Sie sind auf Seite 1von 17

Numerical Heat Transfer, Part B, 58: 287303, 2010

Copyright # Taylor & Francis Group, LLC


ISSN: 1040-7790 print=1521-0626 online
DOI: 10.1080/10407790.2010.528737

INVESTIGATION OF HEAT AND MASS TRANSFER IN


A LID-DRIVEN CAVITY UNDER NONEQUILIBRIUM
FLOW CONDITIONS
Benzi John, Xiao-Jun Gu, and David R. Emerson

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

Computational Science and Engineering Department, STFC Daresbury


Laboratory, Warrington, United Kingdom
Gaseous ow and heat transfer in a lid-driven cavity under nonequilibrium ow conditions
is investigated using the direct simulation Monte Carlo method, from the slip to the
free-molecular regime. The emphasis is on understanding thermal ow features. The impact
of the lid velocity and various degrees of rarefaction on the shear stress and heat ux rates
are analyzed. The role of expansion cooling and viscous dissipation on the heat transfer
mechanism is investigated. Complex heat ow phenomena, such as counter-gradient heat
transfer, are revealed by the simulations which the conventional Navier-Stokes-Fourier
equations are not able to capture, even in the slip-ow regime.

1. INTRODUCTION
Fluid ow at the micro- and nanoscale has received considerable attention due
to the advent of miniaturization and advances in fabrication technology. A basic
understanding of the nature of ow in these devices is considered essential for an
efcient design. It is well known that the Navier-Stokes-Fourier (NSF) equations
cannot accurately describe gaseous ow under rareed conditions [1, 2], especially
in the transition regime and beyond. The degree of rarefaction can be described
by the Knudsen number (Kn), which relates the mean free path of the gas to a
characteristic length scale of a device and is commonly used to understand the different ow regimes that micro-electro-mechanical Systems (MEMS) will encounter.
Unlike high-altitude and vacuum devices, where the low density conditions lead to a
mean free path of the order of meters (and correspondingly high Kn), these rarefaction effects are present under standard atmospheric (STP) conditions. The mean free
path of air at STP is about 100 nm, and the rareed ow conditions that occur in
MEMS are generally caused by the extremely small geometric dimensions involved.
Various macroscopic modeling techniques have been developed over the years
[35] with a view to capturing nonequilibrium ows accurately, especially in the transition regime (Kn > 0.1). This includes the various well-known slip-based models, the
Received 22 April 2010; accepted 6 September 2010.
The authors would like to thank the Engineering and Physical Sciences Research Council (EPSRC)
for their support of Collaborative Computational Project 12 (CCP12).
Address correspondence to Benzi John, Computational Science and Engineering Department,
STFC Daresbury Laboratory, Warrington WA4 4AD, United Kingdom. E-mail: benzi.john@stfc.ac.uk

287

288

B. JOHN ET AL.

NOMENCLATURE

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

d
Kn
L
m
Ma
P
P0
qx
q0
R
Re
s
T
Tw
T0

particle diameter, m
Knudsen number ( k=L)
characteristic length of device, m
particle mass, kg
p
Mach number Uw = cRT
2
pressure, N=m
reference pressure, N=m2
heat ux in the x direction,
W=m2
reference heat ux, W=m2
specic gas constant, J=kg K
Reynolds number ( qUwL=m)
distance along the cavity walls, m
gas temperature, K
wall temperature, K
reference temperature, K

u
Uw
v
Vmp
c
k
m
q
rxy

velocity of gas in the x direction, m=s


wall velocity, m=s
velocity of gas in the y direction, m=s
most
probable molecular velocity
p
2RT , m=s
specic heat ratio
mean free path, m
dynamic viscosity, N s=m2
density, kg=m3
shear stress, N=m2

Subscripts
x, y
coordinates x and y
0
reference value
w
wall quantity
mp
most probable value

