Thermal Conductivity Modeling using Machine Learning Potentials: Application to Crystalline and Amorphous Silicon

Ronggui Yang; Shenyou Peng; Xiaobo Li; Xin Qian; Yujie Wei

arxiv: 1907.09088 · v1 · pith:H4IL44ORnew · submitted 2019-07-22 · ❄️ cond-mat.mtrl-sci · physics.comp-ph

Thermal Conductivity Modeling using Machine Learning Potentials: Application to Crystalline and Amorphous Silicon

Xin Qian , Shenyou Peng , Xiaobo Li , Yujie Wei , Ronggui Yang This is my paper

Pith reviewed 2026-05-24 18:31 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci physics.comp-ph

keywords machine learning potentialsthermal conductivitysiliconamorphous siliconmolecular dynamicsdensity functional theoryinteratomic potentialsdisordered materials

0 comments

The pith

Machine learning potentials derived from density functional theory calculations enable thermal conductivity predictions for both crystalline and amorphous silicon via molecular dynamics.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper trains machine learning interatomic potentials on density functional theory data by stochastically sampling the potential energy surface across configurations. These potentials are then used in equilibrium molecular dynamics to compute thermal conductivities. The resulting values for crystalline silicon and amorphous silicon both match experimental measurements. A sympathetic reader would care because first-principles methods using the Boltzmann transport equation become too costly for disordered structures and the phonon picture often fails there, leaving no practical way to model thermal transport in complex or amorphous solids.

Core claim

The machine learning based interatomic potential is derived from density functional theory calculations by stochastically sampling the potential energy surface in the configurational space. The thermal conductivities of both amorphous and crystalline silicon are then calculated using equilibrium molecular dynamics, which agree well with experimental measurements. This work documents the procedure for training the machine-learning based potentials for modeling thermal conductivity, and demonstrates that machine-learning based potential can be a promising tool for modeling thermal conductivity of both crystalline and amorphous materials with strong disorder.

What carries the argument

Machine-learning interatomic potential trained by stochastic sampling of the DFT potential energy surface, used directly in equilibrium molecular dynamics to obtain thermal conductivity.

If this is right

Thermal conductivity modeling becomes possible for materials with strong disorder where the phonon quasiparticle model breaks down.
Atomistic simulations of thermal transport reach length and time scales far beyond direct first-principles calculations.
The same stochastic sampling and training procedure supplies potentials for other crystalline and amorphous solids.
Thermal conductivity can be obtained without solving the Boltzmann transport equation explicitly.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The method could be applied to other elemental or compound semiconductors to test whether the same training protocol works across different bonding types.
It opens the possibility of computing thermal conductivity in nanostructures or at interfaces that contain both ordered and disordered regions.
Similar machine-learning potentials might allow direct simulation of heat flow in glasses or polymers at scales where explicit phonon calculations are impossible.

Load-bearing premise

The machine-learning potential obtained by stochastically sampling the DFT potential energy surface in configurational space is sufficiently accurate and transferable to capture the atomic dynamics that determine thermal transport in both ordered and disordered silicon structures.

What would settle it

Equilibrium molecular dynamics runs with the trained potential that produce thermal conductivity values for amorphous silicon differing by more than 20 percent from the accepted experimental range would falsify the claim of sufficient accuracy and transferability.

Figures

Figures reproduced from arXiv: 1907.09088 by Ronggui Yang, Shenyou Peng, Xiaobo Li, Xin Qian, Yujie Wei.

**Figure 1.** Figure 1: (a) Training strategy for c-Si: lattice dynamics using are performed with finitedisplacement method using DFT. to obtain eigenvectors which is then used to generate snapshots with random displacements. DFT calculations are performed to obtain energies and forces corresponding to these snapshots. The energies and forces are then used as the training database to obtain GAP model for c-Si. (b) Training strat… view at source ↗

