You are on page 1of 8

Computational Materials Science 152 (2018) 300–307

Contents lists available at ScienceDirect

Computational Materials Science


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

Point defect effects on the thermal conductivity of β -SiC by molecular T


dynamics simulations
⁎ ⁎
Yichen Maoa,b, Yingying Lib, Yangheng Xionga, , Wei Xiaob,
a
School of Power and Mechanical Engineering, Wuhan University, Wuhan 430072, PR China
b
Division of Nuclear Materials and Fuel, State Power Investment Corporation Research Institute, Beijing 102209, PR China

A R T I C LE I N FO A B S T R A C T

Keywords: The point defect effects on phonon thermal conductivity of bulk crystalline β -SiC at 750 K and 1098 K are
Thermal conductivity calculated with reverse non-equilibrium molecular dynamics method. Quantum correction is used to correct the
Point defect corresponding temperatures. By extrapolation and quantum correction, the results of thermal conductivity of
Molecular dynamics bulk β -SiC with point defects are achieved and corrected. For the eight types of point defects studied in this
Reverse non-equilibrium
work, the thermal conductivity of SiC decreases with the increase of the concentration of certain point defects.
SiC
The thermal conductivity drops rapidly at low concentration condition and gradually reaches certain value at
high concentration condition. At high defect concentration condition, the temperature does not affect the
thermal conductivity too much. The additional thermal resistivity is proportional to the concentration of all
types of the point defects we studied. In these point defects, the interstitial SiTC decreases the thermal con-
ductivity of β -SiC the most. On the other hand, the interstitial CTSi affects the thermal conductivity the least.

1. Introduction With experimental methods, Rohde has measured the thermal


conductivity of fast neutron and α -particle irradiated SiC by laser flash
Silicon carbide (SiC) is a promising alternative cladding material for method and demonstrated that the thermal conductivity decreases after
light water reactors due to its good mechanical and thermal properties, irradiation in a wide temperature range [6]. Price measured the
for example, high strength, creep resistance, low thermal expansion, thermal conductivity of fast neutron irradiated SiC by heat-pulse
high thermal conductivity, remarkable oxidation resistance, corrosion method and found that the thermal conductivity decreases; meanwhile,
resistance, and low neutron absorption cross section [1–5]. Moreover, he found that the thermal conductivity of unirradiated samples de-
under the extreme environment (for example, accident condition) its creases with the increase of temperature, and that of the irradiated
mechanical properties and corrosion properties are still good at high samples changes less at different temperatures [7].
temperature and strong neutron radiation conditions. However, some Numerical simulations are also used to study the thermal con-
problems still exist for the application of SiC. First of all, it is a brittle ductivity of SiC. For example, Malakkal et al. simulated the thermal
material and it may break during the running time. Second, since the conductivity of SiC under various temperature conditions by calculating
material defects are induced under irradiation condition, physical Boltzmann transport equation [8]. Equilibrium molecular dynamics is
properties may change due to the defects. For example, thermal con- used to simulate the thermal conductivity of 4H-SiC with different point
ductivity may become lower during the running time. defects [9]. Equilibrium molecular dynamics is also used to simulate the
As a cladding material, thermal conductivity is an essential physical point defect effects (five types of defects) on the thermal conductivity of
property because it is important for the heat transfer and the reactor β -SiC at the concentration of 0.5% [10]. Similar method is used to
safety issues of the nuclear reactor. During the running time, point calculate the effects of the concentration of vacancy and void on the
defects are generated by irradiation effects. The concentration of cer- thermal conductivity of β -SiC [11]. Non-equilibrium molecular dy-
tain point defect is a function of irradiation dose. Different defects may namics is used by Papanikolaou to simulate the thermal conductivity of
affect the thermal conductivity in different ways and the irradiation bulk and SiC nanowires [12] and by Crocombette et al. to study the
dose may also affect the thermal conductivity. As a result, it is inter- thermal conductivity of irradiated β -SiC and the β -SiC with various
esting to investigate the thermal conductivity change by different point point defects of different concentrations [13,14]. As point defects are
defects and different concentrations. important to the thermal conductivity of SiC, the effects of point defects


Corresponding authors.
E-mail addresses: yhxiong@whu.edu.cn (Y. Xiong), xiaowei@snptc.com.cn, xiaowei1@gmail.com (W. Xiao).