method of moments, and the Burnett equations. The NSF equations, modied with
various velocity-slip and temperature-jump boundary conditions, are able to capture
certain ow features in the slip regime (0.001  Kn  0.1). Another modeling
approach is the method of moments, which was originally developed by Grad [4]
as a 13-moment equation model and was used initially for modeling hypersonic
ows. Signicant progress has been made with regard to the method of moments
in modeling nonequilibrium ow in microgeometries [69]. Recent studies show that
the method is able to capture several well-known nonequilibrium phenomena, such
as the bimodal temperature prole in force-driven Poiseuille ow, nongradient heat
ux in Couette ow, and the Knudsen minimum. Although other high-order models,
such as the Burnett equations, can provide improved solutions in the transition
region, they have also been shown to have severe limitations [10, 11].
An important motivation for developing alternative modeling strategies is computational efciency and, while the foregoing methods can provide good approximations, they have only been tested and validated against relatively simple problems
such as Couette, Poiseuille, and Rayleigh ows or for cases involving comparatively
small ow and temperature gradients. In addition, it is unlikely that these approaches
will be able to provide accurate solutions in the upper transition regime (Kn  1),
especially in complex geometries. The direct simulation Monte Carlo (DSMC)
method is an excellent alternative approach, and the method has been widely
employed for modeling gaseous ows under nonequilibrium ow conditions. The
DSMC method, originally developed by Bird [12], is a physical model which is based
on the discrete molecular nature of the system, essentially capturing the physics of a
dilute gas governed by the Boltzmann equation. Applications of the DSMC method
range from simulation of hypersonic ow problems in the upper atmosphere, such as
reentry vehicles, orbital vehicles, etc., to simulation of low-speed gas ows in devices
at the micro- and nanoscale. The method is able to predict ow and heat transfer
aspects in all ow regimes accurately and is therefore considered to be the most
reliable approach for rareed gas simulations. However, for an accurate simulation,

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

NONEQUILIBRIUM HEAT AND MASS TRANSFER

289

there are constraints with regard to the cell size, time step, and number of particles per
cell that need to be strictly adhered to, which makes the DSMC method inherently
computationally intensive in nature. Additionally, statistical noise is relatively high
at low ow speeds, a typical situation for MEMS. In such instances, the number of
particles and the sampling period needed to minimize the statistical noise is high.
The DSMC method can be readily parallelized, and high-delity solutions can be
achieved with the aid of parallelization techniques, as adopted in the current study.
Studies of the driven cavity problem have generally been limited to continuum
ows [13, 14], wherein the problem is often considered to be a benchmark problem
for validation. In the present work, however, the DSMC method is used to investigate rareed ow and heat transfer characteristics of a driven microcavity. The
classical shear-driven (Couette) and pressure-driven (Poiseuille) rareed ows have
been studied extensively [1519] in the literature. In addition, the effects of rarefaction and compressibility on ows in two-dimensional microchannels have been investigated by Sun and Faghri [20]. Gaseous ow and heat transfer in square cavity
enclosures that represent vacuum-packaged MEMS devices have been studied by
Liu et al. [21] and Cai [22], and the driven cavity problem was also examined with
the aid of the lattice Boltzmann method by Nie et al. [23]. Rarefaction effects were
recently investigated by Naris and Valougeorgis [24], who used a linearized
Bhatnagar-Gross-Krook (BGK) kinetic equation to investigate the ow aspects of
a driven cavity over the whole Knudsen number range. More recently, Mizzi et al.
[25] showed comparisons between the NSF equations and the DSMC method for
the driven cavity problem in the slip-ow regime. However, heat transfer characteristics were not considered and are in paucity in the literature. Heat transfer under
rareed ow conditions could depend on a combination of various ow factors such
as thermal creep, viscous dissipation, compressibility, etc., which conventional ow
models cannot reliably predict. Accurate simulations of nonequilibrium heat transfer
can give critical information to designers toward understanding the thermal characteristics and in determining the cooling strategies of micro- and nanodevices. In the
present work, rareed gaseous ow in a lid-driven microcavity is investigated using
the DSMC method for a wide range of ow speeds and Knudsen number. In particular, the focus has been on understanding heat transfer in the cavity under various
nonequilibrium ow conditions.
2. DSMC SIMULATION METHOD
DSMC is a particle method based on kinetic theory for the simulation of dilute
gases. A detailed explanation of the steps involved in the DSMC method can be
found in Bird [12]. The evolution of molecules is determined by tracking their positions and velocities. Starting from a set of initial conditions, the ow develops in a
physically realistic manner by evaluating the collisions between the particles and
interactions of the particles with solid bodies and inow=outow boundaries. In contrast to molecular dynamics (MD), the exact trajectory of each particle is not calculated in DSMC; instead, a stochastic algorithm is used to evaluate the collision
probabilities and scattering distributions based on kinetic theory. As the time step
in DSMC simulations is much smaller than the mean collision time, collisions
between particles can be decoupled from the positional changes of the particle for

