Combustion and Flame 246 (2022) 112397

Contents lists available at ScienceDirect

Combustion and Flame

journal homepage: www.elsevier.com/locate/combustflame

A molecular dynamics study of evaporation mode transition of

hydrocarbon fuels under supercritical conditions
Yifei Gong a, Xiao Ma a,∗, Kai Hong Luo b,c,∗, Hongming Xu a,d, Shijin Shuai a
a
State Key Laboratory of Automotive Safety and Energy, Tsinghua University, Beijing 100084, China
b
Center for Combustion Energy, Department of Energy and Power Engineering, Key Laboratory for Thermal Science and Power Engineering of Ministry of
Education, Tsinghua University, Beijing 100084, China
c
Department of Mechanical Engineering, University College London, Torrington Place, London WC1E 7JE, UK
d
Department of Mechanical Engineering, University of Birmingham, Birmingham B15 2TT, UK

a r t i c l e i n f o a b s t r a c t

Article history: The mode transition of evaporation for single- and multi-component hydrocarbon fuels is investigated
Received 20 April 2022 at the molecular level. This study scrutinizes first the subcritical and supercritical evaporation of n-
Revised 5 September 2022
hexadecane droplets and liquid films by molecular dynamics (MD) simulations. The mode regime map
Accepted 22 September 2022
of n-hexadecane droplets is obtained. Then the mode transition of evaporation of a three-component
droplet and a six-component droplet is studied. A critical dimensionless number τ 0.9P of 0.5 based on the
Keywords: average displacement increment (ADI) of fuel atoms is used to identify the evaporation mode transition
Hydrocarbon fuels of fuels with any type and number of components. It is found that in the diffusion mode of evapora-
Supercritical conditions tion, the entropy becomes the dominant factor in the evaporation of fuels, and the disorder of the fuel
Mode transition of evaporation
molecules increases significantly compared with that in the classic evaporation mode. Compared with the
Molecular dynamics
case of the quiescent droplet, with increasing relative velocity between the droplet and the ambient gas,
Size effect
Relative velocity effect the mode transition becomes easier, although this is a non-linear process. Fuel droplets and liquid films
with different initial sizes are investigated to understand the size effect. In addition, for the same ambi-
ent temperature and pressure, the smaller the normalized specific heat transfer surface area of the fuel
is, the easier the mode transition of evaporation is. A correlation was proposed to compare the possibility
of mode transition of evaporation for single- and multi-component fuels.
© 2022 The Author(s). Published by Elsevier Inc. on behalf of The Combustion Institute.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)

1. Introduction conditions, and the surface tension plays an important role. But
in the supercritical evaporation, the vapor-liquid interface almost
Combustion of hydrocarbon fuels under ever high temperature disappears. Several reviews have discussed the theory and mod-
and ever high pressure conditions is an important measure to im- eling of supercritical droplets and jets [3–6]. According to these,
prove the efficiency of power devices such as liquid rocket engines there are still unanswered key questions about the physics of the
and internal combustion engines. Such temperature and pressure mode transition of evaporation for liquid fuels. CFD simulations
often exceed the critical point of the injected fuel so that supercrit- have been extensively attempted to answer these questions [8,9].
ical combustion takes place in the combustion chamber [1]. Due However, the dramatic changes in physical properties of the
to the drastic heat and mass transfer, the evaporation mode of fu- fluids near its thermodynamic critical point may invalidate the
els may undergo a transition from classical evaporation to diffusive conventional CFD approaches [10], especially for complex evapo-
mixing via the transitional mixing [2]. Therefore, understanding ration systems with multi-fuel components [11]. In particular, the
the evaporation mode transition of fuels and its effect on the sub- basic assumption of vapor-liquid equilibrium (VLE) adopted by
sequent air-fuel mixing and combustion processes has motivated conventional CFD has been shown to be invalid under supercritical
many research efforts [3–6]. and multicomponent conditions [12]. Similar issues have been
Subcritical and supercritical evaporation has significant differ- encountered in the theoretical analysis of supercritical evapora-
ences [7]. There is a sharp vapor-liquid interface under subcritical tion. Dahms et al. [13] applied the linear gradient theory and
proposed a Knudsen-number (Kn) criterion to delineate the spray
regime diagram. When Kn is less than 0.1, the mixing of fuels was
∗
Corresponding author. thought to be in the continuum regime. However, such theoretical
E-mail addresses: [email protected] (X. Ma), [email protected] (K.H. Luo). analyses are usually based on the steady-state hypothesis and

https://doi.org/10.1016/j.combustflame.2022.112397
0010-2180/© 2022 The Author(s). Published by Elsevier Inc. on behalf of The Combustion Institute. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/)
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

fail to account for the unsteady nature of mode transition of CFD techniques [24,27]. As reported [21,27,28], the fluid property
evaporation [14]. Due to the limitations of experimental devices data obtained in MD simulations can provide a reliable benchmark
[2], there are limited experimental data available for the validation for CFD to employ the correct equation of state (EoS) and mixing
of models [2,15,16]. In particular, limited by the intrinsic optical rules of components under subcritical and supercritical conditions.
effects under high-pressure conditions, the response of imaging In addition, the knowledge of droplet evaporation will be expanded
system, and image processing techniques for moving objects [2], by such MD simulations, which is valuable for modern combustion
it is hard to experimentally investigate the dynamic details in the devices [17]. In summary, MD is valuable for further understanding
mode transition from evaporation to diffusion. In recent years, MD the physics of evaporation [17,20,24,26], examining assumptions in
simulations have started to make an impact in this field [17,18]. conventional CFD [14] and providing benchmarks for CFD models
MD is based on Newton’s second law and does not require other [27,28].
thermodynamic assumptions other than the potential function, and Although there have been progresses in understanding mode
the simulation results are statistically treated [19]. MD has unique transition of evaporation using MD [10,14,29], the effect of the rel-
advantages for the study of interfacial transition [20], considering ative velocity between the droplet (or liquid film) and the ambi-
that the thickness of the vapor-liquid interface is usually only a ent gas on the mode transition of evaporation under supercritical
few nanometers [13], which is hard to resolve based on current ex- conditions has not been sufficiently studied. In fact, in a real en-
perimental techniques. Moreover, MD does not need to make any gine combustion chamber, the relative velocity between the fuel
assumptions about the interface, and therefore can easily handle and the ambient gas is very large [16]. In optical experiments, this
the disappearance of the nanoscale interface, which is beyond the velocity effect is hard to investigate because of the difficulties in
conventional CFD [20]. Therefore, MD is ideally suited for investi- tracking the boundary of moving droplets under supercritical con-
gating the physical nature of the evaporation and mode transition. ditions, although there are some useful observations about the at-
It also should be noted that MD has two noticeable limitations. omization regimes in subcritical environments [2]. Based on the
Firstly, MD simulations are computationally expensive and require theoretical analysis of the time-scale of evaporation and breakup,
supercomputing resources, so it is not yet possible to directly sim- Poursadegh et al. [16] found that dense mixing could occur at
ulate the evaporation of macroscopic droplets using MD [17]. Gen- lower ambient pressures and temperatures when the injected fuel
erally speaking, the spatial and temporal scales of such studies are velocity became larger. Ray et al. [30] investigated the effects of
currently limited to a few hundred nanometers and nanoseconds convection velocity on the evaporation of fuel droplets and found
[10], respectively, and the number of molecules studied is limited that the higher velocity gradient near the droplet interface pro-
to millions [20]. In this context, it should be stated how relevant moted the evaporation with ambient pressures ranging from 1 bar
the MD results on evaporation are to the evaporation behaviors oc- to 55 bar. Recently, Liu et al. [29] studied the effect of forced
curring in real combustion chambers [21]. In fact, researchers have convection on the preferential evaporation of light components in
been trying to answer this question since MD was applied to evap- multi-component fuel droplets in CFD simulations and found that
oration studies, including this work. Following the MD work of Lit- there were two different controlling regimes at different convective
tle [22], Kaltz [23] conducted the MD scaling for droplets at a mi- intensities. However, this study was focused on cases at subcriti-
cron level using the Long-Micci method and investigated the va- cal pressures, leaving unanswered questions for supercritical con-
lidity of extrapolating MD results to macroscopic regimes by scal- ditions. Given these situations, it is essential to conduct further in-
ing the intermolecular potential parameters under subcritical and vestigations of the relative velocity effects on the mode transition
supercritical conditions. Qiao et al. [14] demonstrated that MD re- of evaporation.
sults could be used to approximate the transition from subcritical As mentioned, MD can only resolve the evaporation and mode
to supercritical evaporation occurring in macroscopic experiments. transition of droplets/films at nanoscales. However, the droplets
Recently, as reported [24], atomistic simulations have revealed fun- generated by jets after atomization in the combustion chamber of
damental mechanisms that also exist in macroscopic phenomena, an actual engine are usually micron-sized. Therefore, it is neces-
which complements macroscopic experiments. Secondly, the accu- sary to investigate the size effect to bridge MD and experimental
racy and reliability of MD results depend on the interatomic po- studies. In fact, existing studies have inconsistent views on this is-
tentials used [21]. However, as the interatomic potentials are deter- sue. Qiao et al. [14] investigated the size effect of liquid films and
mined by the fundamental interactions between atoms/molecules, found that the results of nanoscale films were size-independent.
there are far less empirical assumptions and uncertainties in inter- They attributed this to interfacial dynamics and thought that the
atomic potentials than in macroscopic models. Thus, the MD sim- interfacial elements were responsible for the phase transition [14].
ulation results are physics-based rather than a priori [21]. Several But in another MD research, Xiao et al. [10] studied the supercrit-
potential models have been successfully developed for complex hy- ical evaporation of n-dodecane droplets with different initial di-
drocarbon molecules [14], which can be divided into two types: ameters (from 21 nm to 43 nm), and found that when the initial
the All-Atom Model (AAM) and the United Atom Model (UAM). droplet size increased, the supercritical transition could occur at
As reported [25], the computational time for MD simulations is lower ambient pressures. Recently similar conclusions were con-
roughly proportional to the square of the total number of inter- ducted by Wang et al. [31]. This was attributed to the fact that
action sites. The AAM regards each atom as one site and is suit- larger droplets had relatively longer evaporation lifetime to achieve
able for systems with a small number of molecules and a higher the mode transition. Clearly, this topic needs further investigations.
requirement for computational accuracy [25]. On the other hand, Actual fuels are usually complex mixtures with hundreds of
the UAM’s treatment of each pair of CH3 and CH2 as one site, components [32]. However, most existing MD studies have only
which keeps a good balance between computational accuracy and considered single-component fuels [20] and two-component fu-
efficiency [10,25,26], is widely used for large molecule systems, els [12,33-35]. Chakraborty et al. [34] studied the evaporation of a
such as hydrocarbons. Jorgensen et al. [25] compared the simula- two-component fuel liquid film in supercritical nitrogen, and found
tion results based on the AAM and the UAM. With suitable pa- that under supercritical conditions, light components and heavy
rameters, the AAM and the UAM could both give highly accurate ones dominated the evaporation simultaneously, which was differ-
descriptions of fluids [25]. Despite these limitations, the physics ent from the preferential evaporation under subcritical conditions.
obtained in MD simulations will be helpful for understanding the Zhang et al. [12] studied the evaporation of a two-component liq-
mechanism of mode transition of evaporation at the atomic level uid film composed of n-heptane and n-dodecane in subcritical and
[10,20,24,26]. MD is a complement to present experimental and supercritical conditions. They concluded that compared with the

