Table 2 Validation of the here-developed MLIP (MLIP-[1]) against theoretical and experimental (exp.) room-temperature elastic constants, Cij

From: Machine-learning potentials for nanoscale simulations of tensile deformation and fracture in ceramics

 

No. of atoms

T

C11

C33

C44

C12

C13

E

B

G

ν

Source

DFT

192

300

640

446

251

62

91

574

244

260

0.106

Ref. 95

AIMD

720

300

588

430

252

79

111

547

244

243

0.126

This work

ML-MD

720

300

588

409

261

85

98

554

236

246

0.113

This work

Exp.

300

660

432

260

48

93

565

244

266

0.099

Ref. 96

Exp.

300

588

503

238

72

84

575

249

255

0.114

Ref. 97

  1. The Cij values for TiB2 (in GPa, at temperature T (K)) are shown together with the polycrystalline bulk modulus, B (in GPa), shear modulus, G (in GPa), Young’s modulus, E (in GPa), and Poisson’s ratio, ν, compared to reference ab initio and experimental (exp.) data. ref. 96 and ref. 97 is for TiB2 single and polycrystal, respectively. AIMD and ML-MD elastic constants were evaluated following ref. 86, based on a second-order polynomial fit of the [0001], [10\(\overline{1}0\)], and [\(\overline{1}2\overline{1}0\)] stress/strain data (C11, C12, C13, C33) and of the (0001)\([\overline{1}2\overline{1}0],(10\overline{1}0)[\overline{1}2\overline{1}0]\), and \((10\overline{1}0)[0001]\) shear stress/strain data (C44), assuming strains up to 4%. For details see the Methodology section. Note that values from ref. 95 were based on 0 K ab initio calculations coupled to phonon-theory assessments of thermal expansion.