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

290

B. JOHN ET AL.

each time step; i.e., the particles are moved as if they do not interact and collision is
considered only after all the particles have moved during the time step.
All the simulations in the present study are performed with a validated parallel
DSMC code originally developed for understanding microow in a head-disk interface (HDI) gap [2628]. In addition to the early validation for the HDI case, the code
has also been independently validated against simulated data from other DSMC
codes for planar Couette ow [8] and force-driven Poiseuille ow [19]. For all the
ow calculations, the gaseous medium is assumed to consist of monatomic molecules
corresponding to that of argon, with mass m 6.63  1026 kg. The variable hard
sphere (VHS) collision model [12] has been used with a reference particle diameter
of d 4.17  1010 m. The VHS model has been employed because it can reproduce
accurately the variation of viscosity with temperature by treating the cross section of
the molecules as a function of the relative translational energy. The particlewall
interactions are assumed to be inelastic and follow the diffuse reection model with
full thermal accommodation. The general guidelines that need to be followed for an
accurate DSMC simulation are that the linear cell dimension, Dx, be less than the
mean free path, and that the time step be less than the mean collision time while
at the same time having about 2030 particles per cell, as found in Bird [12]. The
error in transport coefcients as a function of cell size and time step has been rigorously derived by Alexander et al. [29] and Hadjiconstantinou [30]. The statistical
error associated with the number of particles per cell and the sampling period in particle simulations has been characterized by Hadjiconstantinou et al. [31]. In the
present work, all the guidelines have been rigorously followed to ensure the accuracy
of the computed ow results. The cell size used in the present study is, at most, 0.3
times the mean free path, and in most cases is much smaller than that. The time step
for all the DSMC simulations is taken to be ve times smaller than Dx=Vmp, where
Dx is the smallest cell dimension and Vmp is the most probable molecular velocity. To
minimize statistical noise, about 200 particles per cell have been employed in the
present work. The ow variables are computed as moments, which are derived as
averages during the sampling phase once the system attains a steady state. In the
present study, sampling is carried out over a period of 3 million time steps after
the system has reached a steady state.
3. RESULTS AND DISCUSSION
The conguration of the square driven cavity, of size L, is shown in Figure 1.
The notations A, B, C, and D shown in the gure denote the four corners of the
cavity. The top lid moves with a xed tangential velocity Uw in the positive x direction, and the other walls are stationary. The wall temperature is set to the reference
temperature, i.e., Tw T0 273 K. Any variation in Knudsen number is achieved by
changing the density conditions in the cavity, i.e., the reference pressure P0, at which
the simulation is carried out, differs for each Kn. The mean free path k is dened as
in Eq. (1),
m
k
P0

r
pRT0
2

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

NONEQUILIBRIUM HEAT AND MASS TRANSFER

291

Figure 1. Conguration of driven cavity ow problem.

where m is the viscosity of the gas based on the reference temperature. Computations
have been carried out over a range of lid velocities and various degrees of nonequilibrium ow conditions quantied by the Knudsen number. The impact of lid
velocity and Knudsen number on the ow and, in particular, on the heat transfer
characteristics in the cavity are investigated.

3.1. Impact of Lid Velocity on Flow and Heat Transfer


To assess the ow and heat transfer in the cavity we have considered four different lid velocities, i.e., Uw 10, 50, 100, and 200 m=s. These correspond to Mach
numbers of 0.032, 0.16, 0.32, and 0.64, and to Reynolds numbers of about 1, 5,
10, and 20, respectively. The Knudsen number is kept xed for all the cases in this
part of the study, i.e., Kn 1. Figure 2 shows the computed velocity streamlines
superimposed on pressure contours in the cavity for the different cases considered.
From the contours, it is observed that a signicant drop in pressure is seen toward
the top left corner of the cavity, whereas a signicant rise in pressure is observed
toward the top right corner of the cavity. This can be attributed to the expansion
of gas at the top left corner as it is driven by the moving lid, whereas it is compressed
at the opposite end. At 10 m=s, signicant variation in pressure is observed only at
the top corners of the cavity, while in the larger portion of the domain the pressure
essentially remains constant. With an increase in lid velocity, the pressure contours in
the cavity are no longer symmetrical and signicant variations in pressure are
observed throughout the cavity. The drop and rise in pressure observed at the top
corners of the cavity for the higher-velocity cases are more pronounced. A comparison of the computed velocity streamlines for the different cases considered in Figure 2
indicates that the streamlines become asymmetrical with increasing lid velocity. As the
lid velocity increases, the vortex center is shifted toward the right and also slightly
upward toward the moving wall. The vortex center computed for Uw 10 m=s

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

292

B. JOHN ET AL.

Figure 2. Comparison of computed velocity streamlines overlaid on pressure (N=m2) contours in the driven cavity at Kn 1 and different cases of lid velocity: (a) Uw 10 m=s; (b) Uw 50 m=s; (c) Uw 100 m=s;
(d) Uw 200 m=s.

