Knudsen heat capacity

Gulru Babac and Jason M. Reese

Citation: Physics of Fluids (1994-present) 26, 052002 (2014); doi: 10.1063/1.4872335

View online: http://dx.doi.org/10.1063/1.4872335
View Table of Contents: http://scitation.aip.org/content/aip/journal/pof2/26/5?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Transport processes and new types of boundary Knudsen layers in a gas flows through thin permeable
membranes
AIP Conf. Proc. 1501, 99 (2012); 10.1063/1.4769482

Efficiency derivation for the Knudsen pump with and without thermal losses
J. Vac. Sci. Technol. A 27, 496 (2009); 10.1116/1.3114444

Lattice Boltzmann method for gaseous microflows using kinetic theory boundary conditions
Phys. Fluids 17, 058101 (2005); 10.1063/1.1897010

Surface composition influence on internal gas flow at large Knudsen numbers

AIP Conf. Proc. 585, 911 (2001); 10.1063/1.1407656

Rarefied gas flows in microscales: Applications of RGD in MEMS

AIP Conf. Proc. 585, 463 (2001); 10.1063/1.1407596

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 PHYSICS OF FLUIDS 26, 052002 (2014)

Knudsen heat capacity

Gulru Babac1,a) and Jason M. Reese2
1
Institute of Energy, Istanbul Technical University, Istanbul 34469, Turkey
2
School of Engineering, University of Edinburgh, Edinburgh EH9 3JL, United Kingdom
(Received 15 October 2013; accepted 11 April 2014; published online 8 May 2014)

We present a “Knudsen heat capacity” as a more appropriate and useful fluid prop-
erty in micro/nanoscale gas systems than the constant pressure heat capacity. At these
scales, different fluid processes come to the fore that are not normally observed at the
macroscale. For thermodynamic analyses that include these Knudsen processes, using
the Knudsen heat capacity can be more effective and physical. We calculate this heat
capacity theoretically for non-ideal monatomic and diatomic gases, in particular, he-
lium, nitrogen, and hydrogen. The quantum modification for para and ortho hydrogen
is also considered. We numerically model the Knudsen heat capacity using molecular
dynamics simulations for the considered gases, and compare these results with the
theoretical ones. C 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4872335]

I. INTRODUCTION
The development of thermodynamic and transport models for micro/nanoscale fluid systems
has received a great deal of attention in recent years, e.g., Refs. 1–4. Several physical effects become
important in these systems that are generally ignored in macroscale systems, such as thermal creep,5, 6
quantum size effects,7, 8 and thermosize effects.9–12
Analysis of temperature-driven rarefied gas flows between two reservoirs at temperatures
TH (hot)
√ and TL √(cold), and at pressures pH and pL , leads to the temperature-pressure relation
p H / TH = p L / TL .5, 6 This relation is called the “Knudsen law” and it has been investigated
extensively in the literature for micro/nanoscale systems in connection with thermal creep and the
Knudsen pump or compressor. According to this relation,√for a temperature-driven gas flow at the
micro/nanoscale, the constant parameter in the flow is p/ T, therefore the process can be called a
√ √
p/ T constant process (a Knudsen process), and ∇  p/ T = 0. This is analogous to a constant
pressure process at the macroscale.
The Knudsen law can also affect thermodynamic properties. In some reports of thermodynamic
gas cycles,10, 11 incorporating a Knudsen process within a thermodynamic cycle can open up new
cycle design and introduce some differences from the conventional thermodynamic analysis, for
example, in the constant pressure heat capacities that are generally used in the continuum-fluid
limit.
In this paper, we present a detailed analysis of the heat capacity in a micro/nanoscale gas system.
We show that, instead of the conventional constant pressure heat capacity, a heat capacity that
incorporates the Knudsen process — which we call the “Knudsen heat capacity” — is appropriate.
We present molecular dynamics (MD) simulations of various nanoscale gas flows, and heat capacity
measurements from these simulations for monatomic and polyatomic gases. These measurements
are compared with the theoretical solutions. Our results show that the Knudsen heat capacity is a
more useful fluid property than the conventional heat capacity in gas flows that follow the Knudsen
law.

a) Author to whom correspondence should be addressed. Electronic mail: babac@itu.edu.tr

1070-6631/2014/26(5)/052002/12/$30.00 26, 052002-1 

C 2014 AIP Publishing LLC

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-2 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

