Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

i An update to this article is included at the end

International Journal of Heat and Mass Transfer 147 (2020) 118949

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


journal homepage: www.elsevier.com/locate/hmt

A molecular dynamics study of thermal boundary resistance over solid


interfaces with an extremely thin liquid film
Xiao Liua,b,1, Donatas Surblysb,1,∗, Yoshiaki Kawagoeb, Abdul Rafeq Bin Salemanc,
Hiroki Matsubarab, Gota Kikugawab, Taku Oharab
a
Department of Mechanical and Aerospace Engineering, Graduate School of Engineering, Tohoku University, 6-6, Aramaki Aoba, Sendai 980-0845, Japan
b
Institute of Fluid Science, Tohoku University, 2-1-1 Katahira, Aoba, Sendai 980-8577, Japan
c
Faculty of Mechanical Engineering, Universiti Teknikal Malaysia Melaka, Hang Tuah Jaya, 76100 Durian Tunggal, Melaka, Malaysia

a r t i c l e i n f o a b s t r a c t

Article history: We investigated the characteristics of thermal energy transport over two solid surfaces joined via an
Received 5 February 2019 extremely thin liquid film where the liquid molecules are under the influence of both solid surfaces si-
Revised 17 September 2019
multaneously. Using non-equilibrium molecular dynamics simulations, the thermal resistance between
Accepted 22 October 2019
the two solid surfaces was examined for different thickness of liquid film and different alignment (in-
Available online 2 November 2019
plane orientation) of the two solid surfaces. Both solid surfaces were the (1 1 0) plane of face centered
Keywords: cubic lattice, and two different combinations of alignment, i.e., either parallel or crossed to each other,
Molecular dynamics were examined. The thermal resistance between the solid surfaces was decomposed into the thermal
Solid-liquid interface boundary resistance at the solid-liquid interfaces and the thermal resistance of the liquid film, which
Thermal resistance were analyzed separately. The results showed that when the liquid film thickness is equal or less to four
Thermal interface material molecular dimensions, both the film thickness and the surface alignment have significant influence on the
Thin liquid film
thermal resistance. Specifically, when the liquid film is a single layer of liquid molecules (LLM), the ther-
mal resistance between solid surfaces is extremely low when compared with the cases of more LLM, and
increases with increasing liquid density. In contrast, when the film is composed of two or three LLM, the
solid-liquid interfacial thermal resistance decreases with increasing liquid density, and a discontinuous
increase occurs as the number of LLM changes from two to three. As for the effect of surface alignment,
it was found that the parallel surface alignment gives a lower thermal resistance than the crossed surface
alignment.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction Because it is usual to use liquid-like materials such as grease for


TIM, the thermal resistance between the TIM and the solid walls
Enhancement of heat transfer between solid surfaces, especially can be modeled by a solid-liquid system. Several studies have been
in the case where thermal interface material (TIM) is applied be- carried out by the authors using molecular dynamics simulations
tween the solid surfaces, such as in power modules, has become an to investigate the thermal transport over the solid-liquid interface
important task in thermal engineering. When two solid materials [1–3]. Non-equilibrium molecular dynamics simulations have been
are jointed with some type of TIM in between, the thermal resis- performed in which the system consisted of hot and cold solid
tance at the junction is determined by the thermal conductivity of walls with a liquid film placed between them, and the heat trans-
the TIM and thermal boundary resistance (TBR) of the solid-TIM fer characteristics at the solid-liquid interface for various combina-
interfaces. When the solid surface is rough, the TIM is extremely tion of crystal planes and liquids were investigated. In these stud-
thin in some areas of the solid surface around the real contact area. ies, the size of the liquid film between the solid walls was set suf-
In these areas, only a few layers of TIM molecules are placed be- ficiently large so that the heat transfer at the solid-liquid inter-
tween the solid surfaces. The heat transfer characteristics in such face could be considered as a phenomenon between bulk liquid
cases are examined in the present paper. and a single solid wall. However, when the distance between the
solid walls is very small, both the solid and liquid molecules at
the solid-liquid interfaces are affected simultaneously by the solid

Corresponding author. walls on both sides. Therefore, the interfacial heat transfer charac-
E-mail address: donatas@microheat.ifs.tohoku.ac.jp (D. Surblys). teristics are considered to be different from those in the systems
1
These authors have equally contributed.

https://doi.org/10.1016/j.ijheatmasstransfer.2019.118949
0017-9310/© 2019 Elsevier Ltd. All rights reserved.
2 X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949

with large distances between solid walls. Recently, several stud-


ies for the heat transfer in such cases were reported by Tsai’s and
Quan’s groups [4–6]. These studies have analyzed the change of
interfacial thermal resistance under a different number of layers
of liquid molecules (LLM) by varying the distance between solid
walls. For example, when the liquid film between the solid walls
consists of only a single LLM, the interfacial thermal resistance be-
tween the solid surface and the LLM is much smaller than in other
cases [4,6]. Instead of fixing distance between the solid surfaces,
we chose to control the number of liquid molecules under a con-
stant pressure condition, to investigate the general change in ten-
dencies of the thermal resistance between the solid surfaces. In
this work, we aim to investigate heat transfer characteristics for a
range of distances between the solid surfaces and the liquid layer
conditions. Additionally, in contrast with past studies where a pair
of solid surfaces with the same lattice alignment (in-plane surface
orientation) were applied, a case in which the two solid surfaces
have different surface alignment is also examined in the present
study. In such case, the liquid molecules between the two solid
walls are affected by both of the solid walls in an asymmetric way.
The total TBR between the solid walls is mainly determined by
the TBR at the solid-liquid interfaces, because in the cases investi- Fig. 1. Simulation system with heat flux over liquid film between solid surfaces.
gated in this work the liquid layer is extremely thin and the ther- Heat baths are illustrated by red and blue semi-transparent boxes. (For interpreta-
mal resistance of the liquid is miniscule when compared to the tion of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
TBR at the solid-liquid interfaces. TBR is a measurement of heat
flow resistance which causes temperature discontinuity at the in-
terface when different surfaces are in contact with each other and
large vacuum. The velocity-Verlet method was applied for the time
there have been many studies on TBR at solid-liquid interfaces in
integration with the time step of 1 fs.
the past [7–10]. In this study, thermal resistance over the junc-
The simulations in this work were done in the following
tions of two solid walls with a liquid in between is decomposed
phases: (1) creation of simulation system, (2) system equilibration
to the two TBRs at the solid-liquid interfaces and the thermal re-
to a uniform temperature under constant pressure condition, (3)
sistance of the liquid, and each of these components is analyzed
relaxation of system to a non-equilibrium steady state with sta-
separately. The influence of the thickness of the liquid film and the
tionary heat flow under constant pressure condition, (4) relaxation
alignment of the solid walls on the heat transfer characteristics is
of system to a non-equilibrium steady state of stationary heat flow
examined by utilizing non-equilibrium molecular dynamics simu-
under constant volume condition chosen so that system pressure
lations to analyze the TBR of the system, which consisted of high
would be the same as in phase 3, (5) a production run used for
and low temperature solid walls and a liquid film in between.
analysis, (6) additional run to do vibration of state analysis for sev-
eral selected systems.
2. Molecular dynamics simulation