lies at x=L 0.501, y=L 0.698, whereas that predicted for Uw 200 m=s lies at
x=L 0.539, y=L 0.714.
The variation in shear stress along different sections of the cavity has also been
studied. Changes in shear stress along a horizontal and a vertical line crossing the
center of the cavity are shown in Figure 3. The plots indicate that the shear stress
is largely negative in the cavity, which is due to the moving wall, and its magnitude
reaches a maximum along the moving wall. Increasing the lid velocity results in
larger values of shear stress, and the patterns become asymmetrical.
The velocity components, u and v, in the x and y directions, respectively, are
shown in Figure 4. The u-velocity prole is characterized by velocity slip on the
upper and lower walls. It is interesting to note that the nondimensional slip velocity
is largely unaffected by the signicant change in lid velocity. From Figure 4b, it is
observed that at a lid velocity of 10 m=s, the v-velocity prole is symmetrical about
a vertical line crossing the cavity center. However, at higher lid velocities, the v velocity is no longer symmetrical and is shifted to the right. This is consistent with the
vortex center shifting toward the right of the cavity at higher lid velocities, as
observed from the velocity streamline plots in Figure 2.
Heat transfer under nonequilibrium gaseous ow conditions forms an interesting study. Figure 5 shows plots of heat ux stream traces overlaid on temperature

NONEQUILIBRIUM HEAT AND MASS TRANSFER

293

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

Figure 3. Variation of computed shear stress as a function of lid velocity plotted along (a) a vertical line
and (b) a horizontal line crossing the center of the cavity at Kn 1.

contours at Kn 1 with Uw 10, 50, 100, and 200 m=s. When the lid velocity is high,
the temperature eld and the heat ux vary signicantly. From the temperature contours, it is observed that for all cases considered, a cold region is found toward the
left corner of the cavity, whereas a hot region is observed toward the right corner of
the cavity. More interestingly, it is noted that for all cases, the direction of heat ow
is generally from the cold to the hot region, as illustrated in the heat ux streamline
plots. This represents a counter-gradient heat ux made possible by the rareed ow
conditions. Under nonequilibrium ow conditions, various factors such as expansion
cooling, viscous heat generation, compressibility, and thermal creep could signicantly affect ow and heat transfer. For the driven cavity case, an expansion cooling
(gas temperature T less than wall temperature Tw) occurs at the top left corner of the
cavity due to a drop in pressure which results in heat transfer from the wall to the
gas, whereas viscous heat generation (T > Tw) results in heat transfer from the gas
to the wall toward the right corner of the cavity. The direction of heat transfer is

Figure 4. Computed (a) u-velocity prole plotted along a vertical line and (b) v-velocity prole plotted
along a horizontal line, crossing the center of the cavity at Kn 1.

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

294

B. JOHN ET AL.

Figure 5. Comparison of heat ux stream traces overlaid on temperature contours (K) in the driven cavity
computed at Kn 1 and different cases of lid velocity: (a) Uw 10 m=s; (b) Uw 50 m=s; (c) Uw 100 m=s;
(d) Uw 200 m=s.

governed by both expansion cooling and viscous dissipation. The heat transfer plots
in the driven cavity indicate that thermal energy transfer need not always be from a
hot region to a cold region as continuum theory dictates. A recirculation pattern is
observed from the heat transfer streamlines toward the bottom of the cavity, where
no signicant variations in temperature exist. At Uw 10 m=s the streamlines are
symmetrical, but with increasing lid velocity the streamline patterns begin to lose
their symmetry.
Variations in gas temperature and heat ux along the four walls of the cavity
are shown in Figure 6. In the wall plots, AB, BC, CD, and DA (as shown in
Figure 1) represent the four walls of the cavity, i.e., left, top, right, and bottom walls,
respectively, and s=L denotes the nondimensional distance along the walls starting
from point A. The heat ux qx in these plots is nondimensionalized with respect
to the parameter q0, where q0 mRT0=L. Temperature variations along the walls
are found to be similar to those observed for pressure. The fall and rise in observed
temperature are clearly affected by an increase in lid velocity. Along the left stationary wall, the gas temperature is lower than the wall temperature, i.e., T < Tw, and this
results in expansion cooling, while along the top moving lid and right stationary wall
we have T > Tw, which is due to viscous dissipation. With increasing lid velocity, the

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

NONEQUILIBRIUM HEAT AND MASS TRANSFER

295

Figure 6. Variation of computed (a) gas temperature and (b) heat ux qx as a function of lid velocity
plotted along the four walls of the driven cavity at Kn 1.

