High-efficient ab initio Bayesian active learning method and applications in prediction of two-dimensional functional materials

Abstract

Beyond the conventional trial-and-error method, machine learning offers a great opportunity to accelerate the discovery of functional materials, but still often suffers from difficulties such as limited materials data and...

Two-dimensional semiconductor materials with high stability and electron mobility in group-11 chalcogenide compounds: MNX (M = Cu, Ag, Au; N = Cu, Ag, Au; X = S, Se, Te; M ≠ N)

“…They were two to three times those of the other MNX materials, slightly higher than the one of single-layer MoS 2 (4.7 × 10 5 cm −1 ), 49,50 much higher than those of single-layer t-ZnS 51 and BiOI 52 and were even comparable to organic perovskite solar cells. 53 To obtain more accurate optical absorption properties, we used an advanced method (G 0 W 0 + BSE, 54,55 the GW method considers the multibody interaction by the quasi-particle method, and the BSE equation considers the electron-hole interaction) to calculate the light absorption spectra of two structures of interest (β-AgAuS, β-AgAuSe), as shown in Fig. S7.…”

“…Moreover, the absorption peak in the visible light range was intensive, and even higher than the one in the visible light range of MoS 2 and phosphorene. 54,55 It indicated that β-AgAuS and β-AgAuSe materials were more superior in the application of optoelectronic devices.…”

Exploring the Conformers of an Organic Molecule on a Metal Cluster with Bayesian Optimization

“…In recent years, Bayesian Optimization in combination with DFT has been widely applied to molecules and materials for structure search and property optimization. − In our recent work, we developed an active machine learning procedure for molecular conformer identification and ranking. We first fix the bond lengths and angles and choose the molecular dihedral angles as features to reduce the dimension of the search space.…”

Machine learning of the Γ-point gap and flat bands of twisted bilayer graphene at arbitrary angles

“…As known, machine learning (ML) is a data-driven approach that has been widely used in many important fields of physics, chemistry, and material science. In particular, several ML models [32][33][34][35][36][37][38][39][40] have been recently proposed to effectively evaluate the energy band gap in a high-throughput way. For instance, Pilania et al utilized a cross-validated kernel ridge regression (KRR) model to reliably predict the band gaps of double perovskites.…”

“…[35] For the two-dimensional systems, Ma et al combined active learning with high-throughput ab-initio calculations to screen out hexagonal binary compounds with proper photovoltaic band gaps. [37] To accelerate the discovery of promising photocatalytic and photovoltaic candidates, Sa et al used the crystal graph convolutional neural network (CGCNN) model to reliably predict the gaps of 1035 Janus III-VI vdW heterostructures. [38] In the case of TBG, however, the ML approaches for the accurate prediction of the energy gap at the Γ -point and especially the flat bands are less known so far.…”