Center for Turbulence Research 193

Annual Research Briefs 2016

On interfacial transport in transcritical flows of

liquid fuels into high-pressure combustors
By L. Jofre AND J. Urzay

1. Motivation and objectives

The characteristic pressure and temperature of the burnt gases in the combustor of
Apollo’s Saturn-V F-1 rocket engines were 77.5 bar and 3572 K. The engine operated a
mixture of liquid oxygen (critical pressure Pc = 50 bar; critical temperature Tc = 155
K) and RP-1 fuel (Pc = 21 bar; Tc = 662 K). The typical injection temperatures of RP-
1 and liquid oxygen were 311 K and 97 K, respectively. The propellants were injected
separately into the combustion chamber through a multi-perforated plate that forced
their atomization and mixing by delivering them as mutually impinging jets.
The F-1 rocket engine is a classical example of a contemporary generation of high-
power aero-propulsion devices that pushed the operating conditions above the critical
points of the propellants. The technical command of such high combustion pressures
represented a cornerstone in enabling the large power and specific impulse necessary for
a manned mission to the Moon. However, the extreme operating conditions brought along
several engineering challenges, among which the most critical one for the Apollo Program
proved to be that of combustion instabilities created by the injection configuration (Oe-
felein & Yang 1993). Early analyses of this problem recognized the conceptual difficulties
related to describing atomization, vaporization and combustion of propellants injected
at subcritical temperatures into an environment at a pressure larger than that of their
corresponding critical points (Rosner 1972). This class of trajectories in the thermody-
namic space have been referred to as transcritical in more recent literature (Sirignano &
Delplanque 1999). The deleterious combustion pressure waves present in the first designs
of the F-1 rocket engines were removed by modifying the injection plate with a number of
baffles, since the phenomenon was sensitive to the characteristic distance from the injec-
tion plate to the beginning of the combustion zone. Paradoxically, that distance appears
to be a quantity hardly possible to predict even nowadays with current theoretical and
computational models for reasons explained later in this section.
The study of transcritical dynamics also finds important applications in recent de-
signs of gas turbine engines for jet propulsion. In particular, current trends in ultra-low
emission technologies for aviation industry are gearing combustors toward lean burn and
high pressure ratios. Lean burn aims at decreasing nitrogen oxides by avoiding their peak
production rate at stoichiometry, but requires an intense dilution and mixing of the fuel
with the air entering the combustor. Similarly, high pressure ratios are employed to in-
crease engine power and reduce emissions of carbon oxides and unburnt hydrocarbons.
These pressure ratios yield combustor pressures of order 45 bar at takeoff, while most jet
fuels have critical pressures in the range 15-22 bar. In this way, transcritical conditions
likely develop in the combustor that may have an impact on the fuel-air mixing char-
acteristics because of the alteration of the classic spray atomization dynamics expected
at lower pressures. However, the extent of these effects remains mostly unknown. Recent
194 Jofre & Urzay
experimental observations of similar aspects have been made within the context of diesel
engines by Dahms et al. (2013).
The challenge for predictive calculations of transcritical phenomena is the complexity
of the transitional character of liquid breakup, dispersion and vaporization as conditions
approach the critical point. To understand this, consider first the subcritical limit in which
the liquid is injected into a hot gas environment whose pressure is lower than the critical
pressure of the liquid. In this limit, the liquid-to-gas density ratio is large, and the liquid
atomizes following classic dynamics reported in several studies (Lasheras & Hopfinger
2000). The resulting thickness of the liquid-gas interface is clearly not in the continuum
range and can be taken to be infinitesimally small in hydrodynamic scales. Additionally,
a relatively large amount of energy must be provided by the gas in order to heat up
and vaporize the liquid phase. As a result, the beginning of the vaporization stage is
delayed farther downstream until the liquid has broken up into a sufficiently dilute cloud
of droplets (Sánchez et al. 2015). However, as the pressure is increased above the critical
point of the liquid, the liquid-gas density ratio decreases because of an increase in the
density of the gas environment. The interface becomes thicker as the liquid receives heat
from the combustor environment and its temperature nears the critical temperature.
This is accompanied by a decrease in surface tension and vaporization enthalpy, in a
manner that makes the atomization process to increasingly resemble one at infinite Weber
numbers followed by rapid mixing with the gas environment without significant energy
barrier for vaporization.
Heavy hydrocarbons require increments of temperature of order 300−400 K in order to
reach their critical points. For these fuels, it is conceptually plausible that finite surface-
tension and vaporization effects persist longer in the combustor. However, the problem
becomes exceedingly complex in mixtures of hydrocarbons and typical oxidizers, in that
the resulting phase diagram displays critical-point elevation properties that, depending
on the local composition, may lead to locally subcritical conditions even if the pressure
and temperature are much larger than the corresponding critical values of the separate
components. This may lead to the persistence of relatively sharp interfaces in pressure
and temperature conditions where they were not expected to exist. Note that for single-
component systems consisting of a liquid atomizing in its own vapor, the description of the
dynamics becomes much simpler, in that there is practically no distinction between the
two phases at pressures above the critical point. Accordingly, the interface disappears, and
the surface tension vanishes. The single-component case, however, is not the one found
in most practical applications. Additional factors that prevent further understanding of
transcritical dynamics are the large uncertainties in high-pressure physical properties of
complex mixtures of reactants and combustion products, and the lack of quantitative
experimental diagnostics for model validation in such extreme environments.
This report addresses basic theoretical aspects of flow transport under transcriti-
cal conditions. It is organized as follows. In Section 2, a phase diagram for a typical
hydrocarbon-fueled system is described that illustrates the thermodynamic space of so-
lutions at high pressures and describes critical-point elevation properties. Section 3 is
devoted to a derivation of a set of conservation equations that simultaneously consider
surface-tension effects along with relatively permeable interfaces. Lastly, conclusions are
provided in Section 4. This report builds on recent analyses by Gaillard et al. (2016)
by providing different perspectives of the formulation and additional considerations that
may be of practical use in the computation of hydrocarbon-fueled transcritical flows.
 Interfacial transport in high-pressure transcritical flows of liquid fuels 195
n-dodecane boiling line
1000 nitrogen boiling line
critical line
Garcı́a-Córdova et al.

