ArticlePDF Available

Abstract and Figures

The objective of the current study was to formulate RP-3 kerosene surrogate by emulating fuel properties affecting the physical and chemical processes of the target fuel under the engine relevant conditions. This study utilized two-dimensional gas chromatography with time-of-flight mass spectrometry (GC × GC-TOFMS) and 13 C and 1 H nuclear magnetic resonance (NMR) spectroscopy to characterize the com-positional characteristics of RP-3 fuel, and various standard test methods were applied to measure the physical and chemical properties of the target fuel. Two surrogate fuels (K1, a mixture of five components and K2, a mixture of seven components) were optimally determined through a multi-property regression algorithm by matching carbon types (CTs), distillation curve, cetane number (CN), density, and threshold sooting index (TSI) of the target fuel. The measured and estimated values of both target properties and non-target properties of surrogates were validated against the experimental data of RP-3 kerosene. Ignition delay times (IDTs) of both surrogates were investigated in a heated shock tube and a heated rapid compression machine under engine relevant conditions and validated against the measured results of RP-3. Overall, K1 and K2 both exhibited good matching on the compositional characteristics , physical-chemical properties, and gas phase ignition behaviors with the target fuel. In contrast, the seven-component K2 was more competitive and more comprehensive.
Content may be subject to copyright.
Combustion and Flame 208 (2019) 388–401
Contents lists available at ScienceDirect
Combustion and Flame
journal homepage:ame
Surrogate fuels for RP-3 kerosene formulated by emulating molecular
structures, functional groups, physical and chemical properties
Zhiyong Wu, Yebing Mao, Mohsin Raza, Jizhen Zhu, Yuan Feng, Sixu Wang, Yo ng Qian,
Liang Yu, Xingcai Lu
Key Laboratory for Power Machinery and Engineering of Ministry of Education, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai
200240, PR China
a r t i c l e i n f o
Article history:
Received 15 April 2019
Revised 20 May 2019
Accepted 11 July 2019
Keywo rds:
RP-3 kerosene
Fuel properties
Surrogate fuel formulation
NMR spectroscopy
Ignition delay time
a b s t r a c t
The objective of the current study was to formulate RP-3 kerosene surrogate by emulating fuel proper-
ties affecting the physical and chemical processes of the target fuel under the engine relevant conditions.
This study utilized two-dimensional gas chromatography with time-of-flight mass spectrometry (GC ×
C and
H nuclear magnetic resonance (NMR) spectroscopy to characterize the com-
positional characteristics of RP-3 fuel, and various standard test methods were applied to measure the
physical and chemical properties of the target fuel. Two surrogate fuels (K1, a mixture of five compo-
nents and K2, a mixture of seven components) were optimally determined through a multi-property
regression algorithm by matching carbon types (CTs), distillation curve, cetane number (CN), density,
and threshold sooting index (TSI) of the target fuel. The measured and estimated values of both tar-
get properties and non-target properties of surrogates were validated against the experimental data of
RP-3 kerosene. Ignition delay times (IDTs) of both surrogates were investigated in a heated shock tube
and a heated rapid compression machine under engine relevant conditions and validated against the
measured results of RP-3. Overall, K1 and K2 both exhibited good matching on the compositional charac-
teristics, physical-chemical properties, and gas phase ignition behaviors with the target fuel. In contrast,
the seven-component K2 was more competitive and more comprehensive.
©2019 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
1. Introduction
Computational fluid dynamics (CFD) coupling with detailed
chemical kinetic models of real fuels provides a promising ap-
proach for the development of modern engine by investigating
combustion fundamentals. The real fuels, like gasoline, diesel, and
jet fuel, contain thousands of individual chemical components, and
the compositions of these fuels vary according to manufacturers
and production places. As a result, simulations directed toward
these complex fuels are beyond the computational limitations of
engine level CFD applications. However, the simulation work could
be greatly simplified by replacing real fuels with surrogate fu-
els. Surrogate fuels are formulated with significantly fewer real
fuel components that can emulate the combustion and physical-
chemical properties of target fuels to an acceptable extent.
Surrogate fuels are used to mimic the behaviors of real fuels in
various combustion devices, and the definition and complexity of
the surrogate fuel formulation depend on the intended applications
Corresponding author.
E-mail address: (X. Lu).
[1] . For the conventional petroleum-derived fuels, important phys-
ical properties include volatility (distillation curve), density, vis-
cosity, surface tension, thermal conductivity and sound speed, and
the important chemical and combustion properties include ignition
quality (cetane number (CN)), lower heating value (LHV), TSI and
composition (like H/C ratio, molecular weight (MW) and molecular
structure). Generally, the physical properties determine the evapo-
ration of spray, the fuel–air mixing process and the oil-gas distri-
bution within the combustion chamber, and the chemical proper-
ties directly affect the ignition timing, soot formation, combustion
rate and combustion duration [2–4] . The different combinations of
the target properties lead to different surrogate formulation meth-
ods. Generally, the more properties chosen as the matching targets,
the more complicated the surrogate fuels will be.
