Table 4 Transferability of the here-developed MLIP (MLIP-[4]) in molecular statics (MS) calculations of surface energies, Esurf (J m−2), for low-index surfaces of TiB2

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

 

Surface

Esurf

Source

ML-MS

(0001)

3.80

This work

 DFT

(0001)

3.80

This work

 DFT

(0001)

4.50–4.72

Ref. 98

 DFT

(0001)

4.22–4.24

Ref. 73

 DFT

(0001)

4.14

Ref. 99

 DFT

(0001)

≈ 4.20

Ref. 74

ML-MS

(10\(\overline{1}\)0)

3.98

This work

 DFT

(10\(\overline{1}\)0)

4.12

This work

 DFT

(10\(\overline{1}\)0)

≈ 4.10

Ref. 74

ML-MS

(\(\overline{1}2\overline{1}0\))

3.42

This work

 DFT

(\(\overline{1}2\overline{1}0\))

3.57

This work

 DFT

(\(\overline{1}2\overline{1}0\))

4.20–4.83

Ref. 73

 DFT

(\(\overline{1}2\overline{1}0\))

≈ 4.02

Ref. 74

  1. Our calculations consider stoichiometric TiB2 surfaces, i.e., with 1:2 Ti-to-B ratio.