100
1000
P [bar]
100
10 P [bar]
10

1 1
500
T [K] 300 0.8 1.0
100 0.4 0.6
0.0 0.2 Y [-]

Figure 1. Vapor-liquid equilibrium curves (solid lines) for n-dodecane/nitrogen mixtures colored
by pressure, with Y indicating the mass fraction of n-dodecane. Pure-substance boiling lines for
nitrogen (dashed red) and n-dodecane (dashed blue) are shown, along with their corresponding
critical points (squares). Experimentally measured critical points are denoted by triangles.

2. Phase diagrams and critical-point elevation properties

of hydrocarbon-fueled mixtures
Some insight can be gained into system trajectories leading to transcritical regimes by
studying vapor-liquid equilibrium curves. These correspond to boundaries of the region
within which the system separates into two or more different phases across an interface.
The phase diagram illustrates the thermodynamic space of solutions of the problem,
but does not provide any information about the dynamics. For instance, the liquid-fuel
stream and the hot gas environment may be both represented by two different points
in the phase diagram, but the thermodynamic trajectories of the mixture elements are
solutions to the conservation equations and the associated boundary conditions. In this
section, mixtures of n-dodecane (Pc,1 = 18 bar, Tc,1 = 658 K) and nitrogen (Pc,2 = 34
bar, Tc,2 = 126 K) are studied since they are commonly considered as surrogates of high-
pressure fuel/air mixtures for gas turbines and diesel engines (Manin et al. 2014; Qiu &
Reitz 2015).
The details of the computation of phase envelopes will be omitted here as they are
a classical subject treated in reference textbooks (Firoozabadi 2015). In general, vapor-
liquid equilibrium
PN PNcurves are obtained by computing solutions of the zeroed second-order
variation i=1 j=1 (∂ 2 F/∂ni ∂nj )T,ρ,nk6=i,j ∆ni ∆nj = 0 of the Helmholtz free energy F ,
where ∆ni are non-zero perturbations of moles of species i, ρ is the density, and N is
the number of components of the mixture (N = 2 in this example). Similarly, the calcu-
lation of the critical line, which corresponds to the curve connecting the critical points
for different mixture compositions, follows the methodology introduced by Heidemann
& Khalil (1980). The critical point of multi-component mixtures is the thermodynamic
state at which the bubble point and the dew point converge (which does not necessar-
ily occur at inflection points of isobars), and therefore corresponds to a stable point at
the limit
P ofP thermodynamic stability. They are obtained from the solutions to the equa-
N PN
tion N i=1 j=1 k=1 (∂ 3
F/∂n i ∂nj ∂nk )T,ρ,nℓ6=i,j,k ∆ni ∆nj ∆nk = 0 for the third-order
variation of F .
The equations described above are supplemented with the Peng & Robinson (1976)
196 Jofre & Urzay
P [bar] Y [-]
10000 1.0
Garcı́a-Córdova et al.
0.8
1000 critical line
Y = 0.5
0.6
100 nitrogen
critical point n-dodecane 0.4
critical point
10 liquid two phases
0.2
gas
1 0.0
100 200 300 400 500 600 700 100 200 300 400 500 600 700
T [K] T [K]
Figure 2. Two-dimensional projections of the vapor-liquid equilibrium curves on the tempera-
ture-pressure plane for a n-dodecane/nitrogen system, including description of the diagram for
Y = 0.5 (left panel) along with arrays of vapor-equilibrium curves for different values of Y (right
panel). Refer to caption in Figure 1 for the remaining symbols and lines.

equation of state, which is formally introduced later in Eq. (3.9). The coefficients a
and b of the equation of state depend on the critical temperatures, critical pressures and
acentric factors of the individual mixture components, as well as on the local temperature
and mixture composition. They are obtained by first computing the individual values of
the coefficients for each species, ai and bi , as specified in Harstad et al. (1997), which are
combined using van der Waals mixing rules as
N X
X N N
X √
a= Xi Xj aij , b= Xi b i with aij = (1 − ηij ) ai aj , (2.1)
i=1 j=1 i=1

where Xi is the molar fraction of species i, and η1,2 = 0.1561 is a binary interaction
parameter fitted to experimental data by Garcı́a-Córdova et al. (2011).
The vapor-liquid equilibrium curves resulting from the computations are shown in Fig-
ure 1 in a three-dimensional space formed by P , T and the mass fraction of n-dodecane Y .
The three-dimensional two-phase region, which is enclosed under the surface enveloping
the curves, reaches much larger pressures than the critical pressure of each component.
In practical terms, this is translated into the fact that an n-dodecane liquid jet flow-
ing into a nitrogen ambient at pressures much larger than Pc,1 = 18 bars may undergo
transcritical trajectories that cross the two-phase region. As a result, such flow may dis-
play remnant effects of surface tension and atomization characteristics similar to lower
pressure jets that in principle were not expected to be observed in these thermodynamic
conditions, as shown in experiments by Manin et al. (2014). This critical-point elevation
property is also illustrated by the divergence of the the critical line arriving to the nitro-
gen side, as observed in Figure 2, which indicates that the two-phase region is unbounded
in pressure. Conversely, the critical line starting at the nitrogen critical point meets a
liquid-liquid-gas phase-equilibrium line (indistinguishable from the nitrogen boiling line
in the scales of Figure 2) at an upper critical end point. The three-phase equilibrium line
continues toward lower pressures and temperatures between the boiling lines of the two
 Interfacial transport in high-pressure transcritical flows of liquid fuels 197

1000
Max. T at P = 50 bar
nitrogen environment Max. T at P = 100 bar
900 Max. T at P = 200 bar

800
case 1 case 2
700 critical points
T [K] single-phase region
600

500

400 two-phase region

liquid-fuel stream
300
0.0 0.2 0.4 0.6 0.8 1.0
Y [-]

Figure 3. Examples of linear thermodynamic trajectories (solid lines) superimposed on max-

imum-temperature curves (dashed colored lines) bounding the two-phase region of n-dode-
cane/nitrogen mixtures at nitrogen-environment pressures of 50, 100, and 200 bar. Symbols
denote the thermodynamic conditions of the nitrogen environment (left-hand-side square) and
n-dodecane stream at 363 K (lower right-hand-side square) and 563 K (upper right-hand-side
square). The diamond symbols represent critical points at the corresponding pressure.