II. THE KNUDSEN PROCESS

A temperature-driven gas flow in a macro channel is dominated by hydrodynamic flow behavior:
molecule-molecule collisions determine the gas properties under the applied temperature gradient.
 p = 0, which can be easily shown through
At steady state, this is a constant pressure process, i.e., ∇
the conventional Navier-Stokes equations. In a thermodynamic analysis, the ratio of the amount of
transferred heat energy to the imposed temperature difference along the channel requires the heat
capacity for a constant pressure process, Cp .
On the other hand, a gas flow driven by a temperature gradient in a micro or nano channel can
be dominated by free-molecular, or close to free-molecular, flow behavior in which molecule-wall
collisions mainly determine the gas properties. At steady state, the zero net flux condition results in
the following equation relating the pressure to the temperature in the system:13
1 ∂p 1 ∂T
=γ , (1)
p ∂x T ∂x
where x is the direction along the channel, and γ is a function of the rarefaction parameter and
depends on the geometry, molecule-surface interaction, type of gas etc.14 When Equation (1) is
integrated along the channel from the hot (TH ) to the cold (TL ) temperature ends, and the Knudsen
number variation is ignored, the constant process in this micro or nano domain can be written
 p/T γ ) = 0. The familiar Knudsen equation can be obtained from this equation in the free-
as ∇(  √ 
 p/ T = 0.
molecular limit by setting γ = 0.5 so that ∇

III. KNUDSEN HEAT CAPACITY FOR IDEAL GASES

The ratio of the amount of heat energy transferred, to the imposed temperature difference along
thechannel, is the heat capacity for the constant process in free molecular flows that are defined by
√ 
∇ p/ T = 0. We term this the “Knudsen heat capacity,” C p/√T .
A heat capacity can be calculated by taking the temperature derivative of the entropy, S, in a
constant process, i.e.,
 
dS
Ccons. proc. = T . (2)
dT cons. proc.
The entropy for ideal gases can be obtained from the Helmholtz free energy, F = −NkB T ln Z,
where kB is the Boltzmann constant, N is the number of molecules, and Z is the partition function.
Considering only translational modes, Z = Z trans = V N (2π mk B T )3N /2 /N !h 3N , where m is the
molecular mass, and h is the Planck constant. Therefore, the entropy per molecule, s, can be written
as
   
S dF Ak B T 5/2 5k B
s= =− = k B ln + , (3)
N dT V p 2
where A is a constant given by A = (2π mkB )3/2 /h3 . By substituting Eq. (3) into Eq. (2) and taking
derivatives under the conditions of constant pressure and of a Knudsen process, the heat capacities
per molecule can be obtained as follows:
Cp
cp = = 5k B /2, (4)
N

C p/√T
c p/√T = = 2k B , (5)
N
for monatomic ideal gases. In other words, when a temperature-driven gas flow in the free molecular
flow regime is considered, the heat capacity in the process should be assessed as c p/√T = 2k B ,
instead of cp = 5kB /2, because of the Knudsen law. In free molecular flow-dominated systems, such
as at the micro/nanoscale, using the Knudsen heat capacity should give more accurate results than
using the constant pressure heat capacity that is more appropriate for classical thermofluid systems.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-3 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

As discussed in Sec. II, the pressure drop in a rarefied gas flow depends on γ , through Eq. (1).
The heat capacity can be generalized accordingly, in order to encompass the hydrodynamic regime
to the free molecular flow regime, as
 
5
c p/T γ = − γ kB, (6)
2
where γ is 0 or 1/2 for the hydrodynamic or the free molecular flow regimes, respectively.
The parameter γ depends on many factors, but significantly for our purposes it depends on the
Knudsen number Kn.14 The relation between Kn and γ has been generally discussed in the literature
for two limiting cases: hydrodynamic flow (Kn → 0), and free molecular flow (Kn → ∞).15–17 In
the Kn → 0 limit, the heat capacity is already known to be cp = 5kB / 2. As Kn → ∞, γ converges to
0.5, with the deviation from this value decreasing with increasing Knudsen number. Therefore, for
free molecular flow γ can be set directly to be 0.5 and the Knudsen heat capacity can be obtained as
c p/√T = 2k B , as discussed above.
For the flow characteristic dependency of the heat capacity, we need a more general relation
between Kn and γ that can cover mid-range Knudsen numbers, however, the Kn-γ relations proposed
in the literature are not always successful for these systems.15–17 Because of the lack of a well-
developed analytical expression for the Kn-γ relation over all flow regimes from the hydrodynamic
to the free molecular, we have curve-fitted a relationship from data in the literature,14, 18, 19 as follows.
In Refs. 14, 18, and 19, rarefied gas flows at different Knudsen numbers in different geometries
were investigated using the S-model,14 the BGK model,18 and the hard-sphere gas model.19 The
γ relation can be obtained from logarithmic curve fitting to this published data in the range Kn =
0.1–10, i.e.,
γ = c1 ln (K n) + c2 , (7)
where c1 and c2 are coefficients determined by the curve fitting. The corresponding coefficients and
R2 values for each reported case are presented in Table I. Note that Eq. (7) does not converge in the
limit Kn → ∞, so for this limit case it is better to use other relations in the literature.15–17 However,
here we choose Knudsen numbers in the range Kn = 0.1–10 and continue with Eq. (7).
With Eq. (7), the heat capacity can be written as a function of the Knudsen number Kn, i.e.,
 
