Computational Materials Science 186 (2021) 110074

Contents lists available at ScienceDirect

Computational Materials Science

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

TransOpt. A code to solve electrical transport properties of semiconductors

in constant electron–phonon coupling approximation
Xin Li a, Zhou Zhang a, Jinyang Xi a, David J. Singh b, Ye Sheng a, Jiong Yang a, *,
Wenqing Zhang c, d
a
Materials Genome Institute, Shanghai University, 99 Shangda Road, Shanghai 200444, China
b
Department of Physics and Astronomy and Department of Chemistry, University of Missouri, Columbia, MO 65211, USA
c
Department of Physics and Shenzhen Institute for Quantum Science & Engineering, Southern University of Science and Technology, Shenzhen, Guangdong 518055,
China
d
Guangdong Provincial Key Laboratory of Computational Science and Materials Design, and Shenzhen Key Laboratory of Advanced Quantum Functional Materials and
Devices, Southern University of Science and Technology, Shenzhen, Guangdong 518055, China

A R T I C L E I N F O A B S T R A C T

Keywords: A Fortran code, TransOpt (formerly named Transoptic) is presented. This code calculates electrical transport
Electrical transport coefficients of semiconductor materials based on Boltzmann transport theory in the relaxation time approach
Electronic relaxation time with the recently developed constant electron–phonon coupling approximation. The code interfaces with the
Electron group velocity
Vienna ab initio Simulation Package (VASP). The band structure related scattering phase space is calculated and
Electronic fitness function
used in determining the effective carrier relaxation time. The electronic structure part of the relaxation time is
treated explicitly with using the detailed first-principles, while the electron–phonon coupling matrix (EPCM) part
is treated as a constant. This constant EPCM can be parameterized using the deformation potential method in
semiconductors, and the absolute value of electrical conductivity can thus be obtained. The code can also use full
electron–phonon data from Quantum Espresso. Transport properties, including the electrical conductivity, the
Seebeck coefficient, electronic thermal conductivity, Lorenz number, power factor, and electronic fitness func­
tion, can be calculated by TransOpt. The electron group velocities vnk as a function of the position in the Brillouin
zone, k, can be determined in two different ways: 1) The momentum matrix method, which naturally avoids the
“band crossing” problem, and yields better convergence with the number of first-principles k points and 2) the
gradient method, where vnk is solved by the gradient in reciprocal space of the dispersion relation, which avoids
the need to compute momentum matrix. Several examples are presented to highlight the major features of
TransOpt.

[ ]
1. Introduction 1∑ ∂fμ (εnk , T)
σ αβ (μ, T) = vnkα vnkβ τnk − (1)
V nk ∂εnk
First-principles based predictions of electrical transport property [ ]
such as electrical conductivity σ and Seebeck coefficient (thermopower) 1 ∑ ∂f (ε , T)
Sαβ (μ, T) = σαβ (μ, T)− 1 vnkα vnkβ τnk (μ − εnk ) − μ nk (2)
S have become an important tool in condensed matter physics and eTV nk
∂εnk
materials science, for example as related to thermoelectric materials
[1–6]. It is generally valid when the conduction electron’s mean free Here vnk is electron group velocity corresponding to band index n
path is larger than the unit cell dimension of a crystal, in which case and reciprocal coordinate k, and τnk is electronic relaxation time. T, μ, V,
electrical transport can be regarded as the transport of quasi-particles. fμ, and e are respectively absolute temperature, Fermi level, volume of
Within this approach, the transport properties, such as electrical con­ unit cell, the Fermi-Dirac distribution, and electron charge. εnk is band
ductivity σ and Seebeck coefficient S, can be calculated as following energy.
[4,7], BoltzTraP, which is a widely used program for this purpose [4,5],
calculates transport coefficients based on Boltzmann transport theory.
BoltzTraP can interface with many packages such as Vienna ab initio

* Corresponding author.

https://doi.org/10.1016/j.commatsci.2020.110074
Received 25 July 2020; Received in revised form 15 September 2020; Accepted 17 September 2020
Available online 28 September 2020
0927-0256/© 2020 Elsevier B.V. All rights reserved.
X. Li et al. Computational Materials Science 186 (2021) 110074

