on regular and retrograde condensation in multiphase compressible flows
TRANSCRIPT
Accepted Manuscript
On regular and retrograde condensation in multiphase compressible flows
Lu Qiu, Yue Wang, Rolf D. Reitz
PII: S0301-9322(14)00084-6DOI: http://dx.doi.org/10.1016/j.ijmultiphaseflow.2014.05.004Reference: IJMF 2040
To appear in: International Journal of Multiphase Flow
Received Date: 12 January 2014Revised Date: 28 April 2014Accepted Date: 6 May 2014
Please cite this article as: Qiu, L., Wang, Y., Reitz, R.D., On regular and retrograde condensation in multiphasecompressible flows, International Journal of Multiphase Flow (2014), doi: http://dx.doi.org/10.1016/j.ijmultiphaseflow.2014.05.004
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customerswe are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, andreview of the resulting proof before it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Body Page 1 of 34 Body
On regular and retrograde condensation in multiphase compressible flows
Lu Qiu1,a), Yue Wang1,b) and Rolf D. Reitz1,c) 1Engine Research Center, Department of Mechanical Engineering,
University of Wisconsin-Madison
1500 Engineering Drive, Madison, WI 53706, USA
a): The corresponding author, email address: [email protected] b): Email address: [email protected] c): Email address: [email protected]
Abstract
Two distinctive condensation mechanisms of pure species, regular and retrograde
condensation, are investigated using a one-fluid model and a homogeneous phase equilibrium
model based on entropy maximization. Fluid dynamics simulations are performed to model
the regular condensation process of ethylene in a converging nozzle, and the retrograde
condensation process of a fluorinated compound in a shock tube. To our knowledge, the
present simulations are the first CFD simulations of retrograde condensation processes. The
simulations show reasonably good agreement with available experimental data both
quantitatively and qualitatively, which confirms the consistency between the theory-guided
simulations and experiments. For the supercritical injection problem, condensation is found
to occur after continuous expansion when ethylene is brought into the two-phase region from
a supercritical state. For the shock tube problem, both simulations and experiments show that
condensation occurs after the high pressure reflected shock is formed from the end wall.
Increase in the initial pressure ratio increases the incident shock strength and reinforces
condensation by elevating the liquid volume fraction. A complete liquefaction shock is found
at high incident shock Mach numbers when the compression is strong enough to send the
fluid from the pure vapor to the pure liquid state by crossing the two-phase mixture region.
After condensation, the condensed liquid phase is fully depleted as the pressure wave
expands and the fluid is brought back to the vapor state though a continuous evaporation
Body Page 2 of 34 Body
process.
Key words: liquefaction shock; retrograde condensation; phase transition; shock waves;
two-phase flows; equation of state; phase equilibrium
1. Introduction
Condensation and evaporation are common phase transition phenomena. Condensation
usually occurs for species experiencing an expansion process. Species exhibiting such
characteristics are referred to as regular or simple species, such as water, air and carbon
dioxide, etc. During the expansion process, a supersaturation condition is achieved, and when
the supersaturation exceeds a critical limit (i.e., the critical supersaturation ratio),
spontaneous condensation occurs. This kind of condensation has been experimentally
observed in supersonic nozzles for water [1-3] and in hypersonic wind channels for air [4, 5].
However, condensation due to compression was not studied until the early work of Thompson,
who investigated compression condensation both analytically, using thermodynamic analysis
[6], and experimentally in a shock tube [7] in the 1970s.
According to Thompson and Sullivan [6], compression-induced retrograde condensation
is seen for retrograde species, like benzene and octane, etc. Compression-induced
condensation occurs because these species have a large number of molecular degrees of
freedom (i.e., large heat capacity), and as a result, compression work can be stored as internal
energy with only a small temperature rise [6]. The distinctive and different behavior of
regular and retrograde fluids is illustrated in Figure 1, which presents the temperature-entropy
phase diagram of pure ethylene (C2H4), a regular species, and perfluoroheptane (C7F16), a
retrograde species that is used in the simulations to be presented later. The thermodynamic
properties used for the two species are given in Table 1. In Fig. 1, the upwards arrow
represents an isentropic compression process (through which temperature and pressure
Body Page 3 of 34 Body
increase) while the downwards arrow stands for an isentropic expansion process (through
which temperature and pressure decrease). More specifically, for C2H4, whether it is initially
in the pure liquid or vapor phase, only through an expansion process will the fluid enter the
two-phase region with condensation. In this paper, the two-phase region stands for
vapor-liquid equilibrium and the solid state is not considered. For C7F16, however, there are
two separate scenarios. If the fluid is initially in the vapor phase, only compression can bring
the fluid into the two-phase region with condensation. In contrast, if the fluid is initially in the
liquid phase, only an expansion can bring the fluid into the two-phase region with
evaporation. Understanding these basic condensation mechanisms is critical to comprehend
and explain the different physical problems associated with phase transitions of pure species,
which then serves as the basis for further examinations of the phase transition mechanisms of
mixtures.
Lin et al. [8] examined the injection of a supercritical binary fuel composed of methane
and ethylene. An opaque background in shadowgraph images appeared in the near-nozzle
region, which was thought to be due to the condensation of the fuel. The immediate
occurrence of condensation at the nozzle exit indicates that condensation may have already
occurred inside the nozzle. Indeed, experimental visualization performed using a transparent
injector for pure ethylene showed that the transparent passage became opaque “abruptly” at
the end of the convergent section. Lin et al. [8] attributed the near-spontaneous condensation
to homogeneous nucleation and calculated the nucleation rate and critical nucleus size using
the classical nucleation theory of Frenkel [9]. This non-equilibrium nucleation dynamics
analysis was also later adopted by Star et al. [10]. Based on the phase diagram in Fig. 1(a),
since ethylene is a regular species, its condensation is naturally connected to classical
homogeneous nucleation theory, for which supersaturation due to expansion is the physical
cause.
Body Page 4 of 34 Body
Dettleff et al. [7] experimentally studied retrograde condensation of a fluorinated species
in a shock tube. It was found that condensation could occur, and when it occurred, a clear
liquid phase was seen at the closed end of the tube driven section from photographs. Their
experimental work was a direct proof of the condensation mechanism for retrograde species
as the fluid only undergoes phase change upon compression by the reflected shock.