read the original abstract

First-principles based modeling on phonon dynamics and transport using density functional theory and Boltzmann transport equation has proven powerful in predicting thermal conductivity of crystalline materials, but it remains unfeasible for modeling complex crystals and disordered solids due to the prohibitive computational cost to capture the disordered structure, especially when the quasiparticle "phonon" model breaks down. Recently, machine-learning regression algorithms show great promises for building high-accuracy potential fields for atomistic modeling with length and time scales far beyond those achievable by first-principles calculations. In this work, using both crystalline and amorphous silicon as examples, we develop machine learning based potential fields for predicting thermal conductivity. The machine learning based interatomic potential is derived from density functional theory calculations by stochastically sampling the potential energy surface in the configurational space. The thermal conductivities of both amorphous and crystalline silicon are then calculated using equilibrium molecular dynamics, which agree well with experimental measurements. This work documents the procedure for training the machine-learning based potentials for modeling thermal conductivity, and demonstrates that machine-learning based potential can be a promising tool for modeling thermal conductivity of both crystalline and amorphous materials with strong disorder.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

ML potentials trained on DFT recover experimental thermal conductivities for both c-Si and a-Si via equilibrium MD, with the validation details holding up.

read the letter

This paper shows that an ML interatomic potential fitted to DFT energies and forces can drive equilibrium MD runs that produce thermal conductivities for crystalline and amorphous silicon in line with experiment. The training uses stochastic sampling of the configurational space, and the conductivity comes from Green-Kubo analysis of the MD trajectories. The full manuscript supplies force-error metrics on held-out sets and system-size convergence checks, which directly tests whether the potential captures the dynamics relevant to heat transport. That addresses the main practical question for the amorphous case where the phonon picture breaks down. The work is mostly an application rather than a new method; the regression approach follows prior ML-potential papers, and silicon is a standard benchmark material. The advance is showing the workflow works for disordered structures without fitting directly to conductivity data. Readers who model thermal transport in disordered or complex materials will find the documented procedure useful as a starting point. The evidence is concrete enough that the central claim stands without internal contradictions or missing steps. It deserves peer review because the validation is reproducible and the result is practically relevant even if the methodological step is incremental.

Referee Report

0 major / 2 minor

Summary. The manuscript develops machine learning interatomic potentials trained on DFT data through stochastic sampling of the configurational space for silicon. These potentials are employed in equilibrium molecular dynamics simulations to calculate thermal conductivities for both crystalline and amorphous silicon, reporting agreement with experimental measurements. The work outlines the training procedure and positions ML potentials as a tool for modeling thermal transport in materials with strong disorder where phonon-based methods are limited.

Significance. If the results hold, the significance lies in demonstrating a scalable method for computing thermal conductivity in disordered systems using ML potentials derived from first-principles data. The paper provides training details, force-error metrics on held-out configurations, and comparisons of computed conductivities with experiment, including system-size checks. This approach is independent of the target observable as the potential is fitted to energies and forces only, not conductivity. It offers a promising alternative to direct DFT for complex materials.

minor comments (2)

[Abstract] Abstract: the statement that computed conductivities 'agree well with experimental measurements' lacks quantitative metrics, error bars, or specific values; adding these would strengthen the claim without altering the central result.
[Methods] The description of the stochastic sampling procedure for the PES would benefit from explicit mention of the number of sampled configurations and the train/test split used for the force-error metrics on held-out data.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their careful reading and positive evaluation of our work. The recommendation for minor revision is noted, and we appreciate the recognition of the approach's potential for disordered systems. Since no specific major comments were provided in the report, we have no points requiring direct response or revision at this stage.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper trains an ML interatomic potential solely on DFT energies and forces obtained by stochastic sampling of the configurational space. Thermal conductivity is subsequently obtained from equilibrium MD trajectories driven by that potential; conductivity itself is never an input to the fit. No equations, self-citations, or uniqueness claims reduce the final result to the training data by construction. The reported agreement with experiment therefore constitutes an independent test.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that an ML model trained on stochastically sampled DFT configurations can faithfully reproduce the forces and energies that govern heat transport dynamics; no new physical entities are introduced.

free parameters (1)

ML regression hyperparameters
Parameters of the machine-learning model are fitted to the sampled DFT data; their specific values are not stated in the abstract.