There has been an extensive research focusing on the surrogate
formulation of jet fuels. The status of efforts for jet fuel surrogate
formulation, kinetic model development, and experimental val-
idations prior to 2007 were reviewed by Dagaut and Cathonnet
[5] and Colket et al. [6] . The surrogate formulations and kinetic
models for the commonly used kerosene type fuels, like Jet-A,
Jet-A-1, and JP-8, have been extensively studied to date. Violi et
al. [7] developed two slightly different surrogates by emulating
0010-2180/© 2019 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401 389
ADC advanced distillation curves
BP boiling point
CFD computational fluid dynamics
CKM chemical kinetic mechanisms
CN cetane number
CTs carbon types
DEC decalin
GC ×GC–FID two-dimensional gas chromatography with
flame ionization detection
GC ×GC–TOFMS two-dimensional gas chromatography with
time-of-flight mass spectrometry
HMN 2,2,4,4,6,8,8-heptamethylnonane
IDTs ignition delay times
ISOC isopropylcyclohexane
K1 five-component surrogate
K2 seven-component surrogate
LHV lower heating value
MW molecular weight
NDOD n-dodecane
NMR nuclear magnetic resonance
NPEB n-pentylbenzene
NPED n-pentadecane
NTC Negative Temper ature Coefficient
PHE 2,2,4,6,6-pentamethylheptane
RCM rapid compression machine
SP smoke point
ST shock tube
TBP true boiling point
TEI tetralin
TIPB 1,3,5-triisopropylbenzene
TRIC 1,3,5-trimethylcyclohexane
TSI threshold sooting index
the volatility, sooting propensity, and combustion properties of
JP-8 fuel. The surrogate #2 composed of seven real components
was shown to better match the design targets of JP-8. Dooley et al.
[8,9] built two sets of surrogate fuels for Jet-A kerosene to emulate
the gas phase combustion kinetic phenomena only. The 1st genera-
tion surrogate, which was a mixture of n-decane/isooctane/toluene,
could reproduce the CN and H/C ratio of the target fuel. The 2nd
generation surrogate composed of n-dodecane/isooctane/1,3,5-
trimethylbenzene/n-propylbenzene was generated by matching
CN, MW, H/C ratio, and TSI of the target fuel. The experiments and
simulations were conducted for both surrogates to validate the
gas phase combustion characteristics. Kim et al. [10] developed
two surrogates (UM1 and UM2) using a model-based optimizer to
emulate the fuel properties affecting the spray development and
gas phase ignition of Jet-A-1 kerosene. The eight target properties,
including CN, LHV, H/C, MW, distillation curve, density, viscosity,
and surface tension, were taken into consideration for the opti-
mization process. The estimated values of the target properties and
the ignition delay were compared to the experimental data of the
target fuel. Recently, Yu et al. [11] proposed two surrogates (S1 for
premixed engine combustion and S2 for spay-guided combustion)
for Jet-A kerosene by emulating eight fuel properties, including
physical properties, gas phase chemical properties and TSI.
The methodology of formulation for surrogates of diesel and
gasoline are also relevant to the present study. Mueller et al.
[12,13] proposed a comprehensive diesel surrogate formulation
methodology considering compositional characteristics, volatility,
CN, and density. Some novel methods, including two-dimensional
gas chromatography with flame ionization detection (GC ×GC–FID),
NMR Spectroscopy, and advanced distillation curves (ADC), were
utilized to characterize the target fuels and the surrogates. All the
design properties and various non-design properties were tested
by a series of ASTM standard methods and compared to the ex-
perimental results of the target fuels. Ahmed et al. [14] presented
a computational methodology for formulating surrogates for FACE
gasolines A and C by combining regression modeling with physical
and chemical kinetics simulations. The regression algorithm deter-
mined the optimal surrogate composition to match the fuel prop-
erties of FACE A and C gasoline, including H/C ratio, density, distil-
lation curve, CTs, and research octane number (RON).
Chinese No.3 (RP-3) kerosene, a different type of jet fuel as
compared to Jet-A and JP-8, etc. is the most important civil aviation
fuel in China. Recently, Mao et al. [15] proposed a three-component
surrogate of n-dodecane/iso-cetane/toluene by matching five gas
phase combustion properties of RP-3. The developed kinetic model
of the surrogate was validated against IDTs of RP-3 kerosene over
wide ranges. Yu and Gou [16] constructed a three-component
surrogate composed of n-dodecane/2,5-dimethylhexane/toluene for
RP-3 kerosene based on the functional groups. The estimations of
some physical properties of the surrogate under sub- and super-
critical conditions were validated against the experimental data of
the target fuel. They, however, provided limited experimental val-
idations for IDTs and laminar flame speeds. Moreover, a notice-
able discrepancy was found between the experimental IDTs of RP-3
and the model predictions [15] . Zhang et al. [17] developed a two-
component surrogate that matched the measured H/C ratio of RP-3
only. The surrogates for RP-3 kerosene mentioned above mainly fo-
cused on the gas phase combustion characteristics, while little or
no consideration was given to the sooting tendency and the phys-
ical properties in liquid phase affecting the spray breakup, evapo-
ration, and fuel–air mixing.