5
c p/T =
γ − (c1 ln (K n) + c2 ) k B , (8)
2
which is plotted in Figure 1 for the various coefficients in Table I. The heat capacity is 5/2 in
the hydrodynamic limit, and with increasing Kn the heat capacity decreases. In free molecular
flow conditions, c p/T γ /k B reaches the values at Kn = 10 of 1.99051 for Sharipov,14 2.01725 for
Loyalka,18 and 2.04608 for Ohwada et al.19 These are reasonably close to the theoretical value of
c p/T γ /k B = 2 that we calculated above.
In Secs. IV A and IV B, this heat capacity will be investigated in detail for both monatomic
and diatomic real gases. In Sec. V the Kn-dependency of the heat capacity is also analyzed using
molecular dynamics simulations of gas flow in a nanoscale channel. Molecular dynamics simulates
atomistic or molecular interactions and movements directly through Newton’s laws. For the inter-
molecular interactions, the Lennard-Jones (LJ) potential is often used. Some phenomena that are
usually ignored in the theoretical calculations, such as the effects of inlets/outlets at the ends of the
domain, can be taken into account in MD simulations.

TABLE I. Fitting coefficients c1 , c2 , and corresponding R2 values, for the Kn-γ relation.

c1 c2 R2

Sharipov14 0.0925 0.2965 0.9782

Loyalka18 0.1026 0.2465 0.9853
Ohwada et al.19 0.0850 0.2582 0.9744

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-4 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

FIG. 1. Various heat capacity predictions varying with Knudsen number, Kn.

In our MD simulation reported in Sec. V below, the temperature gradients are kept in the mid-
to low-temperature range so that the vibrational modes of the molecules are not excited. Therefore,
the vibrational contribution to the heat capacities can be ignored in this analysis. The rotational and
translational contributions are taken into account for diatomic cases, while only the translational
contribution is considered for the monatomic case.

IV. KNUDSEN HEAT CAPACITY FOR REAL GASES

A. Monatomic gases
The Knudsen heat capacity for monatomic real gases is calculated by only considering the
translational motion of the molecules. The partition function for translational motion can be written
as
   
1 2π mk B T 3N /2 N N2
Z = Z trans = V 1 − b(T ) , (9)
N! h2 V
where b(T) is the second virial coefficient that is calculated depending on the intermolecular potential
U(r) through
∞   
−U (r )
b(T ) = 2π 1 − exp r 2 dr. (10)
kB T
0

A LJ potential can be used, viz.,

 12  6 
σ σ
U L J = 4ε − , (11)
Ri j Ri j

where ε is the depth of the intermolecular potential well, σ is the finite distance at which the
intermolecular potential is zero, and Rij is the distance between molecules i and j. The LJ parameters
for the gases we consider are discussed below.
The entropy and Knudsen heat capacity per molecule for real gases are then calculated as

5k B Ak B T 5/2
s= + k B ln − pb (T ) , (12)
2 p

 
ds p 
c p/√T = T √
= 2k B − b (T ) − pT b (T ) , (13)
dT p/ T 2

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-5 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

TABLE II. Constant pressure and Knudsen heat capacities calculated for some monatomic gases.

