Phase Prediction via Crystal Structure Similarity in the Periodic Number Representation DOI

Cem Oran,

Riccarda Caputo,

P. Villars

et al.

Inorganic Chemistry, Journal Year: 2024, Volume and Issue: 63(43), P. 20521 - 20530

Published: Oct. 15, 2024

The periodic number (PN) representation of the chemical systems, introduced by Dmitri Mendeleev, uncovers fundamental principle similarity in a straightforward way. In this framework, rows correspond to principal quantum numbers elements' electronic configurations when considered isolated atoms. This systematic arrangement allows for deeper understanding relationships and patterns among elements. study, we propose novel strategy structure type (prototype) prediction utilizing PN concept identify possible modifications phase stability unexplored systems. Our PN-based crystal (PNcsp) program, which evaluates through neighboring map, provides most probable prototypes unknown unreported given phases binary higher order We applied PNcsp 59 distinct systems whose equimolar are indicated respective diagrams but lack accurate experimental determination. methodology identified 93 these equiatomic phases, 47 exhibit mechanical dynamic stability. Notably, approach discovered 19 entirely novel, fully stable polymorphic thereby expanding known landscape potential materials. Furthermore, demonstrated that method is also effective nonequimolar

Language: Английский

Material transformers: deep learning language models for generative materials design DOI Creative Commons
Nihang Fu, Lai Wei, Yuqi Song

et al.

Machine Learning Science and Technology, Journal Year: 2022, Volume and Issue: 4(1), P. 015001 - 015001

Published: Dec. 21, 2022

Abstract Pre-trained transformer language models (LMs) on large unlabeled corpus have produced state-of-the-art results in natural processing, organic molecule design, and protein sequence generation. However, no such been applied to learn the composition patterns for generative design of material compositions. Here we train a series seven modern (GPT, GPT-2, GPT-Neo, GPT-J, BLMM, BART, RoBERTa) materials using expanded formulas ICSD, OQMD, Materials Projects databases. Six different datasets with/out non-charge-neutral or EB samples are used benchmark performances uncover biases Our experiments show that transformers based causal LMs can generate chemically valid compositions with as high 97.61% be charge neutral 91.22% electronegativity balanced, which has more than six times higher enrichment compared baseline pseudo-random sampling algorithm. also demonstrate generation novelty their potential new discovery is proved by capability recover leave-out materials. We find properties generated tailored training selected sets high-bandgap samples. each own preference terms running time complexity varies lot. our discover set validated density functional theory calculations. All trained code accessed freely at http://www.github.com/usccolumbia/MTransformer .

Language: Английский

Citations

17
Prediction of New Crystal of Titanium Oxides and Property Calculations Using Structure Prediction Software and First-Principles Theory DOI
Jinrong Huo, Pengfei Liu, Yuxin Tang

et al.

The Journal of Physical Chemistry C, Journal Year: 2025, Volume and Issue: unknown

Published: May 7, 2025

Language: Английский

Citations

0
Accelerated Organic Crystal Structure Prediction with Genetic Algorithms and Machine Learning DOI
Amit Kadan, Kevin Ryczko,

Andrew Wildman

et al.

Journal of Chemical Theory and Computation, Journal Year: 2023, Volume and Issue: 19(24), P. 9388 - 9402

Published: Dec. 7, 2023

We present a high-throughput, end-to-end pipeline for organic crystal structure prediction (CSP)─the problem of identifying the stable structures that will form from given molecule based only on its molecular composition. Our tool uses neural network potentials to allow efficient screening and structural relaxation generated candidates. consists two distinct stages: random search, whereby candidates are randomly screened, optimization, where genetic algorithm (GA) optimizes this screened population. assess performance each stage our 21 molecules taken Cambridge Crystallographic Data Centre's CSP blind tests. show search alone yields matches ≈50% targets. then validate potential full pipeline, making use GA optimize root-mean-square deviation between experimentally derived structure. With approach, we able find ≈80% with 10–100 times smaller initial population sizes than when using search. Lastly, run an ANI model is trained small data set extracted in Structural Database, generating ≈60% By leveraging machine learning models predict energies at density functional theory level, has approach accuracy ab initio methods efficiency empirical force fields.

Language: Английский

Citations

8
First-principles calculations of electronic structure, optical and thermodynamic properties of GaBN2, Ga3BN4 and GaB3N4 nitrides DOI
Bo Li,

Yonghua Duan,

Mingjun Peng

et al.

