Journal of Chemical Theory and Computation最新文献

The neural network quantum state (NNQS) method has demonstrated promising results in ab initio quantum chemistry, achieving remarkable accuracy in molecular systems. However, efficient calculation of systems with large active spaces remains challenging. This study introduces a novel approach that bridges tensor network states with the transformer-based NNQS-Transformer (QiankunNet) to enhance accuracy and convergence for systems with relatively large active spaces. By transforming tensor network states into active space configuration interaction type wave functions, QiankunNet achieves accuracy surpassing both the pretraining density matrix renormalization group (DMRG) results and traditional coupled cluster methods, particularly in strongly correlated regimes. We investigate two configuration transformation methods: the sweep-based direct conversion (Conv.) method and the entanglement-driven genetic algorithm (EDGA) method, with Conv. showing superior efficiency. The effectiveness of this approach is validated on H₂O with a large active space (10e, 24o) in the cc-pVDZ basis set, demonstrating an efficient routine between DMRG and QiankunNet and also offering a promising direction for advancing quantum state representation in complex molecular systems.

We present an efficient linear-scaling algorithm for evaluating the analytical force and stress contributions derived from the exact-exchange energy, a key component in hybrid functional calculations. The algorithm, working equally well for molecular and periodic systems, is formulated within the framework of numerical atomic orbital (NAO) basis sets and takes advantage of the localized resolution-of-identity (LRI) technique for treating the two-electron Coulomb repulsion integrals. The linear-scaling behavior is realized by fully exploiting the sparsity of the expansion coefficients resulting from the strict locality of the NAOs and the LRI ansatz. Our implementation is massively parallel, and enables efficient structural relaxation based on hybrid density functionals for bulk materials containing thousands of atoms. In this work, we will present a detailed description of our algorithm and benchmark the performance of our implementation using illustrating examples. By optimizing the structures of the pristine and doped halide perovskite material CsSnI₃ with different functionals, we find that in the presence of lattice strain, hybrid functionals provide a more accurate description of the stereochemical expression of the lone pair.

This study introduces correlated flat-bottom elastic network model (CFB-ENM), an extension of our recently developed flat-bottom elastic network model (FB-ENM) for generating plausible reaction paths, i.e., collision-free paths preserving nonreactive parts. While FB-ENM improved upon the widely used image-dependent pair potential (IDPP) by addressing unintended structural distortion and bond breaking, it still struggled with regulating the timing of series of bond breaking and formation. CFB-ENM overcomes this limitation by incorporating structure-based correlation terms. These terms impose constraints on pairs of atom pairs, ensuring immediate formation of new bonds after breaking of existing bonds. Using the direct MaxFlux method, we generated paths for 121 reactions involving main group elements and 35 reactions involving transition metals. We found that CFB-ENM significantly improves reaction paths compared to FB-ENM. CFB-ENM paths exhibited lower maximum DFT energies along the paths in most reactions, with nearly half showing significant energy reductions of several tens of kcal/mol. In the few cases where CFB-ENM yielded higher energy paths, most increases were below 10 kcal/mol. We also confirmed that CFB-ENM reduces computational costs in subsequent precise reaction path or transition state searches compared to FB-ENM. An implementation of CFB-ENM based on the Atomic Simulation Environment is available on GitHub for use in computational chemistry research.

Physics-informed neural networks (PINN) is a machine learning (ML)-based method to solve partial differential equations that has gained great popularity due to the fast development of ML libraries in the past few years. The Poisson-Boltzmann equation (PBE) is widely used to model mean-field electrostatics in molecular systems, and in this work we present a detailed investigation of the use of PINN to solve the linear PBE. Starting from a multidomain PINN for the linear PBE with an interface, we assess the importance of incorporating different features into the neural network architecture. Our findings indicate that the most accurate architecture utilizes input and output scaling layers, a random Fourier features layer, trainable activation functions, and a loss balancing algorithm. The accuracy of our implementation is on the order of 10^-2-10^-3, which is similar to previous work using PINN to solve other differential equations. We also explore the possibility of incorporating experimental information into the model, and discuss challenges and future work, especially regarding the nonlinear PBE. We are providing an open-source implementation to easily perform computations from a PDB file. We hope this work will motivate application scientists into using PINN to study molecular electrostatics, as ML technology continues to evolve at a high pace.

Simulating water accurately has been a challenge due to the complexity of describing polarization and intermolecular charge transfer. Quantum mechanical (QM) electronic structures provide an accurate description of polarization in response to local environments, which is nevertheless too expensive for large water systems. In this study, we have developed a polarizable water model integrating Charge Model 5 atomic charges at the level of the second-order Mo̷ller-Plesset perturbation theory, predicted by an accurate and transferable charge neural network (ChargeNN) model. The spontaneous intermolecular charge transfer has been explicitly accounted for, enabling a precise treatment of hydrogen bonds and out-of-plane polarization. Our ChargeNN water model successfully reproduces various properties of water in gas, liquid, and solid phases. For example, ChargeNN correctly captures the hydrogen-bond stretching peak and bending-libration combination band, which are absent in the spectra using fixed charges, highlighting the significance of accurate polarization and charge transfer. Finally, the molecular dynamical simulations using ChargeNN for liquid water and a large water droplet with a ∼4.5 nm radius reveal that the strong interfacial electric fields are concurrently induced by the partial collapse of the hydrogen-bond network and surface-to-interior charge transfer. Our study paves the way for QM-polarizable force fields, aiming for large-scale molecular simulations with high accuracy.