Calculated constant pressure

LJ parameters20 Molecular mass heat capacity Calculated Knudsen heat capacity
Gas ε/kB (K) σ (Å) m/10−27 (kg) cp /m (J/kg K) (cp/√T )/m (J/kg K)

He 10.2 2.576 6.646 5190 4152

Ne 35.7 2.789 33.509 1029 823
Ar 124 3.418 66.336 521 417
Kr 225 3.498 139.153 249 199
Xe 229 4.055 218.012 159 127

where the pressure defined for real gases is p = nkB T/(1 − nb(T)). Here, b (T) and b (T) are the
first and second derivatives of b(T) with respect to T, respectively. If we were to consider a macro
channel gas flow under the same conditions, we should use the constant pressure heat capacity for
real gases, which is defined per molecule as
 
ds 5
cp = T = k B − pT b (T ) . (14)
dT p 2

Equations (13) and (14) can be compared for monatomic real gases. We perform this comparison
for Ne, Ar, Kr, Xe, and He gases at standard conditions (i.e., at 300 K and 1 atm), and this is presented
in Table II. (Values are presented per unit mass, by dividing Eqs. (13) and (14) by the molecular
mass.)
As can be seen from Table II, the constant pressure heat capacity in the hydrodynamic regime is
some 20% higher than the Knudsen heat capacity in the free molecular regime. So using the constant
pressure heat capacity instead of the Knudsen heat capacity when considering micro/nanoscale free
molecular (or close to free molecular) flows may introduce a significant error into the results and
thermodynamic analyses.
In Table II, cp and cp/√T represent the two limiting transport cases. The heat capacity for real
gases can be generalized using γ in order to cover slip and translational flows as well, i.e.,
 
5
c p/T γ = − γ k B − pγ b (T ) − pT b (T ) . (15)
2
If Eq. (7) is substituted into Eq. (15), with coefficients from Table I, the best match with the calculated
Knudsen heat capacities in Table II is obtained for He and Ne gases. Using the coefficients of
Sharipov’s data14 from Table I, the values of heat capacity are within about 5% for both the low and
high Knudsen number cases; for the coefficients from Loyalka,18 heat capacities are within 2% for
low Knudsen numbers and about 4.5% for high Knudsen number cases; and for the coefficients from
Ohwada et al.,19 within 2.5% for low Knudsen numbers, and within 7.5% for high Knudsen numbers.
The best match seems to be obtained using the coefficients from Loyalka’s18 data, especially for low
Kn. For high Kn, both Sharipov’s14 and Loyalka’s18 data capture the heat capacities within about
5%. The differences between the results slightly increase for Ar, Kr, and Xe because of deviation
from the ideal gas approximation for larger atoms.

B. Diatomic gases
For diatomic gases, the rotational contribution to the partition function should also be considered,
i.e.,21

Z r ot = (2l + 1) exp [−l (l + 1) /T ], (16)
l=0,1,...

where  is the characteristic temperature for rotation and l is the rotational energy level. If T  
the summation can be replaced with an integral, and the rotational partition function is Zrot = T/.21
The rotational contribution to the conventional constant pressure heat capacity makes cp = cp, trans

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-6 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

+ cp, rot = 5kB /2 + kB = 7kB /2 for an ideal gas. The characteristic temperature  is generally low
for many gases, except for H2 . Therefore, the T   assumption can often be used21 and under this
assumption the entropy and Knudsen heat capacity for real diatomic gases take the following forms:

7k B Ak B T 7/2
s= + k B ln − pb (T ) , (17)
2 p
 
ds p 
c p/√T =T √
= 3k B − b (T ) − pT b (T ) . (18)
dT p/ T 2
The constant pressure heat capacity for diatomic real gases is
 
ds 7
cp = T = k B − pT b (T ) . (19)
dT p 2

These can be generalized using γ , as before, to give

 
7
c p/T γ = − γ k B − pγ b (T ) − pT b (T ) . (20)
2
The heat capacities for common gases such as N2 and O2 can then be calculated for free molecular
flows and hydrodynamic
  flows by using Eqs.  (18) and
 (19), respectively. For 300 K and 1 atm,

this results in c p/√T = 890 J/kg K and c p/√T = 779 J/kg K; c p N2 = 1039 J/kg K and
 N2 O2
c p O2 = 909 J/kg K, where the required Lennard-Jones parameters are σ N2 = 3.3 Å and (ε/k B ) N2 =
36 K; σ O2 = 3.1062 Å and (ε/k B ) O2 = 43.183 K.22
For H2 gas, T   cannot be assumed, therefore, Eqs. (17)–(20) cannot be used. The rotational
partition function in Eq. (16) should be rewritten by considering ortho and para hydrogen (which
correspond to symmetric and antisymmetric wave functions for even and odd values of l), and this
is given in Ref. 21 as
 
Z r ot−H2 = (2l + 1) exp [−l (l + 1) /T ] + 3 (2l + 1) exp [−l (l + 1) /T ] . (21)
l=even. l=odd

