Article
Computer Science, Interdisciplinary Applications
Yixiao Chen, Linfeng Zhang, Han Wang, E. Weinan
Summary: DeePKS-kit is an open-source software package for developing machine learning based energy and density functional models. It supports multiple methods and provides simple and customized tools. The paper also provides an example of developing a chemically accurate model for water clusters.
COMPUTER PHYSICS COMMUNICATIONS
(2023)
Article
Physics, Multidisciplinary
Felix Biggs, Benjamin Guedj
Summary: This paper introduces a new class of partially-aggregated estimators and reformulates a PAC-Bayesian bound for signed-output networks, resulting in a directly optimisable, differentiable objective and a generalisation guarantee. Empirical results show competitive generalisation guarantees compared to other methods for training such networks.
Article
Chemistry, Physical
Subrata Jana, Lucian A. Constantin, Prasanjit Samal
Summary: We propose a realistic density functional approximation based on a semilocal indicator that exhibits good screening properties. The local band model shows remarkable density scaling behaviors and is applicable to various atoms. We introduce the LDAg correlation functional, which correctly calculates the correlation energy of atoms and shows improvement in ionization potential.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Materials Science, Multidisciplinary
N. D. Woods, M. T. Entwistle, R. W. Godby
Summary: The study computed the exact exchange-correlation (xc) kernel f(xc) of linear response time-dependent density functional theory at different frequencies, revealing that its frequency dependence is largely influenced by its analytic structure. Despite the presence of singularities at certain frequencies, it was found that f(xc) is approximately independent within certain frequency ranges, and the key differences between the exact f(xc) and its approximations were analyzed.
Article
Multidisciplinary Sciences
John P. Perdew, Adrienn Ruzsinszky, Jianwei Sun, Niraj K. Nepal, Aaron D. Kaplan
Summary: Strong correlations within a symmetry-unbroken ground-state wavefunction may manifest in approximate density functional theory as symmetry-broken spin densities or total densities, arising from soft modes of fluctuations such as spin-density or charge-density waves. An approximate density functional that breaks symmetry can be more revealing than an exact functional that does not, with examples including the stretched H-2 molecule, antiferromagnetic solids, and the static charge-density wave/Wigner crystal phase of a low-density jellium. Time-dependent density functional theory quantitatively shows that the static charge-density wave is a soft plasmon, with the frequency of a related density fluctuation dropping to zero.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2021)
Article
Materials Science, Multidisciplinary
Sebastian Dick, Marivi Fernandez-Serra
Summary: By utilizing an end-to-end differentiable implementation of the Kohn-Sham self-consistent field equations, the study introduces a highly accurate neural network-based exchange and correlation (XC) functional for electronic density. The research evaluates the model against various families of XC approximations and establishes a strong linear correlation between energy and density errors to define a new XC functional quality metric, improving the ranking of different approximations.
Article
Computer Science, Interdisciplinary Applications
Marie Dumaz, Reese Boucher, Miguel A. L. Marques, Aldo H. Romero
Summary: Density functional theory is widely used in characterizing the electronic structure of materials and is considered one of the most successful theories in materials science. This paper measures the specific impact of this theory through citation records of solid-state first principle ab initio packages, showing exponential growth in publications and how different electronic structure packages support various scientific communities. The analysis also reveals interesting observations such as connections between countries where packages are developed and used, and evidence of specialization in software packages despite similar capabilities.
Article
Chemistry, Physical
Ivor Loncaric, Maite Alducin, J. Inaki Juaristi
Summary: Despite the success of density functional theory, there are still cases where current functionals fail to describe certain materials. This study adds to the list by showing that the interaction between O-2 and Ag(110) cannot be properly described by popular functionals. The authors provide clues for the development of a functional that can accurately describe this and similar systems.