Simulation Package (VASP) [8], Quantum Espresso (QE) [9–11], and interfaced with VASP, to study the electrical transport properties,
WIEN2k [5,12]. Here, we mainly focus on its application of electrical including Seebeck coefficient, electrical conductivity, electronic thermal
transport properties in VASP. BoltzTraP uses the constant relaxation conductivity, Lorenz number, power factor, and fitness function [24] is
time approximation (CRTA) as the default solution, which treats τnk as a reported. TransOpt can use both the momentum matrix method and
constant. This then assumes that the energy dependence of the relaxa­ gradient method to determine the electron group velocities. More
tion time is negligible compared to that of the band structure in Eq. (1) importantly, the constant electron–phonon coupling approximation
and (2). With this approximation, the S is independent of the relaxation (CEPCA) is proposed and applied in TransOpt by default. As will be
time τnk. In many cases, CRTA gives good estimations of S [13–15]. shown later, the treatment of τnk improves the accuracy of electrical
However, the conductivity, σ and other coefficient, such as the electrical transport properties.
contribution to the thermal conductivity, have to be calculated with The paper is organized as follows. In Section 2, we summarize the
respect to the relaxation time under the CRTA [16–19], introducing most relevant methods used in TransOpt. Section 3 presents several case
uncertainty and limiting the predictions that can be made. On the other studies by TransOpt to highlight the major features of the code. In
hand, the simplicity of the approximation, and the fact all the needed Section 4, the relevant input parameters are described. Section 5 is a
input is readily obtained from first-principles makes the approach very summary.
amenable to use in high-throughput calculations, where it has been
widely used [14,20–23]. In addition to BoltzTraP, BoltzWann [6] also 2. Code implementation
uses the CRTA. The interpolation methods of BoltzTraP and BoltzWann
are different. BoltzTraP uses a smooth interpolation of the electronic 2.1. Constant electron–phonon coupling approximation (CEPCA)
bands based on symmetrized planewaves. This suppresses oscillations of
the transport curves, allows for arbitrarily dense, but necessarily uni­ Here we only consider the scattering of electron–phonon coupling.
form, k-grids and allows analytic calculation of the vnk from the inter­ The complex scattering rate under electron–phonon interaction can be
polation coefficients. BoltzWann implements band interpolation using expressed by,
Wannier functions.
1 2π∑ λ 2 {[ ] ( )
The accuracy of σ under the CRTA is limited by the undetermined τnk, = |g | fmk’ + nqλ δ εmk’ − εnk − ℏωqλ δk+q,k’
which depends on the scattering mechanisms at play in a given material. τnk ℏ mk’λ mk’,nk
. (6)
Based on the Matthiessen rule, τnk can be written as, [ ] ( ) }
+ 1 + nqλ − fmk’ δ εmk’ − εnk + ℏωqλ δk− q,k’
1 1 1 1 1 1
= + + + + ⋯, (3)
τnk τa τnpo τpo τi τd where |gmk’,nk
λ
| is the EPCM. The two δ-functions in curly braces describe
the absorption and emission of a phonon ωqλ. nqλ denotes the phonon
where τ−a 1, τ−npo
1
, τ−po1, τ−i 1, and τ−d 1 respectively represent the scattering number under the Bose-Einstein distribution. Since the energy of a
rates due to acoustic phonons, non-polar optical phonons, polar optical phonon is far less than that of an electron, ħωqλ can be neglected under
phonons, impurities, and defects. Scattering due to impurity and defect the elastic scattering approximation. Meanwhile, at high temperatures,
scattering is generally temperature independent and proportional to the 2nqλ + 1 can be reduced to 2nqλ. Thus the τnk can be simplified as,
impurity and defect densities. Thus, at high temperatures, the influence
of temperature dependent phonon scattering mechanisms becomes 1 2π ∑ λ
(7)
2
= |g | ⋅2nqλ δ(εnk − εmk’ )δk+q,k’
dominant. As shown in Eqs. 6–8 in section 2, the scattering of electro­ τnk ℏ mk’λ mk’,nk
n–phonon coupling is generally determined by the electron–phonon
Eq. (7) shows that τnk is mainly determined by the EPCM, the phonon
coupling matrix (EPCM), electronic structure, and the number of pho­
number, and the scattering phase space. Here, CEPCA takes the EPCM as
nons nqλ (which is temperature dependent). Apparently, treating τnk as a
an effective constant, independent to band energy, and explicitly cal­
constant might cause uncertainty for σ and power factor when
culates the band-related scattering phase space. For the nqλ, CEPCA as­
comparing the results among different materials.
sumes the high temperature limit, thus nqλ is proportional to T. τnk can
Another problem of commonly used approaches arises with large
be rewritten as,
unit cells, as might be used for modeling alloys, and complex band
structures. In this case, there are many bands and associated band CT ∑
τ−nk1 = δ(εnk − εmk’ ) (8)
crossings and anti-crossings. These can be prohibitive particularly if V ’
mk
hybrid functionals are employed, as may be needed in some materials.
The so-called “band crossing” problem is because the electron group where C is the undetermined constant relating to the EPCM. Under
velocity evaluation is defined as the gradient in reciprocal space of the CEPCA, τnk is related to band energy, and Seebeck coefficient S is no
dispersion relation, longer independent of τnk (though it is independent of the constant C).
1 For semiconductors, C can be parameterized under the long wavelength
vnk = ∇k εnk , (4) limit for acoustic phonons, i.e., the deformation potential (DP) method
ℏ
[25]. Thus τnk can be expressed by using DP and Young’s modulus,
where ħ is the reduced Planck constant. The gradient method requires
2πkB TE2def ∑
analytic interpolations or information about adjacent k-points for nu­ τ−nk1 = δ(εnk − εmk’ ). (9)
merical differentiation. One solution addressing the band crossing VℏG ’
mk
problem is the momentum matrix method [12] (BoltzTraP2 interfacing
Here Edef is the DP of the band edge state, and G is the Young’s
with WIEN2k also implements the momentum matrix method),
modulus. The values of the two parameters in CEPCA can be written in
1 the input file finale.input (see Table 1 and Table 2 of below).
vnk = < ψ nk |̂
p |ψ nk >, (5)
me Although values of EPCM are energy-dependent, due to the fact that