The present study aims to design a surrogate fuel for RP-3
kerosene that is capable of emulating the physical properties, gas
phase chemical properties, functional groups, and the sooting ten-
dency of the target fuel. Novel approaches were utilized to charac-
terize the compositional characteristics, including GC ×GC-TOFMS
C and
H NMR spectroscopy. Various standard test methods
were applied to characterize the physical and chemical properties
of RP-3 fuel, including density, viscosity, surface tension, distilla-
tion curve, CN, LHV, MW, H/C ratio, and smoke point (SP). In or-
der to address the need for a jet fuel surrogate for CFD simulation
that can be applied to capture both physical and chemical pro-
cesses under engine relevant conditions, the CTs, CN, distillation
curve, density and TSI were chosen as target properties in the sur-
rogate formulation. A five-component surrogate (K1) and a seven-
component surrogate (K2) were generated in an automatic way by
employing a regression optimization model. Both surrogates were
produced on a lab scale to validate the target and non-target prop-
erties of RP-3 kerosene. Additionally, the IDTs of the two surrogates
were measured in a heated shock tube (ST) and a rapid compres-
sion machine (RCM), and the experimental data of surrogates was
compared to the experimental results of the target fuel.
2. Material and method
2.1. Targe t-f uel characterization
RP-3 kerosene is selected as the target fuel in this study.
Table 1 therein provides many of the properties of RP-3 as mea-
sured by a series of standard test methods. The main proper-
ties of RP-3 are quantified including both the target properties
and the non-target properties. All the properties exhibited will be
validated for accurately emulating properties of RP-3 and all the
measurements for surrogates will also be conducted utilizing the
same standard methods under the same conditions. In addition, in
390 Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401
Tabl e 1
Measured properties of RP-3 kerosene.
Parameters Te st Methods RP-3
CN ASTM D613 43.3
Smoke point ASTM D1322-18 24 mm
LHV GB/T 384-1981(2004) 42.4 MJ/kg
a ASTM D4052-18 Temperature dependent
a ASTM D445-17a Temperature dependent
Surface tension
b GB/T 5549-2010 25.8 mN/m
Distillation curve
c ASTM D86 and D2887 Results in Section 3.2
n-alkanes 25.1 wt%
iso-alkanes 42.7 wt%
1-ring-cycloalkanes 10.7 wt%
2-ring-cycloalkanes 7.1 wt%
aromatics 14.4 wt%
H/C 1.963
MW 165.3 g/mol
a Test results shown in Section 2.2 .
b Limited by the test apparatus, only the surface tension at 25 °C was measured.
c Result of the simulated distillation curve measured by ASTM D2887 is taken
from [18] .
d Results taken from [15] .
order to accurately quantify the target fuel composition, two novel
approaches are employed to characterize RP-3. First, GC ×GC-
TOFMS technique is used to reveal the molecular structural fea-
tures of the target fuel. This is critical for identifying and selecting
candidate compounds which are representative of the actual con-
stituent molecules of the target fuel. Then the
C and
spectroscopy are employed to determine the content of CT that is
one of the constraints in the surrogate-composition-determination
2.1.1. GC ×GC-TOFMS
GC ×GC-TOFMS is a molecular-level composition analysis
method that realizes a high-resolution separation and identifica-
tion of the components contained in the fuel sample [19] . The Pe-
gasus 4D GC ×GC-TOFMS instrument utilized in this study applies
a non-polar column for the first dimension separating the com-
pounds by boiling points and a polar column for the second di-
mension separating the compounds by polarity. The detailed oper-
ating conditions for GC ×GC-TOFMS analysis are listed in Table S1
in the supporting information. The data processing is conducted on
a ChromaTOF software, and the MAINLIB, NIST_MASS, NIST_MASS2,
NIST_RI and REPLIB library database are used to determine what
the detected compounds should be.
A GC ×GC-TOFMS chromatogram for RP-3 kerosene is shown
in Fig. 1 . A total of 2476 peaks are detected in the RP-3 sam-
ple, corresponding to each of the colored circles in the chro-
matogram. It indicates that there are more than 2476 individual
compounds contained in the RP-3 kerosene, since each peak nom-
inally corresponds to at least one compound. The area of each
circle is proportional to the mass fraction of the corresponding
compound. The compounds in RP-3 are classified into the cat-
egory of n-alkanes/iso-alkanes/cycloalkanes/alkylbenzenes/indanes
or tetralins/2-ring aromatics. The n-alkanes detected range from n-
hexane (C
) to n-octadecane (C
) and the first dimension
time of chromatogram is converted into temperature scale accord-
ing to the correlation established between the boiling points of n-
alkanes and their 1st retention times. The GC ×GC-TOFMS test re-
sult reveals the structural characteristics of the compounds in RP-
3 and the boiling range of RP-3. The information can be instruc-
tive for the palette compounds selection for surrogates. Among the
n-alkanes class, n-dodecane (C
) constitutes the highest fuel
fraction in RP-3 and it has been frequently employed as a surrogate
component for various jet fuels. Indanes/tetralins class in the chro-
matogram also shows a main species group that should be consid-
ered in the palette compound selection. Several bicyclic-aromatic
compounds are detected, which exhibit the highest unsaturation
in the RP-3 kerosene.