One of the simulation systems used in the present study is 2.1. Initial system construction
shown in Fig. 1. The system consists of a liquid film (of argon)
between two parallel solid walls (of platinum). The interaction be- This subsection describes phase 1 of the simulation. The two
tween solid atoms was treated as Morse potential given by solid walls were a FCC crystal with nearest neighbor distance set
    to 0.277 nm and a (1 1 0) surface that is normal to the z direc-
φ ri j = D e−2α (ri j −r0 ) − 2e−α (ri j −r0 ) , (1) tion and in contact with the liquid. For systems where the liquid
molecules formed two or more layers, each wall consisted of 8 lay-
where D = 6.617 × 10−20 J, α = 1.85 × 1010 m, r0 = 2.774 × 10−10 m ers in the z direction, as no size effect was observed even when in-
[11], and rij is the distance between ith and jth atoms. The inter- creasing solid wall layer number to 78. This differs from previous
action between liquid molecules was treated as Lennard Jones (LJ) reports, such as by Yenigun and Barisik [14], where a much big-
potential given by ger size dependence was demonstrated, and we suspect that the
 12  6 size-insensitivity in our case might be due to a choice of Langevin
  σ σ thermostat in simulation phases 4–6. On the other hand, when
φLJ ri j = 4ε − , (2)
ri j ri j the liquid molecules formed a single layer, a substantial size effect
in the interfacial heat conductance was observed when increasing
where the energy parameter is ε = 1.6533 × 10−21 J and the dis- wall layer number from 8 to 78 layers, while no such effect was
tance parameter is σ = 3.4236 × 10−10 m [12]. The interaction be- present when changing from 78 to 156. Therefore, the number of
tween solid and liquid molecules was also modeled by LJ poten- layers in a single wall for systems with a single liquid layer were
tial, where the energy parameter is ε = 1.0927 × 10−21 J and the set to 78. Several of the systems that were used to investigate this
distance parameter is σ = 2.94 × 10−10 m [13]. A cut-off distance size effect are displayed in Fig. S1 of the supplementary material.
of 12 Å was used for all the intermolecular interactions. The peri- In the present study, two different combinations of the atomic
odic boundary conditions were applied in the x and y directions. alignment for the left and right solid surfaces were examined as is
Strictly speaking, periodic boundary condition was also applied in shown in Table 1, where the numbers of solid and liquid molecules
the z direction, but as the empty area outside of the walls was big are also listed. The first case was the parallel surface alignment
enough so that there were no interactions crossing the z periodic (PSA) in which the lattice-scale grooves on the (1 1 0) crystal plane
boundary, the system was essentially surrounded by an infinitely were parallel to those on the other solid wall. The second case was
X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949 3

Table 1
Surface configuration of the pairs of the solid walls and the number of molecules used in the simulations. The xy cross sectional area of the present systems is
3.878 × 3.878 nm2 .

Wall on the left side Wall on the right side Liquid between walls # of solid atoms in a wall # of liquid molecules

Case 1: 1 layer 14 × 10 × 78 90,102,114


Parallel surface alignment (PSA) 2 layers 14 × 10 × 8 140,150,160,192,210
3 layers 225,250,282
4 layers 372
FCC(1 1 0) FCC(1 1 0) Bulk 1520
Case 2: 1 layer 14 × 10 × 78 90,102,114
Crossed surface alignment (CSA) 2 layers 14 × 10 × 8 140,150,160,192,210
3 layers 225,250,282
4 layers 372

FCC(1 1 0) FCC(1 1 0) Bulk 1520

the crossed surface alignment (CSA) in which the right side wall pressed liquid phase for all cases and was applied as described in
was rotated by 90° along the z axis from the state of PSA, i.e., the Section 2.2.1. This resulted in a uniform normal pressure in regards
surface alignments of the left and right walls were cross-like to to solid walls (z component in Fig. 1) throughout the whole sys-
each other. Because a perfect FCC crystal would produce slightly tem, i.e. both liquid and solid walls were compressed at 10 MPa.
different√ sizes in the x and y directions, y dimension was scaled The pressure control scheme is described in detail in the following
by 1.4/ 2 for both PSA and CSA systems before wall rotation to subsection. Afterwards, in order to generate a steady heat flux for
produce equal system size of 3.878 nm in both directions. the NEMD simulations, two heat baths, displayed in Fig. 1, were
The number of liquid molecules in each system is displayed coupled to the second outermost layer in each solid wall, where
in Table 1. Because we imposed a constant pressure and mean the temperature of the left solid wall heat bath was set to 135 K
temperature condition (phases 2–4), i.e. the wall distance was not and that of the right one was set to 105 K, and the mean temper-
fixed during the equilibration and relaxation phases of 2 and 3, ature of approximately 120 K was maintained. The heat bath tem-
adjusting the number of liquid molecules between the solid walls peratures were chosen to produce a large enough heat flux val-
was necessary to obtain liquid films with different thickness and ues and a distinct temperature profile over the interface while also
different number of layers. Based on the results of preliminary making sure the liquid molecules remain in liquid phase, as is cus-
simulations, the initial distance between the solid walls was set tomary in similar NEMD simulations [2,15,16]. Only a single solid
to be (a) 0.4 nm, (b) 0.8 nm, (c) 1.1 nm and (d) 1.4 nm for the wall layer was used for each heat bath, as no change in interfa-
cases of the number of the liquid molecules (a) N = 90, 102 or cial thermal conductance was observed even when the number of
114, (b) N = 140, 150, 160, 192 or 210, (c) N = 225, 250 or 282, and layers coupled to each thermostat was increased to 20 with larger
(d) N = 372, respectively. These numbers of liquid molecules can systems described in Section 2.1. At this stage, heat bath temper-
be expressed in area number density, number of molecules per a ature was controlled via velocity scaling due to a technical limita-
unit xy cross section, as (a) 5.98 nm−2 , 6.78 nm−2 or 7.58 nm−2 , (b) tion. The system was relaxed for 15 ns under a constant pressure
9.31 nm−2 , 9.97 nm−2 , 10.64 nm−2 , 12.77 nm−2 or 13.96 nm−2 , (c) condition of 10 MPa (phase 3).
14.96 nm−2 , 16.62 nm−2 , or 18.75 nm−2 and (d) 24.74 nm−2 , respec- During these phases the walls were allowed to move and the
tively. Here, the distance between the two solid walls was defined mean positions of the outermost wall layers were determined from
as the distance between the averaged locations of the innermost the last 10 ns of phase 3 to be used in later phases.
(surface) atom layers of the solid walls. Setting of the distance be-
tween the solid walls and liquid molecule numbers to (a)-(d) and 2.2.1. Implementation of constant pressure condition
executing the equilibration and relaxation protocol described in To realize the constant pressure condition, the pressure in the z
later subsections (for phases 2–4), resulted in one to four LLM, re- direction was controlled with the method proposed by Lupkowski
spectively. Additionally, larger systems that would eventually con- [17] for 15 ns before the data production run. Virtual walls with
tain a liquid phase close to the state of saturated liquid were also masses equal to 10 platinum atoms were placed outside the sys-
created by setting the number of liquid molecules to 1520 and ini- tem, and a constant force was applied to the virtual walls. The in-
tially positioning the solid walls at a distance of 6 nm. In this case, teraction potential energy between a single virtual wall and the
bulk liquid away from the influence of the solid walls existed in solid atoms, E (Z), is given by
the middle of the liquid region. π ρ D 
1