2
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

mixing ratio of fuel components, the ambient temperature and Table 1

Potential energy parameters of molecules of ambient gasses [10,42,43].
pressure had greater impacts on the mode transition of evapora-
tion [12]. In a rare example, Gong et al. [20,26] studied molecular atom ε /kB [K] σ [Å] q [e]
interactions between individual components in multi-component N(in N2 ) 36.42 3.32 0
fuel droplets and conducted Voronoi tessellation analyses for fu- C(in CO2 ) 28.129 2.757 +0.6512
els. However, many key questions remain unanswered concerning O(in CO2 ) 80.507 3.033 −0.3256
the evaporation behaviors of multi-component fuels. O(in H2 O) 78.18 3.166 −0.8476
H(in H2 O) 0 0 +0.4238
Motivated by the above analysis, this study scrutinizes the
evaporation of n-hexadecane droplets and liquid films in subcrit-
ical and supercritical conditions by MD simulations. The objectives 
of this work are to study the mode transition mechanism of evap- the standard Lorentz-Berthelot combining rules [10], εi j = εii · ε j j
oration for hydrocarbon fuels and propose a criterion to quantita- and σi j = (σii + σ j j )/2. Aromatic molecules were treated as rigid
tively identify the mode transition of evaporation to fill the gaps bodies [20]. In addition, the Coulombic pairwise interaction be-
of the CFD modeling and macroscopic experiments. In Section 2, tween charged molecules was described by:
the research methods are described. The results and discussion are   1 qi q j
presented in Section 3. The evaporation mode transition of a three- U Coul ri j = (2)
4π ε0 ri j
component droplet and a six-component droplet is studied in de-
tail. A critical dimensionless number τ 0.9P , based on the average where U Coul represents the Coulombic pairwise interaction, ε0 de-
displacement increment (ADI) of fuel atoms, is proposed to iden- notes the permittivity of free space. qi and q j denote the charges
tify the mode transition of evaporation of fuels. Based on this cri- on the interactional charged atoms. Referring to previous studies
terion, the mode regime map of n-hexadecane droplets is obtained. [20], the coulomb interactions of fuel molecules and linear di-
The mixing characteristics of fuels in different evaporation modes atomic molecules were ignored in the simulation. The potential
are discussed. Relative velocity effects are also investigated. Fuel energy parameters of molecules of ambient gasses are listed in
droplets and liquid films with varied initial sizes are simulated by Table 1. kB is the Boltzmann constant. Based on the SHAKE algo-
MD to understand the size effect. Finally, conclusions are drawn in rithm, the bond length of nitrogen molecules is fixed as 1.106 Å
Section 4, with a focus on multi-component effects. [10]. The bond lengths of CO2 (lC – O ) and H2 O (lH – O ) are 1.149 Å
[42] and 1 Å [43], respectively. The bond angles of CO2 and H2 O
2. Research methods are 180° and 109.47° [43], respectively.
The experiments by Crua et al. [2] are representative work in
2.1. Interatomic potentials the field of supercritical evaporation of fuels. The fuel in their ex-
periments evaporated in the combustion exhaust gas [2]. However,
All MD simulations in this paper were performed using nitrogen is the most common background gas for fuel evaporations
the Large-scale Atomic/Molecular Massively Parallel Simulator in MD simulations, including this study. In order to investigate the
(LAMMPS). Potential functions are the basis of MD simulations difference of evaporation of fuels in the nitrogen and the combus-
and describe all intermolecular interactions [34]. As mentioned, tion exhaust gas, the evaporation of n-hexadecane in the combus-
the UAM’s treatment of each pair of CH3 and CH2 as one site, tion exhaust gas was also studied. Referring to the experiments by
which keeps a good balance between computational accuracy and Crua et al. [2], the composition of the combustion exhaust gas was
efficiency [10,25,26], is widely used for large molecule systems, determined by the complete combustion of a mixture consisting of
such as hydrocarbons. According to previous reports [36,37], com- acetylene (C2 H2 ), oxygen (O2 ) and nitrogen (N2 ) in certain ratios:
pared with other UAMs [12,38-40], such as the Optimized Poten-
tials for Liquid Simulations United Atom (OPLS-UA) [12] model, the C2 H2 +2.5O2 +2.5XN2 = 2CO2 +H2 O + 2.5XN2 (3)
Transferable Potentials for Phase Equilibria United Atom (TraPPE- where X is the dilution ratio. For convenience, nitrogen and com-
UA) model is more suitable for the simulation of long-chain bustion exhaust gas were marked as “A1” and “A2”, respectively.
alkane molecules and has been widely used in previous studies When X = 8 in Eq. (3), the composition of A2 was determined
[10,20,24,26,41]. This study focuses on fuel molecules with longer by the exhaust gas produced by the combustion of C2 H2 . Thus, A2
chain lengths, so the TraPPE-UA model was chosen for simulations was composed of 87 mol% nitrogen, 8.7 mol% carbon dioxide, and
here. The 12–6 Lennard-Jones (LJ) potential was chosen to describe 4.3 mol% water, similar to the composition of the combustion ex-
the non-bonded interaction between pseudo-atoms separated by haust gas in the experiment [2] (the mole fractions of the three
more than three bonds or belonging to two different molecules: gasses were 89.7 mol%, 6.5 mol%, and 3.8 mol%, respectively).
 12  6  In an earlier publication of the authors [26], the potential en-
  σi j σi j ergy function of n-hexadecane was verified by comparisons of
U LJ
ri j = 4εi j − (1)
ri j ri j VLE data and evaporation rate constants with experiments [44,45].
Moreover, the present potential models for multi-component fuels
where U LJ denotes the non-bonded LJ interaction potential, ri j is have also been validated by the distillation calculation and evapo-
the distance between the two interacting pseudo-atoms, εi j is ration curves of fuels [20,26]. Results [20,26] have been shown to
the energy parameter and σi j is the size parameter. The three- be in good agreement with experimental data [46,47].
component fuel investigated [26] consists of 9.24 mol% isooc-
tane, 30.21 mol% n-dodecane and 60.55 mol% n-hexadecane. 2.2. Simulation configurations
The six-component fuel investigated [20] consists of 13.16 mol%
toluene, 13.81 mol% n-decane, 22.30 mol% n-dodecane, 24.60 In this study, twelve initial configurations of fuel vaporiza-
mol% n-tetradecane, 14.66 mol% n-hexadecane and 11.47 mol% n- tion were constructed, including ten initial configurations for n-
octadecane. It is worth mentioning that all of these fuels are sur- hexadecane, one for the three-component droplet and one for
rogate fuels for diesel [20,26]. All the LJ parameters, the bond- the six-component droplet. The initial configuration of the n-
stretching, bond bending and bond torsion parameters for alka- hexadecane evaporation system is shown in Fig. 1, which includes
nes simulated here can be found in authors’ previous publications droplets and films. Spray combustion is widely used in high-
[20,26]. The LJ parameters for all unlike atoms were determined by pressure combustion devices such as diesel engines and rocket en-

3
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

Fig. 1. Initial configurations of the n-hexadecane droplet/film: (a) Quiescent n-hexadecane droplet (for case M), (b) Non-quiescent n-hexadecane droplet (for case M) and (c)
n-Hexadecane film (for case F2). Purple particles indicate n-hexadecane molecules and yellow particles indicate nitrogen molecules. Molecules of ambient gasses surround
the fuel, which are not shown here in (a) and (b).

gines [10,17,45]. Many researchers in this field have developed fun- as the droplet. The evaporation lifetime was defined based on the
damental theories about fuel droplet evaporation [3–6], treating it temporal variation of the square diameter (D2 ) of the droplet [10].
as a single entity or contained in the spray [17]. It is of great sig- To avoid the overlap of molecules, the nitrogen molecules which
nificance to study the mode transition of evaporation of droplets initially were located in the same region with the droplet in the
in such combustion applications [10]. However, studying the con- simulation box were deleted. The initial temperature of the fuel
figuration of liquid films is also necessary for research in this field. for all cases was set to 363 K, which was near that of the fuel
In the MD study by Qiao et al. [14], the evaporation of the liq- in real engines before injection [10]. The size of the simulation
uid film is size-independent. The configuration of the liquid film domain was 80 nm × 80 nm × 80 nm. Periodic boundary condi-
seems to be more suitable for studying the dynamics of the gas- tions were applied in all three dimensions. The simulations for fuel
liquid interface [12,48]. Besides, the configuration of the liquid film evaporation were conducted using the micro-canonical ensemble
is also widely used to verify intermolecular potentials in MD by (NVE, which means constant atom number N, constant volume V,
comparing VLE data obtained with experiments [12,14,26]. Com- and constant energy E). The region outside of the sphere with a di-
bining studies of droplets and liquid films will be helpful for an- ameter of 79.8 nm was named “thermostat region”, where molec-
swering the size effect of the transition of evaporation modes and ular velocities were rescaled every time step using a speed reset
bridging MD to practical combustion applications. method [20]. A fuel molecule would be deleted when it reached
In the quiescent case for a droplet shown in Fig. 1a, a single the thermostat region, for simulating the evaporation of the fuel
suspended n-hexadecane droplet was located in the center of a cu- taking place in an infinite space [10].
bic simulation box, with a nitrogen ambient surrounding it. Before Figure 1b shows the configuration of the non-quiescent case.
the simulations of fuel evaporation, the fuel droplet and the nitro- The initializations of quiescent and non-quiescent cases were the
gen ambient were simulated separately to respectively reach their same except few details. To study relative velocity effects, in the
own thermodynamic equilibrium state using the canonical ensem- non-quiescent configuration, the center of mass of the droplet
ble (NVT, which means constant atom number N, constant volume was set to move along the x-direction at a constant speed. How-
V, and constant temperature T). After that, the two were combined ever, the ambient gas had no macroscopic translational speed. Cor-
together for the simulations of fuel evaporation. The initial diam- respondingly, the thermostat region was different. As shown in
eter for the n-hexadecane droplet was 26.0 nm. The boundary of Fig. 1b, the region for evaporation was a rectangular block with
the droplet or the film was defined as the contour surface where a size of 80 nm × 57.68 nm × 57.68 nm to keep the volume of
the fluid density was equal to the average of the maximum and thermostat region in this configuration the same as that in the qui-
the minimum densities of the evaporation system. The diameter of escent case. Referring to the experiments by Crua et al. [2], three
the droplet was defined as that of a sphere of the same volume initial relative velocities of droplets were investigated, which were

4
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

Table 2
Simulation details of n-hexadecane droplet/films.