2.1.2. NMR spectroscopy
NMR spectroscopy is an atomic-level analytical technique that
could realize reproducible, accurate, and non-destructive character-
ization of petroleum fractions in a short time [20] . The
C and
NMR spectroscopy are used to classify the carbon atoms in RP-3
kerosene. The RP-3 sample dissolved in a deuterated chloroform
CDCl3 solvent is tested on a 600 MHz nuclear magnetic resonance
and the carbon spectra and the proton spectra are processed using
a MestReNova v9.0.0 software.
The eleven distinct CTs shown in Fig. 2 are classified according
to the carbon-to-carbon bonds within the molecules. The classifi-
cation scheme was previously applied to diesel [12] and gasoline
[14] characterization. The spectral region integral of the chemical
shifts in the NMR spectra is conducted based on the previous stud-
ies of Basu et al. [21] and Japanwala et al. [22] . The NMR spectral
integrals and the elemental analysis results are used to assign the
mole fraction of each CT, and each CT has an estimated uncertainty
of ±3 mol% [23] . Once the CT mole fractions are determined, the
C and H mole fraction could be estimated from the results of CT
analysis. The quantitative CT results are displayed in Section 3.1.1 .
The NMR-derived H content is found to be within 0.14 wt% of the
measured value.
2.2. Surrogate formulation method
2.2.1. Targe t property selection
The objective of this study is to formulate RP-3 surrogate that
could comprehensively emulate the compositional characteristics,
physical properties, gas phase chemical properties, and sooting
tendency. In order to address the need for a RP-3 surrogate that
is capable of emulating the properties mentioned above, the re-
lated properties should be simultaneously kept as matching targets
in the surrogate formulation. In this study, the five chosen target
properties are distillation curve, density, CTs, CN, and TSI.
The selected five target properties could cover all aspects of
the design objectives of this study, including composition, physi-
cal properties, chemical properties, and soot propensity. The dis-
tillation curve is a graphical depiction of the boiling temperature
of a mixture plotted against its distillate volume fraction, and it
is a decisive factor in the fuel evaporation behavior [24] . In this
study, the simulated distillation curve of RP-3 measured by ASTM
D2887 has been chosen as the fitting target by applying a TBP (true
boiling point) curve fitting approach proposed by Reiter et al. [25] .
The density, viscosity and surface tension of liquid fuels are strong
functions of the spray prediction under engine conditions [2,26] .
Here, among these three properties, only the density is selected
as a matching target in the optimization process in order to re-
duce the complexity of calculation, and the other two will be val-
idated as non-target properties. Based on the definition of CTs in
Section 2.1.2 , the CTs represents the compositional characteristics
that are also closely related to the H/C ratio and hydrocarbon class
distribution [12] , and these compositional characteristics strongly
affect the flame phenomena such as adiabatic flame temperature,
flame velocity and diffusive extinction limits [9] . The CN is selected
to quantify the ignition quality of the target fuel, surrogates and
candidate compounds. The sooting tendency of jet fuels is charac-
terized by the indicator of TSI that is closely related to the soot
formation as well as soot emissions in the combustors [27] . Thus,
the five target properties are chosen to properly address the effects
of liquid fuel properties on spray and auto-ignition characteristics,
and we are expecting that the non-target properties of surrogates
Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401 391
Fig. 1. GC ×GC-TOFMS chromatogram for RP-3 kerosene.
Fig. 2. CTs classification based on the carbon-to-carbon bonds.
could also fall into a reasonable range, if the target properties can
be well matched.
2.2.2. Candidate compound selection
The surrogate component selection is based on the considera-
tion of both pros and cons from multiple perspectives, including
target properties, non-target properties, availability, purity, cost,
safety, and availability of detailed and/or reduced chemical kinetic
mechanisms (CKM) [13] . Some principles in the selection process
are summarized as follows:
(1) The candidate compounds should be the representatives in
the target fuel and all the components in the surrogate could
reasonably cover the main hydrocarbon types contained in
the target fuel.
(2) The surrogate components with known properties, commer-
cial and economic availability could realize the formulation
target, and the CKM of these compounds are readily avail-
able or promising to be developed.
(3) The candidate compounds could realize the surrogate de-
sign target in a reasonable complexity in order to reduce the
complexity of future CKM development, and generally the
number of surrogate components in the previous studies is
no more than 12 [6] .
The representative compounds and hydrocarbon classes in
RP-3 have been identified through the high-resolution technique
392 Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401
Fig. 3. Candidate compounds for surrogate ge neration.
Tabl e 2
Surrogate compounds and their properties.
compound name Abbre. CAS no. H/C Ratio MW (g/mol) BP
( °C) CN
b LHV (MJ/kg)
d Sur.Ten.