where me, ̂
p , and ψ nk are mass of free electron, momentum operator, and Table 1
wave function, respectively. This method only needs the information Input parameters of finale.input.
from the k-point in question, which naturally avoids the band crossing
E-fermi Estart dE NE T CBM Nsis NELE
problem.
Here, a Fortran code named TransOpt (formerly named Transoptic), DP G Sigma

2
X. Li et al. Computational Materials Science 186 (2021) 110074

Table 2
The meaning of the input parameters.
d(ε − εref )
DP = (10)
d(lnV)
name Unit meaning Example

E-fermi eV Zero point of the energy 8.05

Here, ε is the energy of valence band maximum (VBM) or conduction
Estart eV Starting energy point w.r.t. E-fermi − 0.50 band minimum (CBM), εref is the reference level, and V is the volume of
dE eV Energy step 0.005 the calculated cell.
NE Numbers Number of energy steps 2001
T K Absolutely temperature 750
CBM Number Conduction band minimum index 0 3. Test cases
Nsis eV Scissors energy below CBM 0
NELE Number Total number of electrons 96 3.1. CoSi: Abnormal temperature dependence of Seebeck coefficient
DP eV Deformation potential 4.40
G GPa Young’s modulus 95.00
Sigma eV Gaussian smearing factor 0.50
CEPCA considers the energy dependent τnk, which is more accurate
than CRTA, especially for systems with strong energy dependent scat­
tering phase space. Take CoSi as an example, previous study has re­
−
∂fμ (εnk ,T)
is close to be a δ function and that bands a few kBT away from ported that it has an abnormal temperature dependence of Seebeck
∂εnk
the Fermi level have little contribution to electrical transport properties coefficient. This abnormal situation is not accurately predicted by the
(see Eq. (1) and (2)), CEPCA can still be applied to semiconductor ma­ CRTA method, but it can be captured by DOS-1 method (DOS is the
terials for their electrical transport properties. In order to obtain accu­ density of states) [29]. In this work, the temperature dependence of
rate values for electrical transport, the correct EPCM elements only for Seebeck coefficient for CoSi is calculated by TransOpt, under CEPCA and
∂fμ (εnk ,T) the momentum matrix method, in order to predict the abnormal Seebeck
the few contributing bands near the Fermi levels (large − ∂εnk ) are
coefficient.
necessary. It is reasonable to assume a constant EPCM element for these For CoSi, the calculated S from 100 K to 700 K is shown in Fig. 1c.
bands. The results under CEPCA are in good agreement with the experimental
Furthermore, if given the constant C, only the band structure infor­ data [30,31], which have n-type negative S. For CRTA, however, the S is
mation is required under CEPCA. The computational cost is far less than positive at room temperature, which is qualitatively inconsistent with
that of full electron–phonon coupling calculation, and comparable to the experimental data. This is due to the unique band structure of CoSi.
CRTA. The commonly used packages for first-principles high- Below the Fermi level (Fig. 1b), bands are usually heavy, while above
throughput screening electrical transport properties is based on the the Fermi level, there are massless (Dirac-like) bands [29]. The heavy
CRTA method. CEPCA, with the relaxation time calculated to some bands induce larger DOS and scattering phase space, which greatly re­
extent under the same computational cost, could serve as a more duces the relaxation times of carriers in the bands below the Fermi level.
advanced alternative for the calculations of electrical transport prop­ Thus, the Seebeck coefficient is strongly influenced by the contribution
erties in high-throughput applications. from the massless bands with long relaxation times above the Fermi
Actually, there are many simplified methods to deal with the EPCM, level, causing the negative sign. As Xia et al. mentioned, DOS-1 method
such as the electron phonon averaged (EPA) method, which replacing considers the influence of DOS change on the electron–phonon coupling