effect of both expansion cooling and viscous dissipation signicantly increases,


impacting the direction of heat transfer. At Uw 200 m=s, the rise in temperature
is more pronounced than the predicted fall in temperature. This indicates that at
very high lid velocities, viscous dissipation plays a more signicant role in comparison to expansion cooling, and heat transfer becomes predominantly directed from
the top moving lid to the stationary right and bottom walls as shown in Figure 5d.
Predictably, the magnitude of the heat ux also rises with an increase in lid velocity.
The maximum variations in heat ux are observed in regions near the intersection
of the top moving wall with the right stationary wall. In regions along the other
walls the heat ux remains reasonably constant.

3.2. Impact of Knudsen Number on Flow and Heat Transfer


The impact of rarefaction on the ow and heat transfer in the driven microcavity has been assessed by carrying out numerical simulations for a wide range of
Knudsen numbers from the slip regime through to the free molecular regime. Only
selected results will be shown here, i.e., Kn 0.1, 0.5, 1, 2, and 8. In this part of the
study, the lid velocity has been xed at Uw 50 m=s. Figure 7a shows the u-velocity
component along a vertical line crossing the cavity center as a function of Kn. As
expected, the velocity slip increases with any increase in Kn. Also shown in
Figure 7b is the v-velocity component along a horizontal line crossing the center
of the cavity at different values of Knudsen number. The v-velocity is expected to
be close to zero at both the left and right stationary walls of the cavity for ow near
the continuum regime. For the case of nonequilibrium ows considered here, it is
noted that with increasing Kn, there is an increase in the v-velocity-slip which attains
a nite value. The ow is almost symmetric about the vertical line crossing the center
of the cavity for all cases of Kn, even at a lid velocity of 50 m=s. With an increase in
Knudsen number, the computed vortex center is found to be shifted downward

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

296

B. JOHN ET AL.

Figure 7. Computed (a) u-velocity prole plotted along a vertical line and (b) v-velocity prole plotted
along a horizontal line, crossing the center of the cavity at Uw 50 m=s.

slightly away from the moving wall. The vortex center computed at Kn 0.1 lies at
x=L 0.511, y=L 0.741, whereas that predicted at Kn 8 lies at x=L 0.510,
y=L 0.685.
The variation of shear stress in the driven cavity as a function of Kn has
also been studied, and comparisons are made along various sections of the cavity.
In this part of the study, the shear stress has been normalized with respect to r0,
where r0 mUw=L. Figure 8a shows the shear stress variation along a vertical line
crossing the center of the cavity, while Figure 8b indicates how it varies along a
horizontal line crossing the center of the cavity. From these gures it is seen that
for large Kn, the shear stress reduces signicantly. This can be attributed to the

Figure 8. Variation of computed shear stress as a function of Kn plotted along (a) a vertical line and (b) a
horizontal line crossing the center of the cavity of Uw 50 m=s.

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

NONEQUILIBRIUM HEAT AND MASS TRANSFER

297

increased dependency of shear stress on the density with increasing Kn. At large
values of Kn (i.e., Kn >> 1) and as the ow approaches the free molecular regime,
the shear stress tends toward zero. These trends in the variation of shear stress
with Kn are also conrmed in Figure 9, which shows how it varies along the four
walls of the cavity. The magnitude of the shear stress reaches its maximum near
the region of intersection of the moving lid with the right stationary wall, but as
we approach the free molecular regime, the comparative change is substantially
reduced.
To assess the impact of nonequilibrium ow conditions on the thermal characteristics in the driven cavity, heat ux streamlines have been studied for ows over a
wide range of Kn. Figure 10 shows plots of heat ux stream traces overlaid on temperature contours at Uw 50 m=s for Kn 0.1, 0.5, 1, and 8. From the contours, the
observed fall and rise in temperature in regions at either end of the cavity increases as
the Knudsen number increases. For all Kn considered (which are in the nonequilibrium ow regime), the direction of heat transfer is from the cold region to the
hot region. At Kn 0.1, signicant variations in temperature are restricted to
regions near the top left and top right corners, while in the major portion of the
domain, the temperature remains fairly constant. A heat transfer stagnation point
is observed at the bottom left corner of the cavity at Kn 0.1. For higher values
of Knudsen number, temperature variations are observed throughout the cavity.
With increasing Kn, the role of expansion cooling on the heat transfer mechanism
increases, as T < Tw in a larger region near the left wall. At Kn 8, the gas temperature is less than the wall temperature in the entire region near the left wall. This,
together with viscous dissipation of the gas toward the right wall, governs the heat
transfer for ow at high Kn. The heat transfer plots indicate that with an increase
in rarefaction, both expansion cooling and viscous heat generation play a greater

Figure 9. Variation of computed shear stress as a function of Kn plotted along the four walls of the driven
cavity at Uw 50 m=s.