−2α (Z−r0 )


2 −α (Z−r0 )

E (Z ) = Z + e − 4 Z + e (3)
2α 2 2α α
2.2. Equilibration and relaxation under constant pressure condition where the parameters D, α and r0 are the same as the parameters
used in the interaction between solid atoms, and ρ is the num-
This subsection describes simulation phases 2 and 3. For all of ber density of platinum. Consequently, the pressure of the system
the systems, after the initial placement of liquid molecules and could be controlled by the constant force exerted on the virtual
before starting the NEMD with heat flux, systems were initially walls. To ensure that the liquid molecules were always in a liquid
equilibrated at 120 K (phase 2), which is approximately the mid- state, the target value of the pressure determined by the constant
point between the critical temperature and the triple point tem- force was set to 10 MPa.
perature of the liquid employed in the present study, with veloc-
ity scaling of the whole system for 4 ns under constant pressure 2.3. Relaxation and production run under constant volume condition
condition of 10 MPa, which is about twice that of critical pres-
sure. For a better illustration of the magnitude of the values: if the This subsection describes the simulation phases 4, 5 and 6.
liquid molecules were substituted with water, the corresponding After the pressure control during phase 3, the outmost layers
control temperature and pressure would be approximately 460 K were fixed using the mean position values obtained a priori and
and 40 MPa. The control pressure was chosen to guarantee a com- described in the previous subsection. The temperature control
4 X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949

method of the solid walls was changed from velocity scaling to


Langevin thermostat with the damping factor of 0.01 fs−1 . After at
least 5 ns relaxation run (phase 4), the data production run was
carried out to measure heat flux and temperature distribution for
7 ns in case of two and more LLM and 50 ns in case of single LLM
(phase 5). The actual pressure of the systems during phase 5 was
computed via the method of planes [18] with control surfaces lo-
cated at the solid-liquid interfaces, and was confirmed to be ap-
proximately the same as the control pressure with a standard error
of mean at most at 10 MPa. Although the standard error of pressure
is somewhat large, it was confirmed that the effect on the results
of heat transfer was insignificant.
During the production run (phase 5), the control surfaces for
the measurement of heat flux were set at the midpoint between
the surface layer of the solid atoms and the layer of the solid
atoms next to the surface layer on both sides, and at the solid-
liquid interfaces. Here, the solid-liquid interface was defined as the
midpoint between the surface layer of the solid atoms and the LLM
adjacent to that. Additionally, for the systems with more than two
LLM, control surfaces were also set at the midpoints between LLM.
All the control surfaces were parallel to the xy plane. Heat flux J
passing through the control surface Sxy was calculated by the fol-
lowing equation [19].
  Fig. 2. Number density distributions of solid and liquid molecules in the case of
1 1 vi,z 1 PSA. The number of liquid molecules determine the amount of LLM and are (a) 102,
JSxy = mv2i + φi j t + (b) 192, (c) 282, (d) 372. The green and red lines represent the number densities
i
2 2
j
|vi,z | 2 i of solid and liquid molecules, respectively. (For interpretation of the references to
   zi j colour in this figure legend, the reader is referred to the web version of this article.)
× Fi j · vi + v j   , (4)
zi j  LLM, while some of the liquid molecules are inside the solid sur-
j>i
face. Apparently, there is inherent instability in single LLM systems
where the first term on the right hand side represents the energy due to the non-uniform distribution of liquid molecules and large
transport by molecules with kinetic energy and potential energy pressure fluctuations experienced by the solid walls, as described
passing through the control surfaces during the infinitely short in the paragraph about pressure control scheme in the previous
time t, and the second term represents energy propagation be- section. However, the number of the defects was very low and the
tween molecules via the intermolecular forces between pairs of overall interfacial resistance and heat transfer characteristics were
molecules sandwiching the control surface. F and v are the in- not substantially influenced by these surface defects. The density
termolecular force and the molecular velocity vector, respectively, distributions for all single LLM systems are given in Fig. S2 of the
while m and z are the mass of the molecule and the intermolecu- supplementary material.
lar distance in the z direction, respectively and φ is intermolecular The relationship between the area number density of liquid
potential. Subscripts i and j indicate the molecules i and j, respec- molecules and the distance between the solid surfaces is shown
tively. Furthermore, the second term can be decomposed into x, y in Fig. 3. Here, the area number density of the liquid molecules
and z directions in the following manner [2] was calculated by dividing the total number of liquid molecules in
Fi j · vi = Fi j,x vi,x + Fi j,y vi,y + Fi j,z vi,z . (5) the system by the cross-sectional area of the xy plane of the simu-
lation box. The distance between the solid surfaces was defined as
For several systems, additional production runs (phase 6) were the difference between the averaged positions of the solid atoms in
conducted for vibrational density of states analysis described in the innermost layers on both sides. The relationship between the
Section 3.7. For each selected system, a 10 0 0 sequential runs, each area number density of liquid molecules and the distance between
lasting 10 ps were conducted to obtain the mean velocity autocor- the solid surfaces is non-linear and shows stepwise increases ac-
relation function in the span of 10 ps. cording to increase in the number of liquid molecules. When the
number of LLM is unchanged, the distance between the solid walls
3. Simulation results increases slightly as the area number density of liquid molecules
increases. On the other hand, when the stepwise change in the
3.1. Structure of the liquid between the solid walls number of LLM occurs, the distance between the walls changes
dramatically. In other words, the distance between the solid walls
Four typical distributions of number density of solid and liquid is mostly dependent on the number of LLM. Moreover, in the case
molecules in the system with PSA configuration of the solid sur- of CSA, even under the same conditions, the distances between the
faces, which are stable under the same pressure applied in the z solid walls are slightly larger than those in the case of PSA. The
direction, are shown in Fig. 2. There are one to four LLM between reason for this is that the arrangement of liquid molecules is dif-
the solid walls in Fig. 2(a-d), respectively. The number of LLM de- ferent due to the different surface alignments between PSA and
pends on the number of liquid molecules between the solid walls. CSA surfaces. The influence of the surface configuration of the solid
When there is a single LLM between the solid walls, as is shown in wall disappears in case of four LLM.
Fig. 2(a), the peak value in the number density of liquid molecules
is much higher than those of other cases in Fig. 2(b-d). As the 3.2. Temperature distribution and definition of thermal resistance
number of LLM increases, the peak value of number density of the
liquid molecules decreases gradually as liquid molecules are less Two typical temperature distributions in the cases of one and
constrained with more LLM. It should be noted that in the case of two LLM with PSA surface configuration are shown in Fig. 4. Al-
a single LLM shown in Fig. 2(a), some solid atoms are inside the though the temperature gradients and heat flux values given in
X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949 5

