Journal of Physics Communications

PAPER • OPEN ACCESS Related content

- The physics of confined flow and its
On the numerical simulation of rarefied gas flows application to water leaks, water
permeation and water nanoflows: a review
in micro-channels Wenwen Lei, Michelle K Rigozzi and David
R McKenzie

- Simulation of gas flows in micro/nano

To cite this article: Chariton Christou and S Kokou Dadzie 2018 J. Phys. Commun. 2 035002 systems using the Burnett equations
Fubing Bao, Xia Yu and Jianzhong Lin

- Non-equilibrium lattice Boltzmann model

G. H. Tang, Y. H. Zhang, X. J. Gu et al.
View the article online for updates and enhancements.

Recent citations
- Investigating enhanced mass flow rates in
pressure-driven liquid flows in nanotubes
Alexandros Stamatiou et al

- Recasting Navier–Stokes equations

M H Lakshminarayana Reddy et al

This content was downloaded from IP address 186.155.11.103 on 01/04/2020 at 23:16

 J. Phys. Commun. 2 (2018) 035002 https://doi.org/10.1088/2399-6528/aab066

PAPER

On the numerical simulation of rarefied gas flows in micro-channels

OPEN ACCESS
Chariton Christou and S Kokou Dadzie
RECEIVED
12 September 2017
School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, EH14 4AS, Scotland, United Kingdom

REVISED
E-mail: c.christou@hw.ac.uk
6 February 2018
Keywords: fluid mechanics, Korteweg, Navier–Stokes
ACCEPTED FOR PUBLICATION
19 February 2018
PUBLISHED
1 March 2018 Abstract
Classical continuum methods break down for rarefied gas flows in micro-channels. Two non-
Original content from this
work may be used under
continuum hydrodynamic models are investigated here: compressible Korteweg fluid-like and Bi-
the terms of the Creative Velocity (Volume Diffusion) hydrodynamic models. The pressure driven rarefied gas flow in a
Commons Attribution 3.0
licence. rectangular micro-channel in the whole range of Knudsen number is numerically considered. The
Any further distribution of Korteweg model shows improved results in comparison with standard Navier–Stokes up to Knudsen
this work must maintain
attribution to the number of unity. The Bi-Velocity method is found to allow a match between experiments and
author(s) and the title of
the work, journal citation
numerical results over the full range of Knudsen number.
and DOI.

1. Introduction

Rarefied gas flows are modelled by solving the Boltzmann equation which is valid for the full range of Knudsen
number [1]. Application of the Boltzmann equation and associated kinetic methods such as the Direct
Simulation Monte Carlo (DSMC) remains computationally expensive and their deployment to solving complex
engineering cases (such as those involving multiphase flows) are still to be explored [2]. The inappropriateness of
the traditional continuum flow model, namely the Navier–Stokes equations in rarefied gases and the need for
additional contributions to its constitutive equations was discussed early by Maxwell [3]. Van der Waals theory
of capillarity led to the observation that a liquid-vapour interface actually represents a rapid but smooth
transition of physical quantities between two fluids [4]. Based on this, Korteweg obtained constitutive equations
for the fluid flow equations within which phase transition phenomena occur by incorporating in the stress
tensor contributions from the gradient of density [5].
The development of extended continuum models in gases has shifted since. Higher order hydrodynamics
equations are derived using Chapman-Enskog expansion [6]. Grad’s moments method [7] and associated
regularizations (e.g., R13 equations) are the second route usually explored [8]. Several issues related to these
higher order (Burnett type) equations are now well-known and detailed in the literature [9, 10]. García-Colín
et al claim that problems with Burnett equations are in reality limitations of those methods and highlighted the
need of alternative approaches [11]. Other routes to ameliorate predictions by the Navier–Stokes at large
Knudsen numbers by the use of slip boundary condition corrections have also been extensively
explored [12, 13].
The inappropriateness of Navier–Stokes system to describe temperature field in a stationary gas in the
hydrodynamic limit has been shown by Sone et al [14]. These authors proved that corrections to the standard
constitutive equations to involve Korteweg diffuse interface type stress tensor are required to acquire
appropriate description of heat transfer processes even in the vanishing Knudsen number limit [14]. An
alternative approach to extend the continuum fluid model appeared more recently in the literature based on the
concept of two different conceptual velocities [15, 16]. According to this theory, several fluid velocities can be
distinguished: (a) the mass velocity Um which is linked to the classical notion of mass motion and (b) the volume
velocity Uv associated with changes in the local volume of the fluid material. This model is termed Bi-Velocity
(or Volume Diffusion) hydrodynamics. In classical Navier–Stokes and other continuum flow descriptions one
deals with one velocity (Um).