This paper investigates the condensation and evaporation characteristics of pure species
using a one-fluid model accompanied with a homogeneous phase equilibrium model. In
Section 2 the computational models for the fluid dynamics and thermodynamics are
introduced. Model validations are presented in Section 3, where the simulation results are
compared with available experimental data for both the regular and retrograde condensation
processes. Summary and conclusions are made at the end.
2. Computational Model
2.1 The Fluid Solver
The open-source computational fluid dynamics (CFD) code KIVA-3V Release 2 [11] was
used as the basic fluid solver. This code been widely applied for multi-dimensional diesel
spray combustion simulations (e.g., [12, 13]). Briefly stated, the solver uses an Arbitrary
Lagrangian Eulerian (ALE) approach to deal with moving boundaries such as the piston
motion. The standard KIVA code solves the conservation equations at each time step in three
consecutive steps. In Phase A, the source terms due to spray and combustion are solved (not
relevant to the current simulations). In Phase B the cell is moved with the local fluid velocity,
and the physical quantities (e.g., pressure, temperature and velocity), except for convection
effects, are solved. Finally, in Phase C (rezoning stage) the cell boundaries are mapped back
to the location where they should be from the specified mesh motion.
The current simulations consider real gas effects, as opposed to the ideal gas relation in
the standard KIVA code. The compressible flow solver in the KIVA code solves the partial
Body Page 5 of 34 Body
differential equation for the internal energy as [14]:
(1)
is the mixture density, is the velocity vector, is the pressure, and is the
turbulence dissipation rate. is the heat flux due to heat conduction and diffusion:
(2)
is the specific enthalpy of species i. and are the heat conduction coefficient and
diffusion coefficient, respectively. In the simulations to be presented below, and are
calculated through the dimensionless Prandtl and Schmidt numbers, and the default values
0.9 and 0.56 are used.
The energy equation is solved in three steps in Phase B, and each step is either a constant
volume or constant pressure process [14-16]. The focus here is the last step occurring at
constant pressure that is used to determine the B-state internal energy from a previous t-state
through:
(3)
with the diffusion term calculated using
(4)
is a parameter used to ensure numerical stability and hence it depends on local flow
situations. is the cell mass updated in Phase B. is the normal area vector of the cell
face a [14]. is the temperature determined from the t-state. After some manipulations as
detailed on Pages 10 and 11 in Ref. [15], the temperature in the B-state is updated through:
Body Page 6 of 34 Body
(5)
where superscript “n” refers to a previous time level. When ideal gas is assumed for the
mixture, as in the standard KIVA code, and in above equation can be easily
calculated. To consider the real gas effects, they must be calculated according to the selected
equation of state (EOS) model. When phase transition occurs, as will be discussion below
shortly, the two-phase mixture properties are determined using algebraic averages. For
instance, the mixture molar enthalpy is calculated as the molar average of the two
phases:
(6)
where is the mole fraction of the liquid phase. In this treatment, both the liquid and
vapor phases are treated as a homogeneous mixture at equilibrium.
Since only one set of conservation equations is solved for a two-phase mixture, this is
essentially a one-fluid model [17-19]. In a two-phase model, such as the volume of fluid (VoF)
method, the liquid/vapor interface is tracked by solving an extra transport equation of, for
instance, the liquid volume fraction. In the present one-fluid model, interfaces are not
resolved between the two phases in a cell. This means that capillary pressure effects are not
considered. However, the existence of phase interfaces can still be represented through the
liquid volume fraction. As the fluid solver updates the conditions of each cell, its
thermodynamic state (e.g., liquid volume fraction) also changes. As a result, a cell’s state can
change from a single-phase to a two-phase mixture, depending on the local conditions, and
hence the phase transition process is captured. In the present homogenous equilibrium model,
the two phases are assumed to have the same velocity, temperature and pressure, no subgrid
Body Page 7 of 34 Body
scale thermal and pressure gradients or relative motion exists. On the contrary, in two-fluid
model treatments, calculations of interfacial heat, mass and momentum transfer rates requires
the introduction of additional models.
2.2 The Thermodynamic Solver
In the standard KIVA code, the state of a cell is determined after the mass, momentum
and energy equations are solved. By assuming that thermodynamic equilibrium is attained
instantaneously, the cell temperature is determined directly from the updated internal energy
because it is only a function of temperature for ideal gases. Pressure is then calculated using
the ideal gas law. From the perspective of classical thermodynamics, because no finite time
scale is involved, the mixture in each computation cell immediately reaches a global
thermodynamically stable state. To consider real gas effects, the thermodynamic solver must
be modified accordingly, as now the internal energy is a function of both temperature and
pressure.
It should be noted that phase transition is predicted if the mixture potential energy can be
further lowered through phase separation. As a result, the phase equilibrium model is soundly
based on fundamental thermodynamic principles (i.e., the second law of thermodynamics).
For a specified thermodynamic system, its global stable state is the one with minimal
thermodynamic potential. For instance, when temperature, pressure and mole numbers are
specified, the global stable state has minimum Gibbs free energy [20]. Of course, once the
equilibrium state is determined, phase transition dynamics analysis requiring non-equilibrium
thermodynamics can be formulated. This is because any kinetic process should evolve toward
the equilibrium state and hence the equilibrium solution serves as a reference solution. It
should also be emphasized that the present equilibrium solution includes not only saturation
densities, temperature and pressure, etc., but also the equilibrium phase compositions.
The global relationship between the fluid and thermodynamic solvers can be seen from
Body Page 8 of 34 Body
either Fig. 3 in Ref. [21] or Fig. 1 in Ref. [22]. This one-fluid phase equilibrium model
assumes velocity, temperature and pressure equilibrium, so the rates of interface heat and
mass transfer are infinitely fast. After the fluid solver advances the specific internal energy
and species density for each computational cell in time, the thermodynamics solver is applied
to determine the state of each cell (e.g., temperature and pressure). For a mixture with
specified internal energy, volume and mole number of each component, the thermodynamic
equilibrium state has the maximum entropy [20]. Applying the entropy maximization
principle to solve practical isochoric-isoenergetic flash problems was proposed by Castier [23]
and its applications can be found in Refs. [22, 24, 25].