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

298

B. JOHN ET AL.

Figure 10. Comparison of heat ux stream traces overlaid on temperature contours (K) in the driven cavity computed at Uw 50 m=s and different cases of Kn: (a) Kn 0.1; (b) Kn 0.5; (c) Kn 1; (d) Kn 8.

role in determining the heat transfer characteristics, as the heat ux vectors become
more predominantly from the stationary left wall to the right wall.
Comparisons of pressure and temperature along a horizontal line crossing the
center of the cavity are shown in Figure 11. Although the trend in the variation of
pressure is the same as that for temperature, actual changes in temperature values
are minimal compared to those for pressure, indicating that pressure values are
most affected by the change in number density. The variations in gas temperature
and heat ux along the four walls of the cavity, as a function of Kn, are shown in
Figure 12. The fall and rise in observed temperature (as well as temperature jump)
clearly increase with an increase in Kn. Interestingly, from the heat ux plot along
the walls, it is observed that the variations in heat ux are noticeably higher at
lower Kn. With any increase in Kn, the magnitude of the heat ux becomes signicantly lower, and as the ow approaches the free molecular regime, the heat ux
values are fairly uniform and approach zero. This trend in the variation of heat
ux with increasing Kn is similar to that observed for the variation in shear stress.
Again, this can be attributed to the increased dependency of heat ux on the
density as Kn increases.

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

NONEQUILIBRIUM HEAT AND MASS TRANSFER

299

Figure 11. Predicted (a) pressure and (b) temperature proles as a function of Kn plotted along a horizontal line crossing the center of the cavity of Uw 50 m=s.

3.3. Comparisons with the Navier-Stokes-Fourier Model


It is generally regarded that the NSF equations, modied with various velocityslip and temperature-jump boundary conditions, are able to capture ow features
accurately in the slip regime. However, in certain instances, the NSF model does
not capture several well-known nonequilibrium phenomena, such as the bimodal
temperature prole in force-driven Poiseuille ow [19] and nongradient heat ux
in Couette ow [8]. A study has been carried out to test whether the NSF model,
modied with second-order slip and temperature-jump boundary conditions, can
capture critical ow features accurately in the slip regime for the driven cavity
problem. An in-house NSF and higher-moment equation solver, THOR [7, 8], was

Figure 12. Variation of computed (a) gas temperature and (b) heat ux qx as a function of lid velocity
plotted along the four walls of the driven cavity at Uw 50 m=s.

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

300

B. JOHN ET AL.

Figure 13. Comparison of (a) u-velocity prole along a vertical line and (b) v-velocity prole along a
horizontal line crossing the center of the cavity predicted by the DSMC and NSF methods at
Kn 0.05 and Uw 50 m=s.

used to obtain the NSF solutions. Comparisons have been made at two Knudsen
numbers, Kn 0.02 and Kn 0.05. Figure 13 compares the velocity components
computed by the DSMC and NSF methods at Kn 0.05 and Uw 50 m=s. Good
agreement is obtained for the velocity components, as expected in the slip regime.
Thermal characteristics computed by the two methods at Kn 0.05 are compared
in Figure 14. It is observed that the temperature eld predicted by the NSF equations
at Kn 0.05 is completely different from that predicted by DSMC. The expansion
cooling phenomena is not observed from the NSF data, with no temperature drop
predicted at the top left corner of the cavity. The temperature, in general, is underpredicted by the NSF model and the peak temperature is found to be close to the
center-right of the moving lid. More importantly, the direction of heat ow predicted

Figure 14. Comparison of heat ux stream traces overlaid on temperature contours (K) in the cavity
predicted by (a) NSF and (b) DSMC at Kn 0.05 and Uw 50 m=s.

