Dahms

On the transition between two-phase and single-phase interface dynamics in
multicomponent fluids at supercritical pressures

Rainer N. Dahms and Joseph C. Oefelein

Citation: Physics of Fluids (1994-present) 25, 092103 (2013); doi: 10.1063/1.4820346
View online: http://dx.doi.org/10.1063/1.4820346
View Table of Contents: http://scitation.aip.org/content/aip/journal/pof2/25/9?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in
Experimental study of elliptical jet from sub to supercritical conditions
Phys. Fluids 26, 044104 (2014); 10.1063/1.4871483

Direct numerical simulation of the near-field dynamics of annular gas-liquid two-phase jets
Phys. Fluids 21, 042103 (2009); 10.1063/1.3112740

Subcritical to supercritical mixing
Phys. Fluids 20, 052101 (2008); 10.1063/1.2912055

Quantitative x-ray phase-contrast imaging of air-assisted water sprays with high Weber numbers
Appl. Phys. Lett. 89, 151913 (2006); 10.1063/1.2358322

Anomaly of gas drag force on liquid droplets in a turbulent two-phase flow produced by a mechanical jet sprayer
at intermediate Reynolds numbers
J. Appl. Phys. 97, 114901 (2005); 10.1063/1.1905777

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
129.2.129.146 On: Wed, 14 May 2014 20:52:34
PHYSICS OF FLUIDS 25, 092103 (2013)
On the transition between two-phase and single-phase
interface dynamics in multicomponent uids
at supercritical pressures
Rainer N. Dahms
a)
and Joseph C. Oefelein
Combustion Research Facility, Sandia National Laboratories,
Livermore, California 94551, USA
(Received 26 February 2013; accepted 15 August 2013; published online 12 September 2013)
A theory that explains the operating pressures where liquid injection processes tran-
sition from exhibiting classical two-phase spray atomization phenomena to single-
phase diffusion-dominated mixing is presented. Imaging from a variety of experi-
ments have long shown that under certain conditions, typically when the pressure of
the working uid exceeds the thermodynamic critical pressure of the liquid phase,
the presence of discrete two-phase ow processes become diminished. Instead, the
classical gas-liquid interface is replaced by diffusion-dominated mixing. When and
how this transition occurs, however, is not well understood. Modern theory still lacks
a physically based model to quantify this transition and the precise mechanisms that
lead to it. In this paper, we derive a new model that explains how the transition
occurs in multicomponent uids and present a detailed analysis to quantify it. The
model applies a detailed property evaluation scheme based on a modied 32-term
Benedict-Webb-Rubin equation of state that accounts for the relevant real-uid ther-
modynamic and transport properties of the multicomponent system. This framework
is combined with Linear Gradient Theory, which describes the detailed molecular
structure of the vapor-liquid interface region. Our analysis reveals that the two-phase
interface breaks down not necessarily due to vanishing surface tension forces, but
due to thickened interfaces at high subcritical temperatures coupled with an inherent
reduction of the mean free molecular path. At a certain point, the combination of
reduced surface tension, the thicker interface, and reduced mean free molecular path
enter the continuum length scale regime. When this occurs, inter-molecular forces
approach that of the multicomponent continuum where transport processes dominate
across the interfacial region. This leads to a continuous phase transition from com-
pressed liquid to supercritical mixture states. Based on this theory, a regime diagram
for liquid injection is developed that quanties the conditions under which classical
sprays transition to dense-uid jets. It is shown that the chamber pressure required
to support diffusion-dominated mixing dynamics depends on the composition and
temperature of the injected liquid and ambient gas. To illustrate the method and
analysis, we use conditions typical of diesel engine injection. We also present a com-
panion set of high-speed images to provide experimental validation of the presented
theory. The basic theory is quite general and applies to a wide range of modern
propulsion and power systems such as liquid rockets, gas turbines, and reciprocating
engines. Interestingly, the regime diagram associated with diesel engine injection
suggests that classical spray phenomena at typical injection conditions do not occur.
C
2013 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4820346]
a)
Rndahms@sandia.gov. URL: http://crf.sandia.gov/index.php/combustion-research-facility/about-us/contact-us/crf-staff/.
1070-6631/2013/25(9)/092103/24/$30.00 C
2013 AIP Publishing LLC 25, 092103-1
129.2.129.146 On: Wed, 14 May 2014 20:52:34
092103-2 R. N. Dahms and J. C. Oefelein Phys. Fluids 25, 092103 (2013)
FIG. 1. Nonreacting shear-coaxial liquid-nitrogenhelium injector operating at (a) 1.0 MPa and (b) 6.0 MPa. T
N2
= 97 K, T
He
= 280 K into GHe at T = 300 K. Reprinted with permission from W. Mayer, A. Schik, B. Vieille, C.
Chaveau, I. G okalp, D. Talley, and R. Woodward, J. Propul. Power 14, 835 (1998). Copyright 1998, American Institute of
Aeronautics and Astronautics.
I. INTRODUCTION
Liquid injection in systems where the working uid exceeds the thermodynamic critical pressure
of the liquid phase is not well understood. Depending on pressure, injected jets can exhibit two
distinctly different sets of evolutionary processes. At lowsubcritical pressures, the classical situation
exists where a well-dened molecular interface separates the injected liquid from ambient gases and
causes the presence of surface tension. Interactions between dynamic shear forces and surface
tension promote primary atomization and secondary breakup processes that evolve from a dense
state, where the liquid exists as sheets laments or lattices intermixed with sparse pockets of gas; to
a dilute state, where drop-drop interactions are negligible and dilute spray theory can be used. As
ambient pressures approach or exceed the critical pressure of the liquid, however, the situation may
become quite different. Under these conditions, interfacial diffusion layers can develop, apparently
as a consequence of both vanishing surface tension and locally diminishing gas-liquid interfaces.
The lack of inter-molecular forces and a distinct interfacial structure promotes diffusion dominated
mixing processes prior to atomization. As a consequence, injected jets evolve in the presence of
exceedingly large but continuous thermo-physical gradients in a manner that is markedly different
from the classical assumptions.
Mayer et al.,
13
were one of the rst to show the distinct changes described above. These
ow visualization studies have illustrated the trends for liquid-nitrogengaseous-helium (LN
2
GHe) and liquid-oxygengaseous-hydrogen (LOXH