of 140 atoms, while in case of liquid each layer consisted of at


least 90 liquid molecules, and when averaged over the production
run in phase 5, resulted in a standard error of mean at most of
0.3 K. A somewhat abrupt change in the solid temperature profile
can be observed at layers closest to the solid-liquid interfaces. Such
non-linear behavior has been observed in previous MD simulations
by both our and other groups in various solid lattices [2,16], and
has been attributed to phonon scattering at the interface. Although
it is technically difficult to directly observe such effect in reality,
non-Fourier heat transfer effects are reported for micro and nano-
scale systems as well as very short time scales [20]. When there
are two LLM shown in Fig. 4(b), a large temperature jump at the
solid-liquid interface is apparent. It shows that the solid-liquid in-
terface reduces heat transfer, which is recognized as TBR at the
solid-liquid interface. On the other hand, in the case of single LLM,
the temperature jump at the solid-liquid interface becomes very
small, and the temperature distribution over the solid and the liq-
uid is approximately continuous even across the solid-liquid inter-
Fig. 3. Relationship between the area number density of the liquid molecules and face. This indicates that TBR is miniscule in this case. The jump in
the distance between the solid surfaces for PSA and CSA systems. Ty at the solid-liquid interface is smaller than those in Tx and Tz
for all PSA systems. In the case of PSA, the motion of the liquid
molecules in the y direction is perpendicular to the lattice-scale
grooves on the (1 1 0) solid surface as shown in Table 1, and thus
the motion of the liquid molecules in the y direction can transfer
kinetic energy to the solid efficiently due to the roughness on the
solid surface, which results in smaller jump of Ty .
The thermal resistance between the two solid surfaces, Rtotal ,
consists of three terms as follows.

Rtotal = Ri,L + Rliquid + Ri,R (6)


where Rliquid is the thermal resistance of the liquid, and Ri ,L , Ri ,R
are the TBR at the left and right solid-liquid interfaces, respectively.
The three thermal resistances, Rliquid , Ri ,L and Ri ,R, were calculated
from temperature difference and the heat flux through the system
as follows:
T
R= (7)
Jz
where Jz is the heat flux. For Rliquid , Tliquid is the temperature
drop across the liquid film and for Ri ,L and Ri ,R , Ti ,L and Ti ,R are
the respective temperature jumps at the interfaces. To keep consis-
tency with the macroscopic model of interfacial thermal resistance,
T values were computed by extrapolating temperature profiles
in a linear regime to the solid-liquid interfaces [21]. Positions of
the solid-liquid interfaces were defined as the midpoint between
the averaged positions of the innermost solid surface layer and the
LLM adjacent to the solid surface. For solid in systems with two or
Fig. 4. Decomposed temperature distributions for PSA in the case of (a) a single more LLM, linear extrapolation was conducted via the temperature
LLM with 102 liquid molecules, (b) two LLM with 192 liquid molecules. The black
dashed lines represent solid-liquid interfaces and dotted lines represent number
values of the three layers in the middle of the solid wall, while for
density, indicated as n in the legend. Standard error of mean of each temperature one LLM systems layers from 30-th outermost to 10-th innermost
point is at most 0.3 K. were used. For liquid, linear extrapolation was conducted by using
the temperature values of every LLM.

Section 3.5 are very high when compared to real systems, they 3.3. Total thermal resistance between the two solid surfaces
can still be considered to be in a linear response regime, and are
comparable to that of previous works [2,15,16]. These large tem- The total thermal resistance between the two solid surfaces,
perature gradient and heat flux values also give the benefit of re- Rtotal as defined in Section 3.2, is shown in Fig. 5 for all the simu-
duced uncertainty in analysis results. Only the one LLM systems lated cases. It is plotted as a function of the area number density
are of slight concern, as they have substantially larger heat flux of liquid molecules in Fig. 5(a), and as a function of distance be-
due to lower TBR, as will be demonstrated in Section 3.3. How- tween the two solid walls in Fig. 5(b). The relation between the
ever, even in this case, we confirmed that significantly different area number density of liquid molecules and the distance between
heat flux produces consistent TBR values, which indicates that the the two solid walls is presented in Fig. 3.
system is still in a linear response regime. The temperature of each Except for the case with a single LLM between the solid walls,
layer of solid and liquid molecules was calculated separately and the total thermal resistance is of the order of 10−7 m2 ·K/W. The
decomposed into Tx , Ty and Tz according to the degrees of free- case of the largest systems containing liquid bulk with the area
dom of molecular motion. In case of the solid, each layer consisted number density of 100 nm−2 and the distance between solid walls
6 X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949

Fig. 6. Thermal resistance of the liquid, Rliquid , as a function of the distance between
the solid surfaces, L, with a fitted line (blue) and the theoretical line of L/λ (green),
where λ is thermal conductivity of the liquid. (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this
article.)

Thermal conductivity, λ, was determined to be 0.0858 W/m·K from


the present simulation with the following equation:
Fig. 5. (a) Total thermal resistance as a function of the area number density of liq-
uid molecules. (b) Total thermal resistance as a function of distance between solid Jz
surfaces. λ= − (8)
(dT /dz )
where, the heat flux Jz and temperature gradient dT/dz were ob-
of 6 nm exhibits approximately 50% higher Rtotal when compared
tained from the largest PSA system containing liquid bulk with
with other cases, which mainly results from the increase of the
solid surfaces approximately 6 nm apart. This thermal conductivity
thermal resistance for the liquid, Rliquid . It is observed in the com-
value is comparable to that of argon previously obtained in both
parison between the cases with PSA and CSA that Rtotal in CSA is
computational and experimental works [22–25]. Temperature gra-
mostly higher than in PSA, which is especially apparent for the
dient was obtained via linear fitting of the temperature profile in
two or three LLM cases. This is mainly due to surface configura-
the region where the distance from solid surfaces was more than
tion of the solid walls, which will be discussed in later sections
1.9 nm.
by analyzing the distribution of heat flux and vibrational density
It is found in Fig. 6 that thermal resistance of the liquid is pro-
of states. In contrast, thermal resistance for a single LLM system is
portional to the distance between the walls and can be approxi-
extremely small when compared with the cases of more LLM and
mated quantitatively by the macroscopic response of thermal resis-
it agrees with the results investigated by Tsai’s and Quan’s groups
tance as Rliquid = L/λ. This appears to be valid even for molecularly
[4,6]. Single LLM systems will be examined in greater detail in later
thin liquid films made of two LLM.
sections. Finally, molecular mechanism of the characteristics of to-
The TBRs of PSA systems at the left and right solid-liquid in-
tal thermal resistance mentioned above will also be discussed in
terfaces, Ri,L and Ri,R , are plotted in Fig. 7(a) as functions of the
the following sections in greater detail.
area number density of liquid molecules, and Fig. 7(b) as functions
of the distance between the solid walls. In the cases of a single
3.4. TBR at the solid-liquid interfaces and in the liquid film
LLM, TBR at the interface has very small non-zero values, which
are much smaller than those in the cases of two or more LLM.
As was discussed in the previous section, the total thermal re-
When there are two or three LLM, TBR at the interface decreases
sistance between the two solid walls is decomposed to TBRs at
as the area number density of liquid molecules increases. This is
the two solid-liquid interfaces, Ri ,L and Ri ,R , and the thermal re-
due to the increase in the number of liquid molecules which take
sistance in the liquid film, Rliquid . In the systems presented in this
part in thermal energy transfer over the interface. When the num-
paper, when there is no liquid bulk and the liquid film thickness
ber of LLM changes from two to three, TBR at the interface sharply
is small, Rliquid is small when compared to Ri ,L and Ri ,R , and ac-
increases up to the value close to the maximum value of the
counts for only a small part of the total thermal resistance. Fig. 6
two LLM case, because the volume density of liquid molecules de-
shows Rliquid as a function of the distance between the solid walls.
creases sharply. Afterwards, TBR decreases again as the area num-
In the case of a single LLM, the thermal resistance of the liquid,
ber density of liquid molecules increases. In contrast, a different
Rliquid , is treated as zero because no temperature drop in the liquid
tendency is observed in the cases of a single LLM. This special case
is assumed, i.e., Tliquid = 0. The data point for the distance of ap-
will be discussed later in Section 3.6.
proximately 6 nm represents the thermal resistance of the largest
PSA system that contains liquid bulk. The blue line in Fig. 6 rep-
resents the linear fit for the data for the cases with two or more 3.5. Comparison between PSA and CSA
LLM. The green line represents the well-known macroscopic ther-
mal resistance due to heat conduction in a material given by L/λ, In this section, we discuss the effect of the surface alignment,
where L and λ are thickness and thermal conductivity of the ma- as shown in detail in Table 1, by comparing the results between
terials. Here, the distance between the two solid surfaces, which is PSA and CSA systems. Fig. 8(a) and (b) shows temperature distri-
slightly larger than the actual thickness of the liquid, is used as L. butions decomposed into Tx , Ty and Tz according to the degrees
X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949 7

perature jumps at the solid-liquid interface among Tx , Ty and Tz ,


and Ty has the smallest value in the both cases of PSA and CSA.
The right side of PSA exhibits the same characteristics. This is ex-
pected, as these three solid walls, left sides of PSA and CSA and the
right side of PSA, have an identical configuration of surface atoms.
On the other hand, the right side of CSA shows different charac-
teristics from the above. Ty exhibits the largest temperature jumps
at the solid-liquid interface and Tx the smallest one. This is consis-
tent with the fact that the right solid wall of CSA is configured by
rotating the left wall by 90 degrees around the z axis. On the right
wall surface of CSA, the lattice-scale grooves of the (1 1 0) crystal
plane lie in the direction of y, so the motion of liquid molecules
in the x direction is more effective when transferring thermal en-
ergy, as discussed in Section 3.2. In summary, the component of
molecular motion that is effective when transferring thermal en-
ergy beyond the solid-liquid interface changes from y to x compo-
nent along the thermal energy transfer direction from the left solid
wall to the right one in the case of CSA. This shift of the effective
component of molecular motion does not occur in the case of PSA.
Fig. 8(c) and (d) shows heat flux obtained at each control surface
according to Eq. (4) in the case of PSA and CSA, respectively. In
the figure, Jtotal is the total heat flux, and Jx , Jy and Jz are compo-
nents of heat flux, i.e., the contributions of each degree of free-
dom of molecular motion to the total heat flux according to Eqs.
(4) and (5). Note that strictly speaking, the sum of Jx , Jy and Jz is
equal to the second term of Eq. (4) and not Jtotal . However, at the
Fig. 7. TBR of PSA at the solid-liquid interfaces as functions of (a) the area number
density of liquid molecules, (b) the distance between the two solid walls. control surfaces placed in these systems, the contribution of the
first term of Eq. (4), i.e., the contribution of the molecular trans-
port, is negligible because the number of molecules that cross the
of freedom of molecular motion for PSA and CSA systems, respec- control surface is approximately zero. In Fig. 8(c) for PSA where
tively. Fig. 8(a) is the same as Fig. 4(b). It is observed that the tem- the solid surfaces are configured in a parallel fashion, the distri-
perature distributions of each degree of freedom in the left half are bution of the components of heat flux between the solid walls is
almost identical for both PSA and CSA. Tx exhibits the largest tem- almost symmetrical. Jy and Jz make the largest contribution to the

Fig. 8. (a) Temperature distribution in the case of parallel surface alignment (PSA). (b) Temperature distribution in the case of crossed surface alignment (CSA). (c) The
distribution of heat flux in the case of PSA. (d) The distribution of heat flux in the case of CSA. The number of liquid molecules is 192 for all the cases presented here. The
black dashed lines represent solid-liquid interfaces.
8 X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949