pure components, thereby suggesting that the crossing of the three-phase equilibrium
region is only relevant at pressures lower than Pc,2 = 34 bar and across a very narrow
range of temperatures around Tc,2 = 126 K. The phenomena of divergence of the critical
line and occurrence of three-phase equilibria are typical in mixtures classified as class-
II/type-III according to the analysis of van Konynenburg & Scott (1980). This group
of mixtures, to which other n-alkane/nitrogen systems also belong, is characterized by
individual components with very different critical temperatures.
It should be stressed that the computation of critical points in complex mixtures
involves a number of assumptions and model parameter values that find little justification
on physical grounds. For instance, the mixing rules (2.1) correspond to an ad-hoc molar
weighting of the individual coefficients ai and bi , whose expressions depend on calibrated
interaction parameters and measured critical points of the pure substances (Harstad et
al. 1997). However, it is of some interest to note that the resulting divergent trend of
the critical line computed from the vapor-liquid equilibrium agrees well with the values
experimentally obtained by Garcı́a-Córdova et al. (2011), as shown in Figures 1 and 2.
The phase diagram facilitates the understanding of the thermodynamic trajectories
involved in the injection of hydrocarbon fuels into high-pressure environments. As an
illustration, consider the examples of linear thermodynamic trajectories followed by mix-
ture elements in the problem of a liquid n-dodecane jet injected in a nitrogen environment
at 900 K, which are provided in Figure 3. Two fuel injection temperatures, corresponding
to 363 K (case 1) and 563 K (case 2), are studied, along with three nitrogen-environment
pressures, namely, 50, 100, and 200 bar. The trajectories are superimposed on maximum-
temperature curves bounding the two-phase region at each pressure. Note that the tra-
198 Jofre & Urzay
jectories resulting from integration of the conservation equations may not be generally
linear (Matheis & Hickel 2016).
For all pressure values considered in case 1, the mixture elements start as compressed
liquids in the n-dodecane stream. As heat is supplied from the surrounding gas, the mix-
ture elements enter the two-phase region where they necessarily separate into liquid and
vapor phases by an interface where surface-tension forces operate. The mixture elements
eventually exit the two-phase region and change phase to a supercritical state while mix-
ing with the surrounding nitrogen gas. On the other hand, in case 2, the intersection
with the two-phase region is completely avoided for the largest pressure value considered
in the nitrogen environment. The resulting thermodynamic path involves no sustained
separation of the two components through an interface. In these conditions, atomization
and mixing are solely limited by the rate of supercritical mass diffusion.
Figure 3 demonstrates that the fluid dynamical description of problems where transcrit-
ical conditions are traversed requires the treatment of emerging and vanishing interfaces
in the same flow field depending on the local thermodynamic conditions of the mixture
elements as they move across the flow field. A set of modifications to the Navier-Stokes
equations are presented below that attempt to enable this treatment.

3. Conservation equations for transcritical flows

This section describes a general formulation of the conservation equations and asso-
ciated transport fluxes based on diffuse-interface theory for multi-component flows. The
development begins by outlining the main characteristics of the theory, and continues
with the general conservation equations along with derivations of the transport fluxes
from thermodynamic considerations.

3.1. Non-local thermodynamic effects

The theoretical foundations of the diffuse-interface approach were first established for
single-component systems in thermodynamic equilibrium by van der Waals (1893). It
was later extended to study binary mixtures near the critical point by Cahn & Hilliard
(1958). More recently, the approach has been coupled to the equations of fluid motion
for single-component systems (Anderson et al. 1998). Thermodynamic investigations of
transition between two- and single-phase states using the diffuse-interface approach for
inert and chemically reacting multi-component systems have also provided understanding
of the behavior of propellants in high-pressure environments (Dahms & Oefelein 2013;
Gaillard et al. 2016).
The diffuse-interface theory rests upon the fact that the two-phase region within the
liquid-vapor equilibrium curve in the phase diagram is thermodynamically unstable, in
that no stable thermodynamic state exists that describes a spatially homogeneous mix-
ture of liquid and vapor. This is perhaps easily visualized by bringing a single-component
fluid in a closed vessel to its vapor pressure. The substance tends to separate in two phases
of unequal density bounded by a thin transition layer where capillary forces become im-
portant. The general results of the diffuse-interface theory are aimed at describing the
mechanics of the transition layer as well as the fluxes of energy and mass across it. The
former correspond to familiar surface-tension forces emerging from the resistance of the
interface to get deformed, while the latter represent vaporization and diffusive mixing.
The description of the structure of the transition layer requires consideration of non-
local thermodynamic potentials, where the non-locality is represented by gradients of
 Interfacial transport in high-pressure transcritical flows of liquid fuels 199
selected variables. Additional considerations based on the second principle of thermody-
namics typically preclude non-locality to be expressed only in terms of composition or
density gradients (e.g., see Dunn & Serrin (1985) for details on the mathematical justifi-
cation). The analysis is facilitated when the degree of non-locality is assumed to be small,
with the characteristic length of the composition gradients being large compared to in-
termolecular distances, which typically limits the theory to situations when the interface
is relatively thick compared to the molecular mean free path, as in conditions near and
above the critical point. In this limit, the disturbances of the local thermodynamic state
are proportional to the square of the composition gradients in the first approximation
(Cahn & Hilliard 1958). For instance, the non-local corrections to the specific values of
Helmholtz free energy f , internal energy e and entropy s are
N N N N
1 XX 1 XX e
fNL = f + κij ∇ρi ∇ρj , eNL = e + κ ∇ρi ∇ρj ,
2ρ i=1 j=1 2ρ i=1 j=1 ij
N N
1 XX s
sNL = s + κ ∇ρi ∇ρj , (3.1)
2ρ i=1 j=1 ij

where ρ is the mixture density, ρi is the partial density of species i, and N is the number
of species. Additionally, κij , κeij and κsij are gradient-energy coefficients, which can be
computed directly as a function of collision parameters from kinetic-theory considerations
of interactions between molecules in regions subjected to macroscopic density gradients
(e.g., see Pismen (2001) and Chapter 1 in Rowlinson et al. (2002)).
Since the gradient-energy coefficients κij are related to the interface thickness and
surface tension, their precise characterization is central to the predictions of the diffuse-
interface theory. However, appropriate formulations of this parameters are lacking, and
√
most investigations utilize relations of the type κij = κii κjj = κji for the cross-influence
coefficients i 6= j, along with empirical correlations for the individual coefficients κii such
as (Lin et al. 2007)
      2
κii 8/3 T T
ln NA = κ0,i + κ1,i ln 1 − + κ2,i ln 1 − (3.2)
ai bi 2/3 Tc,i Tc,i