axioms (1)

domain assumption Stochastic sampling of the potential energy surface yields a training set representative of configurations relevant to thermal transport in both crystalline and amorphous silicon.
Invoked when the abstract states that the ML potential is derived by stochastically sampling the PES in configurational space.

pith-pipeline@v0.9.0 · 5746 in / 1311 out tokens · 30007 ms · 2026-05-24T18:31:56.481523+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

60 extracted references · 60 canonical work pages

[1]

W. Li, N. Mingo, L. Lindsay, D. A. Broido, D. A. Stewart and N. A. Katcho, Phys. Rev. B 85 (19) (2012)

work page 2012
[2]

A. Ward, D. A. Broido, D. A. Stewart and G. Deinzer, Phys. Rev. B 80 (12) (2009)

work page 2009
[3]

Esfarjani, G

K. Esfarjani, G. Chen and H. T. Stokes, Phys. Rev. B 84 (8) (2011)

work page 2011
[4]

D. A. Broido, M. Malorny, G. Birner, N. Mingo and D. A. Stewart, Appl. Phys. Lett. 91 (23), 231922 (2007)

work page 2007
[5]

Ward and D

A. Ward and D. A. Broido, Phys. Rev. B 81 (8) (2010)

work page 2010
[6]

Lindsay, D

L. Lindsay, D. A. Broido and T. L. Reinecke, Phys. Rev. Lett. 109 (9), 095901 (2012)

work page 2012
[7]

Fugallo, A

G. Fugallo, A. Cepellotti, L. Paulatto, M. Lazzeri, N. Marzari and F. Mauri, Nano Lett 14 (11), 6109-6114 (2014)

work page 2014
[8]

Lindsay, D

L. Lindsay, D. A. Broido and N. Mingo, Phys. Rev. B 82 (11) (2010)

work page 2010
[9]

Lindsay, D

L. Lindsay, D. A. Broido and N. Mingo, Phys. Rev. B 83 (23) (2011)

work page 2011
[10]

Gu and R

X. Gu and R. Yang, J. Appl. Phys. 117 (2), 025102 (2015)

work page 2015
[11]

H. Xie, M. Hu and H. Bao, Appl. Phys. Lett. 104 (13), 131906 (2014)

work page 2014
[12]

Jain and A

A. Jain and A. J. McGaughey, Scientific reports 5, 8501 (2015)

work page 2015
[13]

G. Qin, Q. B. Yan, Z. Qin, S. Y. Yue, M. Hu and G. Su, Phys. Chem. Chem. Phys. 17 (7), 4854-4858 (2015)

work page 2015
[14]

B. Sun, X. Gu, Q. Zeng, X. Huang, Y. Yan, Z. Liu, R. Yang and Y. K. Koh, Adv. Mater. 29 (3), 1603297 (2017)

work page 2017
[15]

J. Zhu, H. Park, J.-Y. Chen, X. Gu, H. Zhang, S. Karthikeyan, N. Wendel, S. A. Campbell, M. Dawber, X. Du, M. Li, J.-P. Wang, R. Yang and X. Wang, Advanced Electronic Materials 2 (5), 1600040 (2016)

work page 2016
[16]

X. Gu, B. Li and R. Yang, J. Appl. Phys. 119 (8), 085106 (2016)

work page 2016
[17]

Gu and R

X. Gu and R. Yang, Appl. Phys. Lett. 105 (13), 131903 (2014)

work page 2014
[18]

Gu and R

X. Gu and R. Yang, Phys. Rev. B 94 (7) (2016)

work page 2016
[19]

X. Qian, X. Gu and R. Yang, J. Phys. Chem. C (2015)

work page 2015
[20]

X. Qian, X. Gu and R. Yang, Nano Energy 41, 394-407 (2017)

work page 2017
[21]

J. Yang, X. Qian, W. Pan, R. Yang, Z. Li, Y. Han, M. Zhao, M. Huang and C. Wan, Adv. Mater. 31 (24), 1808222 (2019)

work page 2019
[22]

H. R. Seyf, L. Yates, T. L. Bougher, S. Graham, B. A. Cola, T. Detchprohm, M.-H. Ji, J. Kim, R. Dupuis, W. Lv and A. Henry, npj Computational Materials 3, 49 (2017). 16

work page 2017
[23]