(mN/m) CKM Avail able ?
n-dodecane NDOD 112-40-3 2.17 170.3 216.3 82.5 44.11 7 24.93 Yes [31,32]
n-pentadecane NPED 629-62-9 2.13 212.4 270.6 98 43.98 8 26.74 Yes [31,32]
2,2,4,6,6-pentamethylheptane PHE 13475-82-6 2.17 170.3 177.8 9 44.14 15.4 21.58 Yes [33]
2,2,4,4,6,8,8-heptamethylnonane HMN 4390-04-9 2.125 226.4 246.3 15 43.85 22 24.8 Yes [34,35]
1,3,5-trimethylcyclohexane TRIC 1839-63-0 2 126.2 140.0 30.5 43.36 6 23.04 Yes [36]
isopropylcyclohexane ISOC 696-29-7 2 126.2 154.4 35 43.4 9 25.7
decalin DEC 91-17-8 1.8 138.2 191.4 46.5 42.58 22 32.17 Yes [37,38]
n-pentylbenzene NPEB 538-68-1 1.45 148.2 203.1 18 41.65 50 29.23 Yes [39]
1,3,5-triisopropylbenzene TIPB 717-74-8 1.6 204.3 236.3 2.9 41.99 80 29.25
tetralin TEL 119-64-2 1.2 132.2 207.2 8.9 40.53 70 33.16 Yes [40]
a Boiling points at 0.1 MPa [41] .
Cetane Number [42] .
LHV and liquid Surface Ten sio n at 25 °C [43] .
Threshold Sooting Index [4,29,44,45] .
GC ×GC-TOFMS. Based on the test results, the measured proper-
ties and the selected target properties, five compounds are se-
lected as the candidates for surrogate K1 and seven compounds
are selected for surrogate K2. The structure and use of each of
the selected species are shown in Fig. 3 . The surrogate K1 con-
tains five compounds, including n-dodecane (NDOD), 2,2,4,4,6,8,8-
heptamethylnonane (HMN), isopropylcyclohexane (ISOC), decalin
(DEC), and n-pentylbenzene (NPEB). The five compounds selected
for K1 substantially conform to the above three principles. For
K2, three of the five compounds of K1 are replaced by five
new compounds that are added to improve its ability to match
the composition and properties. The seven components contained
in K2 are: n-dodecane (NDOD), n-pentadecane (NPED), 2,2,4,6,6-
pentamethylheptane (PHE), 1,3,5-trimethylcyclohexane (TRIC), de-
calin (DEC), 1,3,5-triisopropylbenzene (TIPB) and tetralin (TEI).
Among them, PHE is selected to replace HMN as the iso-alkane
compound to reduce the content of the aliphatic C (CT 11 ) . TRIC
and NPED are selected to facilitate better matching of the light end
of the distillation curve and the heavy end of the distillation curve,
respectively. TIPB possessing multiple substituents is added to im-
prove the matching on CT3 and it is the representative of the mul-
tiply substituted, low-cetane, and high-boiling mono-aromatics. Fi-
nally, TEI is added to be the representative of the indanes/tetralins
class found in RP-3. The candidate-compound procurement infor-
mation is provided in Table S2 in the supplemental material. All
the compounds show a reasonable cost, while the price of the
branched-alkane compound PHE is relatively higher than others,
and PHE is still introduced as a surrogate component in K2 because
it can be more representative than HMN of the types of branched
alkanes in jet fuels and provide better agreement with the target
fuel [28] .
The properties of the candidates are summarized in Table 2
and Fig. 4 . The CT distributions of the compounds are exhibited
in Table 3 and this information is required for determining the CT
Tabl e 3
CT distributions of the candidate compounds.
Carbon Types (CT)
1 2 3 4 5 6 7 8 9 10 11
NDOD 2 10 0 0 0 0 0 0 0 0 0
NPED 2 13 0 0 0 0 0 0 0 0 0
PHE 7 2 1 0 0 0 0 0 0 0 2
HMN 9 3 1 0 0 0 0 0 0 0 3
TRIC 3 0 0 3 3 0 0 0 0 0 0
ISOC 2 0 1 5 1 0 0 0 0 0 0
DEC 0 0 0 8 0 2 0 0 0 0 0
NPEB 1 4 0 0 0 0 5 1 0 0 0
TIPB 6 0 3 0 0 0 3 3 0 0 0
TEL 0 0 0 4 0 0 4 0 2 0 0
mole fractions in surrogate fuels. The data sources of the candi-
date information have been indicated in the table below. Some es-
timated values have been marked in the table, for example, the
LHV and surface tension (at 25 °C) of PHE are estimated as the
average of HMN and iso-octane based on the study of Won et al.
[29] . The LHV of TRIC has been estimated equal to that of methyl-
cyclohexane due to unavailability of specific value in the literature.
For the temperature-dependent properties, density and viscosity
of the surrogate candidates are displayed in a wide temperature
range while the surface tension at 25 °C is merely exhibited. This
is because the surface tension of RP-3 is measured only at 25 °C,
which is the limitation of our test apparatus.
The last column of Table 2 shows the availability of the CKM
of each candidate compound (a dash indicates that the mecha-
nism is currently unavailable). The CKM of most selected com-
pounds (8/10) exist up to now, and the compounds (ISOC and TIPB)
whose CKM are unavailable now are also selected in order to bet-
ter realize the matching targets. Though one of the most impor-
tant criteria for the selection of the components is the availability
Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401 393
Fig. 4. Temperature-dependent (a) density and (b) kinematic viscosity of the RP-3
kerosene and the candidate compounds.
of matured CKM, this criterion sometimes may cause noticeable
discrepancies in the matching of target properties, since the forced
selection of some candidate compounds outside the molecular size
distribution of the target fuel due to the limited availability of
CKM [28] . To address this issue, some new candidate components,
which have the potential to improve the agreement with target
fuels regardless of the availability of CKM, can be selected, and
some recent studies have introduced some new compounds to
reach their matching goals [12,13,28,30] . In this study, several new
compounds have been chosen as candidate components to develop
the next generation RP-3 surrogate fuel that can emulate various
physical and chemical properties simultaneously. In addition, in or-
der to address the need of complete CKM for surrogates, our fu-
ture work will focus on experimental kinetic studies and chemical
mechanism development of the new components, and finally de-
velop a complete CKM for RP-3 surrogate fuel.