Article
Biochemical Research Methods
Zechen Wang, Liangzhen Zheng, Sheng Wang, Mingzhi Lin, Zhihao Wang, Adams Wai-Kin Kong, Yuguang Mu, Yanjie Wei, Weifeng Li
Summary: Recently reported machine learning- or deep learning-based scoring functions (SFs) have shown promising performance in predicting protein-ligand binding affinities and have great application prospects. However, accurately differentiating between highly similar ligand conformations, including the native binding pose, remains challenging. Thus, this study presents a fully differentiable, end-to-end framework (DeepRMSD+Vina) for ligand pose optimization based on a hybrid scoring function and traditional AutoDock Vina. The evaluation results demonstrate that DeepRMSD+Vina outperforms most existing SFs and shows high potential in drug design and discovery.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Chemistry, Physical
Yan Oueis, Viktor N. Staroverov
Summary: This article investigates the construction of a unique local real-space potential that is precisely equivalent to a given operator under certain conditions. Utilizing the principles, a method is developed to construct exchange-correlation potentials from canonical Kohn-Sham orbitals.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2022)
Article
Chemistry, Physical
Wenna Ai, Wei-Hai Fang, Neil Qiang Su
Summary: The short-range corrected reduced density matrix functional omega P22 is developed to utilize the advantages of functionals in KS-DFT and RDMFT without double-counting, outperforming other 1-RDM functionals in tests of thermochemistry, nonbonded interactions, and bond dissociation energy. Omega P22 shows less systematic error for systems involving fractional spins and can accurately predict the energies for different single and multiple bonds, filling a gap left by commonly used KS-DFT and RDMFT functionals.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Johannes Gedeon, Jonathan Schmidt, Matthew J. P. Hodgson, Jack Wetherell, Carlos L. Benavides-Riveros, Miguel A. L. Marques
Summary: This article presents a solution to the problems in density functional theory, namely the explicit dependency of the functionals on the particle number and the derivative discontinuity at integer particle numbers. They propose training a neural network as a universal functional that exhibits piece-wise linearity between integer particle numbers and reproduces the derivative discontinuity of the exchange-correlation energy.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Xin Li, Haojie Lei, Li Zhang, Mingzhong Wang
Summary: This paper explores interpretable Deep Reinforcement Learning (DRL) by representing policy using Differentiable Inductive Logic Programming (DILP). The research focuses on the optimization perspective of DILP-based policy learning and proposes using Mirror Descent for policy optimization. The theoretical and empirical studies verify the effectiveness of the proposed approach.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Environmental Sciences
Jianfeng Zhang, Chunyu Liu, Yu Wu, Xinyu Li, Jiejing Zhang, Jing Liang, Yongguang Li
Summary: This study investigates the adsorption of tetracycline (TC) on a prepared polycationic straw (MMS) and provides insights into its mechanism. The results show that the adsorption is a spontaneous, monolayer reaction involving electrostatic interaction and hydrogen bonds. Machine learning prediction confirms the feasibility and offers a novel strategy for reducing the cost of removing other pollutants.
ENVIRONMENTAL POLLUTION
(2024)
Article
Chemistry, Physical
Sara Giarrusso, Aurora Pribram-Jones
Summary: The paper compares the contributions of Hartree-Fock and Kohn-Sham theories to correlation energies in the asymmetric Hubbard dimer model. It also tests the performance of two different functionals and finds that they work better for the Hartree-Fock reference. This study provides insights into the error that may occur when using the strong-interaction ingredient for the Kohn-Sham reference.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Physics, Multidisciplinary
S. M. Vinko, V. Vozda, J. Andreasson, S. Bajt, J. Bielecki, T. Burian, J. Chalupsky, O. Ciricosta, M. P. Desjarlais, H. Fleckenstein, J. Hajdu, V Hajkova, P. Hollebon, L. Juha, M. F. Kasim, E. E. McBride, K. Muehlig, T. R. Preston, D. S. Rackstraw, S. Roling, S. Toleikis, J. S. Wark, H. Zacharias
PHYSICAL REVIEW LETTERS
(2020)
Article
Physics, Multidisciplinary
O. S. Humphries, P. Allan, C. R. D. Brown, L. M. R. Hobbs, S. F. James, M. G. Ramsay, B. Williams, D. J. Hoarty, M. P. Hill, S. M. Vinko
COMMUNICATIONS PHYSICS
(2020)
Article
Chemistry, Multidisciplinary
Gabriel Perez-Callejo, Sam M. Vinko, Shenyuan Ren, Ryan Royle, Oliver Humphries, Thomas R. Preston, Bruce A. Hammel, Hyun-Kyung Chung, Tomas Burian, Vojtech Vozda, Ming-Fu Lin, Tim Brandt van Driel, Justin S. Wark
APPLIED SCIENCES-BASEL
(2020)
Article
Physics, Fluids & Plasmas
P. Hollebon, J. S. Wark, S. M. Vinko
Summary: The use of excited-state projector augmented-wave potentials can extend calculations of dense plasmas to model core-hole states and calculate the electronic structure of various integer charge configurations in a dense plasma environment. These excited-state potentials show good agreement with finite-temperature all-electron calculations, providing an efficient approach for modeling the structure of strongly-coupled non-equilibrium plasmas and their interaction with intense X-rays from first principles.