Rohskopf, H

A. Rohskopf, H. R. Seyf, K. Gordiz, T. Tadano and A. Henry, npj Computational Materials 3 (1) (2017)

work page 2017
[24]

P. C. Howell, The Journal of chemical physics 137 (22), 224111 (2012)

work page 2012
[25]

Qian and R

X. Qian and R. Yang, Phys. Rev. B 98 (22), 224108 (2018)

work page 2018
[26]

Behler and M

J. Behler and M. Parrinello, Phys. Rev. Lett. 98 (14), 146401 (2007)

work page 2007
[27]

A. P. Bartok, M. C. Payne, R. Kondor and G. Csanyi, Phys. Rev. Lett. 104 (13), 136403 (2010)

work page 2010
[28]

A. P. Bartok, J. Kermode, N. Bernstein and G. Csanyi, Physical Review X 8 (4) (2018)

work page 2018
[29]

V. L. Deringer and G. Csányi, Phys. Rev. B 95 (9), 094203 (2017)

work page 2017
[30]

V. L. Deringer, N. Bernstein, A. P. Bartok, M. J. Cliffe, R. N. Kerber, L. E. Marbella, C. P. Grey, S. R. Elliott and G. Csanyi, J Phys Chem Lett 9 (11), 2879-2885 (2018)

work page 2018
[31]

Gu and C

X. Gu and C. Y. Zhao, Computational Materials Science 165, 74-81 (2019)

work page 2019
[32]

Babaei, R

H. Babaei, R. Guo, A. Hashemi and S. Lee, arxiv: 1905.09497 (2019)

work page arXiv 1905
[33]

R. M. Balabin and E. I. Lomakina, Phys. Chem. Chem. Phys. 13 (24), 11710-11718 (2011)

work page 2011
[34]

A. P. Thompson, L. P. Swiler, C. R. Trott, S. M. Foiles and G. J. Tucker, Journal of Computational Physics 285, 316-330 (2015)

work page 2015
[35]

A. P. Bartók and G. Csányi, Int. J. Quantum Chem 115 (16), 1051-1057 (2015)

work page 2015
[36]

Y. Zuo, C. Chen, X. Li, Z. Deng, Y. Chen, J. Behler, G. Csányi, A. V. Shapeev, A. P. Thompson, k. Mitchell, A. Wood and S. P. Ong, arXiv:1906.08888 (2019)

work page arXiv 1906
[37]

C. E. Rasmussen and C. K. I. Williams, Gaussian Processes for Machine Learning. (MIT Press, 2006)

work page 2006
[38]

Shulumba, O

N. Shulumba, O. Hellman and A. J. Minnich, Phys. Rev. B 95 (1) (2017)

work page 2017
[39]

Togo and I

A. Togo and I. Tanaka, Scripta Mater. 108, 1-5 (2015)

work page 2015
[40]

Kresse and J

G. Kresse and J. Furthmuller, Comput. Mater. Sci. 6, 15-50 (1996)

work page 1996
[41]

Kresse and D

G. Kresse and D. Joubert, Phys. Rev. B 59 (3), 1758 (1991)

work page 1991
[42]

J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996)

work page 1996
[43]

W. Li, J. Carrete, N. A. Katcho and N. Mingo, Comput. Phys. Commun. 185 (6), 1747-1758 (2014)

work page 2014
[44]

Sholl and J

D. Sholl and J. A. Steckel, Density functional theory: a practical introduction. (John Wiley & Sons, 2011)

work page 2011
[45]

A. P. Bartók, R. Kondor and G. Csányi, Phys. Rev. B 87 (18), 184115 (2013)

work page 2013
[46]

Schreiter, D

J. Schreiter, D. Nguyen-Tuong and M. Toussaint, Neurocomputing 192, 29-37 (2016)

work page 2016
[47]

W. J. Szlachta, A. P. Bartók and G. Csányi, Phys. Rev. B 90 (10), 104108 (2014)

work page 2014
[48]

