Table 2 The phase transition temperatures of BaTiO3 obtained by effective Hamiltonian56,57, second principles58, reactive force field (ReaxFF)61, specialized machine learning potential (MLIP)54 and experiments63. The size of the supercell is also displayed. \({T}_{{{{\rm{c}}}},\,{{{\rm{R}}}}-{{{\rm{O}}}}}\), \({T}_{{{{\rm{c}}}},\,{{{\rm{O}}}}-{{{\rm{T}}}}}\), and \({T}_{{{{\rm{c}}}},\,{{{\rm{T}}}}-{{{\rm{C}}}}}\) are the phase transition temperatures of rhombohedral–orthorhombic, orthorhombic–tetragonal, and tetragonal–cubic, respectively

From: A foundation machine learning potential with polarizable long-range interactions for materials modelling

Method

\({T}_{c,\,{{{\rm{R}}}}-{{{\rm{O}}}}}\) (K)

\({T}_{c,\,{{{\rm{O}}}}-{{{\rm{T}}}}}\) (K)

\({T}_{c,\,{{{\rm{T}}}}-{{{\rm{C}}}}}\) (K)

This work (10 × 10 × 10)

145

205

285

Effective Hamiltonian (16 × 16 × 16)56

119

158

257

Effective Hamiltonian* (12 × 12 × 12)57

150 ± 10

195 ± 5

265 ± 5

Second Principles (16 × 16 × 16)58

140

180

224

ReaxFF (6 × 6 × 6)61

N.A.

N.A.

240

MLIP (4 × 4 × 4)54

18.6

91.4

182.4

Experiments63

183

278

403

  1. *Deviations from quantum Monte Carlo simulations.