2
) shear-coaxial injector elements. The two
extremes are shown in Fig. 1. Note that the critical pressure and temperature of nitrogen are p
c, N2
= 3.396 MPa and T
c, N2
= 126.2 K, respectively. The critical pressure and temperature of helium
are p
c, He
= 0.227 MPa and T
c, He
= 5.2 K. When LN
2
is injected at low-subcritical pressures
(Fig. 1(a)) atomization occurs forming a distinct spray. Ligaments are detached from the jet surface
forming spherical drops that subsequently breakup and vaporize. As the chamber pressure exceeds
the thermodynamic critical pressure of the LN
2
(Fig. 1(b)), the number of drops present diminishes.
For this situation, the injected LN
2
jet exhibits diffusion dominated mixing processes.
Research performed since has provided additional insights related to the structure and dynamics
of multiphase ows at high pressures. For example, Chehroudi, Talley, and Coy
4
have performed
a wide range of visualizations where cryogenic liquids at subcritical temperatures were injected
into an environment at supercritical temperatures and various pressures ranging from subcritical
to supercritical values. Pure nitrogen and oxygen were injected into environments composed of
nitrogen, helium, argon, and various mixtures of carbon-monoxide and nitrogen. At low subcritical
129.2.129.146 On: Wed, 14 May 2014 20:52:34
chamber pressures, the jets showed surface irregularities that amplied downstreamexhibiting intact,
shiny, but wavy surface features that eventually broke up. Increasing the pressure at constant jet initial
and ambient temperatures caused the formation of many small drops. As the chamber pressure was
further increased, the transition to a full atomization regime was inhibited near but slightly below the
critical pressure. The jet structure at this point changed and began to resemble a turbulent gas jet with
no detectable drops. The reason was attributed to the reduction of surface tension and enthalpy of
vaporization as the critical pressure of the injected liquid is approached. The initial divergence angle
of the jet was measured at the jet exit and compared with the divergence angle of a large number of
other mixing layer ows, including atomized liquid sprays, turbulent incompressible gaseous jets,
supersonic jets, and incompressible but variable density jets.
Chehroudi et al.
4
plotted the divergence angle for the cases described above over four orders of
magnitude in the gas-to-liquid density ratio. This was the rst time such a plot has been reported over
this large range. Noteworthy is that this is also the rst time that hydrocarbon fuels were included
in the analysis, thus acknowledging that such phenomena might also be present at diesel engine
conditions. At and above the critical pressure of the injected liquid, jet growth rate measurements
agreed quantitatively with the theory for incompressible but variable density gaseous mixing layers.
This provided a quantitative parameter to demonstrate that the similarity between the two ows
extends beyond a mere qualitative physical appearance. Finally, as the pressure was reduced to
progressively more subcritical values, the spreading rate approached that measured by others for
liquid sprays. Chehroudi et al.
4
also developed an analysis based on the characteristic separation
time of the jet to form a drop and the characteristic time of vaporization assuming the validity of
the d
2
-law. The analysis indicated that as the chamber pressure approaches the critical pressure
irregularly shaped surface bulges form on the jet that dissipate before separating into isolated drops.
Many other works have served to corroborate the observations described above.
520
Most have
been done in the context of liquid-rocket propulsion, which involves direct injection of liquid fuel
and/or oxidizer into the combustion chamber. However, the observed trends are equally valid for
other liquid fueled devices. In diesel engines, Siebers
21
investigated liquid phase penetration over a
wide range of conditions using Mie-scattered light imaging. Based on these ndings, he proposed a
scaling law for single-component liquid-phase fuel penetration by applying jet theory to a simplied
model of a spray.
22
Close agreement between the scaling law and measured liquid length data
was demonstrated. The results presented, including the negligible effect of fuel injection pressure
and the linear dependence of orice diameter on liquid penetration, indicated that vaporization
was limited by mixing processes. Building on Siebers
22
mixing-limited scaling law, Musculus and
Kattke
23, 24
developed a transient one-dimensional jet model to investigate the entrainment waves in
decelerating diesel jets. Model predictions of substantially increased mixing during the ramp-down
of injection rate were shown to be responsible for many observed fundamental processes in diesel
engines including over-mixed regions and rapid stagnation of mixtures near the injector. This study
indicated that the underlying uid-mechanical processes of decelerating diesel jets are similar to
those of single-phase transient gas jets. In gas jets, entrainment (compared to the steady jet phase) is
substantially increased after the transient deceleration phase.
2527
Segal et al.
28, 29
presented planar
laser-induced uorescence imaging of peruoroketone jets, which is a 3M product chosen for its
good spectroscopic properties and its low critical point (p
c
= 18.4 atm, T
c
= 441 K). Liquid
peruoroketone was injected at supercritical chamber pressures (20 bar < p < 30 bar) into gaseous
nitrogen. Nitrogen temperatures ranged from subcritical and supercritical values with respect to the
liquid-phase critical temperature (383 K < T < 538 K). At subcritical temperatures, droplet and
ligament formation was observed. At lower injection temperatures, drop formation persisted. At
higher injection temperatures, however, the jet surface became smoother and drops were no longer
detectable.
Despite this work, the absence of drops under some high-pressure conditions still poses many
fundamental questions. For a pure uid, basic theory dictates that surface tension forces become
diminished once the operating pressure exceeds the critical pressure. For multicomponent liquid
injection in typical propulsion and power systems, however, this is not necessarily the case and
modern theory still lacks a rst principle explanation for the observed phenomena. According to
fundamental physics, surface tension forces do diminish when the temperature of a multicomponent
129.2.129.146 On: Wed, 14 May 2014 20:52:34
liquid phase approaches its critical value. However, in many cases, as will be demonstrated, the
enthalpy contained in unburnt ambient gases is not sufcient to heat the gas-liquid interface to this
critical temperature. Additionally, in multicomponent uids, elevated pressure alone also does not
provide a complete physical explanation for the observed trends.
In contrast to pure uids, surface tension forces in multicomponent uids do not diminish
simply because the critical pressure of the liquid phase is exceeded. In 1969 and 1971, Faeth and
co-workers
30, 31
proposed, based on research into bipropellant droplet combustion, that the critical
mixing pressure required for supercritical combustion of hydrocarbons in air to be about 2-3 times
the critical pressure of the hydrocarbon. More recently, however, experimental evidence of substantial
surface tension forces in various binary hydrocarbon/nitrogen mixtures at much higher pressures
(p p
C, L
, p > 500 bar) is provided in the literature.
3234
These experiments show that surface
tension forces slowly decrease with increasing pressure. For example, surface tension forces only
decrease by about 50% for a N
2
C
7
H
8
mixture at T =313.13 K even when the pressure is increased
from 50 to 400 bars; e.g., compare Fig. 7 in Ref. 34. Informed by the fundamental relationship
between pressure and surface tension forces, the increase of pressure in Fig. 1 by p = 50 bar can
therefore only account for a small reduction of surface tension forces (roughly about 20%). Such a
moderate reduction of surface tension due to a pressure increase, however, cannot fully explain the
substantial change in observed liquid injection dynamics as they transition from classical sprays to
dense-uid jets. Here, we showthat such a change is in fact induced by a breakdown of the molecular
two-phase interface.
In this paper, we present a fundamental theory and companion analysis to explain and quantify
the change in the interfacial dynamics that leads to the transition between non-continuum jump
conditions exhibited by classical two phase ows and continuous diffusion dominated mixing
processes that have been widely observed at supercritical pressures. Based on the previous work by
Dahms et al.,
35
a comprehensive model based on multicomponent thermodynamics and Gradient
Theory
36, 37
is derived to analyze the detailed structural characteristics of representative gas-liquid
interfaces that format typical liquid injection conditions. Gradient Theory was established by van der
Waals
38, 39
in 1893 and reformulated later by Cahn and Hilliard
40
in 1958 to compute the physical and
continuous variation of uid properties across a thin molecular vapor-liquid interface. The validity
of this approximation has recently been conrmed by statistical analysis of molecular dynamic
simulations of vapor-liquid interface proles. Comparisons demonstrate excellent agreement with
predictions from Gradient Theory.
4143
The capability of gradient theory to efciently calculate the
physical (mean) molecular structure of vapor-liquid interfaces is one of the main reasons for its
application in this analysis.
In the following, the theoretical framework outlined above is applied at injection conditions
relevant to diesel engine operating conditions as an illustrative example. Our analysis shows that,
under certain high-pressure conditions, the structure of the gas-liquid interface is not determined by
vapor-liquid equilibrium and the presence of surface tension forces. Instead, it enters the continuum
length scale regime due to a combination of reduced mean free molecular path and broadened
interfaces. Under these circumstances, the interface essentially disintegrates. The injected liquid
does not evaporate, but instead diffuses into the ambient gas. After demonstrating the mechanisms
responsible for the transition between two-phase and single-phase interfacial dynamics, we then
quantify how and when it occurs by constructing a regime diagram for liquid injection using
a representative set of conditions. Here, again we use diesel engine operating conditions as a
representative system. In addition, we show a companion set of high-speed images to provide
experimental validation of the presented theory. Interestingly (and contrary to conventional wisdom),
the derived regime diagram suggests that the classical view that spray atomization occurs at typical
diesel engine fuel injection conditions is questionable.
II. MODEL FORMULATION
To establish the theory and analyze the interfacial dynamics described above, we develop a
coupled system of models as follows. First, a detailed property evaluation scheme based on a 32-
term Benedict-Webb-Rubin (BWR) equation of state (EOS) is applied along with nonlinear mixing
129.2.129.146 On: Wed, 14 May 2014 20:52:34
rules to account for multicomponent mixture states over the range of pressures and temperatures
of interest.
44, 45
We refer to this as the real-uid model. Second, the real-uid model is used
in combination with vapor-liquid equilibrium theory to obtain representative gas-liquid interface
states during selected fuel injection conditions. Third, Linear Gradient Theory (LGT) is applied to
construct the detailed molecular interface structure, which is required to calculate local values of
surface tension and interface thickness. Following is a description of this system of models.
A. Real-uid model for multicomponent mixtures
The real-uid property evaluation scheme used here is designed to account for thermodynamic
non-idealities and transport anomalies over a wide range of pressures, temperatures, and mixture
states. The scheme is comprehensive and intricate, thus only a skeletal description can be given here.
Details are presented elsewhere by Oefelein.
44, 45
The extended corresponding states model
46, 47
is employed using a modied 32-term BWR equation of state to evaluate the pressure-volume-
temperature (PVT) behavior of the inherent dense multicomponent mixtures. Use of modied BWR
equations of state in conjunction with the extended corresponding states principle has been shown to
provide consistently accurate results over the widest range of pressures, temperatures, and mixture
states, especially at saturated and near-critical conditions. A major disadvantage of the BWR equa-
tions of state, however, is that they are not computationally efcient. Cubic equations of state can
be less accurate, but are computationally efcient. A summary of the various equations of state and
recommended constants is given by Reid et al.
48
Chap. 3. Because of the higher level of accuracy
and analytical nature of the current work, we employ the BWR equation of state throughout in the
results presented.
Having established an analytical representation for real mixture PVT behavior, thermodynamic
properties are obtained in two steps. First, respective component properties are combined at a xed
temperature using the extended corresponding states methodology to obtain the mixture state at
a given reference pressure. A pressure correction is then applied using departure functions of the
form given by Reid et al.
48
Chap. 5. These functions are exact relations derived using Maxwells
relations (see, for example, VanWylen and Sonntag
49
Chap. 10) and make full use of the real
mixture PVT path dependencies dictated by the selected equation of state. Standard state properties
are obtained using the databases developed by Gordon and McBride
50
and Kee et al.
51
Chemical
potentials and fugacity coefcients are obtained in a similar manner. Likewise, viscosity and thermal
conductivity are obtained using the extended corresponding states methodologies developed by
Ely and Hanley.
52, 53
Mass and thermal diffusion coefcients are obtained using the methodologies
outlined by Bird et al.
54
and Hirschfelder et al.
55
in conjunction with the corresponding states
methodology of Takahashi.
56
B. Vapor liquid equilibrium theory
The real-uid model for multicomponent mixtures is applied to obtain vapor-liquid equilibrium
conditions. Assuming that the liquid (L) is in thermodynamic equilibrium with its vapor (V), the
equilibrium conditions are
T
V
= T
L
,
p
V
= p
L
,
V
i
=
L
i
,
(1)
with T as the temperature, p as the pressure, and
i
as the chemical potential of species i. In contrast
to cubic equations of state, the BWR equation of state has been shown to accurately predict saturated
liquid properties (e.g., density) over the widest range of pressures and temperatures. Therefore,
accurate vapor-liquid equilibrium conditions can be computed, particularly at pressures above the
critical pressure of the liquid phase as performed here, when solubility effects have to be considered.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
C. Linear Gradient Theory for gas-liquid interface structures
Gradient theory provides a widely accepted methodology to calculate detailed interface struc-
tures between gases and liquids.
3640
The foundation of this theory was established by van der
Waals
38, 39
in 1893 and reformulated later by Cahn and Hilliard
40
in 1958. It provides a thermo-
mechanical model of continuous uid media that converts statistical mechanics of inhomogeneous
uids into a nonlinear boundary value problem. At equilibrium (as applied in this paper), the model
has been shown in detail to be equivalent to mean-eld molecular theories of capillarity and it has
been successfully applied to a wide variety of uids: e.g., hydrocarbons and their mixtures, polar
compounds and their mixtures, polymer and polymer melts, vapor-liquid and liquid-liquid inter-
faces. Most recently, Gradient Theory has been successfully compared to Monte Carlo simulations
of vapor-liquid and liquid-liquid interfaces.
4143
The theory proved successful in capturing both
surface tension and details of vapor-liquid interfacial structures.
To simplify the calculations without loss of validity, we employ LGT. LGT is derived from
Gradient Theory by assuming a linearized minimization function of the Helmholtz free energy
density distribution across the vapor-liquid interface for the calculation of the interfacial density
proles. It has proven successful for calculating binary and multicomponent interface states of the
kind considered here.
57, 58
The Helmholtz free energy density for a mixture is expanded as a Taylor
series and truncated at lower spatial derivatives of density
F =
_
_
_
f
0
(
M
) +
i, j
1
2
i j
M,i
M, j
_
_
ds (2)
with F dened as the Helmholtz free energy, f
0
() the Helmholtz free energy density of a ho-
mogeneous uid,
M
the molar density, and s the volume unit. The inuence parameter of the
inhomogeneous uid is denoted and carries the molecular structure information of the interface.
This in turn determines the density gradients response to local deviations of the chemical potentials
from their bulk value. The interfacial prole of a planar multicomponent gas-liquid interface in
equilibrium is then obtained by assuming the inuence parameter to be density-independent
59
and
by minimizing the Helmholtz free energy according to the equation
36, 37
j
1
2
i j
d
M,i
dz
d
M, j
dz
= (
M
)
s
(3)
with z as the normal interface direction and
s
= p
s
as the equilibrium pressure. The grand
thermodynamic potential energy density is dened as
(
M
) = f
0
(
M
)
M,i