for T /Tc,i ≤ 0.95, where Tc,i is the critical temperature value, and NA is the Avogadro’s
number. In Eq. (3.2), the correlation coefficients κ0,i , κ1,i and κ2,i are usually calibrated
based on experimental measurements of surface tension, while the parameters ai and bi
correspond to coefficients of the equation of state, as described in Section 2. Note that
models such as Eq. (3.2) typically yield κii = 0 above the critical temperature of the
corresponding species, as suggested by the fact that the surface tension vanishes for single-
component systems above the critical point. For instance, in the n-dodecane/nitrogen
system described in Section 2, the relevant gradient-energy coefficient becomes that of the
n-dodecane, κ1,1 , since the temperature in the flow is larger than the critical temperature
of nitrogen everywhere (i.e., κ2,2 = κ1,2 = κ2,1 = 0).
Exact expressions relating the gradient-energy coefficients κij , κeij and κsij can be easily
derived by substituting the relations (3.1) into the definition of the local Helmholtz free
energy, f = e − T s, with s = −(∂f /∂T )ρ,ni , thereby yielding
   
∂κij ∂κij
κeij = κij + T and κsij = − . (3.3)
∂T ρ,nk ∂T ρ,nk
200 Jofre & Urzay
The consideration of non-local thermodynamic potentials, as in Eqs. (3.1)-(3.3), leads
to the emergence of interface-related transport fluxes and mechanical stresses in the
conservation equations as described below.

3.2. Conservation equations

The description of thin interfaces and their dynamics in conjunction with the outer
fluid motion in a single Eulerian field requires non-trivial extensions of the Navier-Stokes
conservation equations. In principle, the derivation of these modifications from molecular
considerations and first principles is a difficult task due to the lack of a clear physical
understanding of the molecular structure of fluids across the critical point. In this study,
a phenomenological approach is followed based on a linear augmentation of the deviatoric
part of the stress tensor, τ , and the heat and species diffusion fluxes, q and Ji , with the
corresponding interfacial transport terms K, Q and Ji derived from the diffuse-interface
theory. These, as shown below, can be made to satisfy certain conditions of entropy
maximization that are in accord with the second law of thermodynamics. The resulting
conservation equations for mass, momentum, species and total energy are
∂ρ
+ ∇ · (ρv) = 0, (3.4)
∂t
∂ (ρv)
+ ∇ · (ρv ⊗ v) = −∇PNL + ∇ · (τ + K) , (3.5)
∂t
∂ (ρYi )
+ ∇ · (ρvYi ) = −∇ · (Ji + Ji ) , i = 1, ..., N, (3.6)
∂t
∂ (ρE)
+ ∇ · (ρvE) = −∇ · (PNL v) − ∇ · (q + Q) + ∇ · [(τ + K) · v] , (3.7)
∂t
which describe the continuum dynamics of a multi-phase, multi-component fluid of N
species that moves at a mass-averaged velocity v and has a local density ρ and total
energy E, and which may contain thin interfaces separating P different phases. In this
formulation, Yi is the mass fraction of species i, q = qc + N i=1 hi Ji is the sum of the
heat conduction and the energy flux by inter-diffusion, hi is the partial specific enthalpy,
and PNL a non-local thermodynamic pressure defined as
N N
1 XX
PNL = P − κij ∇ρi ∇ρj . (3.8)
2 i=1 j=1

The convenience of redefining pressure as in Eq. (3.8), will become clear in Section 3.3.
In Eq. (3.8) the local thermodynamic pressure P can be obtained, for instance, from the
cubic equation of state (Peng & Robinson 1976)
R0 T a
P = − , (3.9)
v − b v 2 + 2bv − b2
whose utilization is beneficial at the high P pressures considered here. In the notation,
v = W/ρ is the molar volume, with W = ( i=1 Yi /Wi )−1 the mean molecular weight.
The coefficients a and b, which correspond to mixture-averaged versions of the pure-
substance ones ai and bi as in Eq. (2.1), account for real-gas effects such as finite packing
and increased intermolecular interactions at large densities and pressures. It should be
stressed that the choice of Eq. (3.9) is not central to the diffuse-interface formalism
insofar as it reproduces the multivalued character of the mixture density in conditions
 Interfacial transport in high-pressure transcritical flows of liquid fuels 201
of phase change. Note that several other equations of state are available in the literature
that have similar characteristics (Benedict et al. 1942; Soave 1972).
Chemical conversion sources have been excluded for simplicity from Eq. (3.6). Gas-
phase combustion reactions tend to occur far from interfaces and in regions where the
local mass fraction of fuel vapor is sufficiently small to warrant stoichiometric proportions.
However, this approximation may not be appropriate if thermal decomposition of the
liquid fuel plays an important role in modifying the interface properties.
In the species conservation equation (3.6), the different N components of the mixture
are described by their corresponding mass fractions irrespectively of their phase state.
Note that this is in contrast with traditional treatments of dispersed multi-phase flows,
where the gas and liquid mass fractions are described by their corresponding conservation
equations. In the diffuse-interface formulation, the phases are separated by interfaces in
thermodynamic conditions corresponding to the two-phase region. In those situations,
the interfacial stress tensor K in the momentum equation (3.5) provides information
about the dynamical equilibrium of the separating interface, while the fluxes Q and Ji
modify the transport of heat and mass across the interface accordingly. The high-pressure
characteristics of the transport fluxes are described in detail in Section 3.3.
A complete description of the mixture state requires specification of the analytical
form of the thermodynamic potentials. At high pressures, increasing departures from the
ideal-gas theory are observed, and consequently, derivation of more complex expressions
are necessary. A common approach to express high-pressure real-gas thermodynamic
potentials is to decompose them into the sum of their ideal-gas counterparts (denoted
below by the superscript 0 ) and departure functions that measure deviations with respect
to the ideal-gas behavior (Poling et al. 2001). For instance, the departure function for
the molar enthalpy is
Z T Z P   
0 0 ∂v
h̄ − h̄ = Cp dT + v−T dP, (3.10)
T0 0 ∂T P

where h̄0 and Cp0 are the ideal-gas reference molar values of enthalpy and constant-
pressure heat capacity, with T 0 = 298.15 K. Subsequently, the molar internal energy can
be obtained from the enthalpy definition as