the full EPCM by the average of weak energy dependent EPCM [26]. EPA phase space in evaluating energy dependent relaxation time, and gives
approximation is simple and computationally faster by orders of good agreement of Seebeck coefficient between prediction and experi­
magnitude than the direct Brillouin zone sampling and the EPW inter­ ment.
∑
mk’ δ(εnk − εmk’ ) term of CEPCA also takes into account the
polation scheme. However, it also needs to calculate EPCM on a coarse changing of DOS and scattering phase space. Thus on the prediction on
grid in the Brillouin zone, which is more time-consuming than simple electrical transport properties of systems with strong energy dependent
deformation potential method, as implemented in our CEPCA. scattering phase space, CEPCA is more accuracy than CRTA.
TransOpt also interfaces with QE with the consideration of full
electron–phonon coupling. A modified QE has been composed, with the
electron group velocities and relaxation times under Eq. (6) in the form 3.2. NbFeSb: Deformation potential method for τnk evaluation
of TransOpt.
For σ of semiconductors, TransOpt can also calculate the τnk by
plugging in the DP and Young’s modulus, as shown in Eq. (9). Take
2.2. Electron group velocity based on the momentum matrix method
NbFeSb as an example. The calculated DPs are 2.11 eV for p-type and
4.41 eV for n-type, respectively. The reference levels are the 1 s core
To deal with the band crossing, momentum matrix method is
levels of Fe for p-type, and of Nb for n-type, due to the fact that Fe and
implemented. Electron group velocity file GROUPVEC, which contains
Nb contribute the most to VBM and CBM. The calculated Young’s
all the information of vnk, is printed out by the subroutine getnabij.F
modulus is 209.70 GPa. Electrical transport properties are calculated by
(based on the VASP subroutine optic.F. The former name Transoptic is
the momentum matrix method. Fig. 2b shows the calculated σs of
because of this), which is implemented in VASP. The momentum matrix
NbFeSb with carrier concentration at 3.25 × 1020 cm− 3 (n-type, black
method requires the setting NPAR = 1 in INCAR file. Alternatively,
line) and 19.76 × 1020 cm− 3 (p-type, red line). The carrier concentration
TransOpt code also provides the gradient method.
values are the same as the References [32] and [33]. The calculated σ s
are comparable to the experimental values, which demonstrate the ac­
2.3. Details of first-principles calculations curacy of CEPCA combining with the DP method in reproducing the
experimental σ s.
In this work, all the first-principles calculations are carried out using In general, section 3.1–3.2 give the treatments of τnks in TransOpt,
VASP with the projector augmented wave (PAW) method [1,27]. The and the influences of electrical transport properties for semiconductors.
Perdew-Burke-Ernzerhof (PBE) type generalized gradient approxima­ CEPCA considers the contributions from electronic structures on τnks,
tion (GGA) as the exchange–correlation function is used [28]. The which further affects the temperature dependence of S and σ . For
calculated k-meshes are 61 × 61 × 61 for CoSi, 58 × 58 × 58 for NbFeSb semiconductors, when given the DP and Young’s modulus, the absolute
and VFeSb, 25 × 25 × 25 for Co4Sb12 with primitive cell (16 atoms) and values of τnks and σs can be obtained. This is useful for high-throughput
11 × 11 × 11 for Co4Sb12 with 2 × 2 × 2 supercell (128 atoms). The studies. For example, the study of chalcogenides with diamond-like
deformation potentials are calculated for semiconductors by using, structures shows that chalcogenides with same anions possess similar

3
X. Li et al. Computational Materials Science 186 (2021) 110074

Fig. 1. The predicted Seebeck coefficient of CoSi. (a) Bravais crystal structure and (b) electronic structure of CoSi. (c) The Seebeck coefficient of CoSi calculated by
TransOpt. The experimental data and the calculation based on the DOS-1 method are also shown for comparison [29–31].