i,B
(4)
with
i, B
as the chemical potential of species i in the bulk phase. According to Gradient Theory, the
surface tension is calculated as
=
i j
d
M,i
dz
d
M, j
dz
dz. (5)
To avoid solving a boundary value problem on the innite interval, the equations are transformed
from spatial to density coordinates using a reference component that is assumed to be a monotonic
function across the interface. The transformation rules are as follows:
d
i
dz
=
d
i
d
d
dz
,
d
2
i
dz
2
=
d
_
d
i
d
_
d
d
dz
d
dz
+
d
i
d
d
_
d
dz
_
d
d
dz
.
(6)
129.2.129.146 On: Wed, 14 May 2014 20:52:34
The surface tension and spatial interface dimension z can then be calculated once the species density
proles within the interface are known
=
I,L
_
I,V
_2 ( (
M
)
s
)
i j
d
i
d
I
d
j
d
I
d
I
(7)
and
z = z
0
+
I
_
I,0
i j
d
i
d
I
d
j
d
I
2 ( (
M
)
s
)
d
I
(8)
with subscripts I as the reference component. The cross inuence parameter
ij
is obtained as
i j
= (1
i j
)
j
(9)
with
ij
as the binary interaction parameter set to zero
3234
resulting in the geometric mean
i j
= 0. (10)
The inuence parameter
i
of pure species i is calculated following Lin et al.
36
ln
_

i
a
i
b
2/3
i
N
8/3
A
_
= c
0,i
+c
1,i
ln(t
R
) +c
2,i
[ln(t
R
)]
2
(11)
with c
0
, c
1
, c
2
dened as the pure species correlation coefcients, t
R
= 1 T/T
c
as a reduced
temperature function, N
A
as the Avogadro constant, and a and b as coefcients dependent on the
species critical temperature, critical pressure, and acentric factor, respectively.
The boundary conditions of the bulk vapor and liquid phases, referred to by indices V and L in
Eq. (7), are obtained from real-uid multicomponent vapor-liquid equilibrium calculations. Then,
the characteristic interface thickness
VLE
is calculated following the criterion from Lekner and
Henderson
60
VLE
=
0.9
_
0.1
i j
d
i
d
I
d
j
d
I
2
_
f
0
(
M
)
M,i

i

s
_ d. (12)
In Gradient Theory, the species density proles are obtained in relation to the chosen reference
species density using nonlinear equations. In Linear Gradient Theory, however, a linear relationship
between the reference and the species component is assumed as follows:
d
i
d
I
=