© 2018 The Author(s). Published by IOP Publishing Ltd

J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

Bi-Velocity model as an extension to the classical Navier–Stokes system deviates from the earlier mentioned
extended gas flow equations that are for example based on solving the Boltzmann equation using series
expansion in Knudsen number [11]. Similarities with Korteweg equations are however found [17]. In a recent
review by Gorban and Karlin with an extensive analysis of the Korteweg model it was noted that it is the first post
Navier–Stokes equation which remains within continuum mechanics while capturing non-equilibrium kinetic
level phenomena. They concluded that Korteweg equations ‘give us a hint of how post Navier–Stokes equations
may look like’ [18].
Bi-Velocity model was tested in a mixed convection problem and was found to provide reasonable results
compared to Navier–Stokes for high Knudsen number flows [19].
Ewart et al provided experimental data for the mass flow rate of rarefied gas flows in a micro-channel for the
whole range of Knudsen number [20]. Dadzie and Brenner derived an associated analytical solution for the
rectangular micro-channel helium gas flow using a model of first-order slip boundary conditions [21]. Their
solution (which was based on Bi-Velocity) showed agreement with the experimental data up to Knudsen
number 5. Following this method, Lv et al incorporated an effective volume diffusivity coefficient to show an
agreement of the Bi-Velocity equation with the experimental data to a Knudsen number of 50 [22].
The present paper expands on these investigations. The Korteweg model of compressible flows and the Bi-
Velocity model are considered separately for the numerical simulations of the gas flow in a rectangular micro-
channel.
The paper is organized as follows: In section 2 we present the two new non-continuum hydrodynamic
models as alternative to Navier–Stokes. In section 3 we describe their numerical implementations and associated
boundary conditions. Section 4 presents the numerical results, comparison with experimental data and
discussion. The paper conclusion is drawn in section 5.

2. The two non-continuum hydrodynamics models

It is well accepted that Navier–Stokes equations fail to describe rarefied gas flows. Among other continuum fluid
models that are being investigated are Korteweg fluid-like and Bi-Velocity models [5, 16]. These models have
been reported to account for non-equilibrium effects in other configurations [23].

2.1. Compressible Korteweg fluid-like model

From Van der Waals theory of capillarity authors noted that near a liquid-vapour phase transition a
thermodynamic formulation of a fluid flow equation includes contributions from a density gradient-energy
term. This formulation introduces a capillarity stress tensor in the momentum equation [24]. This stress tensor,
also referred to as Korteweg stress tensor, may be obtained as [25]:
⎛ 1 ⎞
P = ⎜b ∣ r ∣2 + grDr⎟ I + a (r Ä r ) (2.1)
⎝ 2 ⎠

where ρ is the fluid density, I is the identity tensor and the term α(∇ρ ⊗ ∇ρ) is also sometimes referred to as ‘the
Korteweg tensor’. Phenomenological coefficients α, β, γ are material functions dependent upon ρ and called
capillary coefficients. Their exact value and expression are still elusive [26]. From a dimensional analysis, they
may be represented as:
m2
a = b = g = a* (2.2)
r3

where α* is a scalar and μ is the dynamic viscosity. The standard Navier–Stokes set may be modified by adding
the Korteweg stress (equation (2.1)) to its viscous stress tensor. A compressible Korteweg fluid flow system of
equations is then written as:

2.1.1. Conservation of mass

¶r
¶t
+  ⋅ [rU ] = 0 (2.3a)

2.1.2. Conservation of momentum

¶rU
¶t
+  ⋅ [rUU ] +  ⋅ [ pI + P ] = 0 (2.3b)

2
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

Figure 1. An illustration of mass velocity and volume velocity: while the number of particles/molecules are the same for both velocities
when changing in time and space, the volume is assumed to vary for the volume velocity.