Fig. 2. The predicted electrical conductivity of NbFeSb. (a) Bravais crystal structure of NbFeSb. (b) The electrical conductivity of NbFeSb calculated by TransOpt.
The experimental data are also shown for comparison [32,33].

experimental σ ~ hole concentration. More generally, this allows one to 3.3. Co4Sb12: Accurate vnk based on the momentum matrix method
include the effects of bonding and the details of electronic structure on
scattering, which can then improve the accuracy of the calculations The complexity of supercell band structure causes the band number
upon which the high-throughput study is based. In this scenario, it is mislabeled in first-principles calculations, i.e., the band crossing prob­
reasonable to adopt the same DP and Young’s modulus for chalcogen­ lem, more severe. For the gradient method, the band crossing limits the
ides with same anions, and the absolute values of power factors for all accuracy of group velocity and thus electrical transport properties (Eq.
the compounds can thus be calculated [34]. Furthermore, for σ s of (4)) [36]. In order to solve this problem, the momentum matrix method
metals, the C constants can also be fitted from the experimental σs at (Eq. (5)) is implemented in TransOpt to calculate the electrical transport
room or lower temperatures and applied for high temperature transport properties. In this work, the momentum matrix method for Co4Sb12 is
properties, which is sometimes useful for those metals whose high tested both in primitive cell with 16 atoms, and 2 × 2 × 2 supercell with
temperature properties are difficult to measure, e.g., UAlx for nuclear 128 atoms. The parameters, including the DP and Young’s modulus, are
application [35]. taken from reference [37].
Fig. 3 shows the transport properties of the supercell and primitive
cell. Under the gradient method, the absolute values of S and σ of the

4
X. Li et al. Computational Materials Science 186 (2021) 110074

Fig. 3. Electrical transport properties calculated for the primitive cell and supercell of Co4Sb12. (a) Bravais crystal structure and (b) band structure of Co4Sb12. The
comparison of calculated primitive cell and supercell transport coefficients, (c) Seebeck coefficient and (d) electrical conductivity, for Co4Sb12 at 750 K as a function
of carrier concentration.

supercell are smaller than those of the primitive cell. The reductions of | of the momentum matrix method for materials with complex electronic
S| and σ in the supercell calculated by the gradient method is due to the structures. This feature is extremely useful for the evaluations of large
band crossing problem. On the other hand, the prediction of S and σ for cells with trace amounts of defects, such as partially filled Co4Sb12 [38].
the supercell under the momentum matrix method is perfectly align­
ment with that of the primitive cell, which demonstrates the feasibility

Fig. 4. Fitness functions calculated for the VFeSb and NbFeSb. Band structure of (a) VFeSb and (b) NbFeSb. The t function calculations of (c) VFeSb and (d) NbFeSb
at 300 K.

5
X. Li et al. Computational Materials Science 186 (2021) 110074

3.4. VFeSb and NbFeSb: Fitness function Table 3

The control.in file with tree lines.
The electronic fitness function (EFF) t=(σ /τ)S2/N2/3, can efficiently Parameter type Parameter Description Default
identify materials with complex band characters [39], where N is the
Electron velocity Logical (eg: TRUE) T means the gradient F
volumetric DOS. Good thermoelectric materials usually possess complex method method; F means the
electronic structures with a high t function. t function is implemented in momentum matrix method
TransOpt. Take Half-Heusler materials, VFeSb and NbFeSb as examples. The dimension of Three Logical (eg: Indicates the dimension of TTT
The t functions are calculated by TransOpt with the gradient method crystal TRUE TRUE calculated crystal
TRUE)
under CRTA in order to directly compare with the results from Boltz­ If print the full k Logical (eg: TRUE) Unfold the reduced Brillouin F
TraP. As shown in Fig. 4, the t functions of TransOpt and BoltzTraP show space zone
the same energy dependence and similar peak values, which demon­ parameters
strates the effectiveness of the t function module in TransOpt. The
implementation of the t function module can provide the possibility of
Declaration of Competing Interest
using TransOpt to quickly screen promising thermoelectric materials.