Entropy maximum oriented equilibrium calculations for a mixture are very complex as
partial derivatives of thermodynamic properties are needed (e.g., ,
and . , where is the fugacity coefficient). As a result, the more complex the
equation of state model, the more intensive the actual computation in the thermodynamic
solver. For pure species, the entropy maximization principle reduces to the constraints of
thermal, mechanical and phase equilibria. Along with mass and energy conservation, the
following algebraic equations are solved simultaneously [10, 26].
(7)
Superscript “spec” stands for specified inputs to the thermodynamic solver, is the volume
fraction of the vapor phase; is the saturation pressure of the pure species. is the
fugacity calculated for phase equilibrium. In the current simulations, the saturation pressure
line of a pure species is calculated by enforcing the equality of fugacity for vapor-liquid
equilibrium. The results are then saved in a tabulated table for future calls in solving Eq. (7)
Body Page 9 of 34 Body
iteratively. The specific internal energy of the mixture is calculated using the departure
function
(8)
where is the specific internal energy of the ideal gas, and it is only a function of
temperature. The internal energy departure is determined from the equation of state
(EOS) model. The Peng-Robinson EOS (PR EOS) [27] is adopted here due to its relative
simplicity and accuracy in calculating liquid density, and successes in investigating
multiphase flow problems [10, 16, 28, 29]. The PR EOS has the form
(9)
is the molar volume and is the universal gas constant. and are determined
using:
(10)
with
(11a)
An expanded formula [30] was used when the acentric factor is greater than 0.5:
(11b)
The compressibility factor, , an indication of departure from ideal gas behavior, is
calculated using
(12)
is determined using the following relationship:
Body Page 10 of 34 Body
(13)
The Mach number (Ma) is calculated using:
(14)
where is the isentropic sound speed, while is the fluid velocity magnitude. All other
thermodynamic properties, such as the fugacity coefficient, entropy departure, etc., are
calculated using the formulae found in Refs. [31] and [32].
3. Results
Simulation results were compared with experimental data for both regular and retrograde
condensation. The ideal gas properties, including the specific heat capacity and entropy, were
calculated using the NASA 7-coefficient polynomials [33]. Properties (e.g., the critical
properties) of pure species were obtained from the DIPPR database [34]. The reference state
was taken at 298.15 K and 1 bar for both the enthalpy and entropy.
3.1 Regular condensation
Regular condensation of ethylene (C2H4) is first considered. A rectangular nozzle with
exit cross section of 1 mm2 was used in the experimental work of Lin et al. [8]. All four
experimental conditions found in Ref. [10] were considered here, as listed in Table 2. Only
the first condition did not lead to condensation. The downstream chamber condition is
atmospheric, but the injection conditions vary.
Computational domain is shown in Figure 2, and more details of the geometries can be
found in Ref. [10]. Injection is oriented towards the +x direction, and in Figs. 3 and 4,
injection is from left to right. Pressure inflow and pressure outflow boundary conditions were
enforced on the left and right sides, respectively. The standard k-ε turbulence model was used,
Body Page 11 of 34 Body
and adiabatic and no-slip boundary conditions were applied on all walls. A three-block mesh
structure was used, with 220 grids points in the nozzle axis direction (x direction), 20 grid
points in the y direction, and 40 grid points across the nozzle cross-section (z direction),
making the final mesh composed of 176000 cells. Simulations with finer meshes showed that
this resolution gave adequately grid-independent results.
In the experiments [10] pressures were measured along the centerline (see red solid line in
Fig. 2) and a comparison with the predicted pressure distributions is presented in Figure 3. To
better illustrate the effects of phase change, simulation results using the ideal gas law without
phase change are also presented. Both ideal and real gas simulations match the experimental
pressure data well at the entrance of the convergent induction section. The real gas simulation
significantly improves the predictions in the straight nozzle section. It is seen that, at the end
of the convergent section, the ideal gas simulation shows a large pressure drop while the
pressure relaxation is not so strong when real gas effects are considered. At the same time,
with phase change included, the predicted pressure is more close to the experiments in the
near-nozzle area. This effect is especially noteworthy for Cases 2-4 for which a relatively
stronger condensation event occurs, compared to Case 1. It is hence concluded that both real
gas and phase change effects are critical in evaluating supercritical condensation processes
and they must be taken into account in simulations to capture phase transition processes. It is
encouraging to see that the current one-fluid approach gives reasonably good agreement with
the experimental data, given its relative low computational effort.
Figure 4 presents the liquid volume fraction and compressibility factors for the four cases.
For Cases 1 and 2, outside the condensing region the liquid volume fraction is zero, as they
are in a supercritical state with vapor-like density. For Case 1, condensation occurs, but it is
limited to the nozzle exit region. Note that the maximum liquid volume fraction is about
0.5% for Case 1, while the experiments show no visible condensation inside the nozzle. Star
Body Page 12 of 34 Body
et al. [10] showed similar results but the disparity was left unexplained. However, the
appearance of liquid phase is consistent with the experimental observation that condensation
did occur downstream of the nozzle (i.e., in the underexpanded jet outside the nozzle).
Therefore, it is believed that the outlet boundary condition used in the simulations may
contribute to the discrepancy, and such discrepancy would be mitigated if both the internal
and external flows were simulated together to avoid the uncertainty in the boundary
conditions. However, the complexity of expanding flow with imbedded shock waves and
phase change outside the nozzle would make this a more difficult simulation. For Case 2, the
two-phase region starts at the end of the convergent section. The liquid volume fraction keeps
increasing until it reaches a maximum at the entrance of the straight section, and it then
decreases all the way to the nozzle exit.
For Cases 3 and 4, the initial liquid volume fraction is unity since the fluid is in a
supercritical state with liquid-like density. In fact, evaporation occurs, and it is evidenced by
the decrease of the liquid volume fraction in the straight section. For these two cases, the
profiles of compressibility factor and liquid volume fraction are similar. Compared to Case 3,
the evaporation in Case 4 is less intensive, so the liquid volume fraction drops more slowly.
