational band filtered detection.
Applied Physics B, 124:1–5.

106

[Haq et al., 2002] Haq, M., Sheppard, C., Woolley, R., Greenhalgh, D., and Lockett, R.
(2002). Wrinkling and curvature of laminar and turbulent premixed flames. Combustion
and Flame, 131(1-2):1–15.

[Hardalupas and Orain, 2004] Hardalupas, Y. and Orain, M. (2004). Local measurements
of the time-dependent heat release rate and equivalence ratio using chemiluminescent
emission from a flame. Combustion and Flame, 139(3):188–207.

[Higgins et al., 2001] Higgins, B., McQuay, M., Lacas, F., Rolon, J.-C., Darabiha, N., and
Systematic measurements of oh chemiluminescence for fuel-lean,

Candel, S.
high-pressure, premixed, laminar flames. Fuel, 80(1):67–74.

(2001).

[Hossain and Nakamura, 2014] Hossain, A. and Nakamura, Y. (2014). A numerical study
on the ability to predict the heat release rate using ch chemiluminescence in non-sooting
counterflow diffusion flames. Combustion and flame, 161(1):162–172.

[Jenkins and Cant, 2002] Jenkins, K. and Cant, R. (2002). Curvature effects on flame
kernels in a turbulent environment. Proceedings of the Combustion Institute, 29(2):2023–
2029.

[Kim and Pitsch, 2007] Kim, S. H. and Pitsch, H. (2007). Scalar gradient and small-scale

structure in turbulent premixed combustion. Physics of Fluids, 19(11).

[Lachaux et al., 2005] Lachaux, T., Halter, F., Chauveau, C., G¨okalp, I., and Shepherd, I. G.
(2005). Flame front analysis of high-pressure turbulent lean premixed methane–air flames.
Proceedings of the Combustion Institute, 30(1):819–826.

[Lauer and Sattelmayer, 2010] Lauer, M. and Sattelmayer, T. (2010). On the adequacy of
chemiluminescence as a measure for heat release in turbulent flames with mixture gradients.

[Lee and Seo, 2015] Lee, H. and Seo, S. (2015). Experimental study on spectral character-

istics of kerosene swirl combustion. Procedia Engineering, 99:304–312.

[Lemaire et al., 2009] Lemaire, R., Faccinetto, A., Therssen, E., Ziskind, M., Focsa, C., and
Desgroux, P. (2009). Experimental comparison of soot formation in turbulent flames of
diesel and surrogate diesel fuels. Proceedings of the Combustion Institute, 32(1):737–744.

[Li et al., 2010] Li, Z., Li, B., Sun, Z., Bai, X.-S., and Ald´en, M. (2010). Turbulence and
combustion interaction: High resolution local flame front structure visualization using
simultaneous single-shot plif imaging of ch, oh, and ch2o in a piloted premixed jet flame.
Combustion and Flame, 157(6):1087–1096.

[McManus et al., 2021] McManus, T. A., Gandomkar, A., Carter, C. D., and Allison, P. M.
(2021). Topological imaging of turbulent premixed, prevaporized liquid fuel jet flames
using ch (cx) band plif. Proceedings of the Combustion Institute, 38(2):3049–3056.

107

[Meyer et al., 2005] Meyer, T. R., Roy, S., Belovich, V. M., Corporan, E., and Gord, J. R.
(2005).
Simultaneous planar laser-induced incandescence, oh planar laser-induced flu-
orescence, and droplet mie scattering in swirl-stabilized spray flames. Applied optics,
44(3):445–454.

[Mitsingas et al., 2019] Mitsingas, C. M., Hammack, S. D., Mayhew, E. K., Rajasegar, R.,
McGann, B., Skiba, A. W., Carter, C. D., and Lee, T. (2019). Simultaneous high speed
piv and ch plif using r-branch excitation in the c2σ+-x2π (0, 0) band. Proceedings of the
Combustion Institute, 37(2):1479–1487.

[Nori and Seitzman, 2008] Nori, V. and Seitzman, J. (2008). Evaluation of chemilumi-
nescence as a combustion diagnostic under varying operating conditions. In 46th AIAA
Aerospace Sciences Meeting and Exhibit, page 953.

[Nori and Seitzman, 2009] Nori, V. N. and Seitzman, J. M. (2009). Ch chemiluminescence
modeling for combustion diagnostics. Proceedings of the Combustion Institute, 32(1):895–
903.