2.1.3. Conservation of total energy

¶ ⎡1 ⎤ ⎡1 ⎤
⎢ rU + re in⎦⎥ +  ⋅ ⎢⎣ rU U + re in U ⎦⎥ +  ⋅ [( pI + P ) ⋅ U ] +  ⋅ J = 0
2 2 (2.3c )
⎣
¶t 2 2

where the shear stress term Π is given by,

P = - 2m U - (b ∣ r ∣2 + grDr ) I + a (r Ä r ) (2.4)

in which we denote
~
U = (U + U ) - I  ⋅ U
1 1
(2.5)
2 3
~
with U denoting the transpose tensor of ∇U and I the identity matrix. The energetic heat flux is then expressed
as [26]:
Dr
J = - k T - KE r (2.6)
Dt
where k is the thermal conductivity and KE is the gradient (internal) energy coefficient which is assumed to be
constant and T the fluid temperature [26]. In the above equations, U is the fluid mass velocity, p is the scalar
pressure and ein is the fluid specific internal energy.

2.2. Bi-Velocity model

The kinetic level derivation of the Volume Diffusion continuum equations is based on the inclusion of an
additional transport term originating from particle molecular level spatial diffusion [16]. Two different
macroscopic fluid velocities can then be derived based on different averaging methods. Bi-Velocity method
recognizes existence of two (conceptually) different velocities. In one hand, mass velocity Um is proportional to
the mass flux and is found in the continuity equation. On the one hand, volume velocity, Uv, accounts for
variation in volume occupied by the fluid mass. A volume flux density relates the two:
Uv = Um - Jc (2.7)
where
km
Jc = - r (2.8)
r
with km an additional transport coefficient: the molecular (or volume) diffusivity coefficient.
Figure 1 depicts the two different concepts of macroscopic fluid velocity. Flux Jc is viewed as a molecular level
diffusive flux associated with gas molecule concentration and −Jc is the gas volume diffusive flux. Um and Uv are
both macroscopic concepts by definition. In the description of Volume Diffusion, a fluid particle is no longer
perceived as a point mass like in classical mechanics but with a varying volume element associated with it. The
diffusive flux −Jc is oriented in the direction from high density to low density in the case of compression and the
reverse in the case of expansion. With no density variation Jc vanishes.
Bi-Velocity hydrodynamics set of equations can be written as [19, 27]:

3
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

2.2.1. Conservation of mass

¶r
¶t
+  ⋅ [rU ] = 0 (2.9a)

2.2.2. Conservation of momentum

¶rU
¶t
+  ⋅ [rUU ] +  ⋅ [ pI + P ] = 0 (2.9b)

2.2.3. Conservation of total energy

¶ ⎡1 ⎤ ⎡1 ⎤
⎢⎣ rU + re in⎥⎦ +  ⋅ ⎢⎣ rU U + re in U ⎥⎦ +  ⋅ [( pI + P ) ⋅ U ] +  ⋅ Ju = 0
2 2 (2.9c )
¶t 2 2

where the shear stress term Π is now given by,

P = Pv - rJc Ä Jc (2.10)

with Pv = - 2m  [U - Jc ] (2.11)
and the energetic heat flux, Ju, is given by,
Ju = qve + pJc . (2.12)
The specific internal energy of the fluid is ein = (3/2)RT, with R being the specific gas constant. Note that in
the set (2.9) the mass velocity Um is now denoted U as it is equivalent to the traditional mass velocity like in a
classical system (2.3). The Fourier’s law now applies to the entropic heat flux as [16]:
qve = - k T. (2.13)
Here, the volume diffusivity coefficient is related to dynamic viscosity, μ, as:
m
km º a* (2.14)
r
where our investigation led to α* as:
a* = Pr 3 (2.15)
*
This value of the coefficient α was previously observed in other classical rarefied gas flow configurations
[19, 27]. The mean free path is calculated using G. Bird formula, λ = (kbT)/((2πd2p)1/2),where kb is the
Boltzmann constant and d is the molecular diameter [28].
By substituting equations (2.14) and (2.15) into (2.8) we have the final expression of the molecular level
volume (actually, concentration) flux as:
mPr
Jc = - 3r 2 r (2.16)

3. Numerical implementation

Based on the two hydrodynamic models described above we propose two numerical solvers developed on
OpenFOAM platform.