2.2.3. Optimization of surrogate composition
A multi-property regression algorithm is applied to determine
the surrogate composition by minimizing the differences between
the experimental results of RP-3 and the prediction models of sur-
Tabl e 4
Estimation models for the target properties.
Targ et property Estimation method Reference
i =1
i =1
, among that j = 1–11
CN Volume fraction average,
i =1
TSI Mole fraction average,
i =1
Density Volume fraction average,
i =1
Distillation curve TBP curve fitting approach
(x ) =
i =1
( A
i, 1
+ A
i, 2
j, i
is the number of CT
of component i, x
is mole fraction of component i, C
the carbon number of component i, n is the number of components contained in
the surrogate; CN
is cetane number of component i, v
is volume fraction of com-
ponent i; TSI
is the threshold sooting index of component i;
ρT, i
is the density of
component i at temperature T, v
T, i
is the volume fraction of component i at tem-
perature T .
Fig. 5. Distillation curve fitting approach based on the area between stepwise ap-
proximation and smooth distillation curve.
rogates. The similar methods were utilized in the previous studies
for determining surrogate composition [12,30,46] . Table 4 summa-
rizes the estimation models of the target properties and the cor-
responding references. It is worth noticing that most fuel proper-
ties comply with a non-linear mixing rule, and some estimating
models in this study applying liner-mixing rules are just simple
estimates. The distillation curve fitting model in this study em-
ploys a TBP curve fitting approach that has been applied to the
surrogate formulation in the previous studies [25,47] . The details
of the fitting model can be found in the above two studies and it
is briefly described here. The general idea of the distillation curve
fitting method has been illustrated in Fig. 5 . There are several bars
surrounded by the green lines that correspond to the components
contained in surrogate fuel and the bars are sequentially stacked
to form a stepwise approximation model according to the normal
boiling points of the components. The shaded area between the
distillation stepwise approximation model of the surrogate fuel and
the simulated distillation curve of the target fuel is calculated in
Eq. (1) . The value of F ( x ) can be regarded as the average tempera-
ture difference between the distillation behaviors of surrogate and
target fuel and F ( x ) is kept as a fitting criterion in the regression
model to determine whether the surrogate is sufficient to satisfy
the required accuracy for matching volatility.
i =1
i, 1
+ A
i, 2
) (1)
394 Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401
The objective function in the regression algorithm is defined in
Eq. (2) . The S can be regarded as the total difference of the tar-
get properties between the surrogate and the target fuel, and its
value is to be minimized in the iterative optimization process. The
several W ”are the weighing factors that could be adjusted. The
five sub-functions are shown in Eqs. (3) –( 7 ) correspond to the five
target properties, respectively, as the subscripts indicate. The cal-
culated properties denoted “cal” are the estimated values of sur-
rogate introduced in Table 4 and the measured properties denoted
“meas” are the experimental results of the target fuel. T
ave used as
the denominator in Eq. (5) is considered as the average tempera-
ture of the distillation curve of target fuel. And the temperature is
introduced to calculate the difference percentage as the other four
sub functions do. The definition of T
ave is defined in Eq. (8) , and
T ( x ) in the equation is the fitting function of the simulated distilla-
tion curve of the target fuel using the software TableCurve2D V5.01
[49] .
S = W
+ W
+ W
+ W
= 100 C N
= 100 T S I
= 100
i =1
i, 1
+ A
i, 2
a v e
ρ= 100 ρcal
= 100 C T
a v e
dx (8)
The optimization process consists of two parts: initialization
and iterative calculation. Firstly, the weighting factor and the
amount of each selected component are initialized with equal
amounts. Then, the regression model was run to determine a
surrogate-formulation that would match the 11 CTs, CN, distilla-
tion curve points, density and TSI to within 3 mol%, 1. 5 numbers,
±10 °C, 5% and 1 number, respectively, between the target fuel and
the surrogate. These optimization goals were referenced from the
previous studies [13,25] , and whether these matching goals can
be achieved or not depends on the matching ability of the se-
lected candidate compounds. The first step of iterative calculation
is to run the regression algorithm to yield an initial surrogate-
formulation that best matches the design targets with the initial
weighing factors. Comparing the surrogate properties with those of
the target fuel, if a property does not achieve the matching goal,
then the weighting factor of this property would be increased in
the next calculation. Only one weighting factor will be adjusted
in each iterative calculation, so the matching results of the for-
mer calculation can provide clues for the next adjustment. Once
a target property achieves its matching goal, the weighting factor
of this property will be kept in the same level and the iterative
optimization process will continue until all the design goals are
achieved. It is also worth noticing that the results of the optimiza-
tion are not unique and the calculating results depend on the ini-
tial settings.