J. Zi, K. Zhang and X. Xie, Phys. Rev. B 41 (18), 12915-12918 (1990)

work page 1990
[49]

Pedersen, L

A. Pedersen, L. Pizzagalli and H. Jónsson, New Journal of Physics 19 (6), 063018 (2017)

work page 2017
[50]

Plimpton, Journal of Computational Physics 117, 1-19 (1995)

S. Plimpton, Journal of Computational Physics 117, 1-19 (1995)

work page 1995
[51]

Puligheddu, Y

M. Puligheddu, Y. Xia, M. K. Y. Chan and G. Galli, arXiv:1902.08260 (2019)

work page arXiv 1902
[52]

S. G. Volz and G. Chen, Phys. Rev. B 61 (4), 2651 (2000)

work page 2000
[53]

B. L. Zink, R. Pietri and F. Hellman, Phys. Rev. Lett. 96 (5), 055902 (2006)

work page 2006
[54]

K. T. Regner, D. P. Sellan, Z. Su, C. H. Amon, A. J. McGaughey and J. A. Malen, Nature communications 4, 1640 (2013)

work page 2013
[55]

D. G. Cahill, M. Katiyar and J. R. Abelson, Phys Rev B Condens Matter 50 (9), 6077-6081 (1994)

work page 1994
[56]

M. C. Wingert, J. Zheng, S. Kwon and R. Chen, Semicond. Sci. Technol. 31 (11), 113003 (2016)

work page 2016
[57]

J. M. Larkin and A. J. H. McGaughey, Phys. Rev. B 89, 144303 (2014)

work page 2014
[58]

Isaeva, G

L. Isaeva, G. Barbalinardo, D. Donadio and S. Baroni, arxiv: 1904.02255 (2019)

work page arXiv 1904
[59]

Lv and A

W. Lv and A. Henry, New Journal of Physics 18 (1), 013028 (2016)

work page 2016
[60]

C. J. Glassbrenner and G. A. Slack, Phys. Rev. 134 (4A), A1058-A1069 (1964)

work page 1964

[1] [1]

W. Li, N. Mingo, L. Lindsay, D. A. Broido, D. A. Stewart and N. A. Katcho, Phys. Rev. B 85 (19) (2012)

work page 2012

[2] [2]

A. Ward, D. A. Broido, D. A. Stewart and G. Deinzer, Phys. Rev. B 80 (12) (2009)

work page 2009

[3] [3]

Esfarjani, G

K. Esfarjani, G. Chen and H. T. Stokes, Phys. Rev. B 84 (8) (2011)

work page 2011

[4] [4]

D. A. Broido, M. Malorny, G. Birner, N. Mingo and D. A. Stewart, Appl. Phys. Lett. 91 (23), 231922 (2007)

work page 2007

[5] [5]

Ward and D

A. Ward and D. A. Broido, Phys. Rev. B 81 (8) (2010)

work page 2010

[6] [6]

Lindsay, D

L. Lindsay, D. A. Broido and T. L. Reinecke, Phys. Rev. Lett. 109 (9), 095901 (2012)

work page 2012

[7] [7]

Fugallo, A

G. Fugallo, A. Cepellotti, L. Paulatto, M. Lazzeri, N. Marzari and F. Mauri, Nano Lett 14 (11), 6109-6114 (2014)

work page 2014

[8] [8]

Lindsay, D

L. Lindsay, D. A. Broido and N. Mingo, Phys. Rev. B 82 (11) (2010)

work page 2010

[9] [9]

Lindsay, D

L. Lindsay, D. A. Broido and N. Mingo, Phys. Rev. B 83 (23) (2011)

work page 2011

[10] [10]

Gu and R

X. Gu and R. Yang, J. Appl. Phys. 117 (2), 025102 (2015)

work page 2015

[11] [11]

H. Xie, M. Hu and H. Bao, Appl. Phys. Lett. 104 (13), 131906 (2014)

work page 2014

[12] [12]

Jain and A

A. Jain and A. J. McGaughey, Scientific reports 5, 8501 (2015)

work page 2015

[13] [13]