Conical intersections directly mediate the internal energy conversion in photoinduced processes in a wide range of chemical and biological systems. Because of the Brillouin theorem, many conventional electronic structure methods, including configuration interaction with single excitations from a Hartree-Fock reference and time-dependent density functional theory in either the linear response approximation (TDDFT) or Tamm-Dancoff approximation (DFT-TDA), have the wrong dimensionality for conical intersections between the ground state (S₀) and the first excited state (S₁) of the same multiplicity. This leads to unphysical state crossings. Here, we implement and assess the configuration-interaction-corrected Tamm-Dancoff approximation (CIC-TDA) that restores the correct dimensionality of conical intersections by including the coupling between the reference state and the intersecting excited state. We apply the CIC-TDA method to the S₁/S₀ conical intersections in ammonia (NH₃), ethylene (C₂H₄), bithiophene (C₈H₆S₂), azobenzene (C₁₂H₁₀N₂), and 11-cis retinal protonated Schiff base (PSB11) in vacuo. We show that this black-box approach can produce potential energy surfaces (PESs) of comparable accuracy to multireference wave function methods. The method validated here can allow cost-efficient explorations of photoinduced electronically nonadiabatic dynamics, especially for large molecules and complex systems.

Electrostatic preorganization is an exciting mode to understand the catalytic function of enzymes, yet limited tools exist to computationally analyze it. In particular, no methods exist to interpret the geometry, dynamics, and fundamental components of 3D electric fields, E⃗(r), in protein active sites. To address this, we present PyCPET (Python Computation of Electric Field Topologies), a comprehensive, open-source toolbox to analyze E⃗(r) in enzymes. We designed it around computational efficiency and user friendliness with both CPU- and GPU-accelerated codes. Our aim is to provide a set of functions for rich, descriptive analysis of enzyme systems including dynamics, benchmarking, distribution of streamlines analysis in 3D E⃗(r), computation of point E⃗(r), principal component analysis, and 3D E⃗(r) visualization. Finally, we demonstrate its versatility by exploring the nature of electrostatic preorganization and dynamics in three cases: Cytochrome C, Co-substituted Liver Alcohol Dehydrogenase, and HIV Protease. These test systems, along with previous work, establish PyCPET as an essential toolkit for the in-depth analysis and visualization of electric fields in enzymes, unlocking new avenues for understanding electrostatic contributions to enzyme catalysis.

The transport of ions through channels involves multiple rare-event transitions through a web of interconnected intermediates. Extracting open channel mechanisms generally requires quantifying the relative flux through these intermediates in response to a range of electrochemical gradients. Although this is ideally suited to network-based representations like Markov state models (MSMs), the relative contributions from different pathways and the importance of network resolution remain open areas of research. Herein, we use a complementary approach called multiscale responsive kinetic modeling (MsRKM) to explore how the screening of ionic interactions and the competition between multiple mechanistic pathways contribute to channel mechanisms and current profiles of ion channels. We find that explicitly optimizing screened ionic interactions in the MsRKM framework vastly reduces the solution search space, enabling more efficient identification of physically robust solutions. Using a model of the Shaker Kv channel, we demonstrate that even when systems are well described by a single dominant flux pathway, the remaining contributing pathways and off-pathway flux play multiple essential roles, including shifting current profiles and mechanisms in response to different electrochemical gradients. We additionally discover that the current continues to change above the experimentally predicted saturation point. Model systems explain how the degree of dielectric screening influences channel occupancy, the number of contributing pathways, and why current increases or decreases above its experimental saturation point. Our findings emphasize the importance of retaining a full network description to identify and understand ion channel mechanisms.

The isotropic periodic sum (IPS) method provides an efficient approach for computing long-range interactions by approximating distant molecular structures through isotropic periodic images of a local region. Here, we present a novel integration of IPS with the polarizable Gaussian multipole (pGM) model, extending its applicability to systems with Gaussian-distributed charges and dipoles. By developing and implementing the IPS multipole tensor theorem within the Gaussian multipole framework, we derive analytical expressions for IPS potentials that efficiently handle both permanent and induced multipole interactions. Our comprehensive validation includes energy conservation tests in the NVE ensemble, potential energy distributions in the NVT ensemble, structural analysis through radial distribution functions, diffusion coefficients, induced dipole calculations across various molecular systems, and ionic charging free energies. The results demonstrate that the pGM–IPS approach successfully reproduces energetic, structural, and dynamic properties of molecular systems with accuracy comparable to the traditional particle mesh Ewald method. Our work establishes pGM–IPS as a promising method for simulations of polarizable molecular systems, achieving a balance between computational efficiency and accuracy.

The isotropic periodic sum (IPS) method provides an efficient approach for computing long-range interactions by approximating distant molecular structures through isotropic periodic images of a local region. Here, we present a novel integration of IPS with the polarizable Gaussian multipole (pGM) model, extending its applicability to systems with Gaussian-distributed charges and dipoles. By developing and implementing the IPS multipole tensor theorem within the Gaussian multipole framework, we derive analytical expressions for IPS potentials that efficiently handle both permanent and induced multipole interactions. Our comprehensive validation includes energy conservation tests in the NVE ensemble, potential energy distributions in the NVT ensemble, structural analysis through radial distribution functions, diffusion coefficients, induced dipole calculations across various molecular systems, and ionic charging free energies. The results demonstrate that the pGM-IPS approach successfully reproduces energetic, structural, and dynamic properties of molecular systems with accuracy comparable to the traditional particle mesh Ewald method. Our work establishes pGM-IPS as a promising method for simulations of polarizable molecular systems, achieving a balance between computational efficiency and accuracy.