Name S M L G F1 F2 F3

Initial number of molecules in the fuel 9950 19,896 31,998 427,626 8878 13,319 21,075
Initial fuel size, nm 20.4 26.0 30.5 73.0 16.9 25.1 40.0
Simulation box size, nm3 603 803 1003 2203 152 ×200 152 ×200 152 ×200
Thermostat region diameter (V0), nm 59.8 79.8 99.8 219.8 – – –
Ambient temperature, K 750–1500 750–3600 750–1500 750–1500 750–1500 750–1500 750–1500
Ambient pressure, MPa 4–16 4–36 4–16 4–16 4–16 4–16 4–16
Ambient gas (Relative velocity) A1(V0) A1(V0-V2); A2(V0) A1(V0) A1(V0) A1(V0) A1(V0) A1(V0)

0 m/s, 30 m/s and 100 m/s, labeled “V0”,“V1” and “V2” respectively sults can be directly mapped onto macroscopic systems. More dis-
for convenience. It should be noted that, in the simulations here, cussions about the relevance of MD results on evaporation to those
referring to the previous study [23,24,49], the droplet is given de- in real combustion chambers, and the advantages and limitations
sired relative velocity initially. The ambient gas is stationary, that of MD for investigating mode transition of evaporation could be
is, it has no initial bulk velocity. The relative velocity between the found in the Introduction section.
droplet and the ambient gas will decrease slowly [49]. Referring The temperature and pressure for the various cases are initial
to previous MD studies on moving droplets [24,49,50], the evap- values. As mentioned, the simulations for fuel evaporation were
oration simulations here also use the NVE ensemble in the non- performed using the NVE ensemble. The velocities of the molecules
quiescent case. The kinetic energy of the bulk flow is also included located in the thermostat region were rescaled every time step
in the energy term. [20] according to
Figure 1c shows the initial evaporation configuration of the  0.5
liquid film. The initialization of this configuration was similar to 3T Nt kB
Vinew = Viold (4)
that of the droplet in Fig. 1a. The size of the cuboid simula- 2Ek
tion box was 15 nm × 200 nm × 15 nm. The center of mass
where Ek indicates the total kinetic energy of the Nt atoms in the
of the n-hexadecane liquid film was located in the center of the
thermostat region, and T indicates the target ambient temperature.
simulation domain, and the liquid film was surrounded by nitro-
As a consequence, the temperature of this region could be kept at
gen molecules. Two separated thermostat regions with a size of
a constant target value [10]. The calculations of temperature in MD
15 nm × 40 nm × 15 nm were located in two sides of the simula-
simulations are based on the average kinetic energy of molecules
tion box in the y direction. Similarly, fuel molecules would be re-
[34]:
moved when they reached the thermostat region. Periodic bound-
ary conditions were used in all three directions. As a result, the 3 1N

current configuration could simulate the evaporation of an infinite- kB T N = mi v2i (5)

2 2
area fuel film in an open space [41]. Different initial configuration i=1

settings, such as the size of a simulation domain and the thermo- where N is the number of atoms in the target atom group, mi
stat setting, were tested before the formal simulations presented. and vi are respectively the mass and velocity of atom i. i indi-
Five replica runs were performed to reduce or eliminate statistical cates the identifier of the atoms. In the case with relative veloc-
errors. For n-hexadecane, seven initial configurations of droplets or ities, the temperature of a group of atoms is calculated after sub-
liquid films with different initial sizes were constructed to study tracting out the center-of-mass velocity of the group caused by the
the size effect of MD, as detailed in Table 2. It is worth men- bulk movement, temperature only depends on the random portion
tioning that the initial size for droplets is the diameter and that of the molecular velocities, that is molecular thermal velocity [23].
for liquid films is the thickness. It should be noted that MD is After the center-of-mass velocity has been subtracted from each
a first-principle method that describes the fundamental interac- atom, temperature is calculated by Eq. (5). The number of atoms
tions between atoms/molecules, without any assumptions about in the simulation domain varies with time due to evaporation, and
the system’s macroscopic behaviours. Therefore, the pros of MD the number of atoms N and their associated degrees of freedom in
simulations are that MD results (1) are more reliable than con- the calculation group are recalculated each time the temperature is
ventional CFD (with prior assumptions of behaviours such as sub- computed to ensure that the temperature is properly normalized.
critical or supercritical evaporation), and (2) can reveal dynamic The removal of the center-of-mass velocity is essentially comput-
evaporation behaviours and provide quantitative information that ing the temperature after a “bias” has been removed from the ve-
are difficult to obtain by experiments. The cons of MD simulations locity of the atoms. Moreover, in the thermostat region, this bias
are the small droplet sizes attainable. It is thus important to inves- will be subtracted from each atom, thermostatting of the remain-
tigate what evaporation properties are size-dependent and what ing thermal velocity will be performed, and the bias will be added
properties are size-independent. And if some properties are size- back in to keep a constant temperature value. The initial ambient
dependent, can scaling laws be obtained to bridge the size dif- pressure was determined by a combination of the initial ambient
ferences? This is an ongoing investigation but some conclusions temperature and the initial number of nitrogen molecules in the
can already be drawn. For example, the regression rates [22,23] simulation box in NVT simulations. However, the ambient pressure
and squares of droplet diameters [10,26] are size-dependent, but was determined by choosing a suitable initial number of nitrogen
with the method of scaling, MD results can achieve satisfactory molecules in the system in NVE simulations [10,20]. The number
agreement with experimental values. More comparisons of distil- of nitrogen molecules ranges from tens of thousands to millions
lation curves and evaporation rate constants from MD simulations in different cases, which depends on the ambient condition and
with experiments could be found in [10,26]. As reported [10,26], simulation domain size [10,20]. As mentioned, a fuel molecule will
although the evaporation rate is size-dependent [22,23], the evap- be removed when it gets to the thermostat region to simulate the
oration rate constant obtained by MD simulations can agree with evaporation of the droplet taking place in an infinite space [10]. It
the macroscopic experiments [45] under subcritical conditions. is worth mentioning that the droplet accounts for less than 4% of
Fortunately, some properties are size-independent [17], and MD re- the volume of the simulation box, so the fluctuations of ambient

5
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

was obtained via dividing the mixing time by the evaporation life-
time. The equations of AD and ADI are as follows:

k = m 
  k=1 (Xk (t n ) − 0 )2 + (Yk (t n ) − 0 )2 + (Zk (t n ) − 0 )2
AD t n =
m
  AD(t n )
ADR0 t n =
R0
     
ADI t n = AD t n − AD t n−1
  ADI(t n )
ADIR0 t n = (6)
R0

where t n represents the present moment and t n−1 represents

the previous moment. The sampling time interval (t n − t n−1 ) is
0.05 ns. k represents the identifier of the fuel atoms and m rep-
resents the maximum identifier of the fuel atoms (the total num-
ber of fuel atoms) remaining in the simulation domain at time t n .
Fig. 2. Initial configurations of multicomponent fuel droplets (front view): (a) The AD(t n ) represents AD at time t n , and the coordinate of displace-
three-component droplet and (b) The six-component droplet. Nitrogen molecules ment vector of the fuel atom k at time t n is (Xk (t n ), Yk (t n ), Zk (t n )).
surround the droplet, which are not shown here. As stated in Eq. (6), for fuel atom k, the reference position for cal-
culating the displacement magnitude here is its initial position be-
fore the evaporation. In other words, the initial coordinate of the
pressure could be negligible [10,20,26]. For the case of liquid films, displacement vector of fuel atom k is (0, 0, 0). ADI(t n ) represents
the conditions are similar [41]. ADI at time t n . R0 is the initial droplet radius. ADR0 (t n ) represents
Figure 2 shows the initial configurations of multicomponent the dimensionless AD at time t n . ADIR0 (t n ) represents the dimen-
droplets. The detailed size parameters, such as the size of ther- sionless ADI at time t n .
mostat regions, have been marked on Fig. 2. The initialization of Figure 3 reveals the AD and ADI profiles of fuel atoms under
configurations of multi-component droplets was the same as that different ambient conditions. As shown in Fig. 3a, at low ambi-
of the single-component droplet in Fig. 1a. The critical pressure ent pressures and temperatures (8 MPa@750 K, case1), the AD per
and temperature of the fuel have important effects on its evapo- fuel atom gradually increased with time, and the time-dependent
ration. For n-hexadecane, the three-component fuel and the six- profile exhibited a concave shape, meaning that the AD increased
component fuel studied here, the estimated critical temperatures faster and faster with time. Correspondingly, in Fig. 3b, the ADI per
are 1.4 MPa, 1.89 MPa and 2.45 MPa, and the estimated critical fuel atom gradually increased with time until near the end of the
temperatures are 722 K, 697 K and 686 K, respectively [20,26]. Tak- mixing process. However, at high pressures and temperatures (24
ing future operating conditions of rocket engines into account [14], MPa@1350 K, case2), the time-dependent profile of AD was con-
the simulated ambient pressure and ambient temperature for fu- cave initially and then convex, meaning that the profile of growth
els were in the range of 4–36 MPa and 750–3600 K, respectively, rate of AD with time was unimodal, which was proved in Fig. 3b.
which were much higher than those investigated in similar reports Obviously, two different mechanisms dominated evaporation in the
[10,12,14,20,26]. Moreover, evaporation of droplets [10,17,20,26,31] case1 and case2 respectively.
or liquid films [12,14,34] was investigated in isolation in previous In case1, the attractive forces of fuel molecules were always
similar studies. In fact, those studies have not yet reached a uni- strong and only a small amount of ambient gas dissolved into
fied conclusion on the size effect [10,14,31], leaving unanswered the droplets via the droplet surface [20]. For the fuel, the vapor
questions on the mode transition of evaporation. This paper stud- and the liquid phases coexisted, the interaction energy of the fuel
ies droplets and liquid films simultaneously, and proposes a size molecules was dominant in this case, and the distribution of the
factor in Section 3.4 to unify the findings of the two configura- fuel molecules was ordered. With continuous gaseous heat and
tions, which provides new atomic-level insights into the size effect mass transfer, the fuel was transformed from liquid phase to gas
of evaporation and further bridges MD to actual combustion ap- phase via the sharp phase interface in evaporation mode. It is
plications. The time step was 1.0 femtosecond for “S” cases shown worth mentioning the behaviors of liquid phase dominated the
in Table 2. The time step was 2.0 femtosecond in other cases. The profile of ADI, because the liquid phase was much denser than the
total time steps for a case ranged from 10 0 0,0 0 0 to 960 0,0 0 0 de- gas phase. After that, the evaporated fuel molecules via the vapor-
termined by evaporation lifetimes. liquid interface diffused into the ambient gas.
In case 2, since it took time for heat and mass transfer, the
3. Results and discussion surface tension of the droplet still played an important role ini-
tially, the liquid fuel initially transitioned into the gas phase in
3.1. Mode transition mechanism of evaporation for hydrocarbon fuels evaporation mode, and the ADI of the fuel atoms increased with
time. However, with the dramatic transfer of heat and mass un-
It is well known that ambient temperature and pressure have a der high ambient temperatures and pressures, a large number of
significant impact on evaporation modes of fuels. This paper inves- ambient gas molecules dissolved into the fuel droplets, the fuel
tigated the profiles of the average displacement (AD) per fuel atom density decreased rapidly [20]. Meanwhile, the attractive force be-
and ADI per fuel atom with time under varied ambient conditions. tween fuel molecules was greatly reduced, the surface tension de-
The results of configuration M were discussed here as an exam- clined rapidly until it disappeared. The fuel firstly was transformed
ple. In case M, the diameter of the droplet was 26.0 nm. For com- from the liquid phase to the liquid-like phase, further to the gas-
parisons, the displacement and its increment of fuel atoms were like phase, and finally to the gas phase [31,33]. In this process, the
normalized here. The mixing time of fuels was also normalized. molecular disorder rose due to the difference in concentration of
The dimensionless AD and ADI were obtained via dividing the dis- the species and the diffusion mode was dominant. For the fuel,
placement and its increment at each moment by the radius of the the vapor and liquid phases could not coexist, the effect of entropy
droplet (13.0 nm). The dimensionless mixing time, indicated by τ , was decisive. The detailed discussion of entropy will be carried out

6
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

Fig. 3. The temporal variation of the dimensionless AD and ADI per fuel atom (for case M, A1(V0)).

based on Fig. 8. Once the vapor-liquid interface disappeared, the

mode transition of evaporation was fully completed. As the con-
centration difference of species decreased with time, the ADI of the
fuel atoms gradually decreased with time after reaching a peak.
It is crucial to achieve a quantitative distinction between the
two evaporation modes for the experiment and modeling of su-
percritical evaporation. Considering the fluctuation of MD results, a
dimensionless number τ 0.9P was used here which was the smaller
dimensionless mixing time corresponding to 90% of the peak of
the time-dependent ADI profile for fuel atoms. If τ 0.9P is less than
0.5, it is defined as the diffusion mode. If not, it is in evaporation
mode, as shown in Fig. 3b.
As mentioned, the size of the droplets studied here is only
tens of nanometers, and the corresponding evaporation lifetime
is only a few to tens of nanoseconds, which is similar to previ-
ous MD studies [10,14]. Considering scale differences, the corre-
sponding time in MD simulations is usually normalized by, for ex-
ample, the evaporation lifetime [10,12,14,24,51] and a scale factor
[17,21,22]. Kaltz [23] used the Long-Micci method to scale the time Fig. 4. Mode transition map of hexadecane droplets (for case M, A1(V0)). Here “E”
indicates experimental data from Crua et al. [2]. The transition line colored by blue
in MD simulations and investigated the possibility of extrapolating
is from Qiao et al. [14].
MD results to macroscopic systems. In this context, such timescale
analyses used are valid and suitable for MD simulations. Referring
to previous studies [10,12,14,24], the time in this paper is normal- others, as shown in Fig. 4. The experimental data in Fig. 4 were
ized by the evaporation lifetime, and the criterion τ 0.9P here is from Crua et al. [2]. Based on the morphological evolution of mi-
based on the dimensionless normalized time. croscopic droplets in optical experiments, Crua et al. [2] provided
Figure 4 illustrates the mode transition map on the P-T dia- the criteria to clarify the mixing regimes of fuels. In the experi-
gram for n-hexadecane droplets (case M) in the nitrogen ambi- ments, the droplets in the regime of classical evaporation followed
ent. As mentioned, MD simulations have inherent randomness. To the classical evaporation process, and the surface tension played a
overcome this, we have 5 replica runs with different initial con- dominant role [2]. For the droplets in the regime of transitional
figurations for each operating condition and the presented results mixing, initially they followed the classic evaporation. But then
are based on the averages of multiple runs [24]. Referring to pre- due to the rapid reduction of intermolecular forces, the apparent
vious MD work [24], for each presented case, the repeatability of deformation of the droplet was observed, and the mixing of the
results is good and the statistical deviations are small. For ex- droplet was accelerated [2]. As for the regime of diffusion mixing,
ample, for the case of evaporation under 16 MPa and 900 K in the evaporation of droplets initially was controlled by the surface
A1, τ0.9P = 0.49 ± 0.026 based on 5 runs. A fitted transition line of tension. But soon, the droplets lost spherical shapes, and the clas-
τ 0.9P = 0.5 (indicated by the red solid line) was obtained based sical two-phase flow transitioned to the single-phase mixing [2].
on simulation points (T: 750–1700 K, P: 6–20 MPa), and the corre- The regression models were obtained to distinguish the different
sponding regression model using the least squares method was as mixing regimes [2] based on the experimental data using the least
follows: squares method, as shown in Fig. 4. The transition line colored by
P = 130687.34T −1.354 (7) blue was redrawn based on the MD simulations by Qiao et al. [14].
In their study [14], based on the critical mixing point from the VLE
R2 for this fitting model was 0.9708. As the ambient tempera- theory, they proposed a threshold dimensionless transition time of
ture or pressure increases, τ 0.9P gradually decreases, and the dom- 0.35 to separate the subcritical and supercritical regimes, which
inant mode gradually transitions from classic evaporation to dif- was also plotted in Fig. 4.
fusion. Therefore, the operating conditions above the transition As shown in Fig. 4, the transition line with τ 0.9P = 0.5 was gen-
line are dominated by diffusion, and those below are dominated erally in good agreement with the diffusion mixing line obtained
by evaporation. The results here were compared with those from by experiments [2]. The following discussion focused on the dif-

7
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

ferences in the results of several methods. From Fig. 4, the tran- et al. varied from 0.05 (1/20) to 0.1 (1/10), which was smaller than
sition line obtained here can satisfactorily distinguish the exper- that of the droplet studied here (0.231). Nonetheless, Qiao et al.
imental data [2] between transitional mixing and diffusive mix- [14] concluded that the transition line based on liquid films was
ing, with acceptable differences. Specifically, the transition line for size-independent.
diffusion here was slightly higher than that obtained by the ex- In fact, previous studies have proposed the criteria for tran-
periment [2]. In other words, the minimum ambient temperature sition from evaporation to diffusion for liquid fuels. The critical
for transition at constant ambient pressure was slightly higher or locus criterion has been widely used, relying on the calculation
the minimum ambient pressure for transition at constant ambi- of the critical point of a mixture composed of fuels and ambi-
ent temperature was slightly higher in this study. This consistency ent gasses, whose base is the VLE theory [54]. Moreover, this ap-
indicates that although there are huge scale differences between proach relies on numerous thermodynamic assumptions and es-
experiments and MD simulations, atomistic simulations can reveal timations, and fails to resolve complex evaporation systems with
fundamental mechanisms that also exist in macroscopic phenom- multi-components, because the critical points of them are hard to
ena and complement macroscopic experiments, which is consistent be available [11]. As mentioned, the Knudsen number criterion by
with the results on multi-component fuels [24]. The possible rea- Dahms et al. [13] is also proposed to distinguish the mixing regime
sons were as follows. First, the following discussion of the effect of fuels. Nevertheless, the transition from classic evaporation to
of relative velocity (in Fig. 9) showed that when a moving droplet diffusive mixing is assumed to be in steady state in this theory,
evaporated in the ambient gas, compared to the quiescent case failing to resolve the dynamics of mode transition under super-
considered here, the minimum ambient temperature and pressure critical conditions [14]. Qiao et al. [14,34] have tried to indicate
for transition became smaller. The theoretical analyses by Pour- the subcritical-to-supercritical transitions based on the calculation
sadegh et al. [16] led to similar conclusions. Second, the transition of surface tension in MD simulations. However, it has been hard
line shown in Fig. 4 was obtained by the evaporation of quiescent and even impossible to achieve reliable calculation of surface ten-
n-hexadecane droplets with a diameter of 26.0 nm in the nitrogen sion under supercritical conditions [55]. In summary, a reliable and
(for case M, A1(V0)). In optical experiments [2], however, much versatile criterion is urgently needed for supercritical and complex
larger micron-sized droplets were observed. According to the dis- multi-component conditions to distinguish the evaporation mode
cussion of the size effect below (in Fig. 10), with larger size of the of liquid fuels under certain conditions. The criterion of τ 0.9P here
droplet, SSAH decreased, and the minimum ambient temperature is a new attempt to achieve this goal. It should be noted that this
and ambient pressure for the transition tended to decrease. The criterion is based on the profile of ADI of fuel atoms, which is
studies by Xiao et al. [10] and Wang et al. [31] also provided sup- available for the fuel-ambient gas mixing system composed of any
port for this conclusion. Third, the ambient gas was the nitrogen type and number of components, even under highly supercritical
here, while it was the combustion exhaust gas in the experiment conditions.
[2]. According to the following discussion on the multicomponent
effects (in Fig. 11), compared with the nitrogen, the transition of 3.2. Mixing characteristics of fuel droplets in different evaporation
the evaporation mode in the combustion exhaust gas was a little modes
easier, which meant a smaller minimum ambient temperature or
pressure for transition. However, judging from the difference in re- Figure 5 reveals typical snapshots of the molecular distributions
sults, based on the ADI criterion, the results for nanoscale droplets of n-hexadecane droplets in different evaporation modes. In Fig. 5a
slightly larger than here will be in better agreement with the ex- (the non-transition case), the droplet always had a distinct vapor-
perimental values, although size, relative velocity and multicompo- liquid interface and remained spherical during most of the mixing
nent effects still exist. time. However, as shown in Fig. 5b, at higher ambient tempera-
A discussion about the limitations of experimental results tures and pressures, the mode transition of evaporation occurred.
themselves is instructive. Crua et al. [2] investigated the injection Before the transition (Fig. 5b1), the droplets remained spherical
of three single-component fuels into high-temperature and high- with a sharp interface. When mode transition occurred (Fig. 5b2),
pressure ambient gasses using high-speed long-range microscopy. the spherical boundary was already difficult to recognize. After the
However, such optical experiments have intrinsic limitations [2], transition (Fig. 5b3), the droplet was deformed and the original
such as optical effects under high-pressure conditions, the re- vapor-liquid interface was replaced by an expanding mixing layer
sponse of imaging system, and spatial resolution, and it is impos- full of fuel and nitrogen molecules with densities of continuous
sible to investigate actual states of interfacial structures [52]. In changes.
addition, such qualitative observations do not provide quantitative Figure 6 reveals the evaporation histories of n-hexadecane
details such as droplet temperature profiles, and do not directly re- droplets in different evaporation modes. In evaporation mode (4
veal the physics of mode transition of evaporation [10]. Crua et al. MPa@900 K), the profile of fuel molecular number in the droplet
[2] addressed some of these limitations by using high-resolution with time was first convex and then concave. The evaporation rate
microscopy at the University of Brighton [53] and high-speed mi- of fuel molecules gradually increased at first, and then decreased
croscopy at Sandia National Laboratories [15,52]. And, three criteria continuously in the second half of evaporation. In diffusion mode
were proposed to account for optical uncertainty when classifying (16 MPa@1500 K), however, the profile of that with time generally
their experimental results [2]. With these limitations in mind, ex- was concave. The evaporation rate of fuel molecules reached the
periments can provide independent reference data for comparison maximum value soon after the start of mixing, and then decreased
with numerical simulations [2], such as the MD studies by Qiao continuously. Figure 6b shows the histories of the number of newly
et al. [14]. evaporated n-hexadecane molecules leaving from the droplet, cor-
Moreover, the transition line colored by blue from Qiao et al. responding to the evaporation rate defined by the number of fuel
[14] was closer to the experimental transitional mixing line [2], molecules in Fig. 6a. As shown in Fig. 6b, in evaporation mode, the
however, the transition line obtained here was closer to the ex- maximum evaporation rate occurred in the second half of evapo-
perimental diffusion mixing line [2]. This may be due to the fol- ration. In contrast, in diffusion mode, the evaporation rate reached
lowing reasons. The transition line of Qiao et al. [14] was based on a peak soon after the start of mixing. In evaporation mode, the
the MD simulations of liquid films, while that in this paper was density gradient near the droplet interface was larger. Initially, due
based on those of droplets. According to the discussion below on to the large temperature difference between the droplet and the
the size effect (in Fig. 10), SSAH of the liquid film studied by Qiao ambient gas, the droplet temperature increased rapidly with heat