The rotational contribution to heat capacity can then be incorporated into the Knudsen and
constant pressure heat capacities, respectively, and leads to
p
c p/√T = 2k B − b (T ) − pT b (T ) + 2k B T f  (T ) + T 2 k B f  (T ) , (22)
2
5
cp = k B − pT b (T ) + 2k B T f  (T ) + T 2 k B f  (T ) , (23)
2
with the general form of the hydrogen heat capacity
 
5
c p/T γ = − γ k B − pγ b (T ) − pT b (T ) + 2k B T f  (T ) + T 2 k B f  (T ) , (24)
2
where f is a function defined for ortho and para hydrogen,
 

f para (T ) = ln (2l + 1) exp [−l (l + 1) /T ] , (25)
l,even

 

f or tho (T ) = ln 3 (2l + 1) exp [−l (l + 1) /T ] , (26)
l,odd

and f  and f  are the first and second derivatives of f with respect to T, respectively. The heat
capacities for para and ortho hydrogen can then be calculated at p = 1 atm, by taking the sums in
Eqs. (25) and (26), respectively. The results are presented in Figure 2.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-7 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

FIG. 2. Constant pressure (gray line) and Knudsen (black line) heat capacities at 1 atm, for (a) para hydrogen, and (b) ortho
hydrogen.

Hydrogen gas is also simulated with MD in Sec. V; however, the quantum character of rotational
motions is not included in these simulations, so only translational effects can be examined. The
quantum corrections for para and ortho hydrogen calculations are subsequently added into our MD
results using a separate numerical routine in order to obtain the full predictions of hydrogen heat
capacities.

V. MOLECULAR DYNAMICS SIMULATIONS

We use MD simulations of temperature-driven gas flow in a nano channel in order to investigate
the Knudsen heat capacity. The MD simulations are performed using the open source OpenFOAM
software, which incorporates a parallelized non-equilibrium MD solver.23–25 A schematic of the
system we consider is shown in Figure 3.
A 2D nano channel is thermally in contact with hot (temperature of TH ) and cold (TL ) reservoirs.
Berendsen thermostats, with a coupling parameter of 1 ps, are applied to keep the temperatures
constant in the reservoirs.26 As solid surface boundary conditions (BCs), diffusive wall, and diffusive
wall with a linear temperature gradient, are applied along the y and x directions, respectively. The
solid boundary accommodation coefficient is set to be 1. In the z direction, periodic BCs are used to
ensure the effective two-dimensionality of the setup.
Simulations are run for monatomic and diatomic gases, and results are compared with the
theoretical calculations. Helium and nitrogen gases are considered as exemplars of monatomic and
diatomic gases. Hydrogen gas is also considered in a MD simulation, but, as outlined in Sec. IV,
the heat capacity of H2 depends on quantum effects and the ortho/para state, and these are not
generally included in MD. So the contribution from the translational part is simulated using MD.
Then the partition function, Eq. (21), and the para-ortho functions in Eqs. (25) and (26) are defined
in Mathematica: summations are calculated and the rotational contribution to heat capacities is
evaluated. Finally, the MD and Mathematica results are added together according to Eqs. (22)–(24).
We carry out simulations for Kn = 0.2, 0.5, 0.75, 1, 2, 5, 7.5, and 10. For Kn = 0.2, the flow
is in the slip regime,2 not free molecular; therefore, the heat capacity values should be in between
the constant pressure heat capacity and the Knudsen heat capacity. For Kn = 10, the flow is nearly

FIG. 3. Schematic of the nano channel for MD simulations.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-8 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

TABLE III. Molecular parameters for the gases considered in our MD simulations. Values taken from Refs. 27–30.

Positions Lennard-Jones parameters Charge

X (Å) Y (Å) Z (Å) ε/kb (K) σ (Å) q (e)

Helium (He)
He 0 0 0 10.2 2.576 0
Nitrogen (N2 )
N 0.549 0 0 36 3.3 − 0.5075
N − 0.549 0 0 36 3.3 − 0.5075
M 0 0 0 ... ... 1.015
Hydrogen (H2 )
H 0.37 0 0 17.4 2.65 − 0.489
H − 0.37 0 0 17.4 2.65 − 0.489
M 0 0 0 ... ... 0.978

in the free molecular regime, therefore the measured heat capacity should be close to the Knudsen
heat capacity value. Temperature ranges for the simulations are determined according to the critical
temperature of the gases, in order to maintain in gas phases throughout the simulations: 10–300 K
for He, 30–300 K for H2 , and 110–300 K for N2.
In our MD simulations, the molecules interact with each other via LJ and Coulomb potentials,
for which the parameters are given in Table III.27, 30 The N2 and H2 molecules are represented with
3 sites as NNM and HHM, where M is a massless charged site to reproduce the polar nature of the
molecule. The MD time step is around 2 fs, and the number of atoms in the system is kept at around
40 000–45 000 for all three gases. Each simulation runs in parallel on 16 cores for nearly a month.