3.1. Modification to rhoCentralFOAM in OpenFOAM

Numerical implementations of the two sets of hydrodynamics equations are done on the OpenFOAM platform.
These are adaptations of the solver rhoCentralFOAM which comprises of a finite volume (FV) discretization
using semi-discrete, non-staggered central schemes for collocated variables prescribed on a mesh of polyhedral
cells that have an arbitrary number of faces and solved on a three-dimensional unstructured mesh of polygonal
cells. rhoCentralFOAM was initially developed to simulate compressible flows with better shock capturing [29].
The solver is density-based solver and uses a physical form of the equations described in section 2. It initially
starts with solving the inviscid version of the set of conservative equations in an iterative order. Density-weighted
fields are calculated first. The momentum density is calculated from U = ρU and the total energy density

E = rE where the total energy is E = e in + ∣ U2 ∣ . The set of equations are consequently solved for ρ, U and E.
2

Based on these the temperature, T, is evaluated by the subtraction of kinetic energy from the total energy as:

4
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

1 ⎛ E ∣ U ∣2 ⎞
T= ⎜ - ⎟ (3.1)
cv ⎝ r 2 ⎠
where cv is the specific heat capacity at constant volume.
The momentum and heat diffusion are introduced to the solvers by the inclusion of the appropriate diffusive
terms in the governing constitutive equations. U and T are evaluated explicitly since the momentum and energy
equations are solved explicitly as noted in the inviscid version above.
The solver starts with the solution of the momentum equation by solving for U
¶U
¶t
+  ⋅ [UU ] + p = 0 (3.2)

Mass velocity is then updated by solving U = U r. This step is done before solving a diffusion correction
for U
¶rU
-  ⋅ (m U ) +  ⋅ (m Jc ) -  ⋅ (Pexp) = 0 (3.3)
¶t
For Bi-Velocity method the explicit component of the stress tensor is expressed:
2 2
Pexp = m [(U )T - (Jc )T ] - tr(U ) I + tr(Jc ) (3.4)
3 3
In equation (3.3) we note the appearance of the term ∇ · (μ∇Jc) which is neglected for the Korteweg solver.
The Laplacian terms from the deviatoric tensor ∇ · (μ∇U) and ∇ · (μ∇Jc) are implemented implicitly and they
form coefficients within the solution matrices, rather than values in the source vectors. Similar procedure is
followed to implement the Korteweg compressible fluid model where the additional explicit component of the
stress tensor is now:
⎡ 2 ⎤ ⎛ m2 m2 ⎞ m2
Pexp = m ⎢(U)T - tr(U) I⎥ + a* ⎜ 3 ∣ r ∣2 + 2 Dr⎟ I + a* 3 r Ä  (3.5)
⎣ 3 ⎦ ⎝ 2r r ⎠ r
To obtain the solution for the energy equation, similar procedure is followed by solving first for E as:
¶E
+  ⋅ [U (E + p )] +  ⋅ (P ⋅ U ) = 0 (3.6)
¶t
The temperature, T is updated from ρ, U and E from equation (3.1) before solving a diffusion correction
equation for T via:
⎛ ¶(rc v T ) ⎞ ⎛p ⎞
⎜ ⎟ -  ⋅ (k T ) - k m  ⋅ ⎜ r⎟ = 0 (3.7)
⎝ ¶t ⎠ ⎝r ⎠
We note, however, that the experimental rectangular micro-channel case considered here is isothermal and
therefore solutions to the energy equations are not as important as the momentum equations.

3.2. Boundary conditions

Maxwell type first-order slip boundary conditions are adopted to accompany the different hydrodynamics
equations. In more detail, we used the compressible slip boundary conditions as presented below [30]:
⎛ 2 - su ⎞ l 3 Pr(g - 1)
U - Uw = - ⎜ ⎟ tv - j (3.8)
⎝ su ⎠ m 4 gp
where Uw is the wall velocity, σu is the tangential momentum accommodation coefficient and τv is the new
tangential shear stress, τv = S · (n · Π), extracted from the new modified shear stress in (2.4) and (2.10). The
tensor S = I−nn with j = J · S based on the new energetic heat flux in equations (2.6) and (2.12). The unit normal
vector n is defined as positive in the direction of the flow domain. The specific heat ratio is named γ.