ē = h̄ − P v. (3.11)

These expressions are valid for any equation of state. Exact forms of the departure
functions for multi-species mixtures can be found in Miller (2000) for the Peng-Robinson
equation of state.
Similar considerations apply for the molar chemical potential
 
∂G
µ̄i = (3.12)
∂ni T,P,nj6=i

defined as the partial molar of the Gibbs free energy G. The corresponding decomposition
is given by
µ̄i = µ̄0i + R0 T ln ϕi , (3.13)
where µ̄0i (T, P )
is the ideal-gas counterpart. In Eq. (3.13), the departure function involves
the dimensionless fugacity coefficient ϕi = fi /(Xi P ), which represents the ratio of the
fugacity fi to the partial pressure. In particular, for the Peng-Robinson equation of state,
202 Jofre & Urzay

ln ϕ1 [-] ln ϕ2 [-]
40 40

20 P = 100 bar 20 P = 100 bar

0 0

-20 -20
Y = 0.0 Y = 0.0
-40 0.25 -40 0.25
0.5 0.5
-60 0.75 -60 0.75
1.0 1.0
-80 -80
100 300 500 700 900 100 300 500 700 900
T [K] T [K]
Figure 4. Logarithm of the n-dodecane (left panel) and nitrogen (right panel) fugacity coeffi-
cients as a function of temperature and fuel mass fraction for an n-dodecane/nitrogen mixture
at P = 100 bar.

the logarithm of the fugacity coefficient becomes

bi
ln ϕi = (Z − 1) − ln (Z − B)
b
" PN # " √
#
A 2 j=1 Xi aij bi Z + 1+ 2 B
− √ − ln √
, (3.14)
2 2B a b Z + 1− 2 B

where A = aP/(R0 T )2 , B = bP/(R0 T ), and the coefficients a and b are given by Eq. (2.1).
Additionally, Z = P v/(R0 T ) is the compressibility factor, which quantifies the departures
from the reference value Z = 1 corresponding to the ideal-gas equation of state. Figure 4
shows the fugacity coefficients for an n-dodecane/nitrogen mixture at high pressure.
While departures from ideality are largest at low temperatures, the chemical potential
resembles that of the ideal gas for sufficiently large temperatures (e.g., above 900 K).
Similar trends hold up to pressures of order 103 bar.
The transport coefficients also undergo large variations across the phase diagram at
high pressures. The transition from liquid-like to gas-like characteristics prevent the uti-
lization of simple expressions for the evaluation of mixture’s viscosity, thermal conductiv-
ity and diffusion coefficients. Instead, the method by Chung et al. (1988) is typically used
to evaluate viscosity and thermal conductivity as function of T and ρ, whereas diffusion
coefficients can be calculated, for example, following the expressions given in Chapter
11 of Poling et al. (2001) for high-pressure conditions. These coefficients, however, are
currently subject to large uncertainties.

3.3. Transport fluxes

The system of conservation equations (3.4)-(3.7) requires closure expressions for τ , K, qc ,
Q, Ji , and Ji . This is achieved through the method of irreversible thermodynamics by
specifying constitutive relations such that the entropy production is non-negative. This
methodology requires that one finds the conservation equation of entropy guided by the
fact that the source terms are written as a sum of products of fluxes and thermodynamic
 Interfacial transport in high-pressure transcritical flows of liquid fuels 203
forces (Landau & Lifshitz 1987). The formulation is greatly simplified when κij does
not depend on temperature, in such a way that κeij = κij and κsij = 0, as implied by
Eq. (3.3). In view of the experimental correlation (3.2), this is an approximation that
has an unclear physical justification but has however been used in the literature (Pismen
2001; Papatzacos 2000) and will also be followed here. Starting from the second principle
of thermodynamics for a multi-component system
N
X
T ds = de + P d(1/ρ) − (µ̄i /Wi )dYi , (3.15)
i=1

and substituting the relations (3.1), the equation

N
X N
X
T dsNL = deNL + PNL d(1/ρ) − (µ̄i /Wi )dYi − ψi · d(∇ρi )/ρ, (3.16)
i=1 i=1

is obtained, where
N
X
ψi = κij ∇ρj (3.17)
j=1

is an auxiliary variable. A transport equation for the specific entropy sNL can be derived
by taking the material derivative of Eq. (3.16) and substituting the conservation equations
(3.4)-(3.7) into the resulting expression, which yields
( " N
#)
DsNL 1 X
ρ +∇· qc + Q − {ψi [ρi ∇ · v + ∇ · (Ji + Ji )] − T sei Ji + µ
ei Ji } = ṡprod
Dt T i=1
(3.18)
where sei = si + (∇ · ψi )/T and µ ei = µi − ∇ · ψi are generalized versions of the partial
specific entropy and specific chemical potential, respectively, and ṡprod is an entropy
source given by
" N
#
1 X
ṡprod = τ +K− (ρi ∇ · ψi I − ∇ρi ⊗ ψi ) : ∇v
T i=1
" N
#  
X 1
+ qc + Q − {ψi [ρi ∇ · v + ∇ · (Ji + Ji )] − T sei (Ji + Ji ) + hi Ji } · ∇
i=1
T
N
1 X
− (Ji + Ji ) · ∇e
µi , (3.19)
T i=1

with I the identity matrix. In the derivation of Eq. (3.19), use of the thermodynamic
relation T si = hi − µi and of the vector identity
 
D Dρi
ψi · (∇ρi ) = ψi · ∇ − (∇ρi ⊗ ψi ) : ∇v (3.20)
Dt Dt
have been made. Similarly, the gradient of the specific chemical potential in the last
term of Eq. (3.19) can be expanded as ∇e µi = ∇T µei − {si + [∂ (∇ · ψi ) /∂T ]P,nj }∇T ,
where the reciprocity relation (∂µi /∂T )P,nj = −si has been used, with the subindex T
indicating variations at constant temperature.
P Upon substituting this relation along with
the species diffusion-flux conservation N i=1 (Ji + Ji ) = 0 into Eq. (3.19), the entropy
204 Jofre & Urzay
source becomes
" N
#
1 X
ṡprod = τ +K− (ρi ∇ · ψi I − ∇ρi ⊗ ψi ) : ∇v
T i=1
" N
#  
X 1
+ qc + Q − {ψi [ρi ∇ · v + ∇ · (Ji + Ji )] − χi (Ji + Ji ) + hi Ji } · ∇
i=1
T
N
X −1   
1
+ (Ji + Ji ) · −∇T (e µi − µeN ) , (3.21)
i=1
T

where χi = ∇ · ψi − T [∂ (∇ · ψi ) /∂T ]P,nj is an auxiliary variable.