https://doi.org/10.1016/j.commatsci.2018.05.050
Received 14 March 2018; Received in revised form 20 May 2018; Accepted 25 May 2018
0927-0256/ © 2018 Elsevier B.V. All rights reserved.
Y. Mao et al. Computational Materials Science 152 (2018) 300–307

on the thermal conductivity are studied in this work. atom located at one side of the simulated system. This scheme to cal-
In our work, the point defect effects on phonon thermal conductivity culate thermal conductivity is reverse non-equilibrium molecular dy-
of bulk crystalline β -SiC at 750 K and 1098 K are calculated with re- namics method (rNEMD) [27–29]. The computational cost of EMD
verse non-equilibrium molecular dynamics method. By quantum cor- method is much higher than that of NEMD method and rNEMD method
rection, the temperatures are corrected as 658 K and 1036 K. Cell size [28,29]. Since rNEMD method is an high efficient method to calculate
extrapolation and quantum correction are used to achieve the thermal thermal conductivity [27], it is used in our calculations.
conductivity of bulk SiC with point defects and correct the results. Eight In our thermal conductivity calculations all cells are divided into
types of point defects VC , CSi, SiC , CTSi, CTC , C i−C〈100〉, SiTSi and SiTC are 200 slabs. The No.1 slab and the No.101 slab serve as the heat sink (low
studied in this work and the thermal conductivity of SiC decreases with temperature, section C in Fig. 2) and heat source (high temperature,
the increase of the concentration of certain point defects. section H in Fig. 2) respectively. The velocity of the hottest atom in slab
No. 1 and the velocity of the coldest atom in slab No. 101 are exchanged
2. Simulation model and methods every 100 MD steps until the temperature gradient is established and
stable. The spontaneous heat flow between the heat source and heat
2.1. Simulation cell models sink is balanced by the artificial exchange process and a stable tem-
perature distribution will be established eventually. When steady state
In order to simulate the thermal conductivity of cubic β -SiC. Super is established, the total heat flux J can be obtained by the net kinetic
cells with different lengths are generated to extrapolate the thermal energy exchange per second and unit area in Eq. (1),
conductivity of SiC with infinite size. The cross sections of all simula- 2 2
∑ m /2(vhigh−vlow )
tion cells are the same which are composed of 4 × 4 unit cells (corre- n
sponding to 1.74 × 1.74 nm2 ). Here, the lattice constant of β -SiC is J= ,
2At (1)
a0 = 4.359 Å. The axial lengths of the super cells are 100, 150, 200, 250,
and 300a 0 (corresponding to 43.59, 65.39, 87.18, 108.98, and 130.77 where vhigh and vlow represent the highest velocity of the atom in the
nm). The total atoms of these super cells are 12800, 19200, 25600, heat sink and the lowest velocity of the atom in the heat source, re-
32000, and 38400. Periodic boundary conditions are applied in x , y , spectively. Suppose there are n times velocity exchanges happened
and z directions for all simulations. during the simulation time t, the summation of all of the energy ex-
changes is used to evaluate the heat flow. Because the heat flows along
two directions, there is a factor 2 in the denominator. The temperature
2.2. rNEMD method
of each slab inside the super cells is calculated as follow:
nj
The thermal conductivity κ is the property of a material to conduct 3 1
heat. The differential form of Fourier’s law of thermal conduction can nj kB (TMD )j = ∑ mi vi2 ,
2 2 i=1 (2)
be written as J = −κ∇T . Here, J is the heat flux perpendicular to the
section of the system which is proportional to the temperature gradient. where (TMD )j represents the MD temperature of the jth slab, the operator
In general, Boltzmann transport equation [8,15–17] and molecular <> provides the mean value of the kinetic energy during certain si-
dynamics (MD) are two main methods to calculate the thermal con- mulation time period, mi and vi are the mass and velocity of the ith atom
ductivity. Equilibrium molecular dynamics (EMD) [18–20] and non- in the jth slab respectively, nj is the total atom number in the jth slab,
equilibrium molecular dynamics (NEMD) [21–23] are two typical MD and kB is the Boltzmann constant. Once the heat flux and the tem-
methods used to simulate thermal conductivity. For NEMD methods, perature distribution are obtained via Eqs. (1) and (2), the thermal
there are two branches which are homogenous [24,25] and non- conductivity can be calculated through its definition.
homogenous NEMD method [26]. Homogenous NEMD method can also Since point defects may affect the thermal conductivity of SiC, eight
be achieved by two ways. In the first way, with the extensive linear types of point defects are generated in the β -SiC super cells. The point
response theory, the system’s non-equilibrium response induced by an defects studied are VC , CSi, SiC , CTSi, CTC , C i−C〈100〉, SiTSi , and SiTC (see
external perturbation can be used to calculate the thermal conductivity Fig. 1). In order to extrapolate the thermal conductivity of bulk β -SiC
[24]. This way doesn’t generate a temperature gradient. In another with point defects, super cells with four lengths are used in the thermal
way, a heat source and a sink are induced by adding and removing conductivity calculations. The lengths of the super cells with various
energy to generate the temperature gradient, and then the thermal point defects are 100, 150, 200, and 250a 0 along the direction of the
conductivity is calculated by Fourier’s law of heat conduction [25]. The heat flux.
temperature gradient can also be generated by exchanging the velo- For the structures of point defects, some atoms are removed, re-
cities of the fastest atom in the middle of the system and the slowest placed, or added to generate vacancies, anti-sites, or interstitials in the

