A first-principles and machine-learning investigation on the electronic, photocatalytic, mechanical and heat conduction properties of nanoporous C₅N monolayers

Abstract

Phononic thermal transport, mechanical/failure response, electronic, optical and photocatalytic properties of the C5N monolayer are explored using the density functional theory and machine-learning interatomic potentials.

Predicting lattice thermal conductivity via machine learning: a mini review

“…Potential Feature Algorithm Application systems MTP moment tensor 70 linear regression monolayers [77][78][79][80][81][82][83][84]95 , bilayers 85 , heterostructures 20 , perovskites 86 , skutterudites 68,87 , alloys 88 , wurtzite structures 89 , phase change materials 91,92 , complex crystals 93 , etc NNP ACSF 71,72 , digital image 66 , SOAP 18 NN molten salts 21 , polymorphs 103 , near-stoichiometric compounds 106 , high-entropy ceramics 107,108 , ternary salts 109 , nanowires 110 , monolayers 111 , antiperovskites 112 , etc GAP SOAP 18 , two-body and three-body descriptors 117,118 GPR crystalline compounds [113][114][115] , crystals with defects 116 , monolayers 117,118 , amorphous structures 119 , etc Fig. 4 Scheme of active learning bootstrapping iterations for training the Moment Tensor Potential (MTP).…”

“…Table 2 summarizes several important MLPs used for the evaluation of κ L , which includes the Moment Tensor Potentials (MTPs) 70 , the Neural Network Potentials (NNPs) 71,74 , and the Gaussian Approximation Potentials (GAPs) 75 . In particular, the MTPs exhibit an excellent balance between accuracy and computational efficiency 76 , which have been widely used to predict the κ L of various systems, such as monolayers, alloys, and complex compounds 68,[77][78][79][80][81][82][83][84][85][86][87][88][89][90][91][92][93] . In principle, the purpose of training MTP is minimizing the difference between the predicted and DFTcalculated energies (E), forces (f), and stresses (σ) for K atomic configurations: 70,94,95…”

Grain boundary sliding and distortion on a nanosecond timescale induce trap states in CsPbBr3: ab initio investigation with machine learning force field

“…54 These cost-effective ML FFs enable molecular dynamics (MD) simulations for larger systems with longer trajectories compared to ab initio MD (AIMD) simulations. [55][56][57][58] Considering that calculated nonradiative recombination times of the GB systems are in nanoseconds, while AIMD simulations are limited to picoseconds, 39,41 the ML-based MD (MLMD) can bridge this gap and track structure evolutions during the whole charge recombination period. Meanwhile, distinct structural fluctuations are observed in MD simulations of defective perovskites, and the resulting changes in electronic properties can dramatically affect the carrier dynamics.…”

Investigation of the mechanical and transport properties of InGeX3 (X = S, Se and Te) monolayers using density functional theory and machine learning

“…The thermal dissipation behavior of novel semiconductors has attracted much attention. At present, a variety of methods, including first-principles calculations based on density functional theory (DFT), [22][23][24] empirical potential 25,26 and machine learning (ML), [27][28][29][30] are used to predict the mechanical properties and thermal transport performance. The DFT method can provide accurate potential functions of a system, but the computational cost is very high.…”