The liquid volume fraction decreases to 10% and 15% for Cases 3 and 4, respectively.
It is helpful to track the change of state of a fluid element as it travels through the nozzle
passage, such as those shown in Figure 5 for the state evolution along the nozzle centerline in
a phase diagram. Figure 5 (a) shows that for Cases 1 and 2 the fluid at the nozzle entrance is
initially in a vapor-like supercritical state. Because of the pressure gradient across the nozzle,
the fluid continuously expands during the trip to the exit. In the later stages of the expansion
process, when the supersaturation state is large enough, the fluid enters the vapor-liquid
two-phase region and condensation occurs. Since only a small portion penetrates into the
two-phase mixture region with a very feeble condensation, the state is close to the dew point
Body Page 13 of 34 Body
line. The process for Case 2 penetrates even further into the vapor dome, so the condensation
is stronger. For Cases 3 and 4, it is obvious that the evaporation occurs as the fluid crosses the
bubble point line, but because of the different initial state, the ending states are on different
constant-quality lines. Similarly, because the end state for Case 4 is closer to the bubble point
line, the liquid phase volume fraction is larger than that for Case 3.
The present phase-equilibrium approach does not give information about droplet sizes or
number densities, which requires a non-equilibrium approach, such as the study in [10] using
homogeneous nucleation theory. Non-equilibrium models also require that the equilibrium
solutions are available, so the present approach can also be used as a starting point for further
model development. However, it should be noticed that the simulations using classical
nucleation theory in Ref. [10] yield similar results to the present equilibrium solutions.
Specifically, the equilibrium and non-equilibrium simulations yield similar pressure profile
and the liquid volume fraction differs only less than 7% absolute. This information indicates
that the thermodynamic equilibrium assumption is an appropriate first estimation of the
non-equilibrium problem.
Indeed, there is still uncertainty surrounding classical nucleation theory models. For
example, experimental nucleation rates of nitrogen in a supersonic nozzle had been shown to
be underestimated by 13-16 orders of magnitude for the temperature range between (1.6 kPa,
39 K) and (7 kPa, 46 K) [35]. On the other hand, mean field kinetic nucleation theory [36, 37]
was found to do better (with an underestimation of only 2-3 orders of magnitude). In contrast,
classical nucleation theory was found to agree with experiment nucleation rates of methanol
in a supersonic nozzle to within only one order of magnitude [38]. Therefore, the
applicability of classical nucleation theory, such as that used in Ref. [10], needs to be further
validated for use in CFD simulations.
Body Page 14 of 34 Body
3.2 Retrograde condensation
A complete liquefaction shock is a compression discontinuity for which the upstream
state is vapor and the downstream state is pure liquid [7]. For the partial liquefaction shock,
the downstream state is in a two-phase mixture state. Both partial and complete condensation
observed in shock tube experiments are investigated here. To our knowledge, this is the first
time these experiments have been modeled using CFD. The shock tube is composed of a
driver section of 250 cm and a driven section of 242.6 cm. A 1-D domain was selected using
1500 grid points for each section. In the following discussion, “complete condensation” is
used interchangeably with “complete liquefaction shock” and “partial condensation” is used
interchangeably with “partial liquefaction shock.” Together three different initial conditions
are considered for the driven section: 0.51, 0.67 and 0.99 bar. In the shock tube experiments
[7], the driver gas section was filled with nitrogen and the driven section used a fluorinated
compound, perfluoro-dimethycyclohexane (C8F16). Noticing that available thermodynamic
properties of this species are limited, perfluoroheptane (C7F16) was used instead in the present
simulations. Dettleff et al.[7] also argued that the results would not be significantly different
if other heavy hydrocarbons were used, and C7F16 also experiences retrograde behavior (as
seen in Fig. 1 (b)). Its constant volume specific heat at the critical point is calculated to
be 48.5 using the Peng-Robinson equation of state. This is in accord with the conclusion
of Thompson and Sullivan [6] that it is necessary that to permit the possibility
of a complete liquefaction shock. On the other hand, for a partial liquefaction shock, only that
is required.
The shock tube experiment in Ref. [7] was designed to produce liquefaction as the
incident shock reflects from the closed end of the tube. A schematic plot is shown in Fig. 6 to
illustrate the shock tube configuration. The driver section is at State 4 while the driven
section is at State 1. Upon releasing the diaphragm, incident compression shock and
Body Page 15 of 34 Body
expansion waves form. The state behind the incident shock front is State 2. The section
between contact surface and the leading expansion wave is at State 3. Finally, the region
between the reflected shock wave and the contact surface is at State 5.
The effects of the incident shock Mach number (Ms) are shown in Fig. 7. Figure 7 (a)
shows that pressure P2 downstream of the incident shock does not show a strong non-linear
behavior, as would be expected for ideal gases from the normal shock relations. In contrast,
real gas effects lead to an almost-linear relationship. The pressure P5 and temperature T5 are
shown in Figs. 7 (b) and (c). As the shock strength increases, P5 increases too. It is noted that
the profile of P5 does not show liquefaction effects, as there are no obvious changes in its
slope, as also noticed in Ref. [7]. On the contrary, the plot of T5 is a good indication of the
onset of condensation, which shows a “distorted, angular S-form” [7]. The two
discontinuities on the slope of T5 represent the start of partial and complete condensation,
respectively [7].
The distorted “S” shape can be explained as follows. During the transition from the vapor
to the two-phase state at low Ms, partial condensation occurs, so T5 has a sudden increase, as
seen in Fig. 7 (c). Similarly, during the transition from a two-phase state to a single liquid
state or to a supercritical state at higher Ms, heating effects are lost and the slope of T5
decreases, as also seen in Fig. 7 (c). The shock Mach number required for partial and
complete condensation is specifically marked in all the plots of Fig. 7. Figure 7 (d) plots P5
versus T5 with the saturation line. It is clearly seen that the fluid in State 5 experiences a state
change from superheated vapor to a two-phase mixture, and then to a compressed liquid and
finally to the supercritical state.