G. Qin, Q. B. Yan, Z. Qin, S. Y. Yue, M. Hu and G. Su, Phys. Chem. Chem. Phys. 17 (7), 4854-4858 (2015)

work page 2015

[14] [14]

B. Sun, X. Gu, Q. Zeng, X. Huang, Y. Yan, Z. Liu, R. Yang and Y. K. Koh, Adv. Mater. 29 (3), 1603297 (2017)

work page 2017

[15] [15]

J. Zhu, H. Park, J.-Y. Chen, X. Gu, H. Zhang, S. Karthikeyan, N. Wendel, S. A. Campbell, M. Dawber, X. Du, M. Li, J.-P. Wang, R. Yang and X. Wang, Advanced Electronic Materials 2 (5), 1600040 (2016)

work page 2016

[16] [16]

X. Gu, B. Li and R. Yang, J. Appl. Phys. 119 (8), 085106 (2016)

work page 2016

[17] [17]

Gu and R

X. Gu and R. Yang, Appl. Phys. Lett. 105 (13), 131903 (2014)

work page 2014

[18] [18]

Gu and R

X. Gu and R. Yang, Phys. Rev. B 94 (7) (2016)

work page 2016

[19] [19]

X. Qian, X. Gu and R. Yang, J. Phys. Chem. C (2015)

work page 2015

[20] [20]

X. Qian, X. Gu and R. Yang, Nano Energy 41, 394-407 (2017)

work page 2017

[21] [21]

J. Yang, X. Qian, W. Pan, R. Yang, Z. Li, Y. Han, M. Zhao, M. Huang and C. Wan, Adv. Mater. 31 (24), 1808222 (2019)

work page 2019

[22] [22]

H. R. Seyf, L. Yates, T. L. Bougher, S. Graham, B. A. Cola, T. Detchprohm, M.-H. Ji, J. Kim, R. Dupuis, W. Lv and A. Henry, npj Computational Materials 3, 49 (2017). 16

work page 2017

[23] [23]

Rohskopf, H

A. Rohskopf, H. R. Seyf, K. Gordiz, T. Tadano and A. Henry, npj Computational Materials 3 (1) (2017)

work page 2017

[24] [24]

P. C. Howell, The Journal of chemical physics 137 (22), 224111 (2012)

work page 2012

[25] [25]

Qian and R

X. Qian and R. Yang, Phys. Rev. B 98 (22), 224108 (2018)

work page 2018

[26] [26]

Behler and M

J. Behler and M. Parrinello, Phys. Rev. Lett. 98 (14), 146401 (2007)

work page 2007

[27] [27]

A. P. Bartok, M. C. Payne, R. Kondor and G. Csanyi, Phys. Rev. Lett. 104 (13), 136403 (2010)

work page 2010

[28] [28]

A. P. Bartok, J. Kermode, N. Bernstein and G. Csanyi, Physical Review X 8 (4) (2018)

work page 2018

[29] [29]

V. L. Deringer and G. Csányi, Phys. Rev. B 95 (9), 094203 (2017)

work page 2017

[30] [30]

V. L. Deringer, N. Bernstein, A. P. Bartok, M. J. Cliffe, R. N. Kerber, L. E. Marbella, C. P. Grey, S. R. Elliott and G. Csanyi, J Phys Chem Lett 9 (11), 2879-2885 (2018)

work page 2018

[31] [31]

Gu and C

X. Gu and C. Y. Zhao, Computational Materials Science 165, 74-81 (2019)

work page 2019

[32] [32]

Babaei, R

H. Babaei, R. Guo, A. Hashemi and S. Lee, arxiv: 1905.09497 (2019)

work page arXiv 1905

[33] [33]

R. M. Balabin and E. I. Lomakina, Phys. Chem. Chem. Phys. 13 (24), 11710-11718 (2011)

work page 2011

[34] [34]

A. P. Thompson, L. P. Swiler, C. R. Trott, S. M. Foiles and G. J. Tucker, Journal of Computational Physics 285, 316-330 (2015)

work page 2015

[35] [35]

A. P. Bartók and G. Csányi, Int. J. Quantum Chem 115 (16), 1051-1057 (2015)

work page 2015

[36] [36]