i,L

i,V
I,L

I,V
. (13)
Because of the small length scales associated with vapor-liquid interface structures, the one-
dimensional approximation applied in Gradient Theory is valid during transient liquid injections
up to the time of the breakdown of the interface. This period represents the critical time in the
understanding and quantifying the transition between classical two-phase phenomena and dense-
uid jet dynamics. The interface length scale is orders of magnitude smaller than the Kolmogorov
scale. Therefore, interactions between the interface and three-dimensional turbulent ow eld are
negligible. In addition, the curvature of the local interface is assumed to have only a minor inu-
ence on the interface structure and surface tension. This conclusion is supported by a negligible
pressure difference across the vapor-liquid interface and the small length scale ratio associated with
interface thickness and ow induced curvature. After the time of the breakdown of the vapor-liquid
interface, however, the interface substantially grows in thickness and develops into a turbulent
mixing layer. Then, a multidimensional model becomes necessary to appropriately describe the
diffusion-dominated gas-liquid mixing dynamics.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 2. Density contours of pure n-dodecane showing the saturation line, critical point (T
c
658.1 K, p
c
18.17 bar), and
key thermodynamic regimes.
III. RESULTS AND DISCUSSION
To establish the theory for the transition between two-phase and single-phase interface dynam-
ics, we present a comprehensive thermodynamic analysis using representative multicomponent uid
mixtures and conditions. Both single-phase mixing and two-phase interface processes are investi-
gated to understand and quantify at what conditions transition occurs. To achieve this goal, real-uid
vapor-liquid equilibrium conditions are calculated and then combined with LGT to reconstruct the
molecular vapor-liquid interface structure for a wide range of pressures and temperatures. The real-
uid calculations then provide a framework to quantify the precise state of the interface as a function
of specic injection conditions (i.e., pressure, local mixture composition, and temperature). Building
on this analysis, a regime diagram is introduced that quanties the conditions under which classical
sprays transition to dense-uid jets. To illustrate, we use n-dodecane and nitrogen as a representative
liquid-fuel and ambient gas. To provide a baseline pure-uid reference, Fig. 2 presents the density
contours of pure n-dodecane in a pressure-temperature diagram. These data were computed using
the BWR equation of state described in Sec. II A. The diagram shows the saturation line, which
separates the liquid and vapor states. Key thermodynamic regimes are then dened accordingly. The
compressed liquid regime is dened as being supercritical with respect to pressure and subcritical
with respect to temperature. The ideal gas regime is found in regions of supercritical temperature but
subcritical pressure. The supercritical state is then dened for a uid at both supercritical tempera-
ture and pressure. Using the equation of state and corresponding mixing rules, other thermodynamic
properties for mixtures such as the compressibility factor, enthalpy, entropy, heat capacities, and
fugacity coefcients can be obtained in a similar manner. This methodology is used to calculate
vapor-liquid equilibrium conditions as follows.
Figure 3 presents typical vapor and liquid equilibrium compositions for a binary system con-
taining n-dodecane and nitrogen at different operating pressures. The calculations show that the
critical mixing temperature, above which vapor-liquid equilibrium cannot exist anymore, decreases
with increasing pressure due to an increase of dissolved nitrogen into the liquid phase. Figure 4
shows that this decrease in critical mixing temperature scales linearly with pressure. Similarly,
the mole fraction of n-dodecane (denoted X
C12
) at the critical condition decreases asymptotically
with pressure. The combination of the equation of state with vapor-liquid equilibrium calculations
provide the compositions and mixture states of the respective vapor and liquid phases at given tem-
perature and pressure conditions. However, such calculations do not provide information about the
spatial distribution of these compositions across the interface, the structure and size of the interfacial
region, or the value of surface tension within. Without this information, no conclusions can be drawn
regarding the nature of the transitional processes of interest.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 3. Vapor-liquid equilibrium conditions for a binary system of n-dodecane and nitrogen at different pressures as a
function of the mixture fraction. Note that the critical mixing temperature decreases with increasing pressure.
To understand the nature of the interface dynamics, LGT presented in Sec. II C is applied using
the equilibrium phase states as the boundary conditions while assuming a constant temperature
distribution across the interface. The accuracy of the method is demonstrated for systems similar
to those considered here in Fig. 5. This gure presents calculations of surface tension for a binary
iso-octanenitrogen system as a function of pressure and at a constant interface temperature of
T =313 K. It is shown that the computed values agree well with measured values.
61
This agreement
between experiments and calculations is representative to that made by other authors.
57, 58
Respective
values of surface tension for a n-dodecanenitrogen system at the same pressures and temperature
are also shown for comparison. These results show that surface tension forces do decrease with
increasing pressure. However, the results also highlight a major point. In contrast to pure uids, the
value of surface tension forces in multicomponent mixtures does not diminish simply because the
critical pressure of the liquid phase or the critical mixing pressure, as dened by Faeth et al.
30, 31
to be 2-3 times the critical pressure of the hydrocarbon, is exceeded.
In Fig. 6, LGT is applied to reconstruct the density proles of n-dodecane and nitrogen mixtures
across their vapor-liquid interface for different conditions. The calculations show that vapor-liquid
equilibrium conditions are only required to be fullled across a very thin molecular layer. Based
on this analysis, it is justied to use the equilibrium vapor pressure as a boundary condition in
conventional multiphase theory even during the injection transient. This conclusion is consistent
with earlier results presented by Bellan and Summereld.
6264
Outside of these thin layers, the
temperature of the liquid quickly approaches the fuel injection temperature in the liquid core while
the gaseous conditions approach the ambient mixture state. Figure 6 (left) presents mixture density
FIG. 4. Saturation temperature of the liquid and critical mixing temperature and composition resulting from vapor-liquid
equilibrium calculations as a function of pressure for a n-dodecanenitrogen system.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 5. Comparison of measured and calculated values of surface tension for a binary iso-octanenitrogen mixture as a
function of pressure, at a constant temperature of T = 313 K. For comparison, respective values for a n-dodecanenitrogen
system are also shown.
proles across the vapor-liquid interface at T = 450 K and for two different pressures: p = 30 bar
and p = 60 bar, respectively. The interfacial structure broadens with increasing pressure. Similarly,
the density of the vapor increases, while pressure has only a minor effect on the density of the liquid.
The same trend is observed in Fig. 6 (right), but the prole variations caused by a change in pressure
have broadened due to the elevated interface temperature.
The effect of temperature on the interface is studied in more detail in Fig. 7. The left plot
presents mixture density proles at a constant pressure of p = 60 bar and variations in temperature.
The interfacial prole is shown to broaden substantially with increasing temperature. This effect is
quantied in the right plot by calculating representative interface thicknesses using Eq. (12). The
analysis shows that the interface thickness increases exponentially with temperature.
Having established a framework that computes the detailed vapor-liquid equilibrium interface
structures, we now add physical complexity to analyze the non-equilibrium problem of gases and
liquids interacting at different temperatures. We calculate a set of representative mixing lines in
mixture fraction space. The goal is to obtain the relationship between temperature and mixture
fraction for different mixtures of ambient gas and vapor. Here, the enthalpy of vaporization is
included assuming adiabatic mixing using an enthalpy distribution that is linear in mixture fraction.
The adiabatic mixing assumption is supported by earlier works from Siebers,
21, 22
who presented (a)
a mass and energy balance analysis for diesel injection that results in vapor temperatures consistent
FIG. 6. Sensitivity of the vapor-liquid interfacial structure to variations in pressure and temperature. Two different pressures
(p =30 bars and p =60 bars) are investigated at interface temperatures of T =450 K(left) and T =550 K(right), respectively.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 7. Sensitivity of the vapor-liquid interfacial structure to variations in temperature at a constant pressure of p = 60 bar.
The interface broadens substantially with an increase in temperature (left). Representative interface thicknesses are calculated
using Eq. (12) and are shown to increase exponentially with temperature (right).
with the adiabatic mixing assumption and (b) developed a scaling law based on this assumption for
single-component liquid-phase fuel penetration. Results demonstrated a close agreement with liquid
length data obtained from Mie-scattered light imaging. Furthermore, Dec et al.
6567
and Pickett
et al.
68, 69
presented Rayleigh scattering temperature and equivalence ratio measurements in the
gaseous phase of diesel fuel injections. The measurements were consistent with both the predicted
vapor temperatures and the assumptions being made using a linear enthalpy distribution in mixture
fraction space.
Using the assumptions above, the BWR equation of state is applied to calculate the temperature
for each composition determined by the mixture fraction in order to match this prescribed linear
enthalpy distribution. To illustrate, Fig. 8 shows the resulting enthalpy and constant pressure heat
capacity at conditions similar to the Spray A nonreactive case dened as part of the Sandia
National Laboratories Engine Combustion Network (see www.sandia.gov/ECN).
69, 70
N-dodecane
is injected at a temperature T
F
= 363 K into nitrogen at a temperature T
A
= 900 K and a chamber
pressure p = 60 bar. The corresponding mixing temperatures are then given in Fig. 9, where we
show n-dodecane injected at a constant temperature T
F
= 363 K into nitrogen at (a) T
A
= 450 K
and p = 30 bar and (b) T
A
= 900 K and p = 60 bar, respectively.
Superimposing the two mixing lines shown in Fig. 9 onto the corresponding vapor-liquid
equilibrium curves now poses a potentially interesting dilemma. As shown in Fig. 10, the two
different chamber pressures dene the envelope of temperature and mixture conditions where vapor-
liquid equilibrium is possible. Likewise, the adiabatic mixing temperatures dene the envelope of
conditions under which a vapor state is possible. The representative interface state for both cases
is found at the intersection points T
1
and T
2
between adiabatic mixing and vapor equilibrium state.
FIG. 8. Enthalpy and constant pressure heat capacity as a function of mixture fraction for n-dodecane injected as a liquid at
T
F
= 363 K into gaseous N
2
at T
A
= 900 K and p = 60 bars.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 9. Adiabatic mixing temperature distributions for n-dodecane injected at T
F
= 363 K into nitrogen at (a) T
A
= 450 K
and p = 30 bar and (b) T
A
= 900 K and p = 60 bar, respectively.
Interestingly, the analysis shows that the enthalpy of the mixture is not sufcient to heat the liquid
to its critical value. The mixing line required to actually obtain such critical conditions is also
shown in Fig. 10, which illustrates its nonphysical nature. These results suggest that, for many high-
temperature and high-pressure conditions, a distinct two-phase interface should exist. However, this
appears to be in conict with experimental evidence of vanishing interfaces at some supercritical
pressure conditions.
13
To better understand the implications of Fig. 10, our next goal is to reconstruct the details of the
interface itself using LGT. To apply LGT, we must apply appropriate boundary conditions for the
liquid and vapor phases on either side of the interface. Here, we assume that the temperature
across the interface is constant and can be represented by the intersection points T
1
and T
2
in
Fig. 10. Assuming that the pressure at the interface is equal to the chamber pressure, we dene
the two different chamber conditions as the high-temperature interface and low-temperature
interface states, respectively. Variations of the interface condition as a function of chamber pressure
are then investigated.
Figure 11 presents the reduced interface temperature for different chamber pressures. The
reduced interface temperature is dened as the actual temperature of the interface T normalized
FIG. 10. Superimposition of the adiabatic mixing lines shown in Fig. 9 onto the corresponding vapor-liquid equilibrium
curves showing intersection points T
1
and T
2
between adiabatic mixing and vapor equilibrium state temperatures.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 11. Variations of reduced interface temperatures as a function of chamber pressure and temperature.
by its critical temperature T
c
. Previously, T
c
has been shown to decrease linearly with pressure,
as shown in Fig. 4. The analysis here shows that higher chamber pressures lead to higher reduced
interface temperatures. This pressure effect increases at higher ambient gas temperatures.
In addition to boundary conditions, it is necessary to quantify the thickness of the interface
relative to the continuum approximation. The reasons for this will become apparent in what follows.
Figure 12 shows the variation of the molecular mean free path at the interface with increasing
pressure. The molecular mean free path, denoted as , is evaluated using the vapor state composition
x
and by assuming the validity of the ideal gas law in the derivation of the number of molecules per
unit volume as follows:
|x
=
k
B
T
2pd
2
(14)
with k
B
dened as the Boltzmann constant, T the interface temperature, p the bulk pressure, and d
the hard sphere diameter of the average molecule. The average size of a molecule is calculated as
d =
i
X
i
d
i
(15)
with X
i
as the species mole fractions (as provided by the vapor equilibrium composition) and d
i
as
the respective molecular size of each species in the system. This gure shows that the vapor state
shifts toward leaner mixtures with increasing pressure and constant ambient temperature. Thus, the
average molecule size d decreases due to a reduced concentration of large-size fuel components in
the vapor. This pressure effect increases at higher ambient gas temperatures. Furthermore, it was
previously shown that higher pressures also lead to higher interface temperatures, compare Fig. 11.
FIG. 12. Variation of the molecular mean free path at the interface, evaluated for the vapor equilibrium state presented in
Fig. 10, as a function of chamber pressure.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 13. Interfacial density proles and thicknesses for the low-temperature (left) and high-temperature (right) interface
states, as calculated using Linear Gradient Theory. Corresponding mean free paths are shown for reference. Interestingly,
the Knudsen-number criterion indicates that the high-temperature interface falls within the uid mechanic continuum regime
(Kn <0.1) whereas a classical molecular interface is exhibited by the low-temperature interface. Notice that Gradient Theory
is fully valid to calculate the physical mean density prole of molecular vapor-liquid interfaces.
Both of these pressure effects increase the mean free path. However, as shown in Fig. 12, they cannot
compensate for the dominating direct effect of pressure prescribed by Eq. (14). Thus, the molecular
mean free path is substantially shorter in the high-temperature interface than in the corresponding
low-temperature interface.
With the analysis above in place, we can now focus on the interfacial structure of our test
cases using LGT. Figure 13 shows the calculated results. At the high-temperature condition, the
interface shows substantially reduced values of density differences between vapor and liquid states
and is distinctively thicker than at the corresponding low-temperature condition. The relevance
of this thickness is highlighted by comparing it to the molecular mean free path, also shown in
Fig. 13. Fluid particles equilibrate over distances comparable to the molecular length scales. Since
the mean free path at the low-temperature interface is larger than the interface thickness, its interfa-
cial processes are governed by molecular dynamics and the temperature of the vapor T
V
equilibrates
with the temperature of the liquid T
L
to the value of the interface temperature obtained from Fig. 10
(T
2
=T
V
=T
L
). This justies the vapor-liquid equilibrium assumptions. The low-temperature inter-
face exhibits surface tension forces, supports evaporation and latent heat of vaporization phenomena,
and is therefore expected to lead to the classical spray phenomena as suggested by two-phase the-
ory. The situation for the high-temperature interface, however, is reversed. The mean free path has
become more than an order of magnitude shorter than the interface thickness. For this situation,
vapor-liquid interfacial processes are not governed by molecular dynamics anymore. The interface
enters the continuum regime and thus the heat ux across the interface becomes Fickian in nature.
Since the ambient gas temperature is substantially higher than the liquid temperature, the tempera-
ture equilibriumassumption between vapor and liquid (Eq. (1)) at the interface temperature obtained
from Fig. 10 also breaks down (T
1
= T
V
= T
L
). Vapor-liquid equilibrium assumptions and classi-
cal two-phase theory do not apply. Then, the thermodynamic value of surface tension forces (i.e.,
5.5 mN/m for the example here) diminishes as the thickness of this interface substantially
increases.
The two distinctively different interfacial conditions shown in Fig. 13 can be classied using
the Knudsen-number criterion
Kn =