Some discrepancies between the simulation and experiments are observed in Fig. 7. This
could be due to two reasons. First, as mentioned, a slightly different species (i.e., C7F16) was
used in the simulations while C8F16 was used in the experiments. However, the general trends
Body Page 16 of 34 Body
of the experiments are well captured, as seen from the close-to-linear relationship between P2
and Ms, the non-linear relationship between P5 and Ms, and the “S” shape relationship
between T5 and Ms on the occurrence of partial and complete condensation. Second, it should
be noted that flows in a shock tube are intrinsically three dimensional and involve complex
shock wave turbulence boundary layer interactions and flow separations [39-41], and phase
change further complicates the flow. However, the current 1D simplification provides a
reasonable match on the global parameters (e.g., P2, P5 and T5 in Fig. 7).
Additional simulations were performed to characterize the effects of incident shock
strength on the onset of condensation. Figure 8 presents the start of condensation (i.e., partial
liquefaction) at different Ms for the three P1’s. From P1=0.99 to P1=0.51 bar, as the pressure is
reduced, the Ms required for condensation increases. In other words, condensation is
postponed when P1 is reduced. This is expected, as the reflected shock, which depends on the
initial pressure ratio P4/P1, must be strong enough to bring the vapor phase into the two-phase
mixture region. Similarly, for each specified P1, the onset of condensation is advanced when
Ms or P4/P1 becomes larger.
Two special cases were chosen for further analyses. Case 1 has partial condensation
occurring at P1=0.67 bar and P4= 4.0 bar. Case 2 is at P1=0.99 bar and P4= 8.5 bar, under
which strong (actually complete) condensation occurs. For Case 1, the predicted temperature
and liquid volume fraction are plotted in Fig. 9 at t=16.0 ms (a) and t=18.0 ms (b). These two
time instants are after the reflected shock has formed. The liquid volume fraction has already
reached 2% by t=16.0 ms. It is noticed that there is a plateau of temperature in the
condensation region. Because of the phase change, a large temperature gradient appears.
From t=16.0 to t=18.0 ms, the two-phase region is transported along with the pressure wave,
and the previous temperature gradient is disrupted. Compared with the results at t=16.0 ms,
the decrease liquid volume fraction indicates that evaporation continues to occur while the
Body Page 17 of 34 Body
local fluid is experiencing expansion.
Results for Case 2 are presented in Figure 10 at two instants as well. Condensation has
already occurred at t=13.28 ms. But at this time the condensation is so strong that full
condensation, i.e., a complete liquefaction shock, occurs. The shock strength is so high that
100% of the original gaseous fluid is converted to liquid phase. Therefore, a larger local
gradient is found for both the pressure and temperature when compared with Case 1. At
t=15.0 ms, the liquid phase amount has dropped as evaporation occurs, similar to Case 1.
However, since the vapor-liquid interface is distinctive due to the strong liquid volume
fraction gradient, the evaporative cooling leads to sharp change of temperature.
The two numerical simulations above indicate that partial or complete condensation
occurs when the reflected shock forms as the incident shock is reflected from the closed end
of the tube. Because pressure is enhanced through the reflected shock (also seen through the
color scale), the appearance of the liquid phase can only be due to the compression, and not
through expansion (such as the condensation seen in nozzles and wind tunnels, as mentioned
above). Therefore, the present CFD simulation results are consistent with the condensation
mechanism for this retrograde fluid (refer to Fig. 1 (b)). In addition, it is remarked that the
condensation region is limited to lie between the contact surface and the right-hand end of the
tube, where the pressure is high enough to be able to trigger the phase change.
Since P5 is highest upon pressure wave reflection, it is more intuitive to monitor the
dynamic evolution of the properties at the right-hand tube end. The histories of pressure and
liquid volume fraction of two phases are presented in Fig. 11 for the two cases above. For
both cases, the abrupt increase in the pressure profile (e.g., Fig. 11 (a) and (b)) is a clear
indication of the arrival of the incident shock, which is closely followed by the condensation,
given that the shock strength is strong enough. The following pressure plateau indicates the
state behind the reflected liquefaction shock (i.e., P5). The relative length of the plateau is
Body Page 18 of 34 Body
approximately the residence time of the shock, which is in agreement with the experimental
measurements in Ref. [7]. For Case 1, the partial condensation event takes place at t=13.86
ms. After an additional 0.8 ms, the liquid phase occupies 2% of the volume. There is a period
of a further 2.5 ms, during which the state (pressure and liquid phase amount) does not vary
too much, indicating that a quasi-steady state is obtained. This quasi-steady state is the
preferred state in shock tube to provide relatively constant high temperature and pressure
conditions. After t=17.0 ms, evaporation starts and the liquid phase is continuously depleted.
Compared with the feeble condensation above with only 2% liquid volume conversion,
the situation changes dramatically when strong condensation takes place. Firstly,
condensation occurs earlier and it uses 75% less time than for Case 2 to convert all the
gaseous fluid to the pure liquid phase. Secondly, the oscillations seen in the pressure profile
are stronger due to the stronger phase transition events and the quasi-steady state duration is
reduced (Fig. 11(b)). Thirdly, after the evaporation occurs, within about 75% of the time used
in Case 1, the liquid volume fraction decreases from 100% to about 3%. It is also evident that
the following liquid volume fraction increases due to repeated condensation. However,
evaporation still dominates as the remaining liquid phase is converted to vapor promptly, so it
is not specifically marked in Fig. 11 (b).
It is helpful to analyze these dynamic processes at the right-hand end of the tube in a
phase diagram, which is illustrated in Fig. 12 on temperature-entropy diagrams for the two
cases. The numbers in the brackets are the starting and ending times for the state. For both
cases, it is noticed that the fluid evolves between the single phase and the two-phase state.
For Case 1, spontaneous condensation occurs as the fluid is compressed to cross the dew
point line and enter the two-phase state (e.g., Fig. 12 (a)). The fluid stays in the two-phase
mixture state until evaporation occurs. After the evaporation is completed, the fluid enters the
vapor phase again. During the evaporation in the two-phase state, the temperature and
Body Page 19 of 34 Body
pressure decrease along the saturation line. It is also found that the evolution of the fluid in
the two-phase mixture state and the vapor state after t=19.05 ms is close to an isentropic
process.