The authors declare that they have no known competing financial

4. Input parameters
interests or personal relationships that could have appeared to influence
the work reported in this paper.
There are two input files for TransOpt, finale.input (Table 1 and
Table 2) and control.in (Table 3), besides the standard VASP or QE
Acknowledgements
output. The finale.input file defines some parameters for calculating the
electrical transport properties, such as the range of energy, DP, and
This work was supported by the National Key Research and Devel­
Young’s modulus. The control.in file defines which method to be used to
opment Program of China (Nos. 2017YFB0701600, 2018YFB0703600,
calculate the electron group velocity.
and 2019YFA0704901), the Natural Science Foundation of China (Grant
For finale.input. 1) Nsis defines the energy modifications below the
Nos. 11674211, 51632005, and 51761135127), and the 111 Project
CBM (excluding the conduction band). For example, − 0.5 represents the
D16002. W.Z. also acknowledges the support from the Guangdong
energy values, for all energy bands below CBM, minus 0.5 eV. 2) When
Innovation Research Team Project (No. 2017ZT07C062), Guangdong
NELE is defined, the carrier concentrations will be calculated and
∫μ Provincial Key-Lab program (No. 2019B030301001), Shenzhen Munic­
ipal Key-Lab program (ZDSYS20190902092905285), and the Shenzhen
DOS− NELE
printed. The carrier concentrations are defined with − ∞ V , where
∫μ
DOS is the integration of density of states up to the Fermi level. 3) Pengcheng-Scholarship Program.
− ∞
The DP and Young’s modulus are used for the CEPCA. 4) Sigma is the
Gaussian smearing factor of the δ-function in Eq. (8). 5) Examples of References
parameters in Table 2 are those for Co4Sb12 of Fig. 4. For control.in file,
[1] Mads, Brandbyge, José-Luis, Mozos, Pablo, Ordejón, Jeremy, Taylor, Kurt, Stokbro,
there are three lines with logical parameters. The default electron group Density-functional method for nonequilibrium electron transport, Phys. Rev. B 65
velocity method would be the momentum matrix method if there is no (2002), 165401.
control.in file. [2] J.L. Mozos, P. Ordejón, M. Brandbyge, J. Taylor, K. Stokbro, Simulations of
quantum transport in nanoscale systems: application to atomic gold and silver
wires, Nanotechnology 13 (3) (2002) 346–351.
5. Conclusions [3] K. Stokbro, J. Taylor, M. Brandbyge, P. Ordejón, TranSIESTA: A Spice for
Molecular Electronics, Ann. N. Y. Acad. Sci. 1006 (1) (2006) 212–226.
[4] G.K. Madsen, D.J. Singh, BoltzTraP., A code for calculating band-structure
We present a Fortran code TransOpt to calculate electrical transport dependent quantities, Comput. Phys. Commun. 175 (1) (2006) 67–71.
properties along with several examples illustrating its use and capabil­ [5] G.K. Madsen, J. Carrete, M.J. Verstraete, BoltzTraP2, a program for interpolating
ities. TransOpt adopts the CEPCA for the determination of relaxation band structures and calculating semi-classical transport coefficients, Comput. Phys.
Commun. 231 (2018) 140–145.
times based on explicit electronic structures. It greatly improves the [6] G. Pizzi, D. Volja, B. Kozinsky, M. Fornari, N. Marzari, BoltzWann: A code for the