VI. RESULTS AND DISCUSSION

The heat capacity is measured directly from the energy fluctuations in each MD simulation.31
The measurements are made in the middle of the channel to eliminate inlet/outlet effects. The
variations of the heat capacity of He and N2 gases with Knudsen number are presented in Figure 4,
which were obtained by running MD simulations of the system for different Kn numbers; Kn = 0.2,
0.5, 0.75, 1, 2, 5, 7.5, and 10. The profile of the results in Figure 4 is similar to that of the theoretical
results in Fig. 1: the heat capacity decreases logarithmically with increasing Knudsen number.
The heat capacities for He and N2 are also calculated with the model equations derived in
this paper, and these are presented in Figs. 5 and 6, respectively. In each of these figures, the sub-
figures are the heat capacity measurements for the same gases under the same temperature gradient,
but for the different Knudsen numbers Kn = 0.5, 1, 5, and 10. The pressure and temperature
measurements from the MD simulations are used in the model equations; the γ values are calculated

FIG. 4. The variation of the heat capacity of He and N2 with Knudsen number, from MD simulations.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-9 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

FIG. 5. Heat capacity of Helium for different Knudsen numbers, varying with temperature.

through γ = ln (p2 /p1 )/ln (T2 /T1 ) in our MD cases. Our MD results are compared with those in the
literature.14, 18, 19 Table IV presents the various values of γ reported in the literature and obtained
from our MD simulations. By using these in Eq. (15) for the monatomic case and in Eq. (20) for the
diatomic case, the plots of heat capacity versus T are obtained in Figs. 5 and 6.

FIG. 6. Heat capacity of Nitrogen for different Knudsen numbers, varying with temperature.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-10 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

TABLE IV. Reported γ values for various Kn.

Sharipov14 Loyalka18 Ohwada et al.19 Our MD results for He Our MD results for N2

Kn = 0.5 0.2454 0.17 0.2053 0.2092 0.1635

Kn = 1 0.3179 0.2592 0.2825 0.3047 0.2538
Kn = 5 0.4273 0.4166 0.3912 0.4252 0.3974
Kn = 10 0.4532 0.4486 0.4175 0.4562 0.4520

