Article
Multidisciplinary Sciences
Oliver T. Unke, Stefan Chmiela, Michael Gastegger, Kristof T. Schuett, Huziel E. Sauceda, Klaus-Robert Mueller
Summary: SpookyNet is a deep neural network that addresses the issue of electronic degrees of freedom and nonlocality typically ignored in machine-learned force fields. By incorporating chemically meaningful inductive biases and analytical corrections into the network architecture, SpookyNet can improve performance in quantum chemistry and fill important gaps in machine learning models.
NATURE COMMUNICATIONS
(2021)
Article
Biochemistry & Molecular Biology
Harsh Bhatia, Fikret Aydin, Timothy S. Carpenter, Felice C. Lightstone, Peer-Timo Bremer, Helgi I. Ingolfsson, Dwight V. Nissley, Frederick H. Streitz
Summary: Multiscale modeling in structural biology has a long history and aims to overcome the limitations of atomistic molecular dynamics in terms of time and length scales. Modern machine learning techniques, particularly deep learning, have revolutionized various scientific and engineering fields, including multiscale modeling. Deep learning has been successful in distilling information from fine-scale models and defining latent spaces for efficient exploration of conformational space. The integration of machine learning, multiscale simulation, and high-performance computing holds great promise for discovery and innovation in structural biology.
CURRENT OPINION IN STRUCTURAL BIOLOGY
(2023)
Article
Materials Science, Multidisciplinary
Zengqiang Cao, Xiaoyu Huang, Yanqing Wang, Chaoyang Zhang, Xianggui Xue, Guansong He, Hongyan Wang, Yuxiang Ni
Summary: In this study, the thermal stability of explosive HMX was improved by grafting different functional groups (hydroxyl, carboxyl, and butyl) on the surface of graphene. Molecular dynamics simulations were used to study the thermal transport across the graphene-HMX interface. The results showed that a covalent functionalization coverage rate of less than 7.5% was not beneficial for heat transfer, but increasing the coverage rate significantly enhanced the interfacial thermal conductance (ITC). Among the functional groups studied, butyl had the greatest impact on ITC, increasing it by 48.5% compared to pristine graphene-HMX. The results from molecular dynamics simulations were also used to predict the effective thermal conductivity of graphene-HMX composites using an effective medium theory-based model, and the major factors influencing the composite thermal conductivity were identified. This study enhances the understanding of heat transport in HMX composites and provides guidance for the structural design of thermally conductive HMX-based explosives.
JOURNAL OF MATERIALS SCIENCE
(2023)
Article
Chemistry, Physical
Taichi Inagaki, Shinji Saito
Summary: In this paper, a new hybrid method called potential scaling HMC (PS-HMC) is introduced to study complex chemical processes. By modulating the trajectory and gradually flattening the potential energy surface, the PS-HMC method is capable of constructing the canonical ensemble with a multimodal distribution. Applications to different molecular processes demonstrate the feasibility and features of this new method.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Physics, Applied
Ankit Mishra, Pankaj Rajak, Ayu Irie, Shogo Fukushima, Rajiv K. Kalia, Aiichiro Nakano, Ken-ichi Nomura, Fuyuki Shimojo, Priya Vashishta
Summary: This article presents a framework for calculating the refractive index of polymers, which includes methods such as high-throughput computation, machine learning models, and frequency-dependent calculations. The framework has been tested on a computational database.
APPLIED PHYSICS LETTERS
(2023)
Article
Chemistry, Physical
Jung Cho, Yifeng Yun, Hongyi Xu, Junliang Sun, Allen W. Burton, Karl G. Strohmaier, Gene Terefenko, Hilda Vroman, Mobae Afeworki, Guang Cao, Hao Wang, Xiaodong Zou, Tom Willhammar
Summary: The structure of novel medium-pore borosilicate zeolite EMM-25 was determined using continuous rotation electron diffraction (cRED). A new ammonium dication OSDA was found to reduce synthesis time significantly, and structural disorder in the form of swinging zigzag chains was observed in the EMM-25 framework.
CHEMISTRY OF MATERIALS
(2021)
Article
Chemistry, Physical
Min Li, WenCai Lu, John ZengHui Zhang
Summary: The study successfully extended the coarse-graining strategy to ultra-coarse-grained models and effectively parameterized liquid water, accurately predicting some important properties of liquid water. Additionally, two polarizable states of UCG molecules were observed after system equilibration. The results demonstrate that the proposed UCG models can accelerate liquid water simulation effectively.
JOURNAL OF CHEMICAL PHYSICS
(2021)
Article
Chemistry, Physical
Edoardo Cignoni, Lorenzo Cupellini, Benedetta Mennucci
Summary: We propose a machine learning strategy to calculate the excitonic properties of light harvesting complexes. By combining molecular dynamics simulations with ML prediction of the excitonic Hamiltonian, the proposed model can account for geometric fluctuations and electrostatic interactions. The model is trained on chlorophylls but can extrapolate beyond the training set, and its accuracy is demonstrated through simulations of absorption spectra in different environments.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Miguel Angel Gonzalez, Hiroshi Akiba, Oleg Borodin, Gabriel Julio Cuello, Louis Hennet, Shinji Kohara, Edward J. Maginn, Lucile Mangin-Thro, Osamu Yamamuro, Yong Zhang, David L. Price, Marie-Louise Saboungi
Summary: In this study, a systematic diffraction study of water-in-salt electrolytes and water-in-bisalt electrolytes was conducted using high-energy X-ray diffraction and polarized and unpolarized neutron diffraction. The measurements provided detailed information about the short- and intermediate-range order of the solutions. The experimental results were compared with molecular dynamics simulations, highlighting the differences between simulations and suggesting potential improvements for the force fields used in the simulations.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Joshua A. Rackers, Roseane R. Silva, Zhi Wang, Jay W. Ponder
Summary: A new empirical potential, the HIPPO force field, has been developed for efficient molecular dynamics simulation of water, balancing a wide range of water properties across different conditions. This model explicitly correlates water structure, dynamics, and thermodynamics with ab initio energy decomposition, offering comparable accuracy to previous polarizable atomic multipole force fields. The HIPPO water model can serve as a cornerstone for developing similarly detailed physics-based models for other molecular species.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2021)
Article
Chemistry, Physical
Andrea Rizzi, Paolo Carloni, Michele Parrinello
Summary: The method presented extends the theory of targeted free energy perturbation to accurately calculate free energy differences and surfaces at a quantum mechanical level from a cheaper reference potential. Accelerated convergence is achieved by increasing the overlap between target and reference distributions. The method is validated through numerical evaluations in different systems, demonstrating its effectiveness in avoiding systematic errors.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2021)
Article
Chemistry, Physical
Yeyue Xiong, Saeed Izadi, Alexey Onufriev
Summary: The article introduces a globally optimal polarizable water model, OPC3-pol, which accurately simulates water molecules at the atomic scale with minimal computational overheads. The model demonstrates improved computational efficiency and structure stability compared to existing approaches.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2022)
Article
Chemistry, Physical
Ming Ma, Junjie Song, Yi Dong, Weihai Fang, Lianghui Gao
Summary: In this study, a novel coarse-grained force field was developed to reproduce the structural and thermodynamic properties of triglycerides in bulk phase, as well as at air and water interfaces. The force field accurately reproduced the self-assembled network and diverse molecular conformations of triglycerides in water, and correctly predicted experimental macroscopic thermodynamic properties. This work paves the way for studying complex systems involving triglycerides on a larger scale.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Shiji Zhao, Piotr Cieplak, Yong Duan, Ray Luo
Summary: Accuracy and transferability are key properties of molecular mechanical force fields. Polarizable force fields are thought to have advantages in modeling atomic polarization effects. This study assessed the transferability of the electrostatic parameters in polarizable Gaussian multipole (pGM) models and found that they showed improved transferability compared to additive models. The observation that pGM models have better accuracy and transferability than additive models highlights the importance of intramolecular polarization effects. Overall, this study shows that pGM models possess accuracy and transferability, making them suitable for modeling polarization-sensitive biological systems and processes.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Sehr Naseem-Khan, Jean-Philip Piquemal, G. Andres Cisneros
Summary: The description of each separable contribution of the intermolecular interaction is important for the development of polarizable force fields. The Gaussian Electrostatic Model (GEM) has been improved to better compute Coulomb and exchange-repulsion energies as well as to implement a new dispersion formulation. These improvements have led to a better agreement with experimental results and ab initio references, demonstrating the significance of accurate reproduction for each separate contribution.
JOURNAL OF CHEMICAL PHYSICS
(2021)