transferred thermal energy, while Jx makes little contribution. On faces, i.e., Rtotal = Ri ,L + Ri ,R . As shown in Fig. 9(a), regardless of PSA
the other hand, in Fig. 8(d) for CSA where the surfaces of the solid or CSA, Rtotal increases as the area number density of the liquid
walls are configured in a cross-like fashion, the magnitudes of con- molecules increases, which is an opposite tendency to the results
tribution of x and y components to the total heat flux switch along obtained in the cases with two and three LLM.
the direction of the heat flow from left to right. Jy , which is ini- To elucidate the mechanism that causes this tendency, the total
tially the thermal energy transferred through the left solid-liquid heat flux, Jtotal , was decomposed into two components: the com-
interface, decays as it flows in the liquid while Jx grows and even- ponent resulting from the interaction between solid molecules, JSS ,
tually becomes the thermal energy transferred through the right and the component from the interaction between solid and liquid
solid-liquid interface. This shift of thermal energy transfer from Jy molecules, JSL . As shown in Fig. 10, JSS , representing direct heat
to Jx is accompanied by a difference between temperature com- transfer between solid surfaces, contributes to the total heat flux
ponents, Ty − Tx . It is probable that this energy transfer between significantly. In this sense, it can be stated that the total thermal
different components causes the total thermal resistance of CSA to resistance in this case is composed of TBRs at the two solid-liquid
be larger than that of PSA. Meanwhile, when the number of liquid interfaces and a TBR at the solid-solid interface. A substantial ef-
molecules is the same, the volume density of liquid molecules in fect of direct solid-solid thermal energy transfer is present only in
the case of CSA is lower than that of PSA because of a longer dis- systems with a single LLM, because solid-solid interaction strength
tance between the solid walls. This is another reason of a larger decays rapidly when the distance between the two solid surfaces
total thermal resistance in CSA systems. increases. This is one of the mechanisms that permits very low to-
tal thermal resistance as compared with that for the systems with
3.6. TBR over the interface with a single LLM multiple LLM. Another contributing factor is that when there is a
single LLM, the peak value in the distribution of number density
As mentioned in Section 3.3, heat transfer over the interface of liquid molecules is very high, i.e., the strength of adsorbtion on
with a single LLM between the solid surfaces exhibits unique char- the surface of the solid wall is much stronger and liquid molecules
acteristics with extremely low TBR. Fig. 9(a) and (b) shows the to- are much closer to the solid surfaces than in other cases, as can
tal thermal resistance, Rtotal , as functions of the area number den- be seen in Fig. 2, which greatly enhances solid-liquid heat transfer.
sity of the liquid molecules and the distance between the solid Finally, liquid molecules in single LLM systems have a vastly differ-
walls, respectively, for PSA and CSA. As was described in the pre- ent vibration state, described in detail in the next section, which
vious sections, Rliquid is assumed to be zero in the case of a single
LLM. In such case, Rtotal is formally composed of TBRs at the inter-

Fig. 9. The total thermal resistance, Rtotal , as functions of (a) the area number den- Fig. 10. Decomposed heat flux as functions of (a) the area number density of liquid
sity of liquid molecules, (b) the distance between the solid surfaces. molecules, (b) the distance between the solid surfaces.
X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949 9

is more similar to that of solid walls, likely reducing phonon dis- CSA LLM in the y and z components. Liquid molecules in the single
sipation when crossing over the interface, while in systems with layer PSA appear to be greatly constrained in the y and z directions
multiple LLM, the vibration state of liquid molecules is more rem- (see Table 1), which causes high frequency lattice-like vibrations
iniscent to that of liquid bulk. Although JSL increases as the area as observed in Fig. 11(e, i), while no such effect can be observed
number density of the liquid molecules increases, which is similar for the PSA x component or any of the CSA components, where
to the systems with multiple LLM, JSS decreases at a higher rate the LLM spectra are more spread out. It must be noted that simi-
due to increase in the distance between solid surfaces. Therefore, lar high frequency vibrations were observed in single LLM system
Jtotal decreases during the increase of area number density of liquid simulations conducted by Tsai’s group, where they used an FCC
molecules. (1 1 0) plane [4], which more closely corresponds to our PSA sys-
Comparison between the characteristics of PSA and CSA in tems. Note that none of the liquid layer spectra are similar to that
Fig. 9(a) reveals that the total thermal resistance of CSA systems of liquid bulk, where a single peak is located at lower frequency
is higher than that of PSA systems when they are compared at the range, creating a nonzero value at zero frequency [4,6,27], which
same area number density of the liquid molecules. On the other indicates that while CSA does not contain highly localized high-
hand, the total thermal resistances for PSA and CSA are approx- frequency vibrations, it is still a system with vastly different vibra-
imately on a same trend as a function of the distance between tional state from bulk properties. As a general tendency discussed
the solid surfaces. Based on these findings, it is concluded that in Section 3.6, for the single LLM systems, CSA tends to have higher
the distance between the solid surfaces in the case of CSA tends TBR than PSA at the same number of liquid molecules. There does
to be larger when compared to PSA at the same number of liquid appear to be a substantial overlap in VDOS profiles of the LLM and
molecules under the same pressure, which results in the increase solid for the x component of PSA in Fig. 11(a), but further investi-
of total thermal resistance. This is a key mechanism that causes gation is required to determine if this can be directly translated to
higher thermal resistance for CSA as compared with PSA. heat flux contribution or if the substantially high LLM frequencies
in the y and z components in Fig. 11(e, i) have greater effect.
3.7. Comparison of vibrational density of states In a similar fashion to the previous paragraph, Fig. 11(c, d, g, h,
k, l) displays VDOS spectra for the two LLM systems. We first note
To better illustrate the difference between the single and mul- that unlike the single LLM, no drastic differences exist between dif-
tiple LLM systems, as well as the difference between PSA and ferent systems or components in the spectra of liquid layers. They
CSA, standard vibrational density of states (VDOS) analysis method all show a single peak in the low frequency region, which is simi-
[21,26] was applied to the single and two LLM systems. Each LLM lar to what is obtained from a liquid bulk of monatomic molecules
and the solid surface layer closest to the interface were treated [4,6,27], i.e. liquid molecules essentially act similar to molecules in
separately to obtain distinct VDOS distribution for each of them, liquid bulk, and are no-longer constrained to such a high degree as
as displayed in Fig. 11. To be consistent with the heat flux decom- for the single LLM systems, although y and z components have zero
position in Fig. 8, VDOS is also decomposed into x, y and z compo- values at zero frequency, indicating an absence of self-diffusion in
nents. those directions. As demonstrated in Fig. 8(c, d), in case of two
Firstly, for single LLM systems, in Fig. 11 (a, b, e, f, i, j), a very LLM systems, the z component has the largest contribution to the
stark difference can be seen between the spectrum of PSA LLM and overall heat flux. For PSA, second largest contribution comes from