PLASMA PHYSICS AND CONTROLLED FUSION
(2021)
Article
Chemistry, Physical
Muhammad F. Kasim, Susi Lehtola, Sam M. Vinko
Summary: Automatic differentiation is a paradigm shift in scientific programming that simplifies calculations and reduces development time. It has fuelled the growth of machine learning techniques and is also proving valuable in quantum chemistry simulations.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Physics, Multidisciplinary
Jong-Won Lee, Minju Kim, Gyeongbo Kang, Sam M. Vinko, Leejin Bae, Min Sang Cho, Hyun-Kyung Chung, Minseok Kim, Soonnam Kwon, Gyusang Lee, Chang Hee Nam, Sang Han Park, Jang Hyeob Sohn, Seong Hyeok Yang, Ulf Zastrau, Byoung Ick Cho
Summary: Ultrafast dynamics in photoexcited warm dense Cu were visualized using femtosecond x-ray absorption spectroscopy, revealing rich dynamical features related to d vacancies. Improved understanding can be achieved by including recombination dynamics of nonthermal electrons and changes in screening of the excited d block. The population balance between the 4sp and 3d bands is mainly determined by the recombination rate of nonthermal electrons, showing recovery of the underpopulated 3d block on picosecond timescales.
PHYSICAL REVIEW LETTERS
(2021)
Article
Physics, Fluids & Plasmas
H. Poole, D. Cao, R. Epstein, I. Golovkin, T. Walton, S. X. Hu, M. Kasim, S. M. Vinko, J. R. Rygg, V. N. Goncharov, G. Gregori, S. P. Regan
Summary: In this study, a feasibility study was conducted to diagnose the temperature, density, and ionization of the compressed DT shell using spatially integrated, spectrally resolved, x-ray Thomson scattering measurements. Synthetic scattering spectra were generated and analyzed, showing that the plasma conditions of compressed DT shells can be resolved.
PHYSICS OF PLASMAS
(2022)
Article
Physics, Multidisciplinary
Shenyuan Ren, Yuanfeng Shi, Quincy Y. Y. van den Berg, Muhammad F. F. Kasim, Hyun-Kyung Chung, Elisa V. V. Fernandez-Tello, Pedro Velarde, Justin S. S. Wark, Sam M. M. Vinko
Summary: X-ray free-electron lasers have enabled new experimental investigations into extreme conditions of matter via intense x-ray-matter interactions. A proposed numerical method allows for the study of high-energy-density plasmas driven by XFELs on femtosecond timescales, even when both electrons and ions are far from local thermodynamic equilibrium.
COMMUNICATIONS PHYSICS
(2023)
Article
Multidisciplinary Sciences
Pontus Svensson, Thomas Campbell, Frank Graziani, Zhandos Moldabekov, Ningyi Lyu, Victor S. Batista, Scott Richardson, Sam M. Vinko, Gianluca Gregori
Summary: An extension of the wave packet description of quantum plasmas is proposed, allowing for elongation in arbitrary directions. A generalized Ewald summation is used to account for long-range Coulomb interactions, while fermionic effects are approximated using purpose-built Pauli potentials. The numerical implementation shows good parallel support and close to linear scaling in particle number. Comparisons with the common wave packet model employing isotropic states reveal differences primarily in the electronic subsystem, particularly in the electrical conductivity of dense hydrogen.
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
M. F. Kasim, D. Watson-Parris, L. Deaconu, S. Oliver, P. Hatfield, D. H. Froula, G. Gregori, M. Jarvis, S. Khatiwala, J. Korenaga, J. Topp-Mugglestone, E. Viezzer, S. M. Vinko
Summary: Computer simulations are important for scientific discovery, but their accuracy is often limited by slow execution. To accelerate simulations, researchers propose using machine learning to build fast emulators, but obtaining large training datasets can be expensive. A new method based on neural architecture search is presented, which can build accurate emulators even with limited training data. The method is successfully applied in various scientific fields and provides uncertainty estimation for emulators.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2022)