8
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

Fig. 5. Snapshots of local molecular distributions of n-hexadecane droplets (for case M, A1(V0)): (a) Evaporation (8 MPa@750 K) and (b) Diffusion (24 MPa@1350 K). Here
τ stands for the dimensionless mixing time. In Figure 5a, the mode transition does not occur. Figure 5(b1), (b2) and (b3) show the situation before transition, transition and
after transition, respectively. Ambient nitrogen molecules are not shown here.

Fig. 6. Evaporation histories of n-hexadecane droplets (for case M, A1(V0)).

transferring from the gas phase to the liquid phase via the inter-
face [10]. Meanwhile, the evaporation rate became higher, during
which the droplet experienced an initial heat-up period [26,34].
With time, the temperature difference between the droplet and
the ambient gas decreased, and the droplet became smaller. The
evaporation rate of the droplet decreased after reaching a peak.
In diffusion mode, with higher ambient pressures and tempera-
tures, a lot of nitrogen molecules dissolved into the droplet [20],
the density gradient between the liquid core and the ambient gas
decreased [14], and the evaporation rate peaked rapidly and then
decreased. Meanwhile, the evaporation rate was determined by the
combined effect of diffusion coefficient, mass fraction gradient and
temperature gradient [14]. The temporal variation of the square di-
ameter D2 of the n-hexadecane droplet in evaporation mode (4
MPa@900 K) is shown in Fig. 7. It can be found that the evapora-
tion of droplets follows the classical D2 law after an initial heat-up
period, which is consistent with previous studies [10,26].
The free energy, a function of internal energy, temperature and
entropy, is a crucial parameter for measuring the state of a thermo-
Fig. 7. The temporal variation of the square diameter D2 of the n-hexadecane
dynamic system. Specifically, the competition between energy and droplet. The dotted blue line indicates the fitting of D2 profile [10].

9
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

Fig. 8. Temporal variations of the sum of pair entropy and that of its increment for each atom (for case M, A1(V0)).

entropy determines the state of a thermodynamic system, while It should be mentioned that in Figs. 3,6 and 8, the concavity
the temperature determines the relative weight between them. The or convexity of the curves corresponds to different patterns in the
time-dependent profiles of the mean molecular kinetic energy, pair slope change of the curves. In fact, Figs. 3,6 and 8 illustrate the
potential energy, and intermolecular interactions for fuel evapora- physics of different evaporation modes and the mode transition
tion system like here have been presented in our earlier publica- from the perspectives of AD, evaporation rate, and pair entropy,
tion [20] and will not be discussed. The focus here is on the history which reveal different aspects of the same physical process. There-
of entropy of evaporation systems. The pair entropy fingerprint for fore, the physics revealed by these curves are consistent. It was re-
each atom was calculated to investigate the degree of disorder of ported [2,10,14] that even under very high ambient temperatures
the mixing system. The calculation method of the entropy finger- and pressures, the fuel droplet did not achieve the mode transi-
print of each atom was derived from Piaggi et al. [56] and Nettle- tion of evaporation instantly, and the surface tension did not dis-
ton et al. [57]. It should be noted that the pair entropy, which is appear immediately, because heat and mass transfer took time. As
always negative, uses kB as its unit. The lower pair entropy stands mentioned here, the competition between energy and entropy de-
for the more ordered environment. Figure 8 shows the histories termines the state of a thermodynamic system, which is also ap-
of the sum of pair entropy for each atom and that of its corre- plicable for evaporation and mode transition, so the shape of the
sponding increment. The mixture was composed of n-hexadecane curve is determined by a combination effect of energy and entropy.
and nitrogen. As shown in Fig. 8a, although there were some dif- Under this condition, the concavity and convexity of curves and
ferences, the profile of entropy change of the mixture was basically their connection point (inflection point) reflect exactly which fac-
the same as that of the alkane, which meant the entropy change of tor plays a dominant role at that time. In evaporation mode, the
the alkane was dominant. There were significant differences in the surface tension was relatively strong and the interaction energy of
profiles of entropy for evaporation and diffusion modes. In evap- the liquid fuel was dominant compared with its entropy change.
oration mode (4 MPa@900 K), the time-dependent profile of en- The order of fuel molecules in the liquid phase was more or less
tropy of alkanes nearly always was concave and eventually became kept. With continuous heat and mass transfer, the evaporation was
convex. The entropy of alkanes increased gradually with mixing, accelerated, which could be revealed by concavity of the profiles
whose growth rate rose to a peak near the end of evaporation (The slope of profiles increased). In diffusion mode, however, mas-
and then declined, which was indicated by the increment profile sive ambient gas molecules entered into the fuel droplet and fuel
in Fig. 8b. In diffusion mode (16 MPa@1500 K), however, the time- density decreased dramatically [20]. The attractions between fuel
dependent profile of entropy of alkanes nearly always was convex molecules, which was responsible for the molecular order, dramat-
except concave during a short period initially, whose growth rate ically declined. The original order of fuel molecules in the liquid
rose to a peak in the first half of evaporation and then continu- phase was lost because the entropy change caused by the mixing
ously dropped. Special attention should be drawn to the fact that became dominant. The evaporation was first accelerated and then
the physics of the entropy change for the mode transition was con- decelerated due to the reduced density gradient, which could be
sistent with that of evaporation histories shown in Fig. 6 and that revealed by the convexity of the profiles (The slope of profiles de-
of AD and ADI in Fig. 3. This consistency provides a theoretical creased).
support and a further evidence for the ADI criterion. In addition,
from Fig. 8a, the pair entropy for alkane atoms in diffusion mode 3.3. Relative velocity effects
was greater than that in evaporation mode at almost all times. In
diffusion mode, the ambient temperature and pressure were high As mentioned, effects of relative velocity on the mode transi-
enough relative to the critical point of the fuel. Meanwhile, the in- tion of evaporation of fuels have not been well understood. For this
termolecular interactions of fuel molecules declined dramatically purpose, in addition to the quiescent case (V0), we investigated the
[20], the order of molecules of liquid fuel was hardly maintained, evaporation of n-hexadecane droplets with initial relative veloci-
and the vapor-liquid interface no longer existed. The entropy which ties of 30 m/s (V1) and 100 m/s (V2) in the supercritical nitrogen.
caused molecular disorders, instead of energy which maintained Figure 9 shows the relationship between fuel temperature and mo-
molecular orders, became the dominant factor in the mixing pro- lar fraction of the fuel for different droplet relative velocities at an
cess. As a result, the disorder of the fuel molecules increased sig- ambient pressure of 8 MPa. The temperature was calculated from
nificantly compared with that in the two-phase evaporation. the average kinetic energy of the molecules [14]. The estimated

10
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

fact, regardless of droplet sizes or slip velocities, ambient temper-

ature and pressure play a dominant role in the mode transition
of evaporation [2,10,14,24]. The study here supports the argument
that slip has little effect on the mode transition of evaporation,
which is supported by the fact that the mode transition map ob-
tained by MD can agree satisfactorily with macroscopic experimen-
tal results, consistent with conclusions in [10,14,24]. On the other
hand, MD is based on Newton’s second law and does not make any
additional thermodynamic assumptions about the simulated phys-
ical process other than interatomic potentials. The simulation re-
sults are physics-based rather than a priori, which means that even
if the slip effect is present here, it will be naturally included in the
results. Given that the slip velocities and droplet sizes in this paper
are not large enough, future work is needed in this area, which is
beyond the scope of the present study.