Fig. 1. Eight types of point defects in cubic β -SiC studied in this work. VC represents the carbon vacancy; CSi is an anti-site point defect in which a Si site is occupied by
a C atom; SiC is an anti-site defect in which a C site is occupied by a Si atom; CTSi represents a C interstitial in which a C atom is at the center of a Si tetrahedral; CTC is
a C interstitial in which a C atom is at the center of a C tetrahedral; SiTC is a Si interstitial in which a Si atom is at the center of a C tetrahedral; SiTSi is a Si interstitial in
which a Si atom is at the center of a Si tetrahedral; In the dumbbell configuration of C i−C〈100〉 two C atoms share a carbon site and the dumbbell is along the
direction of 〈100〉.

301
Y. Mao et al. Computational Materials Science 152 (2018) 300–307

2500

2000

1500

T MD (K)
Fig. 2. The schematic of a simulation cell. Lz is the length of the super cell. The
simulation cell is divided into 200 slabs. The regions marked H and C represent 1000
the heat source and heat sink formed by the velocity exchange process re-
spectively, in which there is one slab. The heat flows from heat source to sink in
two directions because of the periodic boundary condition applied in the z
500
direction. Region I represents the intermediate block that point defects (va-
cancy, anti-site, interstitial types) are generated, and the region B is used to
0
block the nonphysical phonon scattering caused by the velocity exchange 0 500 1000 1500 2000
process, in which there are 10 slabs.
T (K)

super cell, respectively. The concentrations of vacancy, anti-site, and Fig. 3. The relation between the quantum corrected temperature TMD and the
interstitial types (n v , na, ni ) are defined as the ratios of the number of MD temperature T for β -SiC.
the deleted, replaced, and added atoms to the number of total atoms in
the super cell. Since the velocity exchange process may cause non- will affect the thermal conductivity calculations. For example, the size
physical phonon scattering in the heat source and heat sink [28,30], effect on the thermal conductivity calculation of Si has been extensively
defects are not generated in these regions and the slabs closed to these studied and the results show that the thermal conductivity increases
two areas (section B in Fig. 2). with the increase of the length of the super cell [29,36,37]. Especially,
In our calculations, the simulations are performed by the classic the relationship between the thermal conductivity and the length of the
modeling tool Large-scale Atom/Molecular Massively Parallel super cell can be written as [29]:
Simulator (LAMMPS) [31]. The modified embedded atom method
1 a3 1 4⎞
(MEAM) [32] is employed to simulate the interatomic interactions. = 0 ⎜⎛ + ⎟.

Since the MEAM SiC potential is relatively good for thermal con- κ 4kB v ⎝ l∞ Lz ⎠ (4)
ductivity calculation, it is used in the following simulations.
where a 0 is the lattice constant of the material, kB = 1.38 × J/K is 10−23
the Boltzmann constant, v is the group velocity of an acoustic branch, l∞
2.3. Quantum correction of thermal conductivity
is the phonon mean free path and Lz is the length of the material. The
thermal conductivity of bulk β -SiC with infinite size can be extrapolated
In MD simulation, the temperature is evaluated by the velocities of
with the linear relation between 1/ Lz and 1/ κ .
the atoms based on the equipartition theorem of classical statistical
mechanics in Eq. (2). However, the Debye temperature of SiC is about
2.5. The parameters for MD simulation
1200 K and the quantum corrections to the MD temperature (TMD ) and
the thermal conductivity (κMD ) are required. Because the total system
The time step of all the simulation processes is 0.25 fs. An NVT
energy at TMD of the MD simulation equals the total phonon energy at
ensemble is applied initially to equilibrate the system for 0.1 ns. During
the corresponding quantum corrected temperature T, the relation be-
this stage, the total number of atoms N, the volume of the system V, and
tween TMD and T can be established with the following equation:
the temperature T are kept as constants. The temperatures of the MD
1 1 simulation at this stage are 750 K and 1098 K. Next, the rNEMD method
〈E 〉 = 3(N −1) kB TMD = ∑ ℏωi ⎜⎛ + ⎞
⎟.