3.3. The micro-channel configuration

Dimensions of the channel were accurately set to the original from Ewart et al [20]. The purpose of their
experiments was to complete the database of mass flow rate measurements, obtained for a gas flow in a single
micro-channel (see figure 2) ranging from the hydrodynamics to the near free molecular regime. The average
pressure in the experiments was between 67 000 Pa and 30 Pa. The experimental method that was used to
measure the mass flow rate through the micro-channel involves the use of two large constant volume tanks that
are much larger than the volume of the micro-channel. More details of the experimental procedure can be found
in Ewart et al [20]. Their data (mass flow rate results) are used as a benchmark for our simulations. Fluid
properties and physical coefficients that are used in the simulations are listed in table 1. The mean Knudsen
number is defined: .

5
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

Figure 2. Schematic diagram of the micro-channel configuration. The flow direction is represented by the red arrows for pinlet and
poutlet. The blue circles show the rarefied gas molecules, where L,W ? h.

Table 1. Fluid properties and micro-channel physical coefficients.

W (m) L (m) h (m) Pr μ (Pa s) R (J kg−1 K−1) kλ

4.92 × 10−4 9.39 × 10−3 9.38 × 10−6 0.67 1.975 13 × 10−5 2076.942 p 2

1 m
Knmean = kl 2RT (3.9)
h pinlet + poutlet
2

It can be noted that the length (L) and the width (W) of the micro-channel are extensively greater than its
height (h).
Tangential momentum accommodation coefficient was set to 0.9 for all simulations along the whole range
of Knudsen number in this paper.
To perform cell independency tests on the three numerical methods developed during the present
investigations, three grids composed of 100 × 50 × 50, 200 × 100 × 100 and 300 × 150 × 250 cells are
considered. figure 3 shows the corresponding normalized velocity profiles along the vertical center line of the
channel for all methods. It is observed that numerical results are equivalent for 200 × 100 × 100 and
300 × 150 × 250 grids. As a result, the grid containing 200 × 100 × 100 cells were selected for the reported
results.

4. Results and discussion

We compare numerical predictions by Korteweg, Bi-Velocity, Navier–Stokes and a previously derived analytical
model in [21] with the experimental data of Ewart et al [20]. The following non-dimensional mass flow rate
equation is used for the analysis:
⎡ Wh2 ⎤-1
Gm = M ⎢ ( pinlet - poutlet ) ⎥ (4.1)
⎣ L 2RT ⎦

The mass flow rate, M through the micro-channel is given by:

h
M = W ò0 rUdy (4.2)

Figure 4 presents the comparison between the models.

Compressible Korteweg fluid-like model with its additional capillarity contribution to the shear stress gives
agreement with experiments up to Knudsen number of about unity with minor deviation observed around
Knmean between 0.2–0.6. Bi-Velocity model based on volume diffusivity coefficient as expressed in
equations (2.14) and (2.15) appears to fit the experimental data over the full flow regime, including excellent
agreement in the free molecular regime (10 < Knmean < 100). Dadzie-Brenner solution which was based also on
the Bi-Velocity hydrodynamics model, but with a form of a slip condition and another expression of the volume
diffusivity coefficient agrees with the experimental data for Knmean < 5. Knudsen studied gas flows through
tubes in the transition and free molecular regimes. From his results, the normalized volumetric flow rate showed
a minimum at a Knudsen number near unity [31]: The Knudsen paradox. The present numerical solution by the

6
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

Figure 3. Grid independency test for: (a) Navier–Stokes, (b) Korteweg and (c) Volume Diffusion.

Bi-Velocity model and Dadzie-Brenner solution as shown in figure 4 both capture the Knudsen paradox. This
Knudsen paradox is usually difficult to capture using a Navier–Stokes with slip conditions [13].
In general, figure 4 reveals that the two continuum hydrodynamics models which involved density gradient
expressions in the shear stress constitutive equations are better in predicting this micro-channel gas flow. This
corroborates recent observations by Gorban and Karlin [18].

7
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

Figure 4. Comparison of the mass flow rate using Bi-Velocity and Korteweg against Navier–Stokes, Dadzie-Brenner and experimental
data.