< 0.1 Continuum regime. (16)

Here, is calculated from Eq. (14) and is the corresponding interface thickness calculated from
Eq. (12). Knudsen numbers less than approximately 0.1 characterize the continuum regime,
as governed by the Navier-Stokes equations. Knudsen numbers greater than 0.1 characterize
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 14. Plot (a) is constructed using the same conditions as in Fig. 13 as a reference and shows three different adiabatic
mixing lines, the corresponding vapor-liquid equilibrium curves, and corresponding interfacial Knudsen-numbers. Note
that these conditions have been selected so that the ambient gas density and fuel injection temperature are constant. At T
A
805 Kand p
A
52 bar, the interface Knudsen-number is Kn =0.1, which approximately marks the region where molecular
interfaces transition to the continuum regime. Plot (b) illustrates the consequences. At conditions where a molecular interface
exists, the adiabatic mixing line is only valid up to the point where it intersects with the vapor-liquid equilibrium curve. For
continuous interfaces, however, the adiabatic mixing line is valid over the entire mixture fraction space.
non-continuum molecular scales. Applying this criterion to the results in Fig. 13 suggests for
the low-temperature interface, where Kn 1.5, that classical two-phase theory applies. The same
criterion applied to the high-temperature interface, where Kn 0.05, implies that continuum as-
sumptions are valid and a molecular two-phase boundary does not exist. Instead, the governing
conservation equations apply across the interface. This nding has substantial implications, which
are illustrated in Fig. 14.
Using the same conditions as in Fig. 13 as a reference, Fig. 14(a) shows three different adia-
batic mixing and vapor equilibrium temperature intersection points along with their corresponding
interfacial Knudsen-numbers. Note that these conditions have been selected so that the ambient
density and fuel injection temperature are held at constant values of
A
22 kg/m
3
and
T
F
= 363 K. The Knudsen-numbers are shown to decrease substantially with increasing ambi-
ent pressure and temperatures. At T
A
805 K and p
A
52 bar, the interfacial Knudsen-number is
Kn = 0.1, which approximately marks the region where molecular interfaces transition to the con-
tinuum regime. Figure 14(b) illustrates the consequences. The governing conservation equations,
EOS, and adiabatic mixing line do not apply across molecular interfaces and are only applicable up
to the point of intersection where the two-phase jump condition prescribed by vapor-liquid equilib-
rium occurs. This situation is shown for the 30 bar, low-temperature interface in (b). Conversely,
at conditions that support the development of continuous interfaces (the 60 bar, high-temperature
interface in (b)), the governing conservation equations, EOS, and adiabatic mixing line apply across
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 15. PTZ space for mixture states that span from pure nitrogen (Z = 0) to pure n-dodecane (Z = 1) computed using the
modied 32-term BWR mixture state equation.
the interfacial region and are valid over the entire mixture fraction space. Interfacial processes be-
come part of the continuum eld and are thus affected by convection and Fickian diffusion. For the
30 bar case in (b), the critical temperature of the interface is then governed by vapor-liquid equi-
librium equations, which here is T
C
658 K. For the 60 bar case, however, critical properties are
governed by the corresponding state theory and (in a manner consistent with the theory and devel-
opment of the EOS) occur where the rst and second partial derivatives of pressure with respect
to volume are zero, with the derivatives evaluated at constant temperature. The physics of such
continuous gas-liquid interfaces is analyzed in what follows.
Application of the EOS is not valid within molecular-scale vapor-liquid interfaces due to the
presence of intense density gradients, which are governed by molecular dynamics and not the
continuum approximation. Such molecular dynamics invalidate the fundamental assumption of
homogeneous properties that respective EOSs are designed around. Conversely, the development of
continuous interfaces facilitates the assumption of homogeneity. Since it has now been established
that both the composition and temperature vary continuously across the 60 bar, high-temperature
interface (Kn 0.05) in Figs. 13 and 14, the envelope of thermodynamic mixture states can be
further analyzed by extending the classical PVT phase diagram to a 4D PVTZ phase for mixtures,
where Z is the mixture fraction. The corresponding PTZ space is shown in Fig. 15. Here, density
contours are shown along with the critical points for mixture fractions Z = 0 (pure nitrogen) and
Z = 1 (pure n-dodecane), respectively. The equation of state computes pseudo-pure saturation lines
of the mixture in a manner consistent with the extended corresponding states model. These lines
form the pseudo-pure saturation surface of the mixture shown in Fig. 16. The critical line of the
mixture, which represents the envelope of critical temperatures T = T
c
and pressures p = p
c
for
each mixture state Z is located on the top edge of this surface. Along the critical line of the mixture,
the pseudo-pure uid assumption (x
= x
; where x
is the liquid composition, and x
is the vapor
composition) is fullled at the critical point of the multicomponent system. The thermodynamic
regimes of liquid, compressed liquid, vapor, ideal gas, and supercritical uid are then dened in a
manner analogous to the pure uid denition in Fig. 2.
Figure 17 shows both the critical temperature and critical pressure of the mixture for each
mixture state in Z. These properties have been processed directly from the equation of state using
the criteria
_
p
_
T
T
C
,
C
=
_
2
p
2
_
T
T
C
,
C
= 0. (17)
The critical temperature and pressure for pure n-dodecane and nitrogen are then found at mixture
fraction Z = 1 and Z = 0, respectively. For mixture fractions below Z < Z
0.79, the adiabatic

mixing temperature exceeds the critical temperature of the mixture. Hence, saturated mixtures are
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 16. Pseudo-pure uid saturation surface of the mixture for a n-dodecanenitrogen system that denes the critical line
of the mixture and the envelope of saturation points. Results shown are consistent with the assumptions associated with the
extended corresponding states theory.
not possible when the chamber pressure exceeds a pressure of p 29 bar at a mixture value of
Z
0.79 since the mixture is supercritical at these conditions. At lower pressures, a discrete
molecular interface must exist and diffusion-controlled mixing cannot occur.
The distribution of the adiabatic mixing temperature, and therefore the value of Z
, depends on
the chamber pressure. Figure 18 shows the supercritical mixture pressure as a function of ambient
gas temperature for a set of representative liquid fuel injection temperatures. This quantity decreases
with an increase in both liquid fuel injection and ambient gas temperatures. The high-temperature
case used previously (T
A
= 900 K, T
F
= 363 K) is also highlighted. The analysis shows that the
chamber pressure of p = 60 bar exceeds the supercritical mixture pressure (here p
c,m
= 29.6 bar),
which facilitates a diffusion-controlled mixing pathway. Using these data, the thermodynamic regime
diagram shown in Fig. 19 was constructed to analyze the envelope of mixture states associated with
the high-temperature case. The reduced critical pressure of the mixture is dened as the cham-
ber pressure normalized by the critical pressure of the mixture p
c
(Z) shown in Fig. 17. Super-
critical pressures of the mixture are then dened as p
Chamber
/p
c
(Z) > 1.0. Similarly, supercritical
FIG. 17. Adiabatic mixing temperature superimposed with the corresponding critical temperature and pressure of the mixture
for a n-dodecanenitrogen system. For mixture fractions below Z < Z
0.79, the adiabatic mixing temperature exceeds the

critical temperature of the mixture. Hence, saturated mixtures are not possible when the chamber pressure exceeds a pressure
of p 29 bar at a mixture value of Z
0.79 since the mixture is supercritical at these conditions.