accuracy of electrical transport properties without significant extra cost evaluation of thermoelectric and electronic transport properties with a maximally-
compared to methods that neglect the dependence of the relaxation time localized Wannier functions basis, Comput. Phys. Commun. 185 (1) (2014)
422–429.
on electronic structure. Absolute values of relaxation times can be [7] J. Yang, L. Xi, W. Qiu, L. Wu, X. Shi, L. Chen, J. Yang, W. Zhang, C. Uher, D.
determined by using deformation potentials for semiconductors. In J. Singh, On the tuning of electrical and thermal transport in thermoelectrics: an
addition, both the gradient method and the momentum matrix method integrated theory–experiment perspective, NPJ Comput. Mater. 2 (2016) 15015.
[8] G. Kresse, J. Furthmuller, Efficient iterative schemes for ab initio total-energy
for electron group velocity evaluation are implemented in TransOpt. The
calculations using a plane-wave basis set, Phys. Rev. B 54 (16) (1996)
momentum matrix method efficiently solves the band crossing problem. 11169–11186.
TransOpt also calculates the electronic fitness function, which describes [9] P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli,
the band complexity in relation to thermoelectric behavior. G.L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. de Gironcoli, S. Fabris,
G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri,
L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello,
6. Data availability L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A.P. Seitsonen, A. Smogunov,
P. Umari, R.M. Wentzcovitch, QUANTUM ESPRESSO: a modular and open-source
software project for quantum simulations of materials, J. Phys.-Conden. Matter 21
The data that support the findings of this study are available from the (39) (2009), 395502.
corresponding author upon reasonable request. [10] J. Noffsinger, F. Giustino, B.D. Malone, C.-H. Park, S.G. Louie, M.L. Cohen, EPW: A
program for calculating the electron–phonon coupling using maximally localized
Wannier functions, Comput. Phys. Commun. 181 (12) (2010) 2140–2148.
CRediT authorship contribution statement [11] C. Liu, J. Yang, J. Xi, X. Ke, The origin of intrinsic charge transport for Dirac carbon
sheet materials: roles of acetylenic linkage and electron–phonon couplings,
Xin Li: Writing - original draft, Writing - review & editing. Zhou Nanoscale 11 (2019) 10828.
[12] T. Scheidemantel, C. Ambrosch-Draxl, T. Thonhauser, J.V. Badding, J.O. Sofo,
Zhang: Investigating on the application of TransOpt on metals. Jinyang Transport coefficients from first-principles calculations, Phys. Rev.B 68 (12)
Xi: Modifying Quantum Espresso to interface with TransOpt. David J. (2003), 125210.
Singh: Discussion and participation in preparing the manuscript. Ye [13] T. Fang, X. Li, C. Hu, Q. Zhang, J. Yang, W. Zhang, X. Zhao, D. Singh, T. Zhu,
Complex Band Structures and Lattice Dynamics of Bi2Te3-Based Compounds and
Sheng: Management of the Github page of TransOpt. Jiong Yang: Solid Solutions, Adv. Funct. Mater. 29 (28) (2019) 1900677.
Project administration. Wenqing Zhang: Supervision.