i ⎝2 exp (ℏωi / kB T )−1 ⎠ (3) is used to calculate the thermal conductivity of β -SiC with an NVE
ensemble. In this stage, the total number of atoms N, total volume of the
where ωi is the ith
normal mode frequency, and N is the number of system V, and the total energy E are kept as constants. With rNEMD
atoms in the system [33,34]. method a stable heat flux is generated along z direction by inter-
The relation between quantum corrected temperature T and the MD changing the velocities of the atoms located at the heat source and the
temperature TMD is showed in Fig. 3. The phonon frequency calculation heat sink. This stage lasts for 0.725 ns. Because it takes some time to
is based on the MEAM atomic interaction. Because the cladding is close achieve the stable heat flux, the data of the initial part of this stage is
to the fuel, the temperature of the cladding is higher than that of the ignored and the data of the system with a stable heat flux is used for
coolant. Our MD temperatures of 750 K and 1098 K are corrected as thermal conductivity calculations. Especially, the data in the last 0.575
658 K and 1036 K which are little higher than the cladding temperature ns MD simulation is used for analysis.
of light water reactors (623 K) [35] and the operating temperature of
gas-reactors (1000 K) [11].
3. Simulation results
Because the heat flux in the MD simulation system JMD is equal to
that in the real quantum system J, JMD = −κMD·∇TMD = −κ·∇T = J .
3.1. Thermal conductivity of defect-free bulk β−SiC
Then the corrected thermal conductivity κ can be written as
dTMD
κ = κMD· dT and κMD is the thermal conductivity calculated by MD
In order to plot the temperature profile of the slabs along the z di-
simulation. Thus after the quantum correction of the temperature, the
rection, the mean values of the temperatures of each slab are calculated
thermal conductivity of κMD in the MD simulation should be corrected
with the following equation,
by the factor of dTMD [10].
dT
3,300,000
1
2.4. Finite-size effect of thermal conductivity
〈T (z ) 〉MD =
2, 300, 000
∑ TN (z )MD.
N = 1,000,001 (5)

Because the lengths of the super cells used in our work are shorter Here, 〈T (z ) 〉MD is the mean value of the MD temperatures of the No.z
than the phonon mean free path of β -SiC, the lengths of the super cells slab in the system, TN (z )MD represents the instantaneous MD

302
Y. Mao et al. Computational Materials Science 152 (2018) 300–307

1550 130
Temperature Profile Density Profile works, for example, simulation data via Porter-NEMD method [13] and
by EMD method [11]. The discrepancy between the simulation and the
experimental results may be attributed to the existence of defects in the
1350 129
materials which increases the scattering of phonon and results the de-
crease of the thermal conductivity.

Atoms/slab
T MD (K)

