Density-of-states similarity descriptor for unsupervised learning from materials data DOI Creative Commons
Martin Kubáň, Santiago Rigamonti, Markus Scheidgen

et al.

Scientific Data, Journal Year: 2022, Volume and Issue: 9(1)

Published: Oct. 22, 2022

We develop a materials descriptor based on the electronic density-of-states (DOS) and investigate similarity of it. As an application example, we study Computational 2D Materials Database (C2DB) that hosts thousands two-dimensional with their properties calculated by density-functional theory. Combining our clustering algorithm, identify groups similar structure. introduce additional descriptors to characterize these clusters in terms crystal structures, atomic compositions, configurations members. This allows us rationalize found (dis)similarities perform automated exploratory confirmatory analysis C2DB data. From this analysis, find majority consist isoelectronic sharing symmetry, but also outliers, i.e., whose cannot be explained way.

Language: Английский

Designing semiconductor materials and devices in the post-Moore era by tackling computational challenges with data-driven strategies DOI
Jiahao Xie, Yansong Zhou, Muhammad Faizan

et al.

Nature Computational Science, Journal Year: 2024, Volume and Issue: 4(5), P. 322 - 333

Published: May 23, 2024

Language: Английский

Citations

7
Constructing a built-in electric field by anchoring highly dispersed Zn single atoms on UiO-66-NH2 for efficient CO2 photoreduction DOI

Mingna Chu,

Yang Li, Ximing Chen

et al.

Journal of Materials Chemistry A, Journal Year: 2022, Volume and Issue: 10(44), P. 23666 - 23674

Published: Jan. 1, 2022

Efficient CO 2 conversion has been realized on UiO-66-NH -0.7Zn SAs benefitting from the formation of built-in electric field by anchoring Zn single atoms (SAs) .

Language: Английский

Citations

23
Deep learning approach to genome of two-dimensional materials with flat electronic bands DOI Creative Commons
Anupam Bhattacharya,

Ivan Timokhin,

Ratnamala Chatterjee

et al.

npj Computational Materials, Journal Year: 2023, Volume and Issue: 9(1)

Published: June 8, 2023

Abstract Electron-electron correlations play central role in condensed matter physics, governing phenomena from superconductivity to magnetism and numerous technological applications. Two-dimensional (2D) materials with flat electronic bands provide natural playground explore interaction-driven thanks their highly localized electrons. The search for 2D band has attracted intensive efforts, especially now open science databases encompassing thousands of computed bands. Here we automate the otherwise daunting task classification by combining supervised unsupervised machine learning algorithms. To this end, convolutional neural network was employed identify materials, which were then subjected symmetry-based analysis using a bilayer algorithm. Such hybrid approach exploring allowed us construct genome hosting reveal material classes outside known paradigms.

Language: Английский

Citations

16
Improving Molecular‐Dynamics Simulations for Solid–Liquid Interfaces with Machine‐Learning Interatomic Potentials DOI
Pengfei Hou, Yumiao Tian, Xing Meng

et al.

Chemistry - A European Journal, Journal Year: 2024, Volume and Issue: 30(49)

Published: June 15, 2024

Emerging developments in artificial intelligence have opened infinite possibilities for material simulation. Depending on the powerful fitting of machine learning algorithms to first-principles data, interatomic potentials (MLIPs) can effectively balance accuracy and efficiency problems molecular dynamics (MD) simulations, serving as tools various complex physicochemical systems. Consequently, this brings unprecedented enthusiasm researchers apply such novel technology multiple fields revisit major scientific that remained controversial owing limitations previous computational methods. Herein, we introduce evolution MLIPs, provide valuable application examples solid-liquid interfaces, present current challenges. Driven by solving multitudinous difficulties terms accuracy, efficiency, versatility booming technique, combined with simulation methods, will an underlying understanding interdisciplinary challenges, including materials, physics, chemistry.

Language: Английский

Citations

5
Rational Design of Earth‐Abundant Catalysts toward Sustainability DOI Creative Commons

Jinyang Guo,

Yousof Haghshenas, Yiran Jiao

et al.