Fig. 11. Comparison of vibrational density of states of each LLM or solid layer at the interface between single and double layer PSA and CSA systems. Left and right solid
configurations are described in Table 1. The number of liquid molecules for single and two LLM systems are 102 and 192, respectively.
10 X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949

the y component, and for CSA, the y component for the left side properties in an unambiguous manner, such as applying spectral
and the x component for the right side. These tendencies are re- analysis to heat flux [29]. Furthermore, investigation of more com-
flected by low frequency solid layer peaks observed close to the plex liquid molecules is also important from the perspective of real
low frequency liquid peaks, which are not observed in solid layers application, as well as a basic study. This will be undertaken in the
further from the interface (not shown in the graph). Such induc- next work. The results of these studies will lead not only to clari-
tion of solid interface vibrational states by interfacial liquid has fication of the physical phenomena in such molecular thermofluid
also been observed for SiC-water systems, where low frequency system but also to finding theoretical references for the design of
bands have been found to have greatest contribution to thermal TIM and better solid-solid junctions.
conductance [28]. This is especially clear in the case of CSA in Fig.
11(d, h), where the left solid layer peaks are more prominent for Declaration of Competing Interest
the y component, while the right layer peaks are more prominent
for the x component, which is consistent with Fig. 8(d). Therefore, The authors declare that they have no known competing finan-
while liquid in the double LLM systems is no-longer as constrained, cial interests or personal relationships that could have appeared to
the coupling between solid and liquid is still present. Furthermore, influence the work reported in this paper.
the spectra of LLM in the x and y components for the PSA sys-
tems in Fig. 11(c, g) perfectly overlap, while there is a mismatch Acknowledgment
between that of the CSA systems in Fig. 11(d, h), supporting previ-
ous observations in Section 3.5 that TBR in CSA systems is higher This work was supported by JSPS Kakenhi Grant Number
due to thermal transfer between the x and y components. 17K06182 and JST CREST Grant Number JPMJCR17I2, Japan.

4. Conclusion Appendix A. Supplementary material

Non-equilibrium molecular dynamics simulations concerning Supplementary data to this article can be found online at
heat transfer over extremely thin liquid film between two paral- https://doi.org/10.1016/j.ijheatmasstransfer.2019.118949.
lel solid walls have been carried out. The solid wall was a FCC
References
crystal with its (1 1 0) plane contacting the liquid. Two combina-
tions of solid wall alignment, with respect to the orientation of [1] T. Ohara, D. Torii, Molecular dynamics study of thermal phenomena in an ul-
the lattice-scale grooves on the (1 1 0) surface were used: parallel trathin liquid film sheared between solid surfaces: The influence of the crystal
plane on energy and momentum transfer at solid-liquid interfaces, J. Chem.
surface alignment (PSA) and crossed surface alignment (CSA). By
Phys. 122 (2005), doi:10.1063/1.1902950.
adjusting the number of liquid molecules between the solid walls [2] D. Torii, T. Ohara, K. Ishida, Molecular-scale mechanism of thermal resistance
under a constant pressure condition, systems with different num- at the solid-liquid interfaces: influence of interaction parameters between
ber of layers of liquid molecules (LLM) were constructed. The en- solid and liquid molecules, J. Heat Transfer. 132 (2010), doi:10.1115/1.3211856.
[3] A.R. bin Saleman, H.K. Chilukoti, G. Kikugawa, M. Shibahara, T. Ohara, A molec-
ergy transfer through these solid-liquid-solid systems for various ular dynamics study on the thermal transport properties and the structure of
number of LLM and the solid wall alignments was analyzed and the solid–liquid interfaces between face centered cubic (FCC) crystal planes of
the mechanism that determines thermal resistance was discussed. gold in contact with linear alkane liquids, Int. J. Heat Mass Transf. 105 (2017)
168–179, doi:10.1016/j.ijheatmasstransfer.2016.09.069.
Notably, the phenomena that occur when two contact surfaces are [4] Z. Liang, H.L. Tsai, Effect of molecular film thickness on thermal conduction
very close, such as in extremely thin areas of thermal interface ma- across solid-film interfaces, Phys. Rev. E - Stat. Nonlinear, Soft Matter Phys.
terials (TIM) applied between the two solid surface, were exam- (2011), doi:10.1103/PhysRevE.83.061603.
[5] Z. Liang, H.L. Tsai, Thermal conductivity of interfacial layers in nanofluids, Phys.
ined. Rev. E - Stat. Nonlinear, Soft Matter Phys. 83 (2011) 1–8, doi:10.1103/PhysRevE.
It was found that the main factor determining the distance be- 83.041602.
tween solid surfaces is the number of LLM, rather than the area [6] X. Wang, P. Cheng, X. Quan, Molecular dynamics simulations of thermal
boundary resistances in a liquid between two solid walls separated by a
number density of liquid molecules. When there are two or three nano gap, Int. Commun. Heat Mass Transf. 77 (2016) 183–189, doi:10.1016/j.
LLM between the solid surfaces, the thermal boundary resistance icheatmasstransfer.2016.08.006.
at the solid-liquid interface decreases with the increase of area [7] S. Murad, I.K. Puri, Molecular simulation of thermal transport across hy-
drophilic interfaces, Chem. Phys. Lett. 467 (2008) 110–113, doi:10.1016/j.cplett.
number density of liquid molecules. In such cases, the total ther-
2008.10.068.
mal resistance depends mainly on the alignment of the solid walls [8] M. Barisik, A. Beskok, Boundary treatment effects on molecular dynamics sim-
and the density of the liquid molecules in the layers. It was also ulations of interface thermal resistance, J. Comput. Phys. 231 (2012) 7881–
demonstrated that some coupling exists between the vibrational 7892, doi:10.1016/j.jcp.2012.07.026.
[9] K.M. Issa, A.A. Mohamad, Lowering liquid-solid interfacial thermal resistance
density of states (VDOS) of LLM and the adjacent solid wall lay- with nanopatterned surfaces, Phys. Rev. E - Stat. Nonlinear, Soft Matter Phys.
ers, and the in-plane components of LLM are largely dependent 85 (2012) 3–7, doi:10.1103/PhysRevE.85.031602.
on surface wall alignment. When there is a single LLM between [10] B.H. Kim, A. Beskok, T. Cagin, Molecular dynamics simulations of thermal re-
sistance at the liquid-solid interface, J. Chem. Phys. 129 (2008), doi:10.1063/1.
the solid surfaces, the total thermal resistance between the solid 3001926.
walls is greatly reduced. The total thermal resistance increases as [11] H.J. Castejón, The effect of point defects on the scattering and trapping of rare
the area density of liquid molecules increases, which is an oppo- gases on metallic surfaces, Surf. Sci. 564 (2004) 165–172, doi:10.1016/j.susc.
2004.06.176.
site response to that in the case with multiple LLM. In addition, by [12] E. Ruckenstein, H. Liu, Self-diffusion in gases and liquids, Ind. Eng. Chem. Res.
comparing the results in the case of PSA and CSA, it was concluded 36 (1997) 3927–3936, doi:10.1021/ie9701332.
that the asymmetry of the solid walls causes the reduction of ther- [13] P. Spijker, A.J. Markvoort, S.V. Nedea, P.A.J. Hilbers, Computation of accommo-
dation coefficients and the use of velocity correlation profiles in molecular dy-
mal energy transfer. It was also demonstrated that the VDOS of a
namics simulations, Phys. Rev. E - Stat. Nonlinear, Soft Matter Phys. 81 (2010)
single LLM is vastly different from that of the multiple LLM sys- 1–15, doi:10.1103/PhysRevE.81.011203.
tems or bulk liquid, and has solid-like tendencies, which is thought [14] O. Yenigun, M. Barisik, Effect of nano-film thickness on thermal resistance at
water/silicon interface, Int. J. Heat Mass Transf. 134 (2019) 634–640, doi:10.
to play an important contribution to reducing phonon dissipation
1016/J.IJHEATMASSTRANSFER.2019.01.075.
at the interfaces. This is especially apparent for PSA, where due [15] N. Shenogina, R. Godawat, P. Keblinski, S. Garde, How wetting and adhesion
to stronger liquid molecule constraints, very acute vibrations in a affect thermal conductance of a range of hydrophobic to hydrophilic aqueous
higher frequency range were observed. interfaces, Phys. Rev. Lett. 102 (2009), doi:10.1103/PhysRevLett. 102.156101.
[16] M. Barisik, A. Beskok, Temperature dependence of thermal resistance at the
Although, detailed analysis on heat flux components and the water/silicon interface, Int. J. Therm. Sci. 77 (2014) 47–54, doi:10.1016/J.
VDOS was conducted, further work is necessary to connect the two IJTHERMALSCI.2013.10.012.
X. Liu, D. Surblys and Y. Kawagoe et al. / International Journal of Heat and Mass Transfer 147 (2020) 118949 11