Compared to Case 1, because the shock strength in Case 2 is higher, the fluid is
continuously compressed into the two-phase mixture state, and further into the pure liquid
state in a much shorter period (see Fig. 12 (b)). In fact, only about 0.15 ms is needed for the
fluid to move from the dew point line to reach the bubble point line. In addition, it is noticed
that the compression process wanders its way from the pure vapor to the pure liquid state.
The pressure in the liquid state is higher than the critical pressure and the fluid is considered
to be at the supercritical state with liquid-like density. After t=14.55 ms, the expansion leads
to continuous evaporation, bringing the fluid back into the two-phase mixture region.
Different from before, from t=14.55 to 17.93 ms, the evaporation process that converts all the
liquid to the pure vapor state is more close to isentropic than the previous condensation
process. After t=17.93 ms, the fluid finally returns to the original vapor state and further
expansion is close to isentropic.
Finally, three more cases are selected to illustrate the effects of P4 (equivalently Ms) on
the condensation intensity. Figures 13 (a), (c) and (e) present the global x-t wave diagrams at
P4=4.5, 5.0 and 6.0 bar with constant P1=0.51 bar, respectively. The colors are based on
pressure values but to increase the color contrast, the logarithm of pressure is used. It is seen
that P5 increases with P4. At the same time, condensation becomes more intensive, as seen
from the blow-up plots of liquid volume fraction contours shown in Figs. 13 (b), (d) and (f)
(note the scales are different). More specifically, the maximum liquid volume fraction is
found to be 8%, 25% and 100% for the three conditions considered, corresponding to a
partial, intermediate and strong condensation, respectively.
Body Page 20 of 34 Body
4. Summary and Conclusions
Computational fluid dynamic simulations were performed for two-phase flows in a
converging nozzle and a shock tube to investigate phase transition behavior for both regular
and retrograde condensation of pure species. The fluid solver is based on the KIVA 3v release
2 open source code. But, to consider the real gas effects, the thermodynamic relationship was
updated using the Peng-Robinson equation of state. Accordingly, the thermodynamic solver
was also updated to account for the fact that the internal energy is not only a function of
temperature. A homogeneous phase equilibrium model based on entropy maximization was
used as the phase transition model.
The simulation results show reasonable agreement with available experimental data using
the present one-fluid approach with a realistic equation of state model. They also conform to
the expected behavior from thermodynamic analyses based in temperature-entropy phase
diagrams, and the deduced condensation mechanisms for the different fluids. Therefore, the
present theory-led numerical simulations are consistent with experiments. For ethylene,
regular condensation can only occur due to an expansion process, which brings the fluid to a
supersaturation state until spontaneous condensation occurs. The simulation results were
verified against measured pressure distributions for four conditions. On the other hand,
condensation can occur due to compression with a retrograde fluid. The present shock tube
simulations with a fluorocarbon fluid were found to predict well both qualitatively and
quantitatively when compared with available experimental data. Therefore, the present
approach can be applied to study phase transition behavior for both regular and retrograde
condensation in multiphase compressible flows.
Specifically for the retrograde fluid condensation, the following conclusions can be drawn
from the present simulations and the comparisons with the classical experiments of Dettleff et
al. [7]:
Body Page 21 of 34 Body
(1) Consistent with the experiments, condensation occurs in the reflected shock region at
the right-hand end of the shock tube, where only the retrograde fluid exists.
(2) Partial condensation occurs when the incident shock is strong enough that the fluid
can be compressed from the vapor state to the two-phase mixture state.
(3) Complete condensation occurs when the incident shock is strong enough that the fluid
can be further compressed from the vapor state to the two-phase mixture state, and to
the compressed liquid state or supercritical state with liquid-like properties.
(4) The temperature at State 5 (behind the reflected shock) is a better indicator of
condensation than pressure.
(5) The vapor-to-liquid conversion (i.e., liquid volume fraction) is enhanced when the
incident shock Mach number or P4 increases.
(6) Later after the condensation, the expanding wave leads to continuous evaporation
until the liquid phase is fully depleted.
Acknowledgements
The research work was conducted at the Engine Research Center and sponsored by
Department of Energy (DOE) and Sandia National Laboratories through the Advanced
Engine Combustion Program (MOU 04-S-383) under the management of Dr. Dennis L.
Siebers.
References
[1] Hill PG. Condensation of Water Vapour during Supersonic Expansion in Nozzles, Journal
of Fluid Mechanics 1966;25:593-620.
[2] Pouring AA. Thermal Choking and Condensation in Nozzles, Physics of Fluids
1965;8:1802-10.
[3] Wegener PP, Pouring AA. Experiments on Condensation of Water Vapor by
Homogeneous Nucleation in Nozzles, Physics of Fluids 1964;7:352-61.
[4] Wegener PP, Mack LM. Condensation in Supersonic and Hypersonic Wind Tunnels. in:
Dryden HL, Th. von K (Eds.) Advances in Applied Mechanics, Elsevier, 1958, pp.
307-447.
[5] Daum FL. Air Condensation in a Hypersonic Wind Tunnel, Aiaa J 1963;1:1043-6.
[6] Thompson PA, Sullivan DA. Possibility of Complete Condensation Shock-Waves in
Body Page 22 of 34 Body
Retrograde Fluids, Journal of Fluid Mechanics 1975;70:639-49.
[7] Dettleff G, Thompson PA, Meier EAG, Speckmann H. An experimental study of
liquefaction shock waves, Journal of Fluid Mechanics 1979;95:279-304.
[8] Lin KC, Cox-Stouffer SK, Jackson TA. Structures and Phase Transition Processes of
Supercritical Methane/Ethylene Mixtures Injected Into A Subcritical Environment,
Combustion Science and Technology 2006;178:129-60.
[9] Frenkel J, Kinetic theory of liquids. Oxford, Clarendon Press, 1946.
[10] Star AM, Edwards JR, Lin KC, Cox-Stouffer S, Jackson TA. Numerical simulation of
injection of supercritical ethylene into nitrogen, J Propul Power 2006;22:809-19.
[11] Amsden AA. KIVA-3V, Release 2, Improvements to KIVA-3V. Technical Report
LA-UR-99-915, Los Alamos National Laboratory, 1999.
