Predicting Lattice Vibrational Frequencies Using Deep Graph Neural Networks

Lattice vibrational frequencies are related to many important materials properties such as thermal and electrical conductivity as well as superconductivity. However, computational calculation of vibrational frequencies using density functional theory methods is computationally too demanding for large number of samples in materials screening. Here we propose a deep graph neural network based algorithm for predicting crystal vibrational frequencies from crystal structures. Our algorithm addresses the variable dimension of vibrational frequency spectrum using the zero padding scheme. Benchmark studies on two data sets with 15,000 mixed-structure and 35,552 rhombohedra samples show that the aggregated R ² scores of the prediction reach 0.554 and 0.724. We also evaluate the structural transferability by predicting the vibration frequencies for 239 individual cubic target structures. The R ² scores for more than 40% of the targets are greater than 0.8 and can reach as high as 0.98 for the model trained with mixed samples, while the average mean absolute error is 43.69 Thz showing low transferability across structure types. Our work demonstrates the capability of deep graph neural networks to learn to predict lattice vibration frequency when sufficient number of training samples are available.

High-Throughput Discovery of Novel Cubic Crystal Materials Using Deep Generative Neural Networks

High-throughput screening has become one of the major strategies for the discovery of novel functional materials. However, its effectiveness is severely limited by the lack of sufficient and diverse materials in current materials repositories such as the open quantum materials database (OQMD). Recent progress in deep learning have enabled generative strategies that learn implicit chemical rules for creating hypothetical materials with new compositions and structures. However, current materials generative models have difficulty in generating structurally diverse, chemically valid, and stable materials. Here we propose CubicGAN, a generative adversarial network (GAN) based deep neural network model for large scale generative design of novel cubic materials. When trained on 375 749 ternary materials from the OQMD database, the authors show that the model is able to not only rediscover most of the currently known cubic materials but also generate hypothetical materials of new structure prototypes. A total of 506 such materials have been verified by phonon dispersion calculation. Considering the importance of cubic materials in wide applications such as solar panels, the GAN model provides a promising approach to significantly expand existing materials repositories, enabling the discovery of new functional materials via screening. The new crystal structures discovered are freely accessible at www.carolinamatdb.org.

Perspective on multi-scale simulation of thermal transport in solids and interfaces

Phonon-mediated thermal transport is inherently multi-scale. The wave-length of phonons (considering phonons as waves) is typically at the nanometer scale; the typical size of a phonon wave energy packet is tens of nanometers, while the phonon mean free path (MFP) can be as long as microns. At different length scales, the phonons will interact with structures of different feature sizes, which can be as small as 0D defects (point defects), short to medium range linear defects (dislocations), medium to large range 2D planar defects (stacking faults and twin boundaries), and large scale 3D defects (voids, inclusions, and various microstructures). The nature of multi-scale thermal transport is that there are different heat transfer physics across different length scales and in the meantime the physics crossing the different scales is interdependent and coupled. Since phonon behavior is usually mode dependent, thermal transport in materials with a combined micro-/nano-structure complexity becomes complicated, making modeling this kind of transport process very challenging. In this perspective, we first summarize the advantages and disadvantages of computational methods for mono-scale heat transfer and the state-of-the-art multi-scale thermal transport modeling. We then discuss a few important aspects of the future development of multi-scale modeling, in particular with the aid of modern machine learning and uncertainty quantification techniques. As more sophisticated theoretical and computational methods continue to advance thermal transport predictions, novel heat transfer physics and thermally functional materials will be discovered for the pertaining energy systems and technologies.

The origin of intrinsic charge transport for Dirac carbon sheet materials: roles of acetylenic linkage and electron-phonon couplings

Within the past few years, intriguing graphene Dirac cones have attracted intense interest in novel two-dimensional (2D) Dirac materials as ultrahigh-mobility functional materials. In this work, the phonon-limited charge transport properties of α-graphyne (α-GY), α-graphdiyne (α-GDY), and β-graphyne (β-GY) were investigated using the Boltzmann transport equation within the first-principles framework while considering the electron-phonon coupling (EPC). Despite all three investigated compounds being 2D Dirac carbon materials, each demonstrated distinctly different carrier mobilities by one order of magnitude (2.2 × 104 cm2 V-1 s-1 for α-GY, 2.1 × 103 cm2 V-1 s-1 for α-GDY and 1.9 × 103 cm2 V-1 s-1 for β-GY at room-temperature and a carrier connection of n ∼ 3 × 1012 cm-2). The essential differences in the mobilities of these materials originated from the acetylenic linkage limiting the group velocity and the E2g phonon modes limiting the scattering time. For example, a few uniformly equivalent acetylenic linkages and E2g phonon modes tend to generate high mobilities. A simple mobility relationship was determined using the number of E2g photon modes, allowing for a quick estimation of the mobilities for Dirac materials. α-GY was identified as a promising alternative to graphene for next generation nanoelectronic devices.