1150 128 3.2. Thermal conductivity of β-SiC with various point defects

In this section, eight types of point defects are generated in the β -SiC
950 127 super cells to evaluate their effects on the thermal conductivity as de-
scribed in the Method Section.
Vacancy effect is studied first. Super cells with different vacancy
750 126 concentrations (n v ) are generated to investigate the thermal con-
0 20 40 60 80
Z (nm) ductivity. The temperature profiles of the super cells with different n v
are shown in Fig. 6(a) and the geometry of these super cells are the
Fig. 4. The temperature profile and the atomic density profile of the simulation
same with the box size of 1.74 × 1.74 × 87.18 nm3 . It shows that the
super cell along the z direction. The MD temperature is 1098 K and the box size
of the super cell is 1.74 × 1.74 × 87.18 nm3 . The linear regions of the tempera- slopes of the linear regions of the temperature profiles (temperature
ture profile are used to calculate the temperature gradients. gradients) increase with the increase of n v . As a result, the thermal
conductivity of SiC is decreased by the vacancies. The similar phe-
nomenon happens for the SiC super cells with all types of the point
temperature of the No.z slab at the time step of N. Suppose the stable
defects studied. After extrapolation to achieve the bulk thermal con-
heat flux has reached in the time step range from N = 1, 000, 001 to
ductivity of defective condition, for both temperature conditions, the
N = 3, 300, 000 , the mean values of the MD temperatures during this
thermal conductivity decreases gradually with the increase of the
period of each slab are calculated and the corresponding temperature
concentration of vacancy (see Fig. 6(b)). Especially, it drops dramati-
profile is shown in Fig. 4.
cally at low concentration condition. For example, about n v = 0.3%, the
The temperature profile (black dash line) and the atomic density
thermal conductivity of SiC drops about 70% at 658 K. Even higher
profile (blue solid line) are shown in Fig. 4. Because of the intensive
vacancy concentration does not decrease the thermal conductivity very
phonon scattering near the heat source and heat sink generated by
much. For the sample with high vacancy concentration, the tempera-
dynamic velocity exchange process, non-linear distributions appear in
ture does not affect the thermal conductivity too much, which agrees
these two regions. In the thermal conductivity calculations, the linear
the experimental results [1,7].
regions of the temperature profile (red solid lines) are adopted to
The anti-site defects drop the thermal conductivity of β -SiC dra-
evaluate the temperature gradients [38]. Especially, the two regions in
matically at low concentration condition (see Fig. 7). Especially, SiC
the ranges of the 12–38% length and 62–88% length of the super cell
drops the thermal conductivity at very low concentration condition and
(10.46–33.13 nm and 54.05–76.72 nm ) are used to calculate the tem-
then the defect effects become very weak, that is, the κ vs. na becomes
perature gradients. Due to the thermal expansion, the atomic density at
flat (see Fig. 7(b)).
the heat source is low and that is high at the heat sink (blue solid line).
The thermal conductivity of β -SiC decreases with the increase of
Because of the size effect on the thermal conductivity calculations,
interstitial concentration for the five types of interstitial atoms although
our calculated 1/ κ values as functions of 1/ Lz for different β -SiC super
the slopes of the curves are different (see Fig. 8). The five types of in-
cells under the corrected temperatures of 658 K and 1036 K are plotted
terstitials are CTSi, CTC , C i−C〈100〉, SiTSi and SiTC . The thermal con-
in Fig. 5. With linear fitting, the corrected thermal conductivity of bulk
ductivity of the system with interstitials of CTSi decreases relatively
β -SiC can be derived at 1/ Lz = 0 . That is, κ ≈ 296.24 W/(m·K) at 658 K
slower than that of other samples at 658 K. On the contrary, the thermal
and κ ≈ 180.43 W/(m·K) at 1036 K after quantum correction. The ex-
conductivity of the system with SiTC drops quickly with the increase of
perimental value of thermal conductivity of β -SiC crystal at 1000 K is
the interstitial concentration at low concentration condition. Even-
about 100 W/(m·K) which is lower than our simulation value [13].
tually, the thermal conductivity curves become flat at high concentra-
However, our results agree with the results from other simulation
tion condition.
At 1036 K, the thermal conductivity of the system with CTC de-
0.025 creases slower than that of the other four types at low interstitial
linearfit @ 658 K concentration. A crossover appears for the thermal conductivity curves
linearfit @ 1036 K
of CTSi and CTC (see Fig. 8(B)).
0.02
4. Discussion

4.1. Effect of simulation parameters on the simulation results


0.015
4.1.1. Cross section area effect on the thermal conductivity simulation
Because the length of the simulation cell affects the thermal con-
0.01 ductivity simulation, it is interesting to investigate the cross section
area effect on the thermal conductivity calculation. In order to study
the cross section effect, perfect super cells with different cross section
areas which are perpendicular to the heat flux and with the same length
0.005 are generated. The thermal conductivities for these super cells are listed
0.005 0.01 0.015 0.02 0.025
in Table 1. Although the cross section area is a variable, the thermal
-1
1/L Z (nm ) conductivity does not change too much. Since there is no heat source
and sink in x and y directions and the periodic boundary condition is
Fig. 5. The relationship between 1/ κMD and 1/ Lz . The thermal conductivity of applied in these two directions, phonon scattering in these two direc-
bulk β -SiC can be extrapolated at 1/ Lz = 0 . tions can be ignored. In general, only the length of the super cell affects

303
Y. Mao et al. Computational Materials Science 152 (2018) 300–307

1800
perfect 300
n v = 0.3125% V C @ 658 K
1600 250 V C @ 1036 K
n v = 0.625%
n v = 1.25%
1400 200

(W/mK)
T MD (K)

1200 150

1000 100

800 50

0
600 0 0.2 0.4 0.6 0.8 1 1.2
0 20 40 60 80
Z (nm) n v (%)

Fig. 6. (a) Temperature profiles of the super cells in z direction with n v = 0, 0.3125%, 0.625%, 1.25% at a MD temperature of 1098 K. All the box sizes of the super
cells are 1.74 × 1.74 × 87.18 nm3 . (b) The calculated thermal conductivities (with quantum correction) as functions of the concentration of VC at the corrected
temperatures of 658 K and 1036 K.