[Otsu, 1979] Otsu, N. (1979). A threshold selection method from gray-level histograms.

IEEE transactions on systems, man, and cybernetics, 9(1):62–66.

[Panoutsos et al., 2009] Panoutsos, C., Hardalupas, Y., and Taylor, A. (2009). Numeri-
cal evaluation of equivalence ratio measurement using oh and ch chemiluminescence in
premixed and non-premixed methane–air flames. Combustion and Flame, 156(2):273–291.

[Paul and Najm, 1998] Paul, P. H. and Najm, H. N. (1998). Planar laser-induced fluores-
In Symposium (international) on combustion,

cence imaging of flame heat release rate.
volume 27, pages 43–50. Elsevier.

[Porter et al., 1967] Porter, R., Clark, A., Kaskan, W., and Browne, W. (1967). A study
of hydrocarbon flames. In Symposium (International) on Combustion, volume 11, pages
907–917. Elsevier.

[Renlund et al., 1982] Renlund, A., Shokoohi, F., Reisler, H., and Wittig, C. (1982). Reac-
tion of ethynyl radical with oxygen. chemiluminescent products. The journal of physical
chemistry, 86(21):4165–4170.

[Renou et al., 1998] Renou, B., Boukhalfa, A., Puechberty, D., and Trinit´e, M.

(1998).
Effects of stretch on the local structure of preely propagating premixed low-turbulent flames
with various lewis numbers.
In Symposium (International) on Combustion, volume 27,
pages 841–847. Elsevier.

[Rodrigues et al., 2015] Rodrigues, H. C., Tummers, M. J., van Veen, E. H., and Roekaerts,
D. J. (2015). Spray flame structure in conventional and hot-diluted combustion regime.
Combustion and Flame, 162(3):759–773.

108

[Rosell et al., 2017] Rosell, J., Bai, X.-S., Sjoholm, J., Zhou, B., Li, Z., Wang, Z., Pettersson,
P., Li, Z., Richter, M., and Alden, M. (2017). Multi-species plif study of the structures of
turbulent premixed methane/air jet flames in the flamelet and thin-reaction zones regimes.
Combustion and Flame, 182:324–338.

[Rutland and Trouv´e, 1993] Rutland, C. and Trouv´e, A. (1993). Direct simulations of
premixed turbulent flames with nonunity lewis numbers. Combustion and Flame, 94(1-
2):41–57.

[Sadiki et al., 2005] Sadiki, A., Chrigui, M., Janicka, J., and Maneshkarimi, M.

(2005).
Modeling and simulation of effects of turbulence on vaporization, mixing and combustion
of liquid-fuel sprays. Flow, Turbulence and Combustion, 75:105–130.

[Savard and Blanquart, 2015] Savard, B. and Blanquart, G.

(2015). Broken reaction
zone and differential diffusion effects in high karlovitz n-c7h16 premixed turbulent flames.
Combustion and flame, 162(5):2020–2033.

[Savard et al., 2015] Savard, B., Bobbitt, B., and Blanquart, G. (2015). Structure of a
high karlovitz n-c7h16 premixed turbulent flame. Proceedings of the Combustion Institute,
35(2):1377–1384.

[Savitzky and Golay, 1964] Savitzky, A. and Golay, M. J. (1964). Smoothing and differentia-
tion of data by simplified least squares procedures. Analytical chemistry, 36(8):1627–1639.

[Smith et al., 2011] Smith, G. P., Golden, D. M., Frenklach, M., Moriarty, N. W., Eiteneer,
B., Goldenberg, M., Bowman, C. T., Hanson, R. K., Song, S., Gardiner Jr, W., et al.
(2011). Gri-mech 3.0, 1999. URL http://www. me. berkeley. edu/gri mech, 38.

[Smolke et al., 2018] Smolke, J., Carbone, F., Egolfopoulos, F. N., and Wang, H. (2018).
Effect of n-dodecane decomposition on its fundamental flame properties. Combustion and
Flame, 190:65–73.

[Smooke et al., 1996] Smooke, M., Ern, A., Tanoff, M., Valdati, B., Mohammed, R., Marran,
D., and Long, M. (1996). Computational and experimental study of no in an axisymmetric
laminar diffusion flame. In Symposium (International) on Combustion, volume 26, pages
2161–2170. Elsevier.