The entropy source (3.21) consists of three terms represented by each one of the lines.
They represent entropy production by three different sources: flow gradients, temperature
gradients, and gradients of the chemical potential at constant temperature. Following the
methodology of Onsager (1931) (see also Landau & Lifshitz (1987)), the terms K, Q and
Ji can be computed as follows. Each term of the entropy source PNcan be represented as a
product of a flux φi and a thermodynamic force Fi as ṡprod = i=1 φi · Fi . In the case of
dissipative systems, the second law of thermodynamics requires the entropy production
to be non-negative, ṡprod ≥ 0. As a consequence, the final forms of the unknown fluxes φi
must be such that they satisfy this thermodynamic requirement. One possiblePapproach to
N
warrant this is to assume linear relations between fluxes and forces as φi = k=1 Lik Fk ,
where Lik are phenomenological coefficients that are reciprocal, Lik = Lki , and positive
diagonal, Lii > 0.
The number of couplings between the different types of fluxes is reduced by means of
the Curie principle (Curie 1894), which states that the fluxes and thermodynamic forces
of different tensorial character do not couple. This principle implies that the viscous
stresses depend exclusively on ∇v, while the expressions of the diffusive fluxes of heat
and species are only function of the gradients of temperature and chemical potential.
For instance, in the case of Newtonian fluids, the expression for the viscous stresses
corresponds to τ = η (∇v + ∇v ⊺ ) + (ζ − 2η/3)(∇ · v)I where η and ζ are the shear
and bulk viscosities, respectively. Accordingly, the viscous dissipation τ : ∇v is always a
positive source of entropy in Eq. (3.21).
The interfacial stress K is assumed to be an elastically restoring one, in such a way
that the first term in Eq. (3.21) yields zero-production of entropy for interface-related
terms. As a result, the interfacial stress tensor becomes
N
X
K= (ρi ∇ · ψi I − ∇ρi ⊗ ψi ) . (3.22)
i=1

Although the surface tension does not appear explicitly in Eq. (3.22), its effects are
accounted for in the density gradients. For a quasi-planar interface, an effective surface-
tension coefficient can be defined as
XN X N Z +∞
dρi dρj
σ= κij dξ, (3.23)
i=1 j=1 −∞
dξ dξ

which, in view of Eq. (3.1), is proportional to the excess of free energy contained in the
interface, where ξ is the coordinate normal to the interface (Lin et al. 2007).
 Interfacial transport in high-pressure transcritical flows of liquid fuels 205
The interfacial heat flux is cast into the form
N
X
Q= {ψi [ρi ∇ · v + ∇ · (Ji + Ji )] − χi (Ji + Ji ) + hi Ji } + Θ. (3.24)
i=1

The term Θ is a dissipating component that produces entropy and is computed below.
In order to obtain expressions for the diffusive fluxes of heat and species, it is convenient
to express Eq. (3.21) in the matrix flux-force form
" N −1
# N −1
" N −1
#
X X X
ṡprod = Lqq F 1 + Lqk F k+1 · F 1 + Liq F 1 + Lik F k+1 · F i+1 , (3.25)
k=1 i=1 k=1

where Lqq = L1,1 , Lq,k−1 = L1k (k = 2, ..., N ), Li−1,q = Li1 (i = 2, ..., N ), Li−1,k−1 =
Lik (i, k = 2, ..., N ), F 1 = ∇(1/T ) and F i>1 = −∇T (eµi − µ
eN ) (1/T ). Comparing the
last two terms of Eq. (3.21) with Eq. (3.25) provides the expressions
N
X −1 N
X −1
Lqq Lqk Lqk
qc = − ∇T − ∇T (µk − µN ) , Θ= ∇T ∇ · (ψ k − ψ N ) , (3.26)
T2 T T
k=1 k=1

and
N
X −1 N
X −1
Liq Lik Lik
Ji = − ∇T − ∇T (µk − µN ) , Ji = ∇T ∇ · (ψ k − ψ N ) , (3.27)
T2 T T
k=1 k=1

for the heat and species diffusion fluxes, respectively. Additionally, the relation Lqq =
λT 2 is obtained by analogy with Fourier’s law of heat conduction, with λ the thermal
conductivity.
In Eqs. (3.22), (3.24) and (3.27), the tensor K and the fluxes Q and Ji , with Θ given
in Eq. (3.26), represent the interfacial disturbances to the deviatoric part of the stress
tensor, τ , and to the diffusive fluxes of heat, q and species Ji , respectively. In absence
of interfaces, K = Q = Ji = 0 and PNL = P , thereby leading to a simplification of the
conservation equations (3.4)-(3.7) to their classic Navier-Stokes form. Symmetries in the
diffusive fluxes are illustrated by the presence of the Dufour term in qc due to chemical-
potential gradients as well as the corresponding Soret term in Ji due to temperature
gradients.
The species diffusion flux Ji can be expanded in terms of pressure and composition
gradients by making use of the differential form
  N
X −1  
∂ µ̄k ∂ µ̄k
∇T µ̄k = ∇P + ∇Xj , (3.28)
∂P T,Xi j=1
∂Xj T,P,Xi6=j

and the Gibbs-Duhem equation

 
N
X −1
1 1
∇T µ̄N = ∇P − Xj ∇T µ̄j  , (3.29)
XN c
j=1

with c = ρ/W as the molar density. The combination of Eqs. (3.28)-(3.29) leads to the
206 Jofre & Urzay
relation
N
X −1  −1 
 NX 
Xℓ δℓk ∂ µ̄ℓ
∇T (µk − µN ) = + ∇Xj
WN XN Wk j=1
∂Xj T,P,Xi6=j
ℓ=1
"   N −1   #
1 ∂ µ̄k 1 X Xℓ ∂ µ̄ℓ
+ − + ∇P (3.30)
Wk ∂P T,Xi cWN XN WN XN ∂P T,Xi
ℓ=1

for the chemical-potential gradients, where δjk is the Kronecker delta. Upon substituting
Eq. (3.30) into Eq. (3.27), the species diffusion flux Ji can be expressed as
 