Figure 5 presents the normalized pressure distribution along the streamwise direction for the Bi-Velocity,
Korteweg and Navier–Stokes. The convex curvature is captured by the three methods identically for the low
Knudsen number. Difference between Bi-Velocity and Korteweg is insignificant across the Knudsen number
range. This may be explained by the fact that both method constitutive equations for the shear stresses contain
the density gradient capillarity effects. For a higher Knudsen number that corresponds to the free molecular
regime (Kn = 80), Navier–Stokes profile may be described as unphysical compared to the other two models that
appear to predict a rather monotonic variation in the profiles from the low to the higher Knudsen number.
Figure 6 shows the normalized velocity profiles in a cross-section of the streamwise direction in a position
where the flow is fully developed for all methods. U0 is the average velocity taken at the channel entrance. For
Kn = 0.043 the mass velocity as derived from the Navier–Stokes has a small difference from Bi-Velocity and
Korteweg mass velocity. The velocity profile differs from the parabolic shape expected from a pure
hydrodynamic regime pressure-driven flow. The mass velocity decreases as the Knudsen number increases and
vanishes for the Navier–Stokes at higher Knudsen number (figure 6(c)) which is consistent with the no mass flow
predicted by this model at higher Knudsen number. Bi-Velocity predicts the highest mass velocity at the highest
Knudsen number which is in agreement with the mass flow rate plotted in figure 4.

5. Conclusion

Two extended hydrodynamic models based on Korteweg and Bi-Velocity theories have been investigated for the
simulation of micro-channel gas flows. The simulation results are compared with existing experimental data.
The two methods agree well for pressure and velocity profiles. Regarding the Korteweg model, it has been shown
that it can capture non-equilibrium effects and match the experimental data up to Kn = 1. Korteweg model
diverges from solutions obtained with the Bi-Velocity model at higher Knudsen numbers. The Bi-Velocity
model shows the excellent agreement for the mass flow rate over the whole range of Knudsen number. The
investigation reveals new insights into the expression of the volume diffusivity coefficient. The capability of the
Bi-Velocity model to capture non-equilibrium effects in rarefied gas flows was demonstrated previously. The
present study reinforces this. The results provide further evidences of the possibility of simulating rarefied gas
flow phenomena with relatively high Knudsen number by a pseudo-continuum fluid flow model. Volume
Diffusion model can agree well with the experimental data over the whole range of Knudsen number up into the
free molecular regime. These results trigger further future investigations into the mathematical structure of both
the Korteweg and the Volume Diffusion type of continuum flow equations in describing non-local-equilibrium
flows beyond the classical model. The study not only provides a simple, accurate and reliable numerical method
for simulating rarefied gas flows in micro-channels but also shows how post Navier–Stokes equations may
look like.

8
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

Figure 5. Normalized pressure distribution along the streamwise direction for various Knudsen numbers (a) Kn = 0.043,
(b) Kn = 1.07 (c) Kn = 80.7.

9
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

Figure 6. Normalized mass velocity profiles along the micro-channel in cross section of the flow direction (a) Kn = 0.043, (b) Kn = 1.07
(c) Kn = 80.7.

Acknowledgments

This work is supported by the EPSRC: RCUK under grant: From Kinetic Theory to Hydrodynamics: re-imagining
two fluid models of particle-laden flows, EP/R008027 and EP/R007438.