[Soika et al., 2003] Soika, A., Dinkelacker, F., and Leipertz, A. (2003). Pressure influence on
the flame front curvature of turbulent premixed flames: comparison between experiment
and theory. Combustion and flame, 132(3):451–462.

[Stepowski et al., 1994] Stepowski, D., Cessou, A., and Goix, P. (1994). Flame stabilization
and oh fluorescence mapping of the combustion structures in the near field of a spray jet.
Combustion and flame, 99(3-4):516–IN1.

109

[Tamura et al., 1998] Tamura, M., Berg, P. A., Harrington, J. E., Luque, J., Jeffries, J. B.,
Smith, G. P., and Crosley, D. R. (1998). Collisional quenching of ch (a), oh (a), and no
(a) in low pressure hydrocarbon flames. Combustion and Flame, 114(3-4):502–514.

[Tripathi et al., 2012] Tripathi, M. M., Krishnan, S. R., Srinivasan, K. K., Yueh, F.-Y., and
Singh, J. P. (2012). Chemiluminescence-based multivariate sensing of local equivalence
ratios in premixed atmospheric methane–air flames. Fuel, 93:684–691.

[Tsushima et al., 1998] Tsushima, S., Saitoh, H., Akamatsu, F., and Katsuki, M. (1998).
Observation of combustion characteristics of droplet clusters in a premixed-spray flame
In
by simultaneous monitoring of planar spray images and local chemiluminescence.
Symposium (international) on combustion, volume 27, pages 1967–1974. Elsevier.

[Walsh et al., 1998] Walsh, K., Long, M., Tanoff, M., and Smooke, M. (1998). Experimental
and computational study of ch, ch, and oh in an axisymmetric laminar diffusion flame. In
Symposium (international) on combustion, volume 27, pages 615–623. Elsevier.

[Wang et al., 2010] Wang, H., E Dames, B Sirjean, D. A. S., R Tango, A Violi, J. Y.
W. L., F N Egolfopoulos, D. F. D., R K Hanson, C. T. B., C K Law, W. T., Cernansky,
N. P., Miller, D. L., and Lindstedt, R. P. (September 19, 2010). A high-temperature
chemical kinetic model of n-alkane (up to n-dodecane), cyclohexane, and methyl-, ethyl-
, n-propyl and n-butyl-cyclohexane oxidation at high temperatures, jetsurf version 2.0.
http://web.stanford.edu/group/haiwanglab/JetSurF/JetSurF2.0/index.html.

[Wang et al., 2018] Wang, H., Xu, R., Wang, K., Bowman, C. T., Hanson, R. K., Davidson,
D. F., Brezinsky, K., and Egolfopoulos, F. N. (2018). A physics-based approach to mod-
eling real-fuel combustion chemistry-i. evidence from experiments, and thermodynamic,
chemical kinetic and statistical considerations. Combustion and Flame, 193:502–519.

[Wang et al., 2019] Wang, Y., Jain, A., and Kulatilaka, W. (2019). Co imaging in piloted
liquid-spray flames using femtosecond two-photon lif. Proceedings of the Combustion
Institute, 37(2):1305–1312.

[Xu et al., 2018] Xu, R., Wang, K., Banerjee, S., Shao, J., Parise, T., Zhu, Y., Wang,
S., Movaghar, A., Lee, D. J., Zhao, R., et al.
(2018). A physics-based approach to
modeling real-fuel combustion chemistry–ii. reaction kinetic models of jet and rocket fuels.
Combustion and Flame, 193:520–537.

[Zhang et al., 2014] Zhang, M., Wang, J., Xie, Y., Wei, Z., Jin, W., Huang, Z., and
Kobayashi, H. (2014). Measurement on instantaneous flame front structure of turbu-
lent premixed ch4/h2/air flames. Experimental Thermal and Fluid Science, 52:288–296.

[Zhang et al., 2019] Zhang, W., Wang, J., Lin, W., Guo, S., Zhang, M., Li, G., Ye, J., and
Huang, Z. (2019). Measurements on flame structure of bluff body and swirl stabilized
premixed flames close to blow-off. Experimental Thermal and Fluid Science, 104:15–25.

110

[Zhang et al., 2020] Zhang, W., Wang, J., Lin, W., Mao, R., Xia, H., Zhang, M., and
Huang, Z. (2020). Effect of differential diffusion on turbulent lean premixed hydrogen
enriched flames through structure analysis.
International Journal of Hydrogen Energy,
45(18):10920–10931.