6
X. Li et al. Computational Materials Science 186 (2021) 110074

[14] W. Chen, J.-H. Pohls, G. Hautier, D. Broberg, S. Bajaj, U. Aydemir, Z.M. Gibbs, [26] G. Samsonidze, B. Kozinsky, Accelerated Screening of Thermoelectric Materials by
H. Zhu, M. Asta, G.J. Snyder, B. Meredig, M.A. White, K. Persson, A. Jain, First-Principles Computations of Electron-Phonon Scattering, Adv. Energy Mater. 8
Understanding thermoelectric properties from high-throughput calculations: (20) (2018).
trends, insights, and comparisons with experiment, J. Mater. Chem. C 4 (20) [27] P.E. Blöchl, Projector augmented-wave method, Phys. Rev. B 50 (24) (1994)
(2016) 4414–4426. 17953.
[15] S. Singh, D. Kumar, S. Pandey, Experimental and theoretical investigations of [28] J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made
thermoelectric properties of La0.82Ba0.18CoO3 compound in high temperature simple, Phys. Rev. Lett. 77 (18) (1996) 3865–3868.
region, Phys. Lett. A 381 (36) (2017) 3101–3106. [29] Y. Xia, J. Park, F. Zhou, V. Ozoliņš, High Thermoelectric Power Factor in
[16] A. Mubarak, F. Hamioud, S. Tariq, Influence of Pressure on Optical Transparency Intermetallic Co Si Arising from Energy Filtering of Electrons by Phonon
and High Electrical Conductivity in CoVSn Alloys: DFT Study, J Electron Mater 48 Scattering, Phys. Rev. Appl 11 (2) (2019), 024017.
(4) (2019) 2317–2328. [30] W. Ren, C. Li, L. Zhang, K. Ito, J. Wu, Effects of Ge and B substitution on
[17] A. Reshak, Thermoelectric properties for AA- and AB-stacking of a carbon nitride thermoelectric properties of CoSi, J. Alloy. Compd. 392 (1–2) (2005) 50–54.
polymorph (C3N4), RSC Adv. 4 (108) (2014) 63137–63142. [31] E. Skoug, C. Zhou, Y. Pei, D. Morelli, High thermoelectric power factor in alloys
[18] M. Yeganeh, F. Kafi, A. Boochani, Thermoelectric properties of InN graphene-like based on CoSi, Appl. Phys. Lett. 94 (2) (2009), 022115.
nanosheet: A first principle study, Superlattices Microstruct. 138 (2020), 106367. [32] J. Shen, L. Fan, C. Hu, T. Zhu, J. Xin, T. Fu, D. Zhao, X. Zhao, Enhanced
[19] M. Ne, M. Boujnah, A. Benyoussef, A. El Kenz, Electronic and Electrical thermoelectric performance in the n-type NbFeSb half-Heusler compound with
Conductivity of AB and AA-Stacked Bilayer Graphene with Tunable Layer heavy element Ir doping, Materials Today Physics 8 (2019) 62–70.
Separation, J. Supercond. Novel Magn. 30 (5) (2017) 1263–1267. [33] C. Fu, S. Bai, Y. Liu, Y. Tang, L. Chen, X. Zhao, T. Zhu, Realizing high figure of
[20] J. Carrete, N. Mingo, S.D. Wang, S. Curtarolo, Nanograined Half-Heusler merit in heavy-band p-type half-Heusler thermoelectric materials, Nat. Commun. 6
Semiconductors as Advanced Thermoelectrics: An Ab Initio High-Throughput (2015) 68144.
Statistical Study, Adv. Funct. Mater. 24 (47) (2014) 7427–7432. [34] L. Xi, S. Pan, X. Li, Y. Xu, J. Ni, X. Sun, J. Yang, J. Luo, J. Xi, W. Zhu, X. Li, D. Jiang,
[21] M. Miyata, T. Ozaki, T. Takeuchi, S. Nishino, M. Inukai, M. Koyano, High- R. Dronskowski, X. Shi, G. Snyder, W. Zhang, Discovery of High-Performance
Throughput Screening of Sulfide Thermoelectric Materials Using Electron Thermoelectric Chalcogenides through Reliable High-Throughput Material
Transport Calculations with OpenMX and BoltzTraP, J. Electron Mater 47 (6) Screening, J. Am. Chem. Soc. 140 (34) (2018) 10785–10793.
(2018) 3254–3259. [35] Z. Mei, Y.S. Kim, A.M. Yacout, J. Yang, X. Li, Y. Cao, First-principles study of
[22] F. Ricci, W. Chen, U. Aydemir, G. Snyder, G. Rignanese, A. Jain, G. Hautier, Data thermal conductivities of uranium aluminides, Materialia 4 (2018) 449–456.
Descriptor: An ab initio electronic transport database for inorganic materials, Sci. [36] J. Yang, G. Meisner, D. Morelli, C. Uher, Iron valence in skutterudites: Transport
Data 4 (2017), 170085. and magnetic properties of Co1-xFexSb3, Phys. Rev. B 63 (1) (2001), 014410.
[23] L. Bassman, P. Rajak, R. Kalia, A. Nakano, F. Sha, J. Sun, D. Singh, M. Aykol, [37] Z. Tu, X. Sun, X. Li, R. Li, L. Xi, J. Yang, Thermoelectric transport properties of Ni-,
P. Huck, K. Persson, P. Vashishta, Active learning for accelerated design of layered Pd-, and Pt- doped skutterudites with S-filling as charge compensation, AIP Adv. 9
materials, NPJ Comput. Mater. 4 (2018) 74. (4) (2019), 045325.
[24] https://github.com/yangjio4849/TransOpt.git or http://mgi.shu.edu.cn/ [38] J. Yang, L. Xi, W. Zhang, L. Chen, J. Yang, Electrical transport properties of filled
TransOpt.zip. CoSb 3 skutterudites: A theoretical study, J Electron Mater 38 (7) (2009)
[25] J. Bardeen, W. Shockley, Deformation potentials and mobilities in non-polar 1397–1401.
crystals, Phys. Rev. 80 (1) (1950) 72. [39] G. Xing, J. Sun, Y. Li, X. Fan, W. Zheng, D.J. Singh, Electronic fitness function for
screening semiconductors as thermoelectric materials, Physical Rev. Mater. 1 (6)
(2017), 065405.