3.4. Size effects

Fig. 9. Temperature as a function of molar fraction of n-hexadecane in the case of
different droplet relative velocities (for case M, A1).
To study the size effect, the evaporation of several fuel droplets
and liquid films with different initial sizes was investigated. Refer-
critical temperature and the critical molar fraction of fuels for the ring to previous studies on size effects of MD [10,14,31], the diam-
mixing system of n-hexadecane and nitrogen at 8 MPa were shown eters of the droplets were selected as 20.4 nm, 26.0 nm,30.5 nm
in Fig. 9, whose calculation method could be found in the authors’ and 73.0 nm, respectively. The thicknesses of the liquid films were
previous report [26]. When the profile of temperature-molar frac- 16.9 nm, 25.1 nm and 40.0 nm, respectively. With varying ambi-
tion passes through or is above the critical point of the mixing sys- ent temperatures and pressures, the dimensionless transition time
tem, supercritical evaporation of fuel can occur [14]. As shown in τ 0.9P for every case was calculated, as shown in Fig. 10a. With in-
Fig. 9, under an ambient pressure of 8 MPa, for the quiescent case, creasing ambient temperature and pressure, τ 0.9P for fuels with
when the ambient temperature ranged from 750 K to 1150 K, the different initial sizes decreased, and the mode transition of evap-
mode transition of evaporation did not occur. As the relative veloc- oration became easier. In addition, τ 0.9P of several sizes of fuels
ity increased, the temperature-molar fraction curve shifted toward had the same size relationship for investigated conditions. SSAH
the region with higher temperature and mole fraction of fuels in of several initial sizes of fuels was studied, which was defined as
Fig. 9, although this was a non-linear process, which meant the the ratio of the surface area available for heat and mass trans-
mode transition became easier. In particular, the mode transition fer to the total volume of the fuel. When it comes to the size
occurred when droplets had relative velocities of 30 m/s and even effect, the discussion is actually on the same fuel. From the MD
higher 100 m/s. However, this did not happen in the quiescent case point of view, the evaporation is derived from the movement of
under the same ambient condition. This may be due to the pro- fuel molecules, and the atomic diameter is a key parameter, which
moted gaseous heat and mass transfer [29]. This finding provided could be indicated by the size parameter σ mentioned before. For
a direct evidence for the results of Poursadegh et al. [16] by theo- fuel atoms, the size parameter σ is indicated by the weighted av-
retical analyses. Further work is needed in this area. erage of that of all the united atoms in fuel molecules. In this case,
Another discussion is about the possible slip effect. In a direct σ for the hexadecane is 3.925 nm [26]. The dimensionless size fac-
injection engine, such as a diesel engine, the droplets produced by tor SF is defined by multiplying SSAH by σ , which is shown in
the fuel spray are usually micron-sized [2]. In this condition, it is Fig. 10b. In this condition, SF can be named as a normalized spe-
appropriate to use the continuum assumption and an approxima- cific heat transfer surface area. It is well known that the evapo-
tion of no-slip condition is valid [58]. However, the evaporation of ration of fuels is strongly dependent on the heat and mass trans-
the nanoscale droplet here may conform to the slip regime based fer processes under certain conditions. Hence, there is no doubt
on the Knudsen number [59], where the no-slip condition is in- that SF defined here will play an important role in this process. As
valid. As reported [60], the slip velocity may affect heat and mass shown in Fig. 10, for the same ambient temperature and pressure,
transfer and thus the evaporation rate. Therefore, it is necessary the smaller SF of the fuel, the easier the transition. A plausible rea-
to evaluate the effect of slip velocity on MD results, especially for son for this was that the smaller SF of the fuel, the slower the
the cases with relative velocities. In contrast, the relative droplet evaporation of the fuel, the longer the evaporation lifetime, and
velocity simulated here is smaller, similar to that in dilute regions the more sufficient time for the fuel to achieve the transition. In
of sprays in diesel engines, where the droplet-gas slip velocity is the studies of droplets by Xiao et al. [10] and Wang et al. [31],
small [60]. Kaltz [23] studied the evaporation of nanoscale droplets it was believed that larger droplets were easier to transition be-
with a relative velocity of 500 m/s under a high pressure envi- cause of their longer evaporation lifetime compared to their tran-
ronment of 20 MPa, and found that the slip between nanoscale sition time. Similar conclusions have been reached in studies of
droplets and ambient gasses was almost negligible. This relative droplets from continuum-based simulations [64]. However, in the
velocity is much larger than the velocity studied in this paper, study on liquid films by Qiao et al. [14], the supercritical tran-
which means that the slip velocity in this paper is also negligible. sition was size-independent. Note that their initial configurations
Michaelides et al. [61] found that for moving nanoscale particles, a for fuel evaporation were different. In the studies of Xiao et al.
velocity slip at the interface did not significantly affect the overall [10] and Wang et al. [31], the fuel molecules would be removed
Nusselt number. Masri et al. [62] found that droplets of diameter when they reached the thermostat region, which was not the case
less than 10 μm in spray had similar velocities to the carrier flow. in the study of Qiao et al. [14]. SF proposed here can give unified
Similar findings have also been reported by Mastorakos et al. [63]. analyses for the results of droplets or films with different initial
Nomura et al. [45] mentioned that the fuel droplets in the spray sizes. Although size effects exist, as shown in Fig. 4, the results for
combustor were so fine that even in high pressure environments, medium-sized droplets (for case M) were already in good agree-
the effect of convection on droplet evaporation was negligible. In ment with the experimental data [2]. Based on the above facts, al-

11
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

Fig. 10. Size effects on the mode transition of evaporation of fuels. Here D indicates the diameter for droplets or the thickness for films.

Fig. 11. Comparison of τ 0.9P of several evaporation systems.

Fig. 12. Comparison of Tr∗ Pr.^0.5 of several single-component fuels. The experimen-
tal data were from Crua et al. [2].
though the mode transition maps obtained by droplets or liquid
films with different initial sizes were slightly different, the results
of droplets or liquid films with sizes of tens of nanometers studied
tions. When the ambient gas was nitrogen, in the order of six-
here were acceptable for practical applications. Moreover, in Ref.
component fuel, three-component fuel and single-component fuel,
[10], the droplet size effects on evaporation of single-component
τ 0.9P decreased in turn, and the mode transition of evaporation be-
fuels have been investigated, and no qualitative differences are ob-
came easier. When the fuel was n-hexadecane, due to the ambient
served but there are gradual quantitative changes. Similar conclu-
gas molecules changing from A1 to A2, τ 0.9P slightly decreased,
sions have been drawn in Ref. [31]. Further work is needed in this
and the transition also became easier, which is consistent with a
area, provided computational resources are made available.
previous study on multicomponent fuels [24].
Figure 12 exhibits the mixing regime of Crua et al. [2] for
3.5. Multicomponent effects several single-component fuels (n-heptane, n-dodecane and n-
hexadecane) which evaporated in the exhaust gas. The critical
Although there have been some attempts [12,33,34] for the dif- temperatures of n-heptane, n-dodecane and n-hexadecane were
ferent multi-component evaporation systems, if subtle differences 540 K, 658 K and 722 K; and the critical pressures were 2.74 MPa,
of the evaporation process are ignored, there has not yet a con- 1.82 MPa and 1.4 MPa, respectively. The reduced ambient temper-
sensus concerning which factor plays a decisive role in the mode ature Tr and the reduced ambient pressure Pr were both calcu-
transition of evaporation. In other words, which factor is primarily lated by dividing ambient values by critical values of fuel. Based
responsible for the difference in the mode transition of evapora- on their experimental data, Crua et al. [2] proposed a regression
tion for a mixing system composed of different components? Fol- model, whose expression was Tr∗ Pr.^0.5, to unify the results of
lowing the author’s previous publications [20,26], this work was different single-component fuels. If Tr∗ Pr.^0.5 was larger, it meant
a new attempt to answer this question. Four evaporation systems that the mixing regime of the fuel tended to be the diffusive mix-
were set up to study the multi-component effect, that is, a six- ing. As shown in Fig. 12, Tr∗ Pr^0.5 of the three fuels under the
component fuel droplet (A1), a three-component droplet (A1), n- same ambient temperature and pressure was calculated for inves-
hexadecane droplet (A1), n-hexadecane droplet (A2), respectively. tigating the possibility for mode transition of the three fuels un-
As mentioned, for n-hexadecane, the three-component fuel and der the same ambient conditions, using that of n-hexadecane as
the six-component fuel studied here, the estimated critical tem- a reference. As shown, in the order of heptane, n-dodecane and
peratures are 1.4 MPa, 1.89 MPa and 2.45 MPa, and the esti- n-hexadecane, Tr∗ Pr^0.5 increased in turn, which meant that the
mated critical temperatures are 722 K, 697 K and 686 K, respec- mode transition became easier. In other words, for the three in-
tively [20,26]. Figure 11 illustrates τ 0.9P for several evaporation sys- vestigated fuels, under the same ambient temperature and pres-
tems under the same ambient temperature and pressure condi- sure, n-hexadecane had the largest Tr∗ Pr.^0.5, and was the easiest