111

APPENDIX

These experiments were conducted in collaboration with John Schihl, Ph.D candidate

of Mechanical Engineering at Michigan State University. This appendix provides a brief

overview of the results.

As mentioned before, the captured signal of CH* chemiluminescence with the specific fil-

ter involves background CO2* emission and C2. Figure A.1 shows the chemiluminescence

spectrometry of JP8 at different equivalence ratios.

Figure A.1: Chemiluminescence spectrometry of JP8 at different equivalence ratios

112

(a)

(b)

Figure A.2: Chemiluminescence spectrometry of (a) n-Dodecane, (b) n-Heptane, and (c)
Methane s at different equivalence ratios

(c)

As shown here, other than background CO2 emission, there are regions with C2*

chemiluminescence which can affect the total CH evaluation. Figure A.2 shows the chemi-

luminescence spectrometry of n-Dodecane, n-Heptane, and methane at different equivalence

ratios. Rich fuel flames produce more CH chemiluminescence compared to the lean cases.

However, it is not just CH that is captured and measured. The figure A.3 shows a zoomed-in

view of chemiluminescence in the region of the CH signal around 430 nm for all fuels at an

equivalence ratio of 1.

113

(a)

(b)

Figure A.3: (a) Chemiluminescence spectrometry of liquid fuels and Methane at equivalence
ratio ϕ = 1 (b) zoom in around CH chemiluminescence to observe CO2 and C2* contributions

There are two peaks in the spectrometry around 433 nm and 438 nm. The first one is

CH, and the latter is the C2 signal. Due to the width of the CH filter equipped in the exper-

iment, the C2 chemiluminescence is captured as well. The amount of C2 chemiluminescence

gets higher as the equivalence ratio of liquid fuels increases. Therefore, the total CH signal

received is a superposition of CH, C2, and CO2. First, the effect of background CO2 should

be calculated. Then the contribution of C2* is investigated, and finally, the total effective

CH chemiluminescence will be measured and reported.

114

Figure A.4: Chemiluminescence spectrometry of JP8 around 430 nm wavelength

Figure A.4 shows background CO2 in JP8 spectrometry within regions of CH chemi-

luminescence. To calculate the contribution of CO2 to CH chemiluminescence, we measure

two areas: one below the CH chemiluminescence, and the other one is the area below the

black dotted line. Then, if we subtract the second area from the first area, the total CH

chemiluminescence without background CO2 is measured. Figure A.5 shows the effect of

background CO2 on the CH chemiluminescence captured in experiments. For liquid fuels,

the contribution of CO2 is around 30 to 50 percent, while it’s around 60 to 80 percent for

methane. The CO2 contribution is higher in lean conditions compared to rich flames for all

fuels. Also, liquid fuels show higher CH chemiluminescence and insensitivity to equivalence

ratio.

115

Figure A.5: Ratio of CH chemiluminescence without background CO2 to total CH chemi-
luminescence with consideration of CO2

Now we try to calculate the effect of C2* on the CH chemiluminescence similarly

to what we did with CO2*. Based on Figure A.4, we calculate two areas: one between

the CH chemiluminescence and black dotted line (green and blue arrows together), and the

other one is the area where the blue arrow represents. The total CH* chemiluminescence is

calculated without the contribution of CO2*. Just similar to CO2 measurement, we calculate

two areas and subtract the second one from the first one. The results are shown in Figure

A.6. It seems that the C2* contribution remains constant across equivalence ratios for all

fuels. The variation of CH* with equivalence ratio is proportional to the change in C2* with

an equivalence ratio. Also, all fuels show similar behavior regarding C2* contribution across

different equivalence ratios.

116

Figure A.6: Ratio of CH chemiluminescence without background C2* to total CH chemilu-
minescence with consideration of C2*

Finally, if we do these steps together, we can calculate the pure CH* chemiluminescence

regarding the elimination of CO2 and C2. Figure A.7 shows the percentage of CH contribution

to what is actually captured in experiments, including the effects of CO2 and C2.

117

Figure A.7: Ratio of only CH chemiluminescence to total CH chemiluminescence with
consideration of CO2* and C2*

The CH chemiluminescence improves by increasing the equivalence ratio, especially

for methane. It seems that around 60 to 70 of what we record in experiments is CO2 or C2,

not CH. These results are based on a 10 nm filter around 430 nm wavelength. This suggests

implementing a narrow filter to reduce the participation of C2 and CO2 in chemiluminescence.

118