[12] Reitz RD, Rutland CJ. Development and testing of diesel engine CFD models, Prog
Energ Combust 1995;21:173-96.
[13] Gustavsson J, Golovitchev VI. Spray Combustion Simulation Based on Detailed
Chemistry Approach for Diesel Fuel Surrogate Model. SAE 2003-01-1848, 2003.
[14] Amsden AA, O’Rourke, P. J. and Butler, T. D. KIVA-II: a computer program for
chemically reactive flows with sprays. Technical Report LA-11560-MS, Los Alamos
National Laboratory, 1989.
[15] Trujillo MF, O'Rourke PJ, Torres D. Generalizing the Thermodynamics State
Relationships in KIVA-3V. Los Alamos Technical Report, Los Alamos National
Laboratories, 2002.
[16] Trujillo MF, Torres DJ, O'Rourke PJ. High-pressure multicomponent liquid sprays:
Departure from ideal behaviour, International Journal of Engine Research
2004;5:229-46.
[17] Schmidt D, Rutland C, Corradini M. A Numerical Study of Cavitating Flow Through
Various Nozzle Shapes. SAE 971597, 1997.
[18] Habchi C, Dumont N, Simonin O. Multidimensional simulation of cavitating flows in
diesel injectors by a homogeneous mixture modeling approach, Atomization Spray
2008;18:129-62.
[19] Lee WG, Reitz RD. A Numerical Investigation of Transient Flow and Cavitation Within
Minisac and Valve-Covered Orifice Diesel Injector Nozzles, Journal of Engineering for
Gas Turbines and Power 2010;132:052802.
[20] Callen HB, Thermodynamics and an introduction to thermostatics. 2nd ed., J. Wiley,
New York ; Chichester etc., 1985.
[21] Wang Y, Qiu L, Reitz RD, Diwakar R. Simulating cavitating liquid jets using a
compressible and equilibrium two-phase flow solver, International Journal of Multiphase
Flow 2014;63:52-67.
[22] Qiu L, Reitz RD. Condensation processes in a motoring engine, The Journal of
Supercritical Fluids 2014;90:84-100.
[23] Castier M. Solution of the isochoric-isoenergetic flash problem by direct entropy
maximization, Fluid Phase Equilibr 2009;276:7-17.
[24] Castier M. Dynamic simulation of fluids in vessels via entropy maximization, J Ind Eng
Chem 2010;16:122-9.
[25] Qiu L, Wang Y, Jiao Q, Wang H, Reitz RD. Simulating condensation in a supercritical
Body Page 23 of 34 Body
gas jet. ILASS Americas, 25th Annual Conference on Liquid Atomization and Spray
Systems, Pittsburgh, PA,, May, 2013.
[26] Giljarhus KET, Munkejord ST, Skaugen G. Solution of the Span–Wagner Equation of
State Using a Density–Energy State Function for Fluid-Dynamic Simulation of Carbon
Dioxide, Ind Eng Chem Res 2012;51:1006-14.
[27] Peng DY, Robinson DB. A New Two-Constant Equation of State, Ind Eng Chem Fund
1976;15:59-64.
[28] Zhu GS, Reitz RD. Engine Fuel Droplet High-Pressure Vaporization Modeling, Journal
of Engineering for Gas Turbines and Power 2000;123:412-8.
[29] Zhu G-S, Reitz RD, Aggarwal SK. Gas-phase unsteadiness and its influence on droplet
vaporization in sub- and super-critical environments, Int J Heat Mass Tran
2001;44:3081-93.
[30] Robinson DB, Peng DY. The characterization of the heptanes and heavier fractions for
the GPA Peng-Robinson programs. Research report RR-28, Gas processors association,
1978.
[31] Firoozabadi A, Thermodynamics of hydrocarbon reservoirs. McGraw-Hill, New York,
1999.
[32] Sandler SI, Chemical and engineering thermodynamics. 3rd ed., Wiley, New York,
1999.
[33] Gordon S, McBride BJ. Computer Program for Calculation of Complex Chemical
Equilibrium Compositions and Applications: I. Analysis, NASA Reference Publication
1311 October 1994.
[34] Rowley RL, Wilding WV, Oscarson JL, Giles N, Rowley RJ, Daubert TE et al. DIPPR
Data Compilation of Pure Compound Properties. 2001.
[35] Bhabhe A, Wyslouzil B. Nitrogen nucleation in a cryogenic supersonic nozzle, The
Journal of Chemical Physics 2011;135:244311.
[36] Kalikmanov VI. Mean-field kinetic nucleation theory, J Chem Phys 2006;124.
[37] Kalikmanov VI, Wolk J, Kraska T. Argon nucleation: Bringing together theory,
simulations, and experiment, J Chem Phys 2008;128.
[38] Laksmono H, Tanimura S, Wyslouzil BE. Methanol nucleation in a supersonic nozzle,
The Journal of Chemical Physics 2011;135:-.
[39] Dussauge JP, Dupont P, Debieve JF. Unsteadiness in shock wave boundary layer
interactions with separation, Aerosp Sci Technol 2006;10:85-91.
[40] Wu MW, Martin MP. Analysis of shock motion in shockwave and turbulent boundary
layer interaction using direct numerical simulation data, Journal of Fluid Mechanics
2008;594:71-83.
[41] Humble RA, Elsinga GE, Scarano F, van Oudheusden BW. Three-dimensional
instantaneous structure of a shock wave/turbulent boundary layer interaction, Journal of
Fluid Mechanics 2009;622:33-62.
Body Page 24 of 34 Body
(a)
(b)
Fig. 1. Temperature-entropy phase diagram of pure species. (a): Regular fluid ethylene (C2H4). (b)
Retrograde fluid perfluoroheptane (C7F16). The downwards arrow shows an isentropic expansion process
while the upwards arrow is an isentropic compression process. Also shown are the vapor dome, critical
point and isobaric lines at four pressures (i.e., 0.2, 0.4 and 0.8 Pc,). The supercritical region where both
temperature and pressure are above the critical point is specifically highlighted.
Fig. 2. 3D view of the computational mesh used for pure ethylene injection. Red solid line shows where
pressure transducers are located. Dashed line shows injection direction.