Advanced Materials, Journal Year: 2024, Volume and Issue: unknown

Published: July 31, 2024

Abstract Catalysis is crucial for clean energy, green chemistry, and environmental remediation, but traditional methods rely on expensive scarce precious metals. This review addresses this challenge by highlighting the promise of earth‐abundant catalysts recent advancements in their rational design. Innovative strategies such as physics‐inspired descriptors, high‐throughput computational techniques, artificial intelligence (AI)‐assisted design with machine learning (ML) are explored, moving beyond time‐consuming trial‐and‐error approaches. Additionally, biomimicry, inspired efficient enzymes nature, offers valuable insights. systematically analyses these strategies, providing a roadmap developing high‐performance from abundant elements. Clean energy applications (water splitting, fuel cells, batteries) chemistry (ammonia synthesis, CO 2 reduction) targeted while delving into fundamental principles, biomimetic approaches, current challenges field. The way to more sustainable future paved overcoming catalyst scarcity through

Language: Английский

Citations

5
Discovery of Two-Dimensional Multinary Component Photocatalysts Accelerated by Machine Learning DOI
Hao Jin,

Xiaoxing Tan,

Tao Wang

et al.

The Journal of Physical Chemistry Letters, Journal Year: 2022, Volume and Issue: 13(31), P. 7228 - 7235

Published: July 30, 2022

Searching for novel and high-performance two-dimensional (2D) materials is an important task photocatalytic applications. Although multinary compounds exhibit more diversity in structure properties comparison to binary 2D materials, they are comparatively under-studied. Herein, using a machine-learning (ML) technique high-throughput screening, we develop efficient approach accurately predict multicomponent photocatalysts. Over 4000 monolayers examined, 75 identified Considering our predictions, find that the ternary quaternary A2P2X6 ABP2X6 with A = Cu/Zn/Ge/Ag/Cd, B Ga/In/Bi, X S/Se superior properties, making them promising candidates overall water splitting. Thus, work provides way explore photocatalysts, which could stimulate further theoretical experimental investigations on application

Language: Английский

Citations

22
Accurate Prediction of HSE06 Band Structures for a Diverse Set of Materials Using Δ-Learning DOI
Santosh Adhikari, Jacob M. Clary, Ravishankar Sundararaman

et al.

Chemistry of Materials, Journal Year: 2023, Volume and Issue: 35(20), P. 8397 - 8405

Published: Oct. 16, 2023

We used machine learning (ML) to accurately predict eigenvalues of the hybrid HSE06 functional using computed by less computationally expensive PBE and associated electronic features based on k-point resolved atomic band character. The ML model was trained from only one for each 168 compounds in training set. across all k-points were then predicted a separate set 169 with mean absolute error (MAE) 0.13 eV, representing significant improvement over PBE-computed relative that (MAE = 0.96 eV). These result remarkably accurate predictions structures, projected density states, gaps, even though not explicitly these other properties. Finally, we demonstrate our has similar accuracy both ternary quaternary well outside initial systems 112 160 atoms, demonstrating its potential rapidly HSE06-quality structures complex materials are practically unfeasible HSE06.

Language: Английский

Citations

11
Inverse design for materials discovery from the multidimensional electronic density of states DOI Creative Commons
Kihoon Bang,

J. H. Kim,

Doosun Hong

et al.

Journal of Materials Chemistry A, Journal Year: 2024, Volume and Issue: 12(10), P. 6004 - 6013

Published: Jan. 1, 2024

To accelerate materials discovery, a deep learning method for inverse design of inorganic using multidimensional DOS properties was developed.

Language: Английский

Citations

4
Enhanced learning loop framework accelerates screening of bimetallic catalysts with high oxygen reduction properties in different coordination environments DOI

Pei Song,

Zepeng Jia,

Sen Lu

et al.

International Journal of Hydrogen Energy, Journal Year: 2024, Volume and Issue: 73, P. 305 - 315

Published: June 11, 2024

Language: Английский

Citations

4
Unsupervised representation learning of Kohn–Sham states and consequences for downstream predictions of many-body effects DOI Creative Commons
Bowen Hou, Jinyuan Wu, Diana Y. Qiu

et al.

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Nov. 2, 2024

Representation learning for the electronic structure problem is a major challenge of machine in computational condensed matter and materials physics. Within quantum mechanical first principles approaches, density functional theory (DFT) preeminent tool understanding structure, high-dimensional DFT wavefunctions serve as building blocks downstream calculations correlated many-body excitations related physical observables. Here, we use variational autoencoders (VAE) unsupervised show that these lie low-dimensional manifold within latent space. Our model autonomously determines optimal representation avoiding limitations due to manual feature engineering. To demonstrate utility space wavefunction, it supervised training neural networks (NN) prediction quasiparticle bandstructures GW formalism. The achieves low error 0.11 eV combined test set two-dimensional metals semiconductors, suggesting captures key information from original data. Finally, explore generative ability interpretability VAE representation.

Language: Английский

Citations

4