12
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

to achieve the diffusive mixing. Based on the above discussions, an and number of components. If τ 0.9P was less than 0.5, it indi-
interesting consistency between multi-component fuels and single- cates the diffusion mode. Otherwise, it signals the evaporation
component fuels was found. It is generally believed that the crit- mode. The comparison between mode transition map from ex-
ical pressure and temperature of the fuel have important effects periments and MD simulations revealed that based on the ADI
on its evaporation. Here, Pc1 and Pc2 indicate critical pressures of criterion, the results from droplets with diameters of tens of
Fuel 1 and Fuel 2, respectively. And Tc1 and Tc2 indicate critical nanometers could be in satisfactory agreement with those from
temperatures of Fuel 1 and Fuel 2, respectively. For the three fuels experiments of micro-sized droplets. This consistency indicates
investigated (including two multi-component fuels and one single- that although there are huge scale differences between exper-
component fuel), when the fuel with a larger critical pressure is iments and MD simulations, atomistic simulations can reveal
regarded as Fuel 1, if Pc1/Pc2 > Tc1/Tc2, then Fuel 2 is easier to fundamental mechanisms that also exist in macroscopic phe-
achieve the mode transition of evaporation, compared with Fuel 1. nomena and complement macroscopic experiments. The molec-
This correlation implies that pressure seems to play a more im- ular physics obtained in MD simulations will be helpful for un-
portant role than temperature for mode transition of evaporation. derstanding the mechanism of mode transition of evaporation.
It is noted that the minimum ambient temperature (750 K) stud- In addition, the knowledge of droplet evaporation will be ex-
ied here is greater than the critical temperatures of three fuels in- panded by such MD simulations.
vestigated (722 K, 697 K and 686 K, respectively). As reported by 2) The pair entropy fingerprint for each atom was calculated to
Zhang et al. [12], compared with the fuel composition, the ambi- investigate the degree of disorder of the mixing system. The
ent pressure and temperature were more crucial factors for deter- entropy change of the alkane dominated that of the mixture.
mining the transition time. Qiao et al. [65] found that at an am- In evaporation mode, the time-dependent profile of entropy
bient pressure fixed to the critical pressure of n-heptane, when of alkanes nearly always was concave and eventually became
the ambient temperature was higher than the critical tempera- convex. The entropy of alkanes increased gradually with mix-
ture of n-heptane, the required transition time from evaporation to ing, whose growth rate peaked near the end of evaporation
diffusion decreased extremely slowly with increasing temperature. and then declined. Nonetheless, in diffusion mode, the time-
On the other hand, according to Gong et al. [26], for single- and dependent profile of entropy of alkanes nearly always was con-
multi-component fuels, at an ambient temperature higher than the vex but concave during a short initial period, whose growth
critical temperatures of fuels, when the ambient pressure was far rate peaked before half of evaporation and then continuously
higher than the critical pressure of the fuel, the fuel evaporation dropped. The physics of the entropy change for the mode tran-
lifetime decreased very slowly with increasing pressure, while that sition was consistent with that of evaporation histories and that
decreased rapidly with increasing temperature. Similar results are of AD and ADI. The competition between energy and entropy
reported by Xiao et al. [10] and He et al. [66]. The above discus- determines the state of a thermodynamic system, which is also
sions show that the evaporation lifetime has different sensitivities applicable for evaporation and mode transition, so the shape of
to the changes of pressures and temperatures. Since τ 0.9P is a di- such curves is determined by a combination effect of energy
mensionless time via dividing the absolute time by the evapora- and entropy. Under this condition, the concavity and convex-
tion lifetime, it may lead to the fact that the effect of pressure ity of curves and their connection point (inflection point) re-
on mode transition of evaporation indicated by τ 0.9P is greater flect exactly which factor plays a dominant role at that time.
than that of temperature. As reported [10,14], having the time In the diffusion mode, the entropy became the dominant factor
normalized by the evaporation lifetime is a common method in in the mixing process, and the disorder of the fuel molecules
this field. In fact, experimental results of the single-component increased significantly compared with that in the two-phase
fuels by Crua et al. [2] are also consistent with the above evaporation. Moreover, MD can reveal dynamic evaporation be-
correlation. haviours and provide quantitative information that are difficult
to obtain by experiments. For some size-dependent evaporation
4. Conclusions properties, it is promising to obtain scaling laws to bridge the
size differences. For some size-independent evaporation proper-
The evaporation processes of n-hexadecane droplets in the ties, MD results can be directly mapped onto macroscopic sys-
nitrogen or exhaust gas ambient were studied using MD sim- tems.
ulations. Moreover, three-component droplet and six-component 3) Compared with the case of a quiescent droplet, the mode tran-
droplet evaporation processes were also studied. The ambient pres- sition became easier with increasing relative velocity between
sure ranged from 4 MPa to 36 MPa and the ambient temperature the droplet and the ambient gas, although this was a non-linear
ranged from 750 K to 3600 K, which covered ambient conditions process. This was due to the promoted gaseous heat and mass
for both subcritical and supercritical evaporations. Significant find- transfer and stagnation effects. Moreover, the slip effect pos-
ings of this study included: sibly exists during the evaporation of nanoscale droplets, es-
pecially for moving droplets. However, even if considering real
1) The profiles of the average displacement (AD) and average dis- spray combustors, such as diesel engines, it is estimated that
placement increment (ADI) per fuel atom with time under var- the slip effect on evaporation is really limited and will be nat-
ied ambient conditions were investigated in detail. At low am- urally included in the MD results. For the same ambient tem-
bient pressures and temperatures, the ADI per fuel atom grad- perature and pressure, the smaller the normalized specific heat
ually increased with time until near the end of the mixing transfer surface area (SF ) of the fuel, the easier the mode transi-
process. However, at high pressures and high temperatures, tion. The possible explanation for this is that the smaller SSAH
the time-dependent profile of the ADI had a peak soon af- of the fuel, the slower the evaporation of the fuel, the longer
ter the start of evaporation. Considering scale differences, the the evaporation lifetime, and the more sufficient time for the
time in this paper is normalized by the evaporation lifetime. fuel to achieve the transition. SF proposed here could give uni-
A dimensionless number τ 0.9P was proposed here which was fied analyses for droplets or films with different initial sizes,
the smaller dimensionless mixing time corresponding to 90% of which will be helpful for answering the size effect on the mode
the peak of the time-dependent ADI profile for fuel atoms. A transition of evaporation and bridging MD to practical com-
critical dimensionless number τ 0.9P of 0.5 was used to iden- bustion applications. A correlation was proposed to compare
tify the mode transition of evaporation of fuels with any type the possibility of mode transition of evaporation for single-

13
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

and multi-component fuels. When the ambient temperature ex- [16] F. Poursadegh, J.S. Lacey, M.J. Brear, R.L. Gordon, On the fuel spray transition to
ceeds the critical temperature of fuels, pressure seems to play dense fluid mixing at reciprocating engine conditions, Energy Fuels 31 (2017)
6445–6454, doi:10.1021/acs.energyfuels.7b0 0 050.
a more important role than temperature for mode transition of [17] L. Consolini, S.K. Aggarwal, S. Murad, A molecular dynamics simulation of
evaporation of fuels. droplet evaporation, Int. J. Heat Mass Transfer 46 (2003) 3179–3188, doi:10.
1016/S0017-9310(03)00101-7.
Future work is needed in this area. Considering the large size [18] L.N. Long, M.M. Micci, B.C. Wong, Molecular dynamics simulations of
difference between MD simulations and actual engineering appli- droplet evaporation, Comput. Phys. Commun. 96 (1996) 167–172, doi:10.1016/
0 010-4655(96)0 0 050-1.
cations, MD simulations with larger sizes should be performed, [19] H. Yanagihara, I. Stanković, F. Blomgren, A. Rosén, I. Sakata, A molecular dy-
provided computational resources are made available. More ac- namics simulation investigation of fuel droplet in evolving ambient conditions,
curate intermolecular potentials with lower computational costs Combust. Flame 161 (2014) 541–550, doi:10.1016/j.combustflame.2013.09.002.
[20] Y. Gong, K.H. Luo, X. Ma, S. Shuai, H. Xu, Atomic-level insights into transi-
should be developed further for complex hydrocarbons. More-
tion mechanism of dominant mixing modes of multi-component fuel droplets:
over, effects of relative velocity on mode transition of evaporation from evaporation to diffusion, Fuel 304 (2021), doi:10.1016/j.fuel.2021.121464.
should also be further investigated in detail. [21] T.L. Kaltz, L.N. Long, M.M. Micci, J.K. Little, Supercritical vaporization of liquid
oxygen droplets using molecular dynamics, Combust. Sci. Technol. 136 (1998)
279–301, doi:10.1080/00102209808924174.
Declaration of Competing Interest [22] J.K. Little, Simulation of Droplet Evaporation in Supercritical Environments Us-
ing Parallel Molecular Dynamics, The Pennsylvania State University ProQuest
The authors declare that they have no known competing finan- Dissertations Publishing, 1996.
[23] T.L. Kaltz, Simulation of Transcritical Oxygen Droplet Vaporization Using
cial interests or personal relationships that could have appeared to Molecular Dynamics, The Pennsylvania State University ProQuest Dissertations
influence the work reported in this paper. Publishing, 1998.
[24] Y. Gong, X. Ma, K.H. Luo, H. Xu, S. Shuai, A molecular dynamics study of
Acknowledgements evaporation of multicomponent stationary and moving fuel droplets in mul-
ticomponent ambient gases under supercritical conditions, Energy (2022) 258,
doi:10.1016/j.energy.2022.124838.
Support from the Natural Science Foundation of China (Grant [25] W.L. Jorgensen, D.S. Maxwell, J. Tirado-Rives, Development and testing of
No. 51976100) and the UK Engineering and Physical Sciences Re- the OPLS all-atom force field on conformational energetics and properties
of organic liquids, J. Am. Chem. Soc. 118 (1996) 11225–11236, doi:10.1021/
search Council under the project “UK Consortium on Mesoscale En- ja9621760.
gineering Sciences (UKCOMES)” (Grant No. EP/R029598/1) is grate- [26] Y. Gong, G. Xiao, X. Ma, K.H. Luo, S. Shuai, H. Xu, Phase transitions of multi-
fully acknowledged. This work made use of computational support component fuel droplets under sub- and supercritical conditions, Fuel (2021)
287, doi:10.1016/j.fuel.2020.119516.
by CoSeC, the Computational Science Centre for Research Commu- [27] J.M. Stubbs, Molecular simulations of supercritical fluid systems, J. Supercrit.
nities, through UKCOMES. Fluids 108 (2016) 104–122, doi:10.1016/j.supflu.2015.10.027.
[28] C. Chen, X. Jiang, Transport property prediction and inhomogeneity analysis of
References supercritical n-dodecane by molecular dynamics simulation, Fuel 244 (2019)
48–60, doi:10.1016/j.fuel.2019.01.181.
[1] L. Zhou, W. Zhao, K.H. Luo, M. jia, H. Wei, M. Xie, Spray–turbulence–chemistry [29] L. Luo, Y.C. Liu, Controlling parameters and regimes for preferential vaporiza-
interactions under engine-like conditions, Prog. Energy Combust. Sci. 86 tion of jet fuel droplet with liquid transport and convection, Fuel (2022) 321,
(2021), doi:10.1016/j.pecs.2021.100939. doi:10.1016/j.fuel.2022.123817.
[2] C. Crua, J. Manin, L.M. Pickett, On the transcritical mixing of fuels at diesel [30] S. Ray, V. Raghavan, G. Gogos, Two-phase transient simulations of evaporation
engine conditions, Fuel 208 (2017) 535–548, doi:10.1016/j.fuel.2017.06.091. characteristics of two-component liquid fuel droplets at high pressures, Int.
[3] V. Yang, Modeling of supercritical vaporation,mixing,and combustion process J. Multiphase Flow 111 (2019) 294–309, doi:10.1016/j.ijmultiphaseflow.2018.10.
in liquid-fueled processes in liquid-fueled propulsion systems, Proc. Combust. 002.
Inst. 28 (20 0 0) 925–942, doi:10.1016/S0 082-0784(0 0)80299-4. [31] Z. Wang, L. Zhou, G. Shu, H. Wei, Droplet evaporation and phase transition
[4] J. Bellan, Supercritical (and subcritical) fluid behavior and modeling: drops, modes in supercritical environment by molecular dynamic simulation, Phys.
streams, shear and mixing layers, jets and sprays, Prog. Energy Combust. Sci. Fluids (2021) 33, doi:10.1063/5.0053328.
26 (20 0 0) 329–366, doi:10.1016/S0360-1285(0 0)0 0 0 08-3. [32] T. Kitano, J. Nishio, R. Kurose, S. Komori, Evaporation and combustion of mul-
[5] H.H. Chiu, Advances and challenges in droplet and spray combustion. I. Toward ticomponent fuel droplets, Fuel 136 (2014) 219–225, doi:10.1016/j.fuel.2014.07.
a unified theory of droplet aerothermochemistry, Prog. Energy Combust. Sci. 26 045.
(20 0 0) 381–416, doi:10.1016/S0360-1285(0 0)0 0 016-2. [33] Z. Wang, W. Zhao, L. Zhou, G. Shu, H. Wei, A molecular dynamic study
[6] J.A. Shawn, D. Givler, Supercritical droplet vaporization and combustion stud- of evaporation/supercritical-transition inter-relationship and multicomponents
ies, Prog. Energy Combust. Sci. 22 (1996) 1–28, doi:10.1016/0360-1285(95) interaction for alkane/alcohol droplets, Phys. Fluids 34 (2022), doi:10.1063/5.
0 0 013-5. 0078471.
[7] Z. Wei, Y. Zhou, Y. Cao, Y. Feng, Q. Zhu, Y. Zhang, Experimental study on the [34] S. Chakraborty, L. Qiao, Molecular investigation of sub-to-supercritical transi-
phase inhomogeneity and separability in supercritical state of acetone fuel, tion of hydrocarbon mixtures: multi-component effect, Int. J. Heat Mass Trans-
Fuel (2022) 321, doi:10.1016/j.fuel.2022.123938. fer 145 (2019), doi:10.1016/j.ijheatmasstransfer.2019.118629.
[8] G.S. Zhu, S.K. Aggarwal, Transient supercritical droplet evaporation with em- [35] R. He, P. Yi, T. Li, Y. Zhang, R. Chen, Phase transition of n-heptane/ethanol
phasis on the effects of equation of state, Int. J. Heat Mass Transfer 43 (20 0 0) blends from subcritical to supercritical conditions, Int. J. Heat Mass Tran. 185
1157–1171, doi:10.1016/S0017-9310(99)00197-0. (2022), doi:10.1016/j.ijheatmasstransfer.2021.122405.
[9] G. Zhu, R.D. Reitz, S.K. Aggarwal, Gas-phase unsteadiness and its infuence on [36] C.D. Wick, M.G. Martin, J.Ilja Siepmann, Transferable potentials for phase equi-
droplet vaporization in sub- and super-critical environments, Int. J. Heat Mass libria. 4. United-atom description of linear and branched alkenes and alkylben-
Transfer 44 (2001) 3081–3093, doi:10.1016/S0017-9310(0 0)0 0349-5. zenes, J. Phys. Chem. B 104 (20 0 0) 80 08–8016, doi:10.1021/jp001044x.
[10] G. Xiao, K.H. Luo, X. Ma, S. Shuai, A molecular dynamics study of fuel droplet [37] M.G. Martin, J.I. Siepmann, Transferable potentials for phase equilibria. 1.
evaporation in sub- and supercritical conditions, Proc. Combust. Inst. 37 (2019) United-atom description of n-alkanes, J. Phys. Chem. B 102 (1998) 2569–2577,
3219–3227, doi:10.1016/j.proci.2018.09.020. doi:10.1021/jp972543+.
[11] Z. Falgout, M. Rahm, D. Sedarsky, M. Linne, Gas/fuel jet interfaces under high [38] S.K. Nath, F.A. Escobedo, J.J. de Pablo, On the simulation of vapor–liquid equi-
pressures and temperatures, Fuel 168 (2016) 14–21, doi:10.1016/j.fuel.2015.11. libria for alkanes, J. Chem. Phys. 108 (1998) 9905–9911, doi:10.1063/1.476429.
061. [39] J.D.M. William, L. Jorgensen, Carol J. Swenson, Optimized intermolecular po-
[12] Y. Zhang, M. Jia, P. Yi, Y. Chang, Z. He, Q. Wang, A molecular dynam- tential functions for liquid hydrocarbons, J. Am. Chem. Soc. 106 (1984) 6638–
ics study of binary-component n-alkane fuel vaporization characteristics at 6646, doi:10.1021/ja00334a030.
sub/supercritical nitrogen environments, Proc. Combust. Inst. 38 (2021) 3303– [40] B. Smit, S. Karaborni, J.I. Siepmann, Computer simulations of vapor–liquid
3312, doi:10.1016/j.proci.2020.06.108. phase equilibria of n-alkanes, J. Chem. Phys. 102 (1995) 2126–2140, doi:10.
[13] R.N. Dahms, J.C. Oefelein, On the transition between two-phase and single- 1063/1.469563.
phase interface dynamics in multicomponent fluids at supercritical pressures, [41] G. Xiao, K.H. Luo, X. Ma, S. Shuai, Liquid fuel evaporation under supercriti-
Phys. Fluids 25 (2013) 092103–092127, doi:10.1063/1.4820346. cal conditions, Commun. Comput. Phys. 23 (2018) 1241–1262, doi:10.4208/cicp.
[14] G. Mo, L. Qiao, A molecular dynamics investigation of n-alkanes vaporizing OA-2016-0252.
into nitrogen: transition from subcritical to supercritical, Combust. Flame 176 [42] J.G. Harris, K.H. Yung, Carbon dioxide’s liquid-vapor coexistence curve and crit-
(2017) 60–71, doi:10.1016/j.combustflame.2016.09.028. ical properties as predicted by a simple molecular model, J. Phys. Chem. 99
[15] J. Manin, M. Bardi, L.M. Pickett, R.N. Dahms, J.C. Oefelein, Microscopic inves- (1995) 12021–12024, doi:10.1021/j10 0 031a034.
tigation of the atomization and mixing processes of diesel sprays injected [43] R.S. Taylor, L.X. Dang, B.C. Garrett, Molecular dynamics simulations of the liq-
into high pressure and temperature environments, Fuel 134 (2014) 531–543, uid/vapor interface of SPC/E water, J. Phys. Chem. 100 (1996) 11720–11725,
doi:10.1016/j.fuel.2014.05.060. doi:10.1021/jp960615b.