[17] M. Lupkowski, F. Van Swol, Computer simulation of fluids interacting with [24] S.C. Maroo, J.N. Chung, Molecular dynamic simulation of platinum heater and
fluctuating walls, J. Chem. Phys. 93 (1990) 737–745, doi:10.1063/1.459524. associated nano-scale liquid argon film evaporation and colloidal adsorption
[18] B.D. Todd, D.J. Evans, P.J. Daivis, Pressure tensor for inhomogeneous fluids, characteristics, J. Colloid Interface Sci. 328 (2008) 134–146, doi:10.1016/J.JCIS.
Phys. Rev. E. 52 (1995) 1627–1638, doi:10.1103/PhysRevE.52.1627. 2008.09.018.
[19] T. Ohara, Intermolecular energy transfer in liquid water and its contribution to [25] K. Hyżorek, K.V. Tretiakov, Thermal conductivity of liquid argon in nanochan-
heat conduction: A molecular dynamics study, J. Chem. Phys. 111 (1999) 6492– nels from molecular dynamics simulations, J. Chem. Phys. 144 (2016) 194507,
6500, doi:10.1063/1.480025. doi:10.1063/1.4949270.
[20] A. Singh, E.B. Tadmor, Thermal parameter identification for non-Fourier heat [26] H. Bao, J. Chen, X. Gu, B. Cao, A review of simulation methods in mi-
transfer from molecular dynamics, J. Comput. Phys. 299 (2015) 667–686, cro/nanoscale heat conduction, ES Energy Environ. (2018), doi:10.30919/
doi:10.1016/J.JCP.2015.07.008. esee8c149.
[21] Z. Liang, M. Hu, Tutorial : Determination of thermal boundary resistance by [27] T. Sun, J. Xian, H. Zhang, Z. Zhang, Y. Zhang, Two-phase thermodynamic model
molecular dynamics simulations, 191101 (2018). doi:10.1063/1.5027519. for computing entropies of liquids reanalyzed, J. Chem. Phys. 147 (2017),
[22] J.C.G. Calado, U.V. Mardolcar, C.A. Nieto de Castro, H.M. Roder, W.A. Wake- doi:10.1063/1.5001798.
ham, The thermal conductivity of liquid argon, Phys. A Stat. Mech. Its Appl. [28] C.U. Gonzalez-Valle, B. Ramos-Alvarado, Spectral mapping of thermal trans-
143 (1987) 314–325, doi:10.1016/0378-4371(87)90071-9. port across SiC-water interfaces, Int. J. Heat Mass Transf. 131 (2019) 645–653,
[23] A.J.H. McGaughey, M. Kaviany, Thermal conductivity decomposition and anal- doi:10.1016/j.ijheatmasstransfer.2018.11.101.
ysis using molecular dynamics simulations. Part I. Lennard-Jones argon, Int. J. [29] K. Sääskilahti, J. Oksanen, J. Tulkki, S. Volz, Spectral mapping of heat transfer
Heat Mass Transf. 47 (2004) 1783–1798, doi:10.1016/J.IJHEATMASSTRANSFER. mechanisms at liquid-solid interfaces, Phys. Rev. E 93 (2016) 1–8, doi:10.1103/
20 03.11.0 02. PhysRevE.93.052141.
Update
International Journal of Heat and Mass Transfer
Volume 150, Issue , April 2020, Page

DOI: https://doi.org/10.1016/j.ijheatmasstransfer.2020.119307
International Journal of Heat and Mass Transfer 150 (2020) 119307

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


journal homepage: www.elsevier.com/locate/hmt

Erratum

Erratum to “A molecular dynamics study of thermal boundary


resistance over solid interfaces with an extremely thin liquid film”
[International Journal of Heat and Mass Transfer, 147 (2020), 118949]
Xiao Liu a,b,1, Donatas Surblys b,1,∗, Yoshiaki Kawagoe b, Abdul Rafeq Bin Saleman c,
Hiroki Matsubara b, Gota Kikugawa b, Taku Ohara b
a
Department of Mechanical and Aerospace Engineering, Graduate School of Engineering, Tohoku University, 6-6, Aramaki Aoba, Sendai 980-0845, Japan
b
Institute of Fluid Science, Tohoku University, 2-1-1 Katahira, Aoba, Sendai 980-8577, Japan
c
Faculty of Mechanical Engineering, Universiti Teknikal Malaysia Melaka, Hang Tuah Jaya, 76100 Durian Tunggal, Melaka, Malaysia

The publisher regrets that Fig. 7 was published incorrectly. The correct version of Fig. 7 is indicated below and has been updated on
the original article online.

The publisher would like to apologise for any inconvenience caused.

DOI of original article: 10.1016/j.ijheatmasstransfer.2019.118949



Corresponding author: Donatas Surblys, Institute of Fluid Science, Tohoku University, 2-1-1 Katahira, Aoba, Sendai 980-8577, Japan.
E-mail address: donatas@microheat.ifs.tohoku.ac.jp (D. Surblys).
1
These authors have equally contributed.

https://doi.org/10.1016/j.ijheatmasstransfer.2020.119307
0017-9310/© 2020 Elsevier Ltd. All rights reserved.

You might also like