129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 18. Supercritical mixture pressure, dened as the necessary chamber pressure to support diffusion-controlled mixing,
as a function of the liquid fuel and ambient gas temperatures.
temperatures are found for mixture states below Z
0.79, compare Fig. 17. Thus, the mixing path

exhibits a continuous phase transition from compressed liquid to supercritical states since it never
crosses the mixture saturation regime and the interface dynamics between the liquid jet and the am-
bient gas are governed by diffusion-controlled mixing. Figure 19 also shows contours and iso-lines
of the compressibility factor which have been computed using the BWR equation of state. They em-
phasize the relevance of the application of the real-uid mixture state equation for fuel-rich mixtures
(Z > 0.5). At leaner mixtures, however, the compressibility factor is close to unity where the ideal
gas law becomes valid. This justies the calculation of the molecular mean free path using Eq. (14)
to evaluate the Knudsen-number criterion dened in Eq. (16).
Akey benet of the ndings presented above is that we can now construct a regime diagram that
quanties under what conditions mixtures will transition from two-phase to single-phase behavior.
FIG. 19. Envelope of thermodynamic mixture states for the high-temperature case. The constant chamber pressure of
p = 60 bar exceeds the critical pressure of all mixture states shown in Fig. 17. For mixtures below Z
0.79, the adiabatic