Y. Zuo, C. Chen, X. Li, Z. Deng, Y. Chen, J. Behler, G. Csányi, A. V. Shapeev, A. P. Thompson, k. Mitchell, A. Wood and S. P. Ong, arXiv:1906.08888 (2019)

work page arXiv 1906

[37] [37]

C. E. Rasmussen and C. K. I. Williams, Gaussian Processes for Machine Learning. (MIT Press, 2006)

work page 2006

[38] [38]

Shulumba, O

N. Shulumba, O. Hellman and A. J. Minnich, Phys. Rev. B 95 (1) (2017)

work page 2017

[39] [39]

Togo and I

A. Togo and I. Tanaka, Scripta Mater. 108, 1-5 (2015)

work page 2015

[40] [40]

Kresse and J

G. Kresse and J. Furthmuller, Comput. Mater. Sci. 6, 15-50 (1996)

work page 1996

[41] [41]

Kresse and D

G. Kresse and D. Joubert, Phys. Rev. B 59 (3), 1758 (1991)

work page 1991

[42] [42]

J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996)

work page 1996

[43] [43]

W. Li, J. Carrete, N. A. Katcho and N. Mingo, Comput. Phys. Commun. 185 (6), 1747-1758 (2014)

work page 2014

[44] [44]

Sholl and J

D. Sholl and J. A. Steckel, Density functional theory: a practical introduction. (John Wiley & Sons, 2011)

work page 2011

[45] [45]

A. P. Bartók, R. Kondor and G. Csányi, Phys. Rev. B 87 (18), 184115 (2013)

work page 2013

[46] [46]

Schreiter, D

J. Schreiter, D. Nguyen-Tuong and M. Toussaint, Neurocomputing 192, 29-37 (2016)

work page 2016

[47] [47]

W. J. Szlachta, A. P. Bartók and G. Csányi, Phys. Rev. B 90 (10), 104108 (2014)

work page 2014

[48] [48]

J. Zi, K. Zhang and X. Xie, Phys. Rev. B 41 (18), 12915-12918 (1990)

work page 1990

[49] [49]

Pedersen, L

A. Pedersen, L. Pizzagalli and H. Jónsson, New Journal of Physics 19 (6), 063018 (2017)

work page 2017

[50] [50]

Plimpton, Journal of Computational Physics 117, 1-19 (1995)

S. Plimpton, Journal of Computational Physics 117, 1-19 (1995)

work page 1995

[51] [51]

Puligheddu, Y

M. Puligheddu, Y. Xia, M. K. Y. Chan and G. Galli, arXiv:1902.08260 (2019)

work page arXiv 1902

[52] [52]

S. G. Volz and G. Chen, Phys. Rev. B 61 (4), 2651 (2000)

work page 2000

[53] [53]

B. L. Zink, R. Pietri and F. Hellman, Phys. Rev. Lett. 96 (5), 055902 (2006)

work page 2006

[54] [54]

K. T. Regner, D. P. Sellan, Z. Su, C. H. Amon, A. J. McGaughey and J. A. Malen, Nature communications 4, 1640 (2013)

work page 2013

[55] [55]

D. G. Cahill, M. Katiyar and J. R. Abelson, Phys Rev B Condens Matter 50 (9), 6077-6081 (1994)

work page 1994

[56] [56]

M. C. Wingert, J. Zheng, S. Kwon and R. Chen, Semicond. Sci. Technol. 31 (11), 113003 (2016)

work page 2016

[57] [57]

J. M. Larkin and A. J. H. McGaughey, Phys. Rev. B 89, 144303 (2014)

work page 2014

[58] [58]

Isaeva, G

L. Isaeva, G. Barbalinardo, D. Donadio and S. Baroni, arxiv: 1904.02255 (2019)

work page arXiv 1904

[59] [59]

Lv and A

W. Lv and A. Henry, New Journal of Physics 18 (1), 013028 (2016)

work page 2016

[60] [60]

C. J. Glassbrenner and G. A. Slack, Phys. Rev. 134 (4A), A1058-A1069 (1964)

work page 1964