Tabl e 5
Surrogate fuel compositions.
Palette compounds K1 K2
mol% wt% mol% wt%
n-dodecane 27.24 28.37 22.39 22.41
n-pentadecane 0 0 12.02 15.00
2,2,4,6,6-pentamethylheptane 0 0 31.28 31.31
2,2,4,4,6,8,8-heptamethylnonane 20.25 28.03 0 0
1,3,5-trimethylcyclohexane 0 0 12.00 8.90
isopropylcyclohexane 20.94 16.16 0 0
decalin 18.79 15.87 9.07 7.37
n-pentylbenzene 12.77 11.57 0 0
1,3,5-triisopropylbenzene 0 0 11.14 13.38
tetralin 0 0 2.11 1.64
2.3. Shock tube (ST) and rapid compression machine (RCM)
The gas phase ignition behaviors within low-to-high temper-
ature ranges of the formulated surrogates are investigated in a
heated ST and a heated RCM under engine relevant conditions. The
details of ST and RCM have been described in the earlier publi-
cations [38,50,51] . The uncertainty of ST measurement is mainly
caused by the determination of shock wave velocity and the max-
imum uncertainty is estimated to be ±20%, and the uncertainty
of IDT in RCM is estimated to be less than ±10% which is mainly
caused by the mixture preparation, measurement of pressure and
the determination of temperature. The detailed uncertainty analy-
sis was provided in the above studies. The IDTs of surrogates/air
mixture are measured at = 0.5 and 1 for temperature ranges
from 642 to 1292 K and pressure at 10 bar. The auto-ignition char-
acteristics of surrogates are validated against the experimental data
of RP-3, and the results are shown in Section 3.5 .
3. Results and validations
3.1. Surrogate compositional characteristics
The surrogate compositions formulated using the above pro-
cedure are given in Table 5 . The five-component K1 and seven-
component K2 are displayed in mole fractions and mass fractions.
3.1.1. Carbon types
Figure 6 shows the comparisons of CTs between the target fuel
and the two surrogates. The carbon type mole factions of RP-3 are
quantified by NMR spectroscopy, described in Section 2.1.2 . The re-
sult shows that 92.1 mol% of the total C atoms in RP-3 are con-
tained in CT 1–4. This can be considered to be somewhat consis-
tent with the results of hydrocarbon class analysis in Table 1 . The
CT distribution of surrogate is estimated based on the mole frac-
tion of each candidate component and the distribution of CT in
each candidate compound (shown in Table 3 ). Overall, the average
absolute difference between RP-3 and K1 as calculated from the re-
gression model is 4.1 mol%, while the difference of K2 is 2.6 mol%.
While concerning to the single-CT-matching within these two sur-
rogates, K2 performs better in matching nine out of eleven CTs ex-
cept the primary C (CT 1) and the cycloalkane to alkyl-chain CH
(CT 5). The larger discrepancy CT 1 shown in K2 is caused by the
inclusion of the poly-substituted aromatic compound TIPB which
is introduced to improve the matching on the iso-alkane CH (CT 3)
and the heavy end of the distillation curve. In addition, the high
level of the cycloalkane CH
(CT 4) in K1 indicates that the surro-
gate fuel has a high content of cycloalkane carbon, and thanks to
the poly-substituted cycloalkane compound TRIC replacing ISOC as
a single-ring-cycloalkane component in surrogate K2, the surrogate
fuel K2 reduces the difference of CT4 by about 11.9 mol% compared
Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401 395
Fig. 6. Comparisons of target- and surrogate-fuel compositional characteristics as quantified by carbon type.
to K1. On the whole, K2 performs better on the emulation of the
target property, CTs.
However, the matching of the iso-alkane CH (CT 3) between
RP-3 and both surrogates is not as close as the other CTs. More-
over, both surrogates exhibit rather large discrepancies with the
target fuel on the content of the aliphatic C (CT 11). Although K2
gets some improvement on the matching of CT 3 and CT 11 due
to the addition of component PHE and TIPB, this is far from the
matching goal yet. The CT 11 in K1 and K2 are all derived from
the iso-alkane components, HMN and PHE, respectively. Even so,
these two compounds are chosen because they can be purchased
with high purity and relatively low expense compared to the other
branched paraffins. The typical single methylated iso-alkanes (i.e.,
2-methylalkanes) are regarded as an important iso-alkane group in
conventional fuels [52] . These compounds can improve surrogate
matching on CT 3 and CT 11 when used as candidates, but these
iso-alkanes are not commercially available in high purity and at a
reasonable cost.