mixing temperature is supercritical. Thus, the mixing path exhibits a continuous phase transition from compressed liquid to
supercritical states since it never crosses the mixture saturation regime and the interface dynamics between the liquid jet and
the ambient gas are governed by diffusion-controlled mixing. The contour plot and iso-lines show corresponding values of
the compressibility factor obtained using the BWR equation of state.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 20. Regime diagram for n-dodecane injected at a temperature T
F
= 363 K into gaseous nitrogen at varying ambient
pressures and temperatures. Regimes of classical sprays and diffusion-dominated mixing are found based on the Knudsen-
number criterion, dened in Eq. (16), and the supercritical mixture pressure criterion, dened in Fig. 18, respectively.
A representative diagram is given in Fig. 20, which shows results for n-dodecane injected at a
temperature T
F
= 363 K into gaseous nitrogen at varying ambient pressures and temperatures.
The classical spray regime (highlighted in grey) and diffusion-dominated mixing regime (white)
are found using the Knudsen-number criterion dened in Eq. (16) and, similarly, the supercritical
mixture pressure criterion, dened in Fig. 18. Isolines of different Knudsen-number values are shown
that span a reasonable range of uncertainty. Across this range, the variation in Knudsen-number is
not yet substantial enough to predict the actual interface behavior. To illustrate the relevance of
this diagram, three ambient gas pressure-temperature lines, which span a range of conditions during
different diesel engine compression cycles, are also presented. These lines showrepresentative diesel
engine conditions for (a) turbo-charged, (b) mediumload, and (c) light load operation that result from
varying in-cylinder conditions during the compression stroke. The corresponding initial pressures
and temperatures are (a) p = 2.5 bar, T = 363 K, (b) p = 1.6 bar, T = 343 K, and (c) p = 1 bar,
T = 335 K, respectively. Fuel injection then occurs at full compression conditions. Note that the
Knudsen number changes many orders of magnitude along such compression lines between ambient
(Kn O(10
2
)) and engine relevant conditions (Kn < 0.1). Interestingly, the cylinder pressures at
full compression exceed the supercritical mixture pressure for all of the cases considered. Only under
representative light load operation does there appear to be a chance that classical fuel spray injection
takes place. Thus, contrary to conventional wisdom, our regime diagramsuggests that classical spray
phenomena does not occur at typical diesel injection conditions. Instead, the fuel is injected as a
continuous jet with diminished interfacial structure and surface tension forces leading to diffusion-
dominated mixing dynamics. Such mixing layers are largely affected by non-ideal thermodynamics
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 21. Sensitivity of transition regime to fuel injection temperature. Note that the ambient gas temperature required to
support interfacial transition conditions decreases substantially with increasing fuel injection temperature.
and transport processes and cannot support evaporation phenomena or the formation of liquid
ligaments or drops.
For reference, the critical temperature T
c, Liquid
658.1 K and critical pressure p
c, Liquid
18.17 bar of pure n-dodecane are also presented in Fig. 20 showing that the ambient gas con-
ditions under which classical sprays transition to dense-uid jets are only loosely related to the
liquid phase critical properties. To provide further insights, we show the sensitivity of the transition
region as a function of fuel injection temperature in Fig. 21. Here, isolines of Kn = 0.1 along
with their corresponding supercritical mixture pressures are presented for injection temperatures of
T
F
= 363 K, T
F
= 413 K, and T
F
= 463 K. The analysis shows that the ambient gas temperature
required to support interfacial transition conditions decreases substantially with increasing fuel in-
jection temperature. The observed changes with fuel injection temperature decrease with increasing
pressure and increase with the fuel injection temperature itself.
To corroborate the ndings presented above, a companion set of experiments was performed
using a single-hole research injector to provide experimental validation of the theory. The re-
search injector is a solenoid-actuated common rail system with a nominal orice diameter of
d = 0.090 mm. Long-distance microscopic imaging was applied by Pickett and co-workers
35
to
visualize the instantaneous structures of the liquid core along with ligaments and droplets at the
nozzle exit (if present) at the end of injection. Evidence of surface tension forces at end-of-injection
timings under low-temperature and liquid-phase supercritical pressure conditions were observed.
At high-temperature conditions, however, no drops or ligaments were detected. The most recent
set of images shown here were obtained by Pickett and co-workers
35
using an ultra-fast LED and
diffuser along with a long-distance microscopy lens and a high-speed camera. The imaging system
129.2.129.146 On: Wed, 14 May 2014 20:52:34
FIG. 22. High-speed imaging of liquid injection at two liquid-phase supercritical pressure conditions. The regime diagram
corresponds identically to Fig. 20. Observed trends from these images are in agreement with the presented theory. Classical
spray phenomena are exhibited at the low-temperature condition (bottom) while diffusion-dominated mixing dynamics are
observed at the high-temperature condition (top).
allows for a spatial resolution of 4.7 m and the tracking of ligaments and subsequent drops at
different stages of injection. In contrast to the previous study,
35
imaging was performed downstream
(614 mm) of the nozzle exit during the main injection event. Results are presented in Fig. 22. As can
be seen, the experimental images are in good agreement with the predictions of the presented theory.
Classical spray phenomena persist under liquid-phase supercritical pressure and low-temperature
chamber conditions (p = 29 bar, T
A
= 440 K) while a more diffusive mixing process without
the formation of liquid ligaments or drops is observed under high-temperature chamber conditions
(p = 60 bar, T
A
= 900 K). It is important to note that these two conditions were chosen to obtain
constant ambient density conditions at = 22.8 kg/m
3
. By preserving the ambient density, similar
jet penetration and gas-liquid interaction forces are maintained.
To further corroborate our ndings, results of this theory have recently been successfully applied
in a Large Eddy Simulation (LES) of the high-temperature fuel injection process using a dense-uid
approximation.
71
The simulations successfully reproduced quantitative penetration data and the key
qualitative experimental features such as the ow structure and spatial evolution. Application of
LES will be a future and ongoing focal point.
IV. SUMMARY AND CONCLUSIONS
Past works have suggested that two extremes exist with regard to liquid injection in high-
pressure systems. At lower (typically subcritical) pressures, the classical situation exists where a
well-dened interface separates the injected liquid from ambient gases and causes the presence
of surface tension. Under these conditions, the discontinuous non-continuum interface promotes
primary atomization, secondary breakup, and the resultant spray phenomena that have been well
recognized and are widely assumed in a variety of systems. At high-pressure conditions (typically
supercritical with respect to the liquid), the situation can become quite different. Under these con-
ditions, a distinct gas-liquid interface may not exist. Effects of surface tension appear to become
diminished, and the lack of these inter-molecular forces minimizes or eliminates the formation of
drops and promote diffusion-dominated mixing processes prior to atomization. To understand and
explain this transition in the context of multicomponent mixtures, we have developed a system of
129.2.129.146 On: Wed, 14 May 2014 20:52:34
models that describes detailed thermodynamic properties in multicomponent, multiphase mixtures.
We have coupled this to vapor-liquid equilibrium theory and Linear Gradient Theory to under-
stand the structure of the interface and the related change in interfacial properties such as surface
tension.
Using this system of models, we have developed some of the rst physically based explanations
that quantify the change in the interfacial dynamics that leads to the transition between the classical
non-continuum jump conditions associated with two phase ows and the continuous gas-liquid
interfacial diffusion layers. Key ndings include (a) the enthalpy contained in hot unburnt ambient
gases is not sufcient to heat up the gas-liquid interface to its critical temperature and that (b) the
transition between two-phase and single-phase interface dynamics is not necessarily induced by
diminished surface tension forces alone. In fact, in multicomponent systems, surface tension forces
are almost never fully negligible, in contrast to pure uids, simply because the critical pressure
of the liquid phase or a critical pressure of a mixture is exceeded. Instead, continuous gas-liquid
interfacial diffusion layers were shown to develop because they enter the continuum length scale
regime, as veried by an applied Knudsen-number criterion. Under these circumstances, transport
processes dominate over intermolecular forces. The applied theory indicates the primary mechanism
that causes the transition from two-phase to single-phase behavior (i.e., the transition from spray
to diffusive mixing) occurs through combination of three phenomenological factors: (1) broadening
interface thickness, (2) reduced mean free path, and (3) a reduction in surface tension. The reduction
of the mean free path is mainly attributed to high pressure, while the broadening of the interface is
mainly caused by high subcritical gas-liquid interface temperatures.
To demonstrate the quantitative aspect of the methodology, a regime diagramwas introduced that
quanties the conditions under which classical sprays transition to dense-uid jets as a function of
specic injection conditions. This diagram quanties the relationship between the required pressure,
local mixture composition, and temperature of the injected liquid and ambient gas. Conditions
relevant to diesel engines were used as a representative example. A companion set of high-speed
imaging was also presented to provide experimental validation of the presented theory. Interestingly,
the resultant regime diagram suggests that classical spray phenomena do not occur under typical
diesel injection conditions. This implies that the classical view of spray atomization as a conceptual
model for high-pressure diesel fuel injection phenomena comes into question. Instead, a real-
uid approximation that accounts for the substantial thermodynamic non-idealities and transport
anomalies during the fuel injection process is required.
The theory presented is shown to be consistent with (a) available data obtained by molecular
dynamic simulations, (b) surface tension measurements, which establishes that an intact molecular
vapor-liquid interface may exist under certain extreme-pressure conditions (p >20 p
c, Liquid
), and (c)
high-pressure liquid injection experiments showing classical spray phenomena at low-temperature
ambient conditions and dense-uid jet dynamics at high-temperature ambient conditions, respec-
tively. Given the general nature of the theory, the present approach is being applied to a wide variety
of fuel-oxidizer combinations over a wide range of pressures and temperatures to quantify the tran-
sition regimes at conditions relevant to a variety of modern propulsion and power systems (e.g.,
liquid rockets, gas turbines, and reciprocating engines).
ACKNOWLEDGMENTS
Support for this research was provided jointly by the (U.S.) Department of Energy (DOE);
Ofce of Science (SC), Basic Energy Sciences (BES) program; and the Ofce of Energy Efciency
and Renewable Energy (EERE), Vehicle Technologies (VT) program, under Grant Nos. KC0301020
and VT0401000, respectively. Fundamental development of the real-uid model and foundational
property evaluation schemes for multicomponent hydrocarbon mixtures was supported by the SC-
BESprogram. Application of these tools to advanced engine combustion research and development of
multiphase regime diagrams using Gradient Theory was supported by the EERE-VTprogram. Sandia