N
X −1
M
Ji = −ρ  Dij ∇Xj + DiT ∇T + DiP ∇P  (3.31)
j=1

M
where Dij , DiT and DiP are mass, thermal and pressure diffusion coefficients defined as

N −1 N −1  
M Wi Xi X X Wℓ Xℓ + WN XN δℓk ∂ ln fℓ
Dij = aiN DiN Lik , (3.32)
Lii W Wℓ ∂Xj T,P,Xm6=j
k=1 ℓ=1
kiT
DiT = aiN DiN , (3.33)
T
N −1
" N −1
#
Wi Xi X WN XN 1 X
DiP = aiN DiN L ik V k − + Xℓ V ℓ , (3.34)
R0 T Lii W Wk c
k=1 ℓ=1

where (∂ µ̄k /∂P )T,Xi = V̄k and (∂ µ̄ℓ /∂Xj )T,P,Xi6=j = R0 T ∂ ln fℓ /∂Xj |T,P,Xi6=j have been

used. In Eqs. (3.32)-(3.34), aiN = Wi WN /W 2 and D

2 0 2
iN = W R Lii / c Wi WN Xi XN
2

are prefactors, and kiT = Wi WN Xi XN Liq / W R0 Lii is the thermal-diffusion ratio.

Simplified versions of the expressions given above are provided in Table 1 for binary
mixtures in which κ1,1 is assumed to be the only non-zero gradient-energy coefficient
for the reasons described in Section 3.1 after Eq. (3.2). Additionally, Table 1 lists the
coefficients D1K,M and D1K,T , which are proportional to the gradient-energy coefficient
κ1,1 and are related to species and heat transport across interfaces. The formulation
provided here can be easily simplified to the single-component gradient theory of van
der Waals (1893), which provides the spatial distribution of density across a stationary
quasi-planar vapor-liquid interface. The reduction is based on imposing the mechanical
equilibrium condition −∇PNL + ∇ · K = 0 in the momentum conservation equation (3.5),
with PNL and K being defined in Eqs. (3.8) and (3.22), respectively. This constraint leads
to the gradient-theory equation
 2
1 dρ d2 ρ
P − P0 = κ + κρ , (3.35)
2 dξ dξ 2

where ξ is the coordinate normal to the interface and P0 the thermodynamic pressure
far from the interface. Integration of Eq. (3.35), subject to far-field boundary conditions
for the vapor- and liquid-phase densities, yields ρ(ξ) across the interface. Further details
on the integration of Eq. (3.35) can be found, for instance, in Jofre et al. (2015).
 Interfacial transport in high-pressure transcritical flows of liquid fuels 207

Transport coefficients
∂ ln f1
 
M
Fickian D1,1 = a1,2 D1,2
∂ ln X1 T ,P
X1 W1
 
Barodiffusion D1P = a1,2 D1,2 0 V1 −
R T ρ
k1T
Soret D1T = a1,2 D1,2
T
D1T W R0 T ∂ ln f1
 
Dufour D1F =
W1 W2 X1 (1 − X1 ) ∂ ln X1 T ,P
W1 W2 X1 (1 − X1 )
Interfacial (species) D1K,M = a1,2 D1,2 κ1,1
W R0 T
k1T
Interfacial (heat) D1K,T = a1,2 D1,2 κ1,1
T

Total stress tensor −PNL I + τ + K

1
 
2
Non-local pressure tensor −PNL I = − P − κ1,1 |∇ρ1 | I
2 
2

⊺

Viscous stress tensor τ = η ∇v + ∇v + ζ − (∇ · v)I
3
Interfacial stress tensor K = κ1,1 (ρ1 ∇2 ρ1 − ∇ρ1 ⊗ ∇ρ1 )

Total heat flux q+Q

Fourier, inter-diffusion q = −λ∇T + (h1 − h2 )J1 − ρD1F ∇X1

and Dufour heat fluxes
Interfacial heat flux Q = κ1,1 ∇ρ1 [ρ1 ∇ · v + ∇ · (J1 + J1 )]
−χ1 (J 1 + J1 ) + (h1 − h2 )J1 + ρD1K,T ∇T ∇2 ρ1

Total species diffusion flux J1 + J1

M P T
Fickian, barodiffusion J1 = −ρD1,1 ∇X1 − ρD1 ∇P − ρD1 ∇T
and Soret species diffusion flux

Interfacial species diffusion flux J1 = ρD1K,M ∇T ∇2 ρ1

Table 1. Transport coefficients and fluxes for a binary mixture with κ1,2 = κ2,1 = κ2,2 = 0.

4. Conclusions
This study focuses on theoretical aspects of transcritical dynamics of liquid-fuel streams
injected into high-pressure environments. The mixture displays a critical-point elevation
property by which the two-phase region extends up to pressures much larger than the
critical pressures of the individual components. As a result, linear thermodynamic tra-
jectories emulating typical injection conditions frequently pass through the two-phase
region, thus indicating that the mixture could become separated into liquid and va-
por phases by an interface. A set of modifications to the Navier-Stokes equations for
multi-component flows is proposed based on diffuse-interface theory in order to treat
the emergent and vanishing interfaces in the same flow field. This requires appropriate
alterations of the stress tensor and diffusive fluxes of heat and species. The resulting
transport formulation is particularized for binary mixtures and single-component flows,
the latter recovering the well-known gradient theory of van der Waals (1893).
It should be mentioned that in most practical cases the computational cost of the
208 Jofre & Urzay
numerical resolution of the resulting interfaces would be prohibitive, since they typically
remain small with respect to the hydrodynamic scales (e.g., see resulting thicknesses
based on the one-dimensional gradient theory in Dahms & Oefelein (2013)). Nonetheless,
this should not deter the derivation and understanding of formulations that may be later
used to inspire subgrid-scale interface-modeling approaches. Future work will involve the
integration of these equations in simple canonical problems.

Acknowledgments
This investigation was funded by the US AFOSR, Grant #FA9550-15-C-0037.