3.1.2. Hydrocarbon classes
The predicted GC ×GC-TOFMS chromatograms for K1 and K2
are shown in panels a and b of Fig. 7 , respectively, overlaid on
RP-3 composition (in gray) for reference. The candidate compounds
contained in surrogates are indicated by the labeled, colored cir-
cles. The area and color of the circle correspond to its mass frac-
tion and hydrocarbon class, respectively. The position of the can-
didate circle is predicted based on the elution time of the related
compound found in RP-3 chromatogram. As is shown in Fig. 7 (a),
the surrogate components in K1 fall into the intermediate range
of polarity (retention time from 2.20 to 3.08 s along the y-axis) at
their given boiling points (range from 154.4 to 246.3 °C). While in
Fig. 7 (b), the components in K2 exhibit a wider range of polariz-
ability (retention time between 1.8 8 and 3.76 s along the y -axis)
at their boiling points (range of 140.0–270.6 °C), due to the inclu-
sions of TRIC, PHE, TEL and NPED. Therefore, K2 can better cover
the whole area of the target fuel chromatogram when compared to
Figure 8 provides the hydrocarbon class composition of K1 and
K2 as well as the composition of the target fuel. It is obvious that
the mass fractions of 1-ring cycloalkanes, 2-ring cycloalkanes, and
aromatics in K2 better reproduce the contents of these hydrocar-
bon classes in RP-3. In general, the absolute differences between
RP-3 and the surrogates averaged over the five specified hydro-
carbon classes in Fig. 8 are 7 (for K1) and 5.3 (for K2) wt%, re-
spectively. Though the mass fraction of the hydrocarbon class is
not selected as an optimization target in the regression algorithm,
the distributions of hydrocarbon classes in the two surrogates still
show relatively good agreements with RP-3 kerosene, which indi-
cates that the better match of the target-property CTs can lead to
a better hydrocarbon-class distribution in surrogates. To sum up,
K2 performs better than K1 on emulating the compositional char-
acteristics from the above aspects.
3.2. Distillation curve
In this study, the simulated distillation curve of RP-3 quanti-
fied by ASTM D2887 is the optimization target in the regression
algorithm employing a TBP curve fitting approach. The simulated
distillation is a gas chromatographic method, where a retention
time calibration mixture is used to develop a retention time versus
boiling point curve and the retention times of tested sample are
converted to temperature with reference to the retention time of
the calibration mixture [53] . In order to evaluate the distillation
modeling method, the industry-standard volatility-quantification
technique, ASTM D86 [54] , is applied to measure the distillation
behaviors of the two surrogates and the target fuel. The estimated
uncertainty in the temperatures is less than 0.5 °C and the uncer-
tainty in the volume measurement that was used to obtain the
distillate volume fraction is 0.05 mL in each case.
Figure 9 shows the distillation data as measured for RP-3 and
the two surrogates as well as the temperature differences of the
distillate points. As evident from the distillation results, both sur-
rogates substantially replicate the distillation characteristics of the
target fuel. The absolute average temperature deviation between
the experimental data of the target fuel and the surrogate K1
within whole evaporation range is 5.6 °C, while that difference be-
tween the measured data of RP-3 and K2 is 4.2 °C. Hence, both
surrogates exhibit excellent fit to the target fuel. However, as is
396 Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401
Fig. 7. Predicted distributions in GC ×GC-TOFMS chromatogram of surrogate fuels, (a) K1 and (b) K2, overlaid on RP-3 composition (in gray)
observed in the figure, K1 provides a relatively poorer match to
the target fuel at the two ends of the distillation curve, being too
light at the heavy of the distillation curve and too heavy at the
light end. Owing to the inclusion of NPED and TRIC, K2 facilitates
the match on both ends of the distillation curve, and the biggest
difference of the distillation points within distillation range be-
tween RP-3 and K1 is found to be 16 °C at the end distillate point,
whereas the biggest difference for K2 is observed to be 9.8 °C at
95 vol% point.
3.3. Temperature-dependent properties
The experimental results of temperature-dependent properties
for target fuel and surrogates, as well as the estimated values
Z. Wu, Y. Mao and M. Raza et al. / Combustion and Flame 208 (2019) 388–401 397
Fig. 8. The hydrocarbon class composition in mass fraction of surrogates vs. RP-3
for surrogates are displayed in Fig. 10 . Among the three proper-
ties provided, density is chosen as a target property whereas vis-
cosity and surface tension are non-target properties. The densi-
ties and viscosities in the temperature range from about 273 to
353 K are shown, while surface tension at 25 °C is only provided
due to the limitations of test conditions. The uncertainties shown
in the above measurements of these temperature-dependent prop-
erties were less than 1%. The estimation of viscosity applies the
Kendall Monroe method [55] and the surface tension here is
roughly estimated by the Dalton-type mass-average equation [56] .
As shown in Fig. 10 (a), the average differences of measured
density are 1.07 % between RP-3 and K1 and 3.7% between RP-3 and
K2, though the figure for K2 is a little larger, it still falls in an ac-
ceptable range. The larger difference of density shown in K2 is be-
cause of the higher content of alkanes (n-alkanes and iso-alkanes)
which show relatively lower densities indicated in Fig. 4 (a), while
the fraction of alkanes (about 68.7 wt%) contained in surrogate K2
matches better with the alkane content (about 67.8 wt%) of RP-3.
For viscosity, the measured values of K1 show an average deviation
of 22% from the target fuel, and K2 shows an average deviation of
15%. The average differences of measured surface tension are 7%
between RP-3 and K1 and 10% between RP-3 and K2. Compared to
the target properties, the non-target properties of surrogates show
larger deviations. But comparing to the results in Table 5 in refer-
ence [11] , which summarized the matching results of the currently
available jet fuel surrogates on the properties of viscosity and sur-
face tension, the deviations shown on these two properties in this
study still fall in a