the thermal conductivity calculation and the cross section area does thermal resistivity (W) and the relative additional thermal resistivity
not. (ΔW / W ) are used to analyze. ΔW as functions of the concentration of
vacancies (n v ) at corrected temperatures of 658 K and 1036 K are
4.1.2. Velocity exchange rate effect on the thermal conductivity simulation plotted in Fig. 10(a). ΔW increases nearly linearly with the increase of
For the rNEMD method, the velocity exchange rate may affect the the n v at both temperatures. Since the slope of the red curve is higher
calculated thermal conductivity. In order to check this effect, the than that of the blue one, vacancy effect is a little more serious at low
thermal conductivities are calculated with different exchange rates. temperature condition.
Suppose f represents the velocity exchange rate, in another word ve- Crocombette’s calculated the additional thermal resistivity of β -SiC
locity exchange happens in every f MD steps. Apparently small f means with homogenous NEMD method and argued that the thermal re-
high velocity exchange rate and the corresponding heat flux J is high, sistivity is proportional to the concentration of the defects [14]. The
which induces a large temperature gradient. The temperature profiles slopes of the curves of the ΔW vs n v are about 2 at 300 K and 900 K from
of perfect super cells simulated with different velocity exchange rates of Crocombette’s work. Meanwhile, our calculations show that the corre-
f = 100, 200, 500 at the MD temperature of 1098 K are shown in Fig. 9. sponding slopes are 3.28 at 658 K and 2.91 at 1036 K. Li et al. [10]
The three temperature profiles intersect each other around the MD calculated the thermal conductivity of five types of point defects with a
temperature of 1098 K. Our results show that the slope of the linear fixed defect concentration of 0.5% by EMD method. The slopes of the
region of the temperature profile decreases with the increase of f. curves of the ΔW vs n v are 3.36 and 4.14 at 620 K and 877 K. Samolyuk
The calculated thermal conductivity as a function of the velocity et al. [11] studied the relative additional thermal resistivity of β -SiC
exchange rate f is listed in Table 2. It demonstrates that the velocity with vacancies which are composed of half C vacancies and half Si
exchange rate does not affect the calculated thermal conductivity too vacancies by EMD method. The results show that ΔW / W linearly in-
much. In our calculations, f = 100 is chosen in the thermal conductivity creases with the increase of n v at 1000 K. We calculated the ΔW / W of
simulations. β -SiC with carbon vacancies and our results agree with Samolyuk’s data
(see Fig. 10(b)). The slope of the curve of Samolyuk’s work is 524.80
and ours is 576.01.
4.2. Effects of point defects on the thermal resistivity
Besides the vacancy in SiC, the point defect effects from all other
defects are also studied. The results show that the additional thermal
Point defects generated by the irradiation in reactors may affect
resistivity increases with the concentration of these defects pro-
thermal conductivity of SiC. The thermal resistivities (inverse of the
portionally. The additional thermal resistivity as functions of the con-
thermal conductivity) of the samples with point defects are calculated
centration of C i−C〈100〉 and SiTC are illustrated in Fig. 11. Another
to discuss the point defect effects. Especially, the additional thermal
observation is that the slope of the curve slightly depends on the
resistivity induced by various point defects (ΔW ) to the perfect crystal

Fig. 7. Anti-site point defect effects on the thermal conductivity of β -SiC. Thermal conductivity) of β -SiC as functions of the concentration of (a) CSi and (b) SiC .
Quantum correction is considered.

304
Y. Mao et al. Computational Materials Science 152 (2018) 300–307

Fig. 8. Interstitial effects on the thermal conductivity of β -SiC are studied and five types of interstitials are considered in this work. The thermal conductivity as
functions of the concentration of certain interstitial point defects at (a) 658 K and (b) 1036 K. Quantum correction is considered.

Table 1 Table 2
The thermal conductivities (κ ) calculated with The calculated thermal conductivity as a function of the velocity exchange rate
different super cells. The lengths of the super cells at the corrected temperature of T = 1036 K. f represents the velocity exchange
are fixed which are all 108.98 nm and the cross rate. J represents the corresponding heat flux generated by velocity exchange.
section areas (A) of the super cells are listed in the The thermal conductivity data is corrected by quantum correction.
table below. The corrected temperature in the
f (MD steps/time) J (× 1011 W/m2 ) κ (W/m·K )
simulation is all T = 1036 K. The results of cor-
rected thermal conductivity are compared. 50 7.84 91.36
80 5.50 90.91
A (nm2) κ (W/m·K)
100 4.51 84.11
1.74 × 1.74 86.74 160 3.14 83.56
2.62 × 2.62 82.43 200 2.59 82.94
3.49 × 3.49 78.31 500 1.14 90.24
4.36 × 4.36 80.24

