Table 2 Summary of widely adopted machine learning potentials for indirectly predicting lattice thermal conductivities.

MTP

moment tensor⁷⁰

linear regression

monolayers^{77,78,79,80,81,82,83,84,95}, bilayers⁸⁵, heterostructures²⁰, perovskites⁸⁶, skutterudites^68,87, alloys⁸⁸, wurtzite structures⁸⁹, phase change materials^91,92, complex crystals⁹³, etc

NNP

ACSF^71,72, digital image⁶⁶, SOAP¹⁸

NN

molten salts²¹, polymorphs¹⁰³, near-stoichiometric compounds¹⁰⁶, high-entropy ceramics^107,108, ternary salts¹⁰⁹, nanowires¹¹⁰, monolayers¹¹¹, antiperovskites¹¹², etc

GAP

SOAP¹⁸, two-body and three-body descriptors^117,118

GPR

crystalline compounds^113,114,115, crystals with defects¹¹⁶, monolayers^117,118, amorphous structures¹¹⁹, etc