Vacuum, Journal Year: 2022, Volume and Issue: 208, P. 111745 - 111745

Published: Dec. 12, 2022

Language: Английский

Citations

13
A guide to discovering next-generation semiconductor materials using atomistic simulations and machine learning DOI
Arun Mannodi‐Kanakkithodi

Computational Materials Science, Journal Year: 2024, Volume and Issue: 243, P. 113108 - 113108

Published: May 27, 2024

Language: Английский

Citations

2
Deep reinforcement learning for inverse inorganic materials design DOI Creative Commons
Christopher Karpovich, Elton Pan,

Elsa Olivetti

et al.

npj Computational Materials, Journal Year: 2024, Volume and Issue: 10(1)

Published: Dec. 19, 2024

Language: Английский

Citations

2
Discovery of 2D Materials using Transformer Network‐Based Generative Design DOI Creative Commons
Rongzhi Dong, Yuqi Song, Edirisuriya M. Dilanga Siriwardane

et al.

Advanced Intelligent Systems, Journal Year: 2023, Volume and Issue: 5(12)

Published: Oct. 10, 2023

Two‐dimensional (2D) materials offer great potential in various fields like superconductivity, quantum systems, and topological materials. However, designing them systematically remains challenging due to the limited pool of fewer than 100 experimentally synthesized 2D Recent advancements deep learning, data mining, density functional theory (DFT) calculations have paved way for exploring new material candidates. Herein, a generative design pipeline known as transformer generator (MTG) is proposed. MTG leverages two distinct composition generators, both trained using self‐learning neural language models rooted transformers, with without transfer learning. These generate numerous compositions, which are plugged into established templates predict their crystal structures. To ensure stability, DFT computations assess thermodynamic stability based on energy‐above‐hull formation energy metrics. has found four DFT‐validated stable materials: NiCl 4 , IrSBr, CuBr 3 CoBrCl, all zero values that indicate stability. Additionally, GaBrO NbBrCl below 0.05 eV. exhibit dynamic confirmed by phonon dispersion analysis. In summary, shows significant discovering

Language: Английский

Citations

4
Machine-Learning-Assisted Acceleration on High-Symmetry Materials Search: Space Group Predictions from Band Structures DOI
Bin Xi,

Kin Fai Tse,

Tsz-Fung Kok

et al.

The Journal of Physical Chemistry C, Journal Year: 2022, Volume and Issue: 126(29), P. 12264 - 12273

Published: July 18, 2022

Efficiency of search wanted materials with desired properties is limited by the huge space. By deep learning methods, we demonstrate that space group information can be acquired from band structure inputs to reduce Despite atomic orbital or accidental degeneracies mixed lattice degeneracies, as input yield 96.0% prediction accuracy for cubic systems leads a 25.1-fold acceleration searching speed overall. Additionally, all groups, 82.0% overall 36.9-fold in speed. In addition, valence satisfactory results and may assist structural analysis ARPES results.

Language: Английский

Citations

5
Structurally Constrained Evolutionary Algorithm for the Discovery and Design of Metastable Phases DOI
Busheng Wang, Katerina P. Hilleke, Samad Hajinazar

et al.

Journal of Chemical Theory and Computation, Journal Year: 2023, Volume and Issue: 19(21), P. 7960 - 7971

Published: Oct. 19, 2023

Metastable materials are abundant in nature and technology, showcasing remarkable properties that inspire innovative design. However, traditional crystal structure prediction methods, which rely solely on energetic factors to determine a structure's fitness, not suitable for predicting the vast number of potentially synthesizable phases represent local minimum corresponding state thermodynamic equilibrium. Here, we present new approach metastable with specific structural features, interface this method XtalOpt evolutionary algorithm. Our relies features include crystalline order (e.g., coordination or chemical environment), symmetry Bravais lattice space group) filter parent pool an search. The effectiveness is benchmarked three known systems: XeN$_8$, two-dimensional polymeric nitrogen sublattice, brookite TiO$_2$, high pressure BaH$_4$ phase was recently characterized. Additionally, newly predicted melaminate salt, $P$-1 WC$_{3}$N$_{6}$, found possess energy lower than two proposed recent computational study. presented here could help identifying structures compounds have already been synthesized, developing synthesis targets desired properties.

Language: Английский

Citations

2
The Liverpool materials discovery server: a suite of computational tools for the collaborative discovery of materials DOI Creative Commons
Samantha Durdy,

Cameron J. Hargreaves,

Mark Dennison

et al.

Digital Discovery, Journal Year: 2023, Volume and Issue: 2(5), P. 1601 - 1611

Published: Jan. 1, 2023

The Liverpool materials discovery server (https://lmds.liverpool.ac.uk) provides easy access to six state of the art computational tools. Creation such cloud platforms enables collaboration between experimental and researchers.

Language: Английский

Citations

1