— In the eight types of defects studied in this work, SiTC affects the
thermal conductivity the most seriously. The thermal conductivity
decreases with the increase of its concentration the most rapidly. On
the contrary, CTSi affects the thermal conductivity the least.
— The antisite and interstitial of Si atoms (SiC , SiTSi , SiTC ) decrease
the thermal conductivity more seriously than the counterpart anti-
site and interstitial of C atoms (CSi, CTSi, CTC ) do.
— The impact factor of SiC is close to that of SiTSi . As a result, the
effects from these two defects on the thermal conductivity are si-
milar. Analogously, the impact factor of C i−C〈100〉 is close to that of
VC .
— The temperature does not affect the impact factor very much. For
most of the defects, the impact factors at low temperature are
slightly higher than that at high temperature.

Our temperature effects on the thermal conductivity agree with the


works of Crocombette et al. [14] and Li et al. [10]. Because Cro-
combette’s added a fictitious external force to induce a heat flux
without a temperature gradient, this different simulation method
Fig. 9. The temperature profiles of defect-free super cells in z direction at a MD causes their impact factors to be smaller than ours. On the other hand,
temperature of 1098 K. The size of all the samples is 1.74 × 1.74 × 108.98 nm3 .
our results are more close to that of Li’s [10]. Especially, Li et al. find
The velocity exchange rates of f are 100, 200, and 500 respectively.
that SiTC decreases the thermal conductivity of SiC the most seriously.
Crocombette et al. argue that the impact factors of vacancies and in-
simulation temperature. terstitials are larger than that of antisites. Our results show that the
The additional thermal resistivity depends on the type of the point impact factor of antisite SiC is rather large which agrees with Li’s results
defects and the temperature. The slope of the curve of ΔW vs n is de- [10].
fined as the “impact factor” of the thermal resistivity to evaluate the
defect effects on the thermal resistivity. The impact factors of the eight
types of the point defects are demonstrated in Fig. 12. In this figure, the 5. Conclusions
higher the impact factor is, the more increase of the additional thermal
resistivity is, in other words the more seriously the point defect affects Reverse non-equilibrium molecular dynamics is applied to simulate
the thermal conductivity. For the additional thermal resisitivity of β -SiC the thermal conductivity of β -SiC and MEAM potential is adopted to
in Fig. 12, there are four observations. describe the inter-atomic interaction. Cell size extrapolation and
quantum correction are used to correct the thermal conductivity of bulk

305
Y. Mao et al. Computational Materials Science 152 (2018) 300–307

Fig. 10. The additional thermal resistivity as functions of the concetration of vacancy. (a) The additional thermal resistivity as functions of n v at corrected tem-
peratures of 658 K and 1036 K. (b) The relative additional thermal resistivity as a funciton of n v , which is compared with Samolyuk’s [11] results with EMD method at
1000 K.

Fig. 11. The additional thermal resistivity as functions of the concentration of the point defects of (a) C i−C〈100〉, (b) SiTC at the corrected temperatures of 658 K and
1036 K.

thermal conductivity the least.

Acknowledgments

This work is supported by International Science and Technology


Cooperation Program of China under Contract No. 2015DFA50510. It is
also supported by Project of State Power Investment Corporation under
Contract No. B-ZY04-201701.
The raw/processed data required to reproduce these findings cannot
be shared at this time due to legal or ethical reasons.