REFERENCES
Anderson, D. M., McFadden, G. B. & Wheeler, A. A. 1998 Diffuse-interface
methods in fluid mechanics. Annu. Rev. Fluid Mech. 30, 139–165.
Benedict, M., Webb, G. B. & Rubin, L. C. 1942 An empirical equation for thermody-
namic properties of light hydrocarbons and their mixtures II. Mixtures of methane,
ethane, propane, and n-butane. J. Chem. Phys. 10, 747–758.
Cahn, J. W. & Hilliard, J. E. 1958 Free energy of a nonuniform system. I. Interfacial
free energy. J. Chem. Phys. 28, 258–267.
Chung, T. H., Ajlan, M., Lee, L. L. & Starling, K. E. 1988 Generalized multi-
parameter correlation for nonpolar and polar fluid transport properties. Ind. Eng.
Chem. Fundam. 27, 671–679.
Curie, P. 1894 Sur la symétrie des phénomènes physiques: symétrie d’un champ
électrique et d’un champ magnétique. J. Phys. 3, 393–415.
Dahms, R. N., Manin, J., Pickett, L. M. & Oefelein, J. C. 2013 Understanding
high-pressure gas-liquid interface phenomena in diesel engines. Proc. Combust. Inst.
34, 1667–1675.
Dahms, R. N. & Oefelein, J. C. 2013 On the transition between two-phase and single-
phase interface dynamics in multicomponent fluids at supercritical pressures. Phys.
Fluids 25, 092103.
Dunn, J. E. & Serrin, J. 1985 On the thermomechanics of interstitial working. Arch.
Ration. Mech. Anal. 88, 95–133.
Firoozabadi, A. 2015 Thermodynamics and Applications in Hydrocarbon Energy Pro-
duction. McGraw-Hill.
Gaillard, P., Giovangigli, V. & Matuszewski, L. 2016 A diffuse interface
LOX/hydrogen transcritical flame model. Combust. Theor. Model. 20, 486–520.
Garcı́a-Córdova, T., Justo-Garcı́a, D. N., Garcı́a-Flores, B. E. & Garcı́a-
Sánchez, F. 2011 Vapor-liquid equilibrium data for the Nitrogen + Dodecane sys-
tem at temperatures from (344 to 593) K and at pressures up to 60 MPa. J. Chem.
Eng. Data 56, 1555–1564.
Harstad, K. G., Miller, R. S. & Bellan, J. 1997 Efficient high-pressure state
equations. AIChE J. 43, 1605–1610.
Heidemann, R. A. & Khalil, A. M. 1980 The calculation of critical points. AIChE
J. 26, 769–779.
Jofre, L., Urzay, J., Mani, A. & Moin, P. 2015 On diffuse-interface modeling of
 Interfacial transport in high-pressure transcritical flows of liquid fuels 209
high-pressure transcritical fuel sprays. Annual Research Briefs, Center for Turbu-
lence Research, Stanford University, 55–64.
Landau, L. D. & Lifshitz, E. M. 1987 Course of Theoretical Physics: Fluid mechanics.
Pergamon Press.
Lasheras, J. C & Hopfinger, E. J. 2000 Liquid jet instability and atomization. Annu.
Rev. Fluid Mech. 32, 275–308.
Lin, H., Duan, Y.-Y. & Min, Q. 2007 Gradient theory modeling of surface tension for
pure fluids and binary mixtures. Fluid Phase Equilib. 254, 75–90.
Manin, J., Bardi, M., Pickett, L. M., Dahms, R. N. & Oefelein, J. C. 2014
Microscopic investigation of the atomization and mixing processes of diesel sprays
injected into high pressure and temperature environments. Fuel 134, 531–543.
Matheis, J. & Hickel S. 2016 Multi-component vapor-liquid equilibrium model for
LES and application to ECN Spray A. Proceedings of the Summer Program, Center
for Turbulence Research, Stanford University, pp. 25–34.
Miller, R. S. 2000 Long time mass fraction statistics in stationary compressible
isotropic turbulence at supercritical pressure. Phys. Fluids 12, 2020–2032.
Oefelein, J. C. & Yang, V. 1993 Comprehensive review of liquid-propellant combus-
tion instabilities in F-1 engines. J. Propul. Power 9, 657–677.
Onsager, L. 1931 Reciprocal relations in irreversible processes. I. Phys. Rev. 37, 405–
426.
Papatzacos, P. 2000 Diffuse-interface models for two-phase flow. Physica Scripta 61,
349–360.
Peng, D.-Y. & Robinson, D. B. 1976 A new two-constant equation of state. Ind. Eng.
Chem. Fundam. 15, 59–64.
Pismen, L. M. 2001 Nonlocal diffuse interface theory of thin films and the moving
contact line. Phys. Rev. E 64, 021603.
Poling, B. E., Prausnitz, J. M. & O’Connel, J. P. 2001 The Properties of Gases
and Liquids. McGraw-Hill.
Qiu, L. & Reitz, R. D. 2015 An investigation of thermodynamic states during high-
pressure fuel injection using equilibrium thermodynamics. Int. J. Multiphase Flow
72, 24–38.
Rosner, D. E. 1972 Liquid droplet vaporization and combustion. In Liquid Propellant
Rocket Combustion Instability, Harrje, D. T. and Reardon H. (Eds.) NASA.
Rowlinson, J. S. & Widow, B. 2002 Molecular Theory of Capillarity. Dover.
Sánchez, A. L., Urzay, J. & Liñán, A. 2015 The role of separation of scales in the
description of spray combustion. Proc. Combust. Inst. 35, 1549–1577.
Sirignano, W. A. & Delplanque, J. P. 1999 Transcritical vaporization of liquid fuels
and propellants. J. Propul. Power 15, 806–902.
Soave, G. 1972 Equilibrium constants from a modified Redlich-Kwong equation of state.
Chem. Eng. Sci. 27, 1197–1203.
van Konynenburg, P. H. & Scott, R. L. 1980 Critical lines and phase equilibria in
binary van der Waals mixtures. Philos. Trans. Royal Soc. Lond. A 298, 495–540.
van der Waals, J. D. 1893 The thermodynamic theory of capillarity under the hy-
pothesis of a continuous variation of density. Translated by J. S. Rowlinson 1979 J.
Stat. Phys. 20, 197–244.