In Figure 5, monatomic helium gas results for heat capacity are presented for Kn = 0.5, 1, 5, and
10. Because of the different transport processes, the heat capacity is different for different Kn. For
Kn = 10, Kn = 5, and Kn = 1, the best agreement of our results is obtained with Sharipov’s14 data.
The data of Loyalka18 and Ohwada et al.19 are consistently slightly above the MD results. However,
for Kn = 0.5, the best agreement is obtained with the data of Ohwada et al.19
As a general observation regarding Figures 4 and 5, the heat capacity for He tends to the constant
pressure heat capacity value at low Knudsen number, and tends towards the Knudsen heat capacity
value with increasing Knudsen number, as can be checked from Table II. If the system could be
simulated for lower and higher Knudsen numbers to get closer to hydrodynamic and free molecular
conditions, respectively, the heat capacity values would also be closer to the values for the respective
constant heat capacity and Knudsen heat capacity given in Table II as (cp )/m = 5190 J/kg K and
(cp/ √T )/m = 4152 J/kg K, respectively.
A similar trend can also be seen for nitrogen gas in Figure 6, where comparison is made for
different γ values within Eq. (20), the expression required for diatomic gases. As can be seen in
this figure, the heat capacity converges to its constant pressure value for the low Kn cases, and to
the Knudsen heat capacity as Kn increases. The best agreement between MD results and theory is
obtained with Loyalka’s18 data, except for Kn = 10, where the best agreement is with Sharipov’s14
data (although, for this Kn, the MD and Sharipov results are in any case close to Loyalka’s18 data).
Figures 5 and 6 show that differences in the fluid transport processes introduce differences in the
heat capacities. In general, the results for Kn = 0.5 are close to the constant pressure heat capacity,
cp , while the results for Kn = 10 deviate from cp by around kB /2, and the heat capacity is then
better represented by the Knudsen heat capacity, c p/√T . Further analysis could be done for different
gases, and also more Knudsen numbers for He and N2 , in order to get a detailed database of the heat
capacities.
For hydrogen gas, MD simulations are run for two different temperature ranges: 10–100 K, and
300–500 K. Hydrogen gas in the 10–100 K temperature range can be considered as para hydrogen,
and in 300–500 K it is considered as an ortho:para hydrogen mixture with the ratio 3:1. Therefore,
the simulation results in the low temperature range are analyzed using Eqs. (22)–(25), while they
are analyzed in the high temperature range using Eqs. (22)–(24) and (26). The results are presented
in Figure 7.
The theoretical results presented in Figure 7 for constant pressure heat capacity cp , are obtained
from Eqs. (23), (25), and (26) at 10 kPa pressure. Our MD simulations are run in rarefied conditions;
Kn ≈ 2.5 for the para hydrogen case (10–100 K), and Kn ≈ 10 for the ortho:para (3:1) hydrogen
case (300–500 K). A pressure of 10 kPa in the centre of the simulated channel is obtained in both
cases, making the results comparable. The comparison with the theoretical calculations enables us to
check the difference between the Knudsen heat capacity and the constant pressure heat capacity. As
can be seen in Figure 7, the maximum difference between cp and c p/T γ is obtained for the ortho:para
hydrogen case (3:1 ratio), and is around 0.446 k B /m H2 . The difference is somewhat smaller than
this for the para hydrogen case, especially close to 100 K. Because the ortho hydrogen case is run at
Kn ≈ 10, then γ = 0.446, while γ = 0.27 for the para hydrogen case at a lower Kn ≈ 2.5, to keep
the cases in the same pressure range.
It can also be seen from Eqs. (23) and (24) that the deviation from the constant pressure heat
capacity should be around (γ kB − pγ b (T))/m for real gases, which can be written as γ kB /m for ideal

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-11 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

FIG. 7. Heat capacity of Hydrogen gas. The black lines are for the theoretical representation of para and ortho hydrogen in
the continuum limit. The red lines represent our MD measurements in rarefied flow conditions.

gases. As γ decreases, the difference between the constant pressure and the Knudsen heat capacities
becomes smaller, as we see in our results.

VII. CONCLUSIONS
We have introduced the Knudsen heat capacity for monatomic and diatomic gas systems in
free molecular flow conditions, such as at the micro/nanoscale. A theoretical derivation of the
Knudsen heat capacity has been presented, and validated using Molecular Dynamics simulations
of nanoscale gas flows under a temperature gradient. The relationship between the Knudsen heat
capacity and the well-known constant pressure heat capacity in the continuum-fluid limit can be
written as c p − c p/√T = k B /2 for ideal gases, where kB is Boltzmann’s constant. For real gases,
this relationship changes depending on the intermolecular interaction potential. The Knudsen heat
capacity should in general be used instead of the conventional constant pressure heat capacity to
provide more realistic thermodynamic analysis when considering micro/nanoscale gas flows that
follow the Knudsen law.

ACKNOWLEDGMENTS
Gulru Babac would like to thank the Scientific and Technological Research Council of Turkey
(TUBITAK) – 2214 Program for part-funding this research. The authors thank the reviewers of this
paper for their helpful comments.
1 W. A. Goddard, D. W. Brenner, S. E. Lyshevski, and G. J. Iafrate, Handbook of Nanoscience, Engineering and Technology
(CRC Press, Boca Raton, FL, 2003).
2 G. E. Karniadikis, N. Aluru, and A. Beskok, Micro and Nano Flows: Fundamentals and Simulation (Springer, New York,

2005).
3 G. Shen, Rarefied Gas Dynamics: Fundamentals Simulations and MicroFlows (Springer-Verlag, Berlin, 2005).
4 E. L. Wolf, Nanophysics and Nanotechnology (Wiley-VCH, Weinheim, 2004).
5 M. Knudsen, “Eine revision der gleichgewichsbedingung der gase. Thermische Molekularstromung,” Ann. Phys. 336(1),