Body Page 25 of 34 Body
(a)
(b)
(c)
(d)
Fig. 3. Comparison of pressure profiles with experimental data along the centerline of the rectangular
nozzle for the four conditions listed in Table 2. Injection is from left to right. Refer to Fig. 2 for the
centerline location. (a) Case 1. (b) Case 2. (c) Case 3. (d) Case 4. Also shown are simulation results using
the ideal gas law without phase change.
Body Page 26 of 34 Body
Fig. 4. Calculated liquid volume fraction and compressibility factors of the two phases along the centerline
of the rectangular nozzle for the four conditions listed in Table 2. Injection is from left to right. Refer to Fig.
2 for the centerline location. (a) Case 1. (b) Case 2. (c) Case 3. (d) Case 4.
(a)
(b)
(c)
(d)
Body Page 27 of 34 Body
(a)
(b)
Fig. 5. Evolution of thermodynamic state along centerline of rectangular nozzle on temperature-entropy
phase diagrams for the four cases listed in Table 2. Refer to Fig. 2 for the centerline location. (a) Cases 1
and 2. (b) Cases 3 and 4. Arrow indicates the direction of the state change. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 6. Position-time wave diagram for the shock tube configuration. Test section is near the right end. The
driver section ○4 and driven section ○1 are initially separated by the diaphragm. Contact surface is
shown by the solid (blue) line. ○2 and ○3 represent the state behind the incident shock and contact
surface, respectively.
Body Page 28 of 34 Body
(a)
(b)
(c)
(d)
Fig. 7. Comparison with experimental data as a function of incident shock Mach number. (a) Pressure P2.
(b) Pressure P5. (c) Temperature T5. The two discontinuities of slope represent onset of partial
condensation and complete condensation, respectively. (d) Pressure P5 versus temperature T5.
Fig. 8. Effects of P1 and incident shock Mach number on the onset of condensation.
Body Page 29 of 34 Body
(a)
(b)
Fig. 9. Predicted contact discontinuity, temperature, and liquid volume fraction for partial liquefaction
shock at P1=0.67 bar and P4=4.0 bar. (a) At 16.0 ms. (b) At 18.0 ms. Note that the liquid volume fraction is
enlarged 10 times. Vertical dashed line indicates the right-hand end wall location.
Body Page 30 of 34 Body
(a)
(b)
Fig. 10. Predicted contact discontinuity, temperature, and liquid volume fraction for partial liquefaction
shock at P1=0.99 bar and P4=9.5 bar. (a) At 13.28 ms. (b) At 15.0 ms. Vertical dashed line indicates the
right-hand end wall location.
Body Page 31 of 34 Body
(a)
(b)
Fig. 11. Evolution of liquid volume fraction and pressure at the right-hand end of the shock tube. (a)
Partial liquefaction shock at P1=0.67 bar and P4=4.0 bar. (b) Complete liquefaction shock P1=0.99 bar and
P4=9.5 bar.
(a)
(b)
Fig. 12. Evolution of state at the right-hand end of the shock tube in temperature-entropy phase diagram.
(a) Partial liquefaction shock at P1=0.67 bar and P4=4.0 bar. (b) Complete liquefaction shock P1=0.99 bar
and P4=9.5 bar. Numbers in legend show the time period.
Body Page 32 of 34 Body
x (cm)
Tim
e (
ms)
P1=0.51 bar
P4=4.5 bar
log(P)
-250 -200 -150 -100 -50 0 50 100 150 2000
2
4
6
8
10
12
14
16
18
20
22
-0.2
0
0.2
0.4
0.6
0.8
1
Contact surface
(a)
x (cm)
Tim
e (
ms)
P1=0.51 bar
P4=4.5 bar
liquid volumefraction
220 225 230 235 24010
12
14
16
18
20
22
0.0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Contact surface
(b)
x (cm)
Tim
e (
ms)
P1=0.51 bar
P4=5.0 bar
log(P)
-250 -200 -150 -100 -50 0 50 100 150 2000
2
4
6
8
10
12
14
16
18
20
22
-0.2
0
0.2
0.4
0.6
0.8
1
Contact surface
(c)
x (cm)
Tim
e (
ms)
P1=0.51 bar
P4=5.0 bar
liquid volumefraction
220 225 230 235 24010
12
14
16
18
20
22
0
0.05
0.1
0.15
0.2
Contact surface
(d)
x (cm)
Tim
e (
ms)
P1=0.51 bar
P4=6.0 bar
log(P)
-250 -200 -150 -100 -50 0 50 100 150 2000
2
4
6
8
10
12
14
16
18
20
22
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Contact surface
(e)
x (cm)
Tim
e (
ms)
P
1=0.51 bar
P4=6.0 bar
liquid volumefraction
220 225 230 235 24010
12
14
16
18
20
22
0
0.2
0.4
0.6
0.8
Contact surface
(f)
Fig. 13. Global x-t wave diagram of logarithm of pressure and liquid volume fraction at P1=0.51 bar. (a) and
(b) Partial condensation at P4=4.5. (c) and (d) Intermediate condensation at P4=5.0. (e) and (f) Strong
condensation at P4=6.0.
Body Page 33 of 34 Body
Table 1. Thermodynamic properties of C2H4 and C7F16.
species [K]
[bar]
[-]
C2H4 282.34 50.41 0.0862
C7F16 475.65 16.1 0.5429
Table 2: Simulation conditions for supercritical injection of ethylene.
Case
Injection conditions Chamber conditions
Condensation? Pressure
(bar)
Temperature
(K)
Pressure
(bar)
Temperature
(K)
1 55.18 309.2 1.37 298.0 No
2 55.42 289.0 1.37 298.0 Yes
3 55.64 283.9 1.37 298.0 Yes
4 54.94 281.5 1.37 298.0 Yes
Body Page 34 of 34 Body
Highlights
Highlights Page 1 of 1 Highlights
Highlights:
Regular and retrograde condensation is simulated using phase equilibrium model.
Supercritical nozzle flow of C2H4 with condensation is simulated.
Shock tube experiments with compression-induced condensation are simulated.
Theory-guided simulation is found to be consistent with experiments.