10
J. Phys. Commun. 2 (2018) 035002 C Christou and S K Dadzie

ORCID iDs

Chariton Christou https://orcid.org/0000-0002-1844-0819

References
[1] Cercignani C 1988 The Boltzmann equation The Boltzmann Equation and Its Applications (Berlin: Springer) pp 40–103
[2] Karniadakis G, Beşkök A and Aluru N R 2005 Microflows and Nanoflows: Fundamentals and Simulation. Interdisciplinary Applied
Mathematics (New York, NY: Springer) vol xxi, p 817
[3] Maxwell J C 1878 On stresses in rarified gases arising from inequalities of temperature Proceedings of the Royal Society of London 27
304–8
[4] Van der Waals J D 1893 Thermodynamische theorie der capillariteit in de onderstelling van continue dichtheidsverandering (Amsterdam: J.
Müller)
[5] Korteweg D J 1901 Sur la forme que prennent les équations du mouvement des fluides si l’on tient compte des forces capillaires causées
par des variations de densité considérables mais continues et sur la théorie de la capillarité dans l’hypothese d’une variation continue de
la densité Archives Néerlandaises des Sciences exactes et naturelles 6 6
[6] Chapman S and Cowling T G 1970 The Mathematical Theory of Non-uniform Gases: An Account of the Kinetic Theory of Viscosity,
Thermal conDuction and Diffusion in Gases (Cambridge: Cambridge University Press)
[7] Grad H 1949 On the kinetic theory of rarefied gases Communications on Pure and Applied Mathematics 2 331–407
[8] Torrilhon M and Struchtrup H 2004 Regularized 13-moment equations: shock structure calculations and comparison to Burnett
models J. Fluid Mech. 513 171–98
[9] Bobylev A 1982 The Chapman-Enskog and Grad methods for solving the Boltzmann equation Akademiia Nauk SSSR Doklady 262
71–5
[10] Comeaux K A, Chapman D R and MacCormack R W 1995 An analysis of the Burnett equations based on the second law of
thermodynamics AIAA, Aerospace Sciences Meeting and Exhibit, 33 rd (Reno, NV)
[11] García-Colín L S, Velasco R M and Uribe F J 2008 Beyond the Navier–Stokes equations: burnett hydrodynamics Phys. Rep. 465 149–89
[12] Shu J-J, Teo J B M and Chan W K 2017 Fluid velocity slip and temperature jump at a solid surface Appl. Mech. Rev. 69 020801
[13] Arlemark E J, Dadzie S K and Reese J M 2010 An extension to the Navier–Stokes equations to incorporate gas molecular collisions with
boundaries Journal of Heat Transfer 132 041006
[14] Sone Y et al 1996 Inappropriateness of the heat conduction equation for description of a temperature field of a stationary gas in the
continuum limit: examination by asymptotic analysis and numerical computation of the Boltzmann equation Phys. Fluids (1994-
present) 8 628–38
[15] Brenner H 2012 Beyond Navier–Stokes Int. J. Eng. Sci. 54 67–98
[16] Dadzie S K 2013 A thermo-mechanically consistent Burnett regime continuum flow equation without Chapman–Enskog expansion
J. Fluid Mech. 716 R6
[17] Brenner H 2014 Conduction-only transport phenomena in compressible bivelocity fluids: diffuse interfaces and Korteweg stresses
Phys. Rev. E 89 043020
[18] Gorban A and Karlin I 2017 Beyond Navier–Stokes equations: capillarity of ideal gas Contemp. Phys. 58 70–90
[19] Dadzie S K and Christou C 2016 Bi-velocity gas dynamics of a micro lid-driven cavity heat transfer subject to forced convection Int.
Commun. Heat Mass Transfer 78 175–81
[20] Ewart T et al 2007 Mass flow rate measurements in a microchannel, from hydrodynamic to near free molecular regimes J. Fluid Mech.
584 337–56
[21] Dadzie S K and Brenner H 2012 Predicting enhanced mass flow rates in gas microchannels using nonkinetic models Phys. Rev. E 86
036318
[22] Lv Q et al 2013 Analytical solution to predicting gaseous mass flow rates of microchannels in a wide range of Knudsen numbers Phys.
Rev. E 88 013007
[23] Christou C et al 2016 Effects of volume diffusion in heat transfer in a cavity flow in non-continuum regime AIP Conf. Proc. (AIP
Publishing)
[24] Anderson D and McFadden G B 1997 A diffuse-interface description of internal waves in a near-critical fluid Phys. Fluids 9 1870–9
[25] Heida M and Málek J 2010 On compressible Korteweg fluid-like materials Int. J. Eng. Sci. 48 1313–24
[26] Anderson D M, McFadden G B and Wheeler A A 1998 Diffuse-interface methods in fluid mechanics Annual review of fluid mechanics 30
139–65
[27] Christou C and Dadzie S K 2017 An investigation of heat transfer in a cavity flow in the noncontinuum regime Journal of Heat Transfer
139 092002
[28] Bird G 1983 Definition of mean free path for real gases The Physics of Fluids 26 3222–3
[29] Greenshields C J et al 2010 Implementation of semi-discrete, non-staggered central schemes in a colocated, polyhedral, finite volume
framework, for high-speed viscous flows Int. J. Numer. Methods Fluids 63 1–21
[30] Maxwell J C 1878 On stresses in rarefied gases arising from inequalities of temperature Proc. of the Royal Society of London 27 304–8
[31] Knudsen M 1909 Die Gesetze der Molekularströmung und der inneren Reibungsströmung der Gase durch Röhren Ann. Phys., Lpz. 333
75–130

11