205 (1909).
6 M. Knudsen, “Thermischer molekulardruck der gase in rohren,” Ann. Phys. 338, 1435 (1910).
7 A. Sisman and I. Muller, “The Casimir-like size effects in ideal gases,” Phys. Lett. A 320, 360 (2004).
8 A. Sisman, “Surface dependency in thermodynamics of ideal gases,” J. Phys. A 37, 11353 (2004).
9 W. Nie and J. He, “Performance analysis of a thermosize micro/nano heat engine,” Phys. Lett. A 372, 1168 (2008).
10 G. Babac and A. Sisman, “Thermodynamic cycles based on classical thermosize effects,” J. Comput. Theor. Nanosci. 8,

1720 (2011).
11 G. Babac and A. Sisman, “Classical thermosize effects in degenerate quantum gases,” J. Comput. Theor. Nanosci. 8, 2331

(2011).
12 A. Sisman and G. Babac, “Quantum size effects on classical thermosize effects,” Continuum Mech. Therm. 24, 339 (2012).
13 F. Sharipov and V. Seleznev, “Data on internal rarefied gas flow,” J. Phys. Chem. Ref. Data 27(3), 657 (1998).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06
 052002-12 G. Babac and J. M. Reese Phys. Fluids 26, 052002 (2014)

14 F. Sharipov, “Rarefied gas flow through a long tube at any temperature ratio,” J. Vac. Sci. Technol. A 14(4), 2627 (1996).
15 C. Cercignani and A. Daneri, “Flow of a rarefied gas between two parallel plates,” J. Appl. Phys. 34, 3509 (1963).
16 H. Niimi, “Thermal creep flow of rarefied gas between two parallel plates,” J. Phys. Soc. Jpn. 30, 572 (1971).
17 K. Takao, “Rarefied gas flow between two parallel plates,” in Proceedings of the Second International Symposium on

Rarefied Gas Dynamics held at the University of California, Berkeley, CA, 1960, edited by L. Talbot (Academic Press,
New York, 1961), p. 465.
18 S. K. Loyalka, “Kinetic theory of thermal transpiration and mechanocaloric effect. II,” J. Chem. Phys. 63(9), 4054 (1975).
19 T. Ohwada, Y. Sone, and K. Aoki, “Numerical analysis of the Poiseuille and thermal transpiration flows between two

parallel plates on the basis of the Boltzman equation for hard-sphere molecules,” Phys. Fluids A 1(12), 2042 (1989).
20 R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena (John Wiley & Sons, West Sussex, 2006).
21 A. M. Guenault, Statistical Physics (Routledge, New York, 1988).
22 Y. Boutard, Ph. Ungerer, J. M. Teuler, M. G. Ahunbay, S. F. Sabater, J. Perez-Pellitero, A. D. Mackie, and E. Bourasseau,

“Extension of the anisotropic united atoms intermolecular potential to amines, amides and alkanols: Application to the
problems of the 2004 Fluid Simulation Challenge,” Fluid Phase Equilib. 236, 25 (2005).
23 G. B. Macpherson and J. M. Reese, “Molecular dynamics in arbitrary geometries: Parallel evaluation of pair forces,” Mol.

Simul. 34(1), 97 (2008).

24 G. B. Macpherson, M. K. Borg, and J. M. Reese, “Generation of initial molecular dynamics configurations,” Mol. Simul.

33, 1199 (2007).

25 M. K. Borg, G. B. Macpherson, and J. M. Reese, “Controllers for imposing continuum-to-molecular boundary conditions

in arbitrary fluid flow geometries,” Mol. Simul. 36(10), 745 (2010).

26 P. H. Hünenberger, “Thermostat algorithms for molecular dynamics simulations,” Adv. Polym. Sci. 173, 105 (2005).
27 J. Vrabec, J. Stoll, and H. Hasse, “A set of molecular models for symmetric quadrupolar fluids,” J. Phys. Chem. B 105,

12126 (2001).
28 H. Cheng, A. C. Cooper, G. P. Pez, M. K. Kostov, P. Piotrowski, and S. J. Stuart, “Molecular dynamics simulations on the

effects of diameter and chirality on hydrogen adsorption in single walled carbon nanotubes,” J. Phys. Chem. B 109, 3780
(2005).
29 P. Diep and J. K. Johanson, “An accurate H –H interaction potential from first principles,” J. Chem. Phys. 112(10), 4465
2 2
(2000).
30 T. Schnabel, Molecular Modeling and Simulation of Hydrogen Bonding Pure Fluids and Mixtures (Logos Verlag, Berlin,

2008).
31 D. C. Rapaport, The Art of Molecular Dynamics Simulation (Cambridge University Press, Cambridge, 1997).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.235.251.160 On: Thu, 18 Dec 2014 17:55:06