References

Fig. 12. The impact factors of eight point defects at the corrected temperatures [1] R.J. Price, Nucl. Technol. 35 (1977) 320–336.
of 658 K and 1036 K. [2] M. Ben-Belgacem, V. Richet, K. Terrani, Y. Katoh, L. Snead, J. Nucl. Mater. 447 (1-
3) (2014) 125–142.
[3] Y. Katoh, L.L. Snead Jr., C.H. Henager, T. Nozawa, T. Hinoki, A. Ivekovic, S. Novak,
β -SiC. S.G. De Vicente, J. Nucl. Mater. 455 (1–3) (2014) 387–397.
[4] L.L. Snead, T. Nozawa, Y. Katoh, T.-S. Byun, S. Kondo, D.A. Petti, J. Nucl. Mater.
The point defect effects on the thermal conductivity of β -SiC are
371 (1-3) (2007) 329–377.
investigated. Especially, these defects are: VC , [5] Y. Li, W. Xiao, H. Li, J. Nucl. Mater. 480 (2016) 75–79.
CSi, SiC , CTSi, CTC , C i−C〈100〉, SiTSi , and SiTC . For all of the point defects [6] M. Rohde, J. Nucl. Mater. 182 (1991) 87–92.
[7] R. Price, J. Nucl. Mater. 46 (3) (1973) 268–272.
studied in this work, the thermal conductivity of β -SiC decreases with
[8] L. Malakkal, B. Szpunar, R.K. Siripurapu, J.A. Szpunar, Comput. Mater. Sci. 128
the increase of the concentration of certain point defects. The thermal (Supplement C) (2017) 249–256.
conductivity drops rapidly at low concentration and gradually reaches [9] T. Kawamura, D. Hori, Y. Kangawa, K. Kakimoto, M. Yoshimura, Y. Mori, Jpn. J.
certain value at high concentration condition. At high concentration Appl. Phys. 47 (12R) (2008) 8898.
[10] J. Li, L. Porter, S. Yip, J. Nucl. Mater. 255 (23) (1998) 139–152.
condition, the thermal conductivity is not affected by the temperature [11] G. Samolyuk, S. Golubov, Y. Osetsky, R. Stoller, J. Nucl. Mater. 418 (1-3) (2011)
change. 174–181.
In general, the additional thermal resistivity increases proportion- [12] N. Papanikolaou, J. Phys.: Condens. Matter 20 (13) (2008) 135201.
[13] J.-P. Crocombette, G. Dumazer, N.Q. Hoang, F. Gao, W.J. Weber, J. Appl. Phys. 101
ally with the increase of the defect concentration for all types of point (2) (2007) 023527.
defects we studied. Among the defects, the interstitial SiTC decreases the [14] J.-P. Crocombette, L. Proville, Appl. Phys. Lett. 98 (19) (2011) 191905.
thermal conductivity the most, and the interstitial CTSi affects the [15] L. Shi, D. Yao, G. Zhang, B. Li, Appl. Phys. Lett. 95 (6) (2009) 063102.

306
Y. Mao et al. Computational Materials Science 152 (2018) 300–307

[16] D.A. Broido, M. Malorny, G. Birner, N. Mingo, D.A. Stewart, Appl. Phys. Lett. 91 [27] F. Müller-Plathe, J. Chem. Phys. 106 (14) (1997) 6082–6085.
(23) (2007) 231922. [28] Y. Lee, S. Lee, G.S. Hwang, Phys. Rev. B 83 (12) (2011) 125202.
[17] L. Lindsay, D.A. Broido, T.L. Reinecke, Phys. Rev. B 87 (16) (2013) 1948–1954. [29] P.K. Schelling, S.R. Phillpot, P. Keblinski, Phys. Rev. B 65 (2002) 144306.
[18] S.G. Volz, G. Chen, Phys. Rev. B61 (2000) 2651–2656. [30] C. Oligschleger, J.C. Schön, Phys. Rev. B 59 (1999) 4125–4133.
[19] X.W. Zhou, S. Aubry, R.E. Jones, A. Greenstein, P.K. Schelling, Phys. Rev. B79 [31] S. Plimpton, J. Comput. Phys. 117 (1) (1995) 1–19.
(2009) 115201. [32] M.I. Baskes, Phys. Rev. B 46 (1992) 2727–2742.
[20] S. Berber, Y.-K. Kwon, D. Tománek, Phys. Rev. Lett. 84 (2000) 4613–4616. [33] L.J. Porter, J. Li, S. Yip, J. Nucl. Mater. 246 (1) (1997) 53–59.
[21] J.A. Thomas, R.M. Iutzi, A.J.H. McGaughey, Phys. Rev. B 81 (2010) 045413. [34] S. chuang Wang, X. gang Liang, X. hua Xu, T. Ohara, J. Appl. Phys. 105 (1) (2009)
[22] S. chuang Wang, X. gang Liang, X. hua Xu, T. Ohara, J. Appl. Phys. 105 (1) (2009) 014316.
014316. [35] J.H. Han, K.S. Rheem, J. Nucl. Mater. 217 (1-2) (1994) 197–199.
[23] X. Yang, A.C. To, R. Tian, Nanotechnology 21 (15) (2010) 155704. [36] Y. Yang, X. Liu, J. Yang, Mol. Simul. 34 (1) (2008) 51–56.
[24] D.J. Evans, Phys. Lett. A 91 (9) (1982) 457–460. [37] Q. TANG, Mol. Phys. 102 (18) (2004) 1959–1964.
[25] T. Ikeshoji, B. Hafskjold, Mol. Phys. 81 (2) (1994) 251–261. [38] T. Wang, G.K.H. Madsen, A. Hartmaier, Modell. Simul. Mater. Sci. Eng. 22 (3)
[26] W.G. Hoover, Annu. Rev. Phys. Chem. 34 (1) (1983) 103–127. (2014) 035011.

307

You might also like