NONEQUILIBRIUM HEAT AND MASS TRANSFER

301

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

Figure 15. Comparison of heat ux stream traces overlaid on temperature contours (K) in the cavity
predicted by (a) NSF and (b) DSMC at Kn 0.02 and Uw 100 m=s.

by the NSF model is found to be from the hot region to the cold region. Comparisons at Kn 0.02 and Uw 100 m=s are shown in Figure 15. The lid velocity considered for Kn 0.02 is Uw 100 m=s, with a view to attaining faster convergence
and less statistical uctuations for the DSMC simulations. At Kn 0.02, for which
the ow is in the early slip regime and closer to continuum, the temperature is again
underpredicted by the NSF model. However, the trends in the temperature eld predicted by both the NSF and DSMC methods are consistent, and the heat transfer
direction predicted by both methods is from hot regions to cold regions. It is noted
from the temperature eld that for ow near the continuum regime, expansion cooling phenomena does not occur, as T > Tw in almost the entire region near the left
wall. Here, viscous dissipation alone dictates the heat transfer direction. These
results indicate that signicant nonequilibrium ow conditions could exist even in
the slip regime, which conventional ow models cannot capture. The NSF model,
which is based on a gradient transport mechanism, cannot predict the countergradient heat ux in the driven cavity, as it does not take into account any
nongradient transport phenomena that may become signicant at the onset on nonequilibrium [8]. Alternative modeling techniques must be employed to gain insight
into all the aspects of the ow, including thermal ow patterns.

4. CONCLUSIONS
The direct simulation Monte Carlo method has been employed to investigate
the ow and thermal characteristics in a lid-driven microcavity under nonequilibrium gaseous ow conditions. The impact of lid velocity and Knudsen number
on the ow and heat transfer patterns in the cavity has been studied. It has been
observed that rarefaction effects signicantly reduce the shear stress and heat ux
rates for a xed lid velocity, as they become increasingly dependent on the density
of the ow with increase in Knudsen number. Heat ux streamlines indicate a
counter-gradient heat ux pattern where the direction of ow is from the cold to

302

B. JOHN ET AL.

the hot region in the cavity, from the slip-ow regime through to the free molecular
regime. For ow in the continuum regime, viscous dissipation governs heat transfer,
as the heat transfer is from hot to cold regions. However, for ow under nonequilibrium conditions, it is revealed that both expansion cooling and viscous dissipation
dictate the heat transfer mechanism, as the gaseous heat transfer direction becomes
nonintuitive, i.e., from cold to hot regions. It has been shown in the present work
that if the boundary conditions for the Navier-Stokes-Fourier equations are modied to account for velocity slip and temperature jump, they cannot capture these
nonintuitive thermal ow features, even though they may capture other ow features
such as velocity reasonably well. Our results illustrate that the direct simulation
Monte Carlo approach is essential for this type of computation and provides
valuable insight into all aspects of the ow under nonequilibrium conditions.

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

REFERENCES
1. M. Gad-el-Hak, The Fluid Mechanics of MicrodevicesThe Freeman Scholar Lecture,
J. Fluids Eng., vol. 121, pp. 533, 1999.
2. G. Karniadakis, A. Beskok, and N. Aluru, Microows and Nanoows: Fundamentals and
Simulation, Springer-Verlag, New York, 2005.
3. S. Chapman and T. G. Cowling, The Mathematical Theory of Non-uniform Gases,
Cambridge University Press, Cambridge, 1990.
4. H. Grad, On the Kinetic Theory of Rareed Gases, Commun. Pure Appl. Math., vol. 2,
pp. 331407, 1949.
5. I. Muller and T. Ruggeri, Extended Thermodynamics, Springer-Verlag, New York, 1993.
6. H. Struchtrup and M. Torrilhon, Regularization of Grads 13 Moment Equations:
Derivation and Linear Analysis, Phys. Fluids, vol. 15, pp. 26682680, 2003.
7. X. J. Gu and D. R. Emerson, A Computational Strategy for the Regularized 13 Moment
Equations with Enhanced Wall-Boundary Conditions, J. Comput. Phys., vol. 225,
pp. 263283, 2007.
8. X. J. Gu and D. R. Emerson, A High-Order Moment Approach for Capturing
Non-Equilibrium Phenomena in the Transition Regime, J. Fluid Mech., vol. 636,
pp. 177216, 2009.
9. X. J. Gu, D. R. Emerson, and G. H. Tang, Analysis of the Slip Coefcient and Defect
Velocity in the Knudsen Layer of a Rareed Gas Using the Linearized Moment
Equations, Phys. Rev. E, vol. 81, 016313, 2010.
10. A. V. Bobylev, The Chapman-Enskog and Grad Methods for Solving the Boltzmann
Equation, Sov. Phys. Dokl., vol. 27, pp. 2931, 1982.
11. L. S. Garcia-Colin, R. M. Velasco, and F. J. Uribe, Beyond the NavierStokes Equations:
Burnett Hydrodynamics, Phys. Rep., vol. 465, pp. 149189, 2008.
12. G. Bird, Molecular Gas Dynamics and the Direct Simulation of Gas Flows, Clarendon
Press, Oxford, UK, 1994.
13. S. Hou, Q. Zou, S. Chen, G. Doolen, and A. C. Cogley, Simulation of Cavity Flow by the
Lattice Boltzmann Method, J. Comput. Phys., vol. 118, pp. 329347, 1995.
14. P. N. Shankar and M. D. Deshpande, Fluid Mechanics in the Driven Cavity, Annu. Rev.
Fluid Mech., vol. 32, pp. 93136, 2000.
15. K. Aoki, H. Yoshida, T. Nakanishi, and A. L. Garcia, Inverted Velocity Prole in the
Cylindrical Couette Flow of a Rareed Gas, Phys. Rev. E, vol. 68, 016302, 2003.
16. C. Cercignani and S. Cortese, Validation of a Monte Carlo Simulation of the Plane
Couette Flow of a Rareed Gas, J. Stat. Phys., vol. 75, pp. 817838, 1994.

