On-the-fly machine learning potential accelerated accurate prediction of lattice thermal conductivity of metastable silicon crystals

Tuning the lattice thermal conductivity of Sb2Te3 by Cr doping: a deep potential molecular dynamics study

“…20–24 These machine learning potentials can be employed to study the abundant physical properties of materials, such as the physical properties of materials at finite temperatures, 25 biological and chemical processes of ions, 26 domain-wall and temperature-dependent phase transition, 27 and so on. Meanwhile, these potentials have been widely used to study the thermal transport properties of solids, 28–39 such as GaN/AlN heterostructures with different interfacial morphologies, 37 Mn X Ge Y materials with a wide range of compositions, 35 complex skutterudite compounds, 31,38 and so on. In addition, our previous works have demonstrated that the machine learning potential can be used to study the thermal conductivities of complex systems such as Bi 2 Te 3 with intrinsic point defects 40 and Bi 2 Te 3 /Sb 2 Te 3 superlattices with different period lengths.…”

Tuning the lattice thermal conductivity of Janus SnSSe by interlayer twisting: a machine-learning-based study

“…As an important data-driven technique, machine learning (ML) has been successfully employed to accelerate the discovery of new materials with desired properties. In particular, ML can be utilized to establish accurate interatomic potentials, which are then implemented into MD simulations or phonon BTE to efficiently determine the κ L of complex structures [19][20][21][22][23][24][25][26][27][28]. For example, Qian et al developed a Gaussian approximation potential (GAP) for amorphous silicon, which outperforms empirical potential in the evaluation of atomic forces, and the predicted κ L agrees well with that measured experimentally [20].…”

Machine-learning-assisted discovery of 212-Zintl-phase compounds with ultra-low lattice thermal conductivity

“…[44][45][46][47] For example, several 2D crystals that haven't been previously classied as favorable TE materials were iden-tied through high-throughput screening function, 48 thermoelectric performance for a series of layered IV-V-VI semiconductors was predicted by ML, 49 and accurate prediction of k L for metastable silicon crystals was accelerated by ML potential. 50 The data-driven ML methods provide new opportunities to accelerate the discovery of promising A 2 BX 2 thermoelectric compounds.…”