National Laboratories is a multiprogram laboratory operated by Sandia Corporation, a Lockheed
Martin Company, for the (U.S.) Department of Energy (DOE) under Contract No. DE-AC04-94-
AL85000.
129.2.129.146 On: Wed, 14 May 2014 20:52:34
1
W. Mayer and H. Tamura, Propellant injection in a liquid oxygen/gaseous hydrogen rocket engine, J. Propul. Power 12,
1137 (1996).
2
W. Mayer, A. Schik, B. Vieille, C. Chaveau, I. G okalp, D. Talley, and R. Woodward, Atomization and breakup of cryogenic
propellants under high-pressure subcritical and supercritical conditions, J. Propul. Power 14, 835 (1998).
3
W. Mayer and J. J. Smith, Fundamentals of supercritical mixing and combustion of cryogenic propellants, Prog. Aeronaut.
Astronaut. 200, 339 (2004).
4
B. Chehroudi, D. Talley, and E. Coy, Visual characteristics and initial growth rates of round cryogenic jets at subcritical
and supercritical pressures, Phys. Fluids 14, 850 (2002).
5
B. Chehroudi, Recent experimental efforts on high-pressure supercritical injection for liquid rockets and their implica-
tions, Int. J. Aerosp. Eng. 2012, 1 (2012).
6
M. Oschwald, J. Smith, R. Branam, J. Hussong, A. Schik, B. Chehroudi, and D. Talley, Injection of uids into supercritical
environments, Combust. Sci. Technol. 178, 49 (2006).
7
M. Habiballah, M. Orain, F. Grisch, L. Vingert, and P. Gicquel, Experimental studies of high-pressure cryogenic ames
on the Mascotte facility, Combust. Sci. Technol. 178, 101 (2006).
8
K.-C. Lin, S. Cox-Stouffer, and T. Jackson, Structures and phase transition processes of supercritical methane/ethylene
mixtures injected into a subcritical environment, Combust. Sci. Technol. 178, 129 (2006).
9
S. Candel, M. Juniper, G. Singla, P. Scouaire, and C. Rolon, Structure and dynamics of cryogenic ames at supercritical
pressure, Combust. Sci. Technol. 178, 161 (2006).
10
N. Zong and V. Yang, Cryogenic uid jets and mixing layers in transcritical and supercritical environments, Combust.
Sci. Technol. 178, 193 (2006).
11
J. C. Oefelein, Mixing and combustion of cryogenic oxygen-hydrogen shear-coaxial jet ames at supercritical pressure,
Combust. Sci. Technol. 178, 229 (2006).
12
J. Bellan, Theory, modeling and analysis of turbulent supercritical mixing, Combust. Sci. Technol. 178, 253 (2006).
13
L. C. Selle, N. A. Okongo, J. Bellan, and K. G. Harstad, Modelling of subgrid-scale phenomena in supercritical transitional
mixing layers: An a priori study, J. Fluid Mech. 593, 57 (2007).
14
G. Ribert, N. Zong, V. Yang, L. Pons, N. Darabiha, and S. Candel, Counterow diffusion ames of general uids:
Oxygen/hydrogen mixtures, Combust. Flame 154, 319 (2008).
15
L. Pons, N. Darabiha, S. Candel, G. Ribert, and V. Yang, Mass transfer and combustion in transcritical nonpremixed
counterows, Combust. Theory Modell. 13, 57 (2009).
16
T. Schmitt, Y. M ery, M. Boileau, and S. Candel, Large-eddy simulation of oxygen/methane ames under transcritical
conditions, Proc. Combust. Inst. 33, 1383 (2011).
17
G. Lacaze, B. Cuenot, T. Poinsot, and M. Oschwald, Large eddy simulation of laser ignition and compressible reacting
ow in a rocket-like conguration, Combust. Flame 156, 1166 (2009).
18
R. N. Dahms, M. C. Drake, T. D. Fansler, R. O. Grover, and A. S. Solomon, Detailed simulations of stratied ignition and
combustion processes in a spray-guided gasoline engine using the SparkCIMM/G-equation modeling framework, SAE
Int. J. Engines 5, 141 (2012).
19
R. N. Dahms, M. C. Drake, T. D. Fansler, T.-W. Kuo, and N. Peters, Understanding ignition processes in spray-guided
gasoline engines using high-speed imaging and the extended spark-ignition model SparkCIMM. Part A: Spark channel
processes and the turbulent ame front propagation, Combust. Flame 158, 2229 (2011).
20
R. N. Dahms, M. C. Drake, T. D. Fansler, T.-W. Kuo, and N. Peters, Understanding ignition processes in spray-guided
gasoline engines using high-speed imaging and the extended spark-ignition model SparkCIMM. Part B: Importance of
molecular fuel properties in early ame front propagation, Combust. Flame 158, 2245 (2011).
21
D. L. Siebers, Liquid-phase fuel penetration in diesel sprays, SAE Paper No. 980809, 1998.
22
D. L. Siebers, Scaling liquid-phase fuel penetration in diesel sprays based on mixing-limited vaporization, SAE Paper
No. 1999-01-0528, 1999.
23
M. P. B. Musculus, Entrainment waves in decelerating transient turbulent jets, J. Fluid Mech. 638, 117 (2009).
24
M. P. B. Musculus and K. Kattke, Entrainment waves in diesel jets, SAE Paper No. 2009-01-1355, 2009.
25
B. Hu, M. P. B. Musculus, and J. C. Oefelein, The inuence of large-scale structures on entrainment in a decelerating
transient turbulent jet revealed by large eddy simulation, Phys. Fluids 24, 045106 (2012).
26
H. Johari and R. Paduano, Dilution and mixing in an unsteady jet, Exp. Fluids 23, 272 (1997).
27
J. Bor ee, N. Atassi, G. Charnay, and L. Taubert, Measurements and image analysis of the turbulent eld in an axisymmetric
jet subject to a sudden velocity decrease, Exp. Therm. Fluid Sci. 14, 45 (1997).
28
A. Roy and C. Segal, Experimental study of uid jet mixing at supercritical conditions, J. Propul. Power 26, 1205 (2010).
29
C. Segal and S. A. Polikhov, Subcritical to supercritical mixing, Phys. Fluids 20, 052101 (2008).
30
G. M. Faeth, D. P. Dominicis, J. F. Tulpinsky, and D. R. Olson, Supercritical bipropellant droplet combustion, Proc.
Combust. Inst. 12, 9 (1969).
31
R. S. Lazar and G. M. Faeth, Bipropellant droplet combustion in the vicinity of the critical point, Proc. Combust. Inst.
13, 801 (1971).
32
O. G. Nino-Amezquita, S. Enders, P. T. Jaeger, and R. Eggers, Measurement and prediction of interfacial tension of binary
mixtures, Ind. Eng. Chem. Res. 49, 592 (2010).
33
J. Dechoz and C. Roze, Surface tension measurement of fuels and alkanes at high pressure under different atmospheres,
Appl. Surf. Sci. 229, 175 (2004).
34
S. Liu, D. Fu, and J. Lu, Investigation of bulk and interfacial properties for nitrogen and light hydrocarbon binary mixtures
by perturbed-chain statistical associating uid theory combined with density-gradient theory, Ind. Eng. Chem. Res. 48,
10734 (2009).
35
R. N. Dahms, J. Manin, L. M. Pickett, and J. C. Oefelein, Understanding high-pressure gas-liquid interface phenomena
in diesel engines, Proc. Combust. Inst. 34, 1667 (2013).
129.2.129.146 On: Wed, 14 May 2014 20:52:34
36
H. Lin, Y.-Y. Duan, and Q. Min, Gradient theory modeling of surface tension for pure uids and binary mixtures, Fluid
Phase Equilib. 254, 75 (2007).
37
C. Miqueu, B. Mendiboure, and J. Lachaise, Modeling of the surface tension of multicomponent mixtures with the gradient
theory of uid interfaces, Ind. Eng. Chem. Res. 44, 3321 (2005).
38
J. D. van der Waals, Square gradient model, Verh. Konik Akad. Wet. Amsterdam 1, 8 (1893).
39
J. S. Rowlinson, The thermodynamic theory of capillarity under the hypothesis of a continuous variation of density,
J. Stat. Phys. 20, 197 (1979) (Translation of J. D. van der Waals).
40
J. W. Cahn and J. E. Hilliard, Free energy of a nonuniform system. I. Interfacial free energy, J. Chem. Phys. 28, 258
(1958).
41
E. A. M uller and A. Mejia, Interfacial properties of selected binary mixtures containing n-alkanes, Fluid Phase Equilib.
282, 68 (2009).
42
A. Mejia, J. C. Pamies, D. Duque, H. Seguara, and L. F. Vega, Phase and interface behaviors and type-I and type-V
Lennard-Jones mixtures: Theory and simulations, J. Chem. Phys. 123, 034505 (2005).
43
C. Miqueu, J. M. Miguez, M. M. Pineiro, T. Latte, and B. Mendiboure, Simultaneous application of the gradient theory
and Monte Carlo molecular simulation for the investigation of methane/water interfacial properties, J. Phys. Chem. B
115, 9618 (2011).
44
J. C. Oefelein, Large eddy simulation of turbulent combustion processes in propulsion and power systems, Prog. Aerosp.
Sci. 42, 2 (2006).
45
J. C. Oefelein, Simulation and analysis of turbulent multiphase combustion processes at high pressures, Ph.D. thesis,
The Pennsylvania State University, 1997.
46
T. W. Leland and P. S. Chappelear, The corresponding states principle: A review of current theory and practice, Ind. Eng.
Chem. Fundam. 60, 15 (1968).
47
J. S. Rowlinson and I. D. Watson, The prediction of the thermodynamic properties of uids and uid mixturesI. The
principle of corresponding states and its extensions, Chem. Eng. Sci. 24, 1565 (1969).
48
R. C. Reid, J. M. Prausnitz, and B. E. Polling, The Properties of Liquids and Gases (McGraw-Hill, New York, 1987).
49
G. J. VanWylen and R. E. Sonntag, Fundamentals of Classical Thermodynamics (John Wiley and Sons, Inc., New York,
1986).
50
S. Gordon and B. J. McBride, Computer program for calculation of complex chemical equilibrium compositions, rocket
performance, incident and reected shocks and Chapman-Jouguet detonations, NASA Technical Report No. SF-273,
1971.
51
R. J. Kee, F. M. Rupley, and J. A. Miller, Chemkin thermodynamic data base, Technical Report No. SAND87-8215B,
1990.
52
J. F. Ely and H. J. M. Hanley, Prediction of transport properties. 1. Viscosity of uids and mixtures, Ind. Eng. Chem.
Fundam. 20, 323 (1981).
53
J. F. Ely and H. J. M. Hanley, Prediction of transport properties. 2. Thermal conductivity of pure uids and mixtures,
Ind. Eng. Chem. Fundam. 22, 90 (1983).
54
R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena (John Wiley and Sons, Inc., New York, 1960).
55
J. O. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and Liquids (John Wiley and Sons, Inc., New
York, 1964).
56
S. Takahashi, Preparation of a generalized chart for the diffusion coefcients of gases at high pressures, J. Chem. Eng.
Jpn. 7, 417 (1974).
57
K. A. G. Schmidt, G. K. Folas, and B. Kvamme, Calculation of the interfacial tension of methane-water system with the
linear gradient theory, Fluid Phase Equilib. 261, 230 (2007).
58
Y.-X. Zuo and E. H. Stenby, A linear gradient theory model for calculating interfacial tensions of mixtures, J. Colloid
Interface Sci. 182, 126 (1996).
59
B. F. McCoy and H. T. Davis, Free-energy theory of inhomogeneous uids, Phys. Rev. A 20, 1201 (1979).
60
J. Lekner and J. R. Henderson, Theoretical determination of the thickness of a liquid-vapour interface, Physica A 94,
545 (1978).
61
J. Tang, J. Satherley, and D. J. Schiffrin, Density and interfacial tension of nitrogen-hydrocarbon systems at elevated
pressures, Chin. J. Chem. Eng. 1, 223 (1993).
62
J. Bellan and M. Summereld, Theoretical examination of assumptions commonly used for the gas phase surrounding a
burning droplet, Combust. Flame 33, 107 (1978).
63
J. Bellan and M. Summereld, Model for studying unsteady droplet combustion, AIAA J. 15, 234 (1977).
64
J. Bellan and M. Summereld, Quasi-steady gas phase assumption for a burning droplet, AIAA J. 14, 973 (1976).
65
C. Espey, J. E. Dec, T. A. Litzinger, and D. A. Santavicca, Quantitative 2-D fuel vapor concentration imaging in a ring
D.I. diesel engine using planar laser-induced Rayleigh scattering, SAE Paper No. 940682, 1994.
66
C. Espey, J. E. Dec, T. A. Litzinger, and D. A. Santavicca, Planar laser Rayleigh scattering for quantitative vapor-fuel
imaging in a diesel jet, Combust. Flame 109, 65 (1997).
67
J. E. Dec, A conceptual model for DI diesel combustion based on laser-sheet imaging, SAE Paper No. 970873, 1997.
68
C. A. Idicheria and L. M. Pickett, Quantitative mixing measurements in a vaporizing diesel spray by Rayleigh imaging,
SAE Paper No. 2007-01-0647, 2007.
69
L. M. Pickett, J. Manin, C. L. Genzale, D. L. Siebers, M. P. B. Musculus, and C. A. Idicheria, Relationship between diesel
fuel spray vapor penetration/dispersion and local fuel mixture fraction, SAE Int. J. Engines 4, 764 (2011).
70
L. M. Pickett, C. L. Genzale, G. Bruneaux, L.-M. Malbec, L. Hermant, C. Christiansen, and J. Schramm, Comparison of
diesel spray combustion in different high-temperature, high-pressure facilities, SAE Int. J. Engines 3, 156 (2010).
71
J. C. Oefelein, R. N. Dahms, and G. Lacaze, Detailed modeling and simulation of high-pressure fuel injection processes
in diesel engines, SAE Int. J. Engines 5, 1410 (2012).
129.2.129.146 On: Wed, 14 May 2014 20:52:34

Dahms

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Dahms

Încărcat de

Drepturi de autor:

Formate disponibile

On the transition between two-phase and single-phase interface dynamics in

multicomponent fluids at supercritical pressures

GHe) and liquid-oxygengaseous-hydrogen (LOXH

< 0.1 Continuum regime. (16)

is the liquid composition, and x

0.79, the adiabatic

0.79, the adiabatic mixing temperature exceeds the

0.79 since the mixture is supercritical at these conditions.

0.79, compare Fig. 17. Thus, the mixing path

0.79, the adiabatic

S-ar putea să vă placă și