Downloaded By: [John, Benzi] At: 10:56 15 November 2010

NONEQUILIBRIUM HEAT AND MASS TRANSFER

303

17. D. R. Emerson, X. J. Gu, S. K. Stefanov, S. Yuhong, and R. W. Barber, Nonplanar


Oscillatory Shear Flow: From the Continuum to the Free-Molecular Flow, Phys. Fluids,
vol. 19, 107105, 2007.
18. J. H. Park, S. W. Baek, S. J. Kang, and M. J. Yu, Analysis of Thermal Slip in Oscillating
Rareed Flow Using DSMC, Numer. Heat Transfer A, vol. 42, pp. 647659, 2002.
19. M. M. Mansour, F. Baras, and A. L. Garcia, On the Validity of Hydrodynamics in Plane
Poiseuille Flow, Physica A, vol. 240, pp. 255267, 1997.
20. H. Sun and M. Faghri, Effects of Rarefaction and Compressibility of Gaseous Flow in
Microchannel Using DSMC, Numer. Heat Transfer A, vol. 38, pp. 153168, 2000.
21. H. Liu, M. Wang, J. Wang, G. Zhang, H. Liao, R. Huang, and X. Zhang, Monte Carlo
Simulations of Gas Flow and Heat Transfer in Vacuum Packaged MEMS Devices, Appl.
Thermal Eng., vol. 27, pp. 323329, 2007.
22. C. Cai, Heat Transfer in Vacuum Packaged Microelectromechanical System Devices,
Phys. Fluids, vol. 20, 017103, 2008.
23. X. Nie, G. D. Doolen, and S. Chen, Lattice Boltzmann Simulations of Fluid Flows in
MEMS, J. Stat. Phys., vol. 112, pp. 279289, 2002.
24. S. Naris, and D. Valougeorgis, The Driven Cavity Flow over the Whole Range of the
Knudsen Number, Phys. Fluids, vol. 17, 097106, 2005.
25. S. Mizzi, D. R. Emerson, S. K. Stefanov, R. W. Barber, and J. M. Reese, Effects of
Rarefaction on Cavity Flow in the Slip Regime, J. Comput. Theor. Nanosci., vol. 4,
pp. 817822, 2007.
26. B. John and M. Damodaran, Computation of Head-Disk Interface Gap Micro Flowelds
Using DSMC and Continuum-Atomistic Hybrid Methods, Int. J. Numer. Meth. Fluids,
vol. 61, pp. 12731298, 2009.
27. B. John and M. Damodaran, Parallel Three Dimensional Direct Simulation Monte Carlo
for Simulating Micro Flows, Lect. Notes Comput. Sci. Eng., vol. 67, pp. 9198, 2009.
28. B. John and M. Damodaran, Hybrid ContinuumDirect Simulation Monte Carlo and
Particle-Laden Flow Modeling in the Head-Disk Interface Gap, IEEE Trans. Magn.,
vol. 45, pp. 49294932, 2009.
29. F. J. Alexander, A. L. Garcia, and B. Alder, Cell Size Dependence of Transport Coefcients in Stochastic Particle Algorithms, Phys. Fluids, vol. 10, pp. 15401542, 1998.
30. N. G. Hadjiconstantinou, Analysis of Discretization in the Direct Simulation Monte
Carlo, Phys. Fluids, vol. 12, pp. 26342638, 2000.
31. N. G. Hadjiconstantinou, A. L. Garcia, M. Z. Bazant, and G. He, Statistical Error
in Particle Simulations of Hydrodynamic Phenomena, J. Comput. Phys., vol. 187,
pp. 274297, 2003.

Das könnte Ihnen auch gefallen