14
Y. Gong, X. Ma, K.H. Luo et al. Combustion and Flame 246 (2022) 112397

[44] K.H. Lin H-M, K.-.C. Chao, Gas-liquid equilibria in nitrogen + n-hexadecane [56] P.M. Piaggi, M. Parrinello, Entropy based fingerprint for local crystalline order,
mixtures at elevated temperatures and pressures, Fluid Phase Equilib. 7 (1981) J. Chem. Phys. 147 (2017) 114112, doi:10.1063/1.4998408.
181–185. [57] R.E. Nettleton, M.S. Green, Expression in terms of molecular distribution func-
[45] H. Nomura, T. Murakoshi, Y. Suganuma, Y. Ujiie, N. Hashimoto, H. Nishida, Mi- tions for the entropy density in an infinite system, J. Chem. Phys. (1958) 29,
crogravity experiments of fuel droplet evaporation in sub- and supercritical doi:10.1063/1.1744724.
environments, Proc. Combust. Inst. 36 (2017) 2425–2432, doi:10.1016/j.proci. [58] A. Giusti, M. Kotzagianni, E. Mastorakos, LES/CMC simulations of swirl-
2016.08.046. stabilised ethanol spray flames approaching blow-off, Flow Turbul. Combust.
[46] H. Ghassemi, S.W. Baek, Q.S. Khan, Experimental study on binary droplet evap- 97 (2016) 1165–1184, doi:10.1007/s10494- 016- 9762- 1.
oration at elevated pressures and temperatures, Combust. Sci. Technol. 178 [59] S. Sazhin, Droplets and sprays, 2014. DOI: 10.1007/978-1-4471-6386-2
(2006) 1031–1053, doi:10.1080/00102200500296697. [60] S. Basu, A.K. Agarwal, A. Mukhopadhyay, C. Patel, Droplets and sprays :appli-
[47] Y. Ra, R.D. Reitz, A vaporization model for discrete multi-component cations for combustion and propulsion, 2018. https://doi.org/10.1007/978-981-
fuel sprays, Int. J. Multiphase Flow 35 (2009) 101–117, doi:10.1016/j. 10-7449-3
ijmultiphaseflow.20 08.10.0 06. [61] Z.-.G. Feng, E.E. Michaelides, Heat transfer from a nano-sphere with temper-
[48] F. Rahmani, T. Weathers, A. Hosangadi, Y.C. Chiew, A non-equilibrium molec- ature and velocity discontinuities at the interface, Int. J. Heat Mass Tran. 55
ular dynamics study of subcritical, supercritical and transcritical mixing of (2012) 6491–6498, doi:10.1016/j.ijheatmasstransfer.2012.06.049.
liquid-gas systems, Chem. Eng. Sci. (2020) 214, doi:10.1016/j.ces.2019.115424. [62] S. Starner, J. Gounder, A. Masri, Effects of turbulence and carrier fluid on sim-
[49] Y.R. Zhang, X.Z. Jiang, K.H. Luo, Bounce regime of droplet collisions: a molecu- ple, turbulent spray jet flames, Combust. Flame 143 (2005) 420–432, doi:10.
lar dynamics study, J. Comput. Sci. 17 (2016) 457–462, doi:10.1016/j.jocs.2016. 1016/j.combustflame.2005.08.016.
03.011. [63] J. Kariuki, E. Mastorakos, Experimental investigation of turbulent flames in
[50] Y.R. Zhang, K.H. Luo, Regimes of head-on collisions of equal-sized binary uniform dispersions of ethanol droplets, Combust. Flame 179 (2017) 95–116,
droplets, Langmuir 35 (2019) 8896–8902, doi:10.1021/acs.langmuir.8b04277. doi:10.1016/j.combustflame.2017.01.014.
[51] E.S. Landry, S. Mikkilineni, M. Paharia, A.J.H. McGaughey, Droplet evaporation: [64] S.K. Aggarwal, C. Yan, G. Zhu, Transcritical vaporization of a liquid fuel droplet
a molecular dynamics investigation, J. Appl. Phys. 102 (2007), doi:10.1063/1. in a supercritical ambient, Combust. Sci. Technol. 174 (2002) 103–130, doi:10.
2821753. 1080/00102200290021399.
[52] R.N. Dahms, J. Manin, L.M. Pickett, J.C. Oefelein, Understanding high-pressure [65] S. Chakraborty, H. Ge, L. Qiao, Molecular dynamics simulations of vapor−liquid
gas-liquid interface phenomena in diesel engines, Proc. Combust. Inst. 34 interface properties of n–heptane/nitrogen at subcritical and transcritical con-
(2013) 1667–1675, doi:10.1016/j.proci.2012.06.169. ditions, J. Phys. Chem. B 125 (2021) 6968–6985, doi:10.1021/acs.jpcb.1c03037.
[53] C. Crua, M.R. Heikal, M.R. Gold, Microscopic imaging of the initial stage of [66] X. He, H. Feng, Z. Liu, H. Wang, X. Li, F. Zeng, F. Liu, F. Liu, Numerical simulation
diesel spray formation, Fuel 157 (2015) 140–150, doi:10.1016/j.fuel.2015.04.041. of the evaporation characteristics of a dimethyl ether droplet in supercritical
[54] R.A. Heidemann, A.M. Khalil, The calculation of critical points, AIChE J. 26 environment, Fuel 267 (2020), doi:10.1016/j.fuel.2020.117120.
(1980) 769–779, doi:10.1002/aic.690260510.
[55] A. Ghoufi, P. Malfreyt, D.J. Tildesley, Computer modelling of the surface tension
of the gas–liquid and liquid–liquid interface, Chem. Soc. Rev. 45 (2016) 1387–
1409, doi:10.1039/c5cs00736d.

15