Article
Chemistry, Physical
Tim Gould, Leeor Kronik
Summary: This study reviews two important extensions of Kohn-Sham (KS) theory: generalized KS theory and ensemble KS theory. The former allows for non-multiplicative potential operators while the latter enables the treatment of quantum ensembles, offering improvements in efficiency for practical calculations. The research combines these two extensions and explores a computationally tractable variant for handling complex scenarios.
JOURNAL OF CHEMICAL PHYSICS
(2021)
Article
Chemistry, Physical
Andrew D. Mahler, Lee M. Thompson
Summary: Nonorthogonal approaches to electronic structure methods are gaining renewed attention due to the potential of new forms of nonorthogonal wavefunction Ansatze to overcome computational bottlenecks associated with orthogonal-based methods. The efficiency of nonorthogonal configuration interaction is determined by the basis in which it is performed, highlighting the importance of selecting suitable determinant basis functions. Optimizing basis determinants from an arbitrarily constructed initial set can help address issues such as exponential scaling of determinant space and potential linear dependencies in the basis.
JOURNAL OF CHEMICAL PHYSICS
(2021)
Article
Chemistry, Physical
Jacques K. Desmarais, Alberto Boccuni, Jean-Pierre Flament, Bernard Kirtman, Alessandro Erba
Summary: This article generalizes a previously proposed noncanonical coupled-perturbed Kohn-Sham density functional theory (KS-DFT)/Hartree-Fock (HF) treatment for spin-orbit coupling to infinite periodic systems. The approach is validated and implemented in the CRYSTAL program, and calculations are performed on the tungsten dichalcogenide hexagonal bilayer series. The results show good agreement with reference two-component self-consistent field (2c-SCF) calculations for total energy, electronic band structure, and density variables of spin-current DFT.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Benjamin G. Janesko
Summary: The Perdew-Zunger self-interaction correction (PZSIC) introduces an exact constraint to approximate density functional theory (DFT), but it can paradoxically degrade performance and standard DFT approximations are not systematically improvable. In this study, the adiabatic projection formalism is used to derive PZSIC in terms of a reference system experiencing only electron self-interaction. The generalization to a self-and-someothers interaction introduces correlation and systematically bridges from PZSIC to exact wave function theory without double counting of correlation. The minimal active spaces method accurately handles nearly one-electron, near-equilibrium, and strongly correlated model systems at a modest computational expense.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2022)
Article
Nanoscience & Nanotechnology
Chiara Ribaldone, Silvia Casassa
Summary: In the framework of ab initio simulations, finding energy minimum atomic structures is crucial for studying the properties of a system. Recent studies suggest that the Fast Inertial Relaxation Engine (Fire) algorithm, based on molecular dynamics concepts, performs better than line-search based schemes in finding local minima. This study describes the implementation of Fire in the Crystal ab initio quantum mechanical simulation package and compares its efficiency and performance with the Bfgs quasi-Newton scheme. Results show that Fire converges well and reproduces the minimum energy structures obtained by the Bfgs approach, especially for systems without symmetry and a large number of atoms.
Article
Chemistry, Physical
Frank Hu, Francis He, David J. Yaron
Summary: Quantum chemistry provides valuable information for chemists, but the computational cost limits its applications. Machine learning has emerged as a cost-effective and accurate solution, but it often lacks interpretability. This study demonstrates that semiempirical quantum chemical models can achieve high accuracy and interpretability while learning from large volumes of data.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Multidisciplinary Sciences
Inbar Seroussi, Gadi Naveh, Zohar Ringel
Summary: Deep neural networks (DNNs) have complex structures, making direct microscopic analysis difficult. However, by identifying slow variables that average the behavior of fast variables, the authors have found a similar scale separation in finite deep convolutional neural networks and fully connected networks. They show that DNN layers couple through the second cumulant of their activations and pre-activations, which fluctuate in a Gaussian manner. The resulting thermodynamic theory of deep learning provides accurate predictions and new ways to understand DNNs.
NATURE COMMUNICATIONS
(2023)
Article
Chemistry, Physical
Laurent Lemmens, Xeno De Vriendt, Patrick Bultinck, Guillaume Acke
Summary: In this study, the spin phases and phase transitions during molecular dissociation in the presence of an external uniform magnetic field were explored. It was demonstrated that these spin phases can be interpreted and described by constraining the states to target expectation values of projected spin. The constrained states offer a complete electronic characterization and can be analyzed using standard quantum chemical tools.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2022)
Article
Computer Science, Interdisciplinary Applications
P-G Reinhard, B. Schuetrumpf, J. A. Maruhn
Summary: The SKYAx code is based on the nuclear mean-field model using Skyrme forces, predicting properties of nuclear ground states with axial symmetry and BCS pairing included. It is implemented in Fortran 90 and allows for fast computational speed. The code is restricted to even-even nuclei and is suitable for nuclei well-bound within the nuclear chart.
COMPUTER PHYSICS COMMUNICATIONS
(2021)
Article
Physics, Mathematical
Rodrigo Matos, Jeffrey Schenker
Summary: Using the fractional moment method, it is proven that weakly interacting Fermions at positive temperature in the Hartree-Fock approximation of the disordered Hubbard Hamiltonian exhibit localization, characterized by exponential decay of eigenfunction correlators. This result is valid in any dimension with large disorder and in the one-dimensional case with any disorder. The study also demonstrates Holder continuity of the integrated density of states with respect to energy, disorder, and interaction.
COMMUNICATIONS IN MATHEMATICAL PHYSICS
(2021)
Article
Multidisciplinary Sciences
Johannes T. Margraf, Karsten Reuter
Summary: Density-functional theory is a rigorous framework for describing the ground state properties of atoms, molecules and solids, while a new machine-learning based DFA called Kernel Density Functional Approximation has been developed, which is pure, non-local and transferable, showing remarkable possibilities for diverse interactions and system sizes.
NATURE COMMUNICATIONS
(2021)
Article
Chemistry, Physical
Artem Mitrofanov, Nikolai Andreadi, Vadim Korolev, Stepan Kalmykov
Summary: Actinide chemistry often faces limitations with modern theoretical tools due to high computational costs and relativistic effects. The novel relPBE functional serves as an actinide-fitted version of the PBE0 functional, offering an alternative computational approach.
JOURNAL OF CHEMICAL PHYSICS
(2021)
Article
Chemistry, Physical
Javad Shirani, Sinan Abi Farraj, Shuaishuai Yuan, Kirk H. Bevan
Summary: This work explores the relative accuracy of a hybrid functional in predicting redox properties while satisfying the general form of Koopmans' theorem. Through direct comparison and first-principles calculations, it is found that employing a total energy difference approach can provide theoretically robust estimates of redox energies when using hybrid functionals.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Emily M. Kempfer-Robertson, Andrew D. Mahler, Meagan N. Haase, Piper Roe, Lee M. Thompson
Summary: The nonorthogonal active space decomposition (NO-ASD) methodology is proposed for systems with multiple correlation mechanisms. It reduces the factorial scaling associated with the size of the correlated orbital space by partitioning the wave function and treating different correlation mechanisms with an effective Hamiltonian approach. Despite its nonorthogonal framework, NO-ASD can take advantage of efficient matrix element evaluation. Comparisons with complete active space methods show that NO-ASD has advantages in reducing problem dimensionality and recovering correlation energy.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2022)
Article
Chemistry, Physical
Francesco Ambrosio, Julia Wiktor, Alessandro Landi, Andrea Peluso
Summary: The performance of Koopmans-compliant hybrid functionals in reproducing the electronic structure of organic crystals is examined. The calculated band gaps are in agreement with the GW method but at a lower computational cost. The impact of these results on the transport properties of acene crystals is discussed.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2023)
Article
Chemistry, Multidisciplinary
Robert Izsak, Christoph Riplinger, Nick S. Blunt, Bernardo de Souza, Nicole Holzmann, Ophelia Crawford, Joan Camps, Frank Neese, Patrick Schopf
Summary: Quantum computers are expected to be useful in simulating strongly correlated chemical systems, but careful selection of orbital spaces is crucial. This study presents a scheme for automatically selecting an active space and combines quantum phase estimation and variational quantum eigensolver algorithms to accurately describe the environment. The protocol outlined here can be applied to chemical systems of any size, including those beyond the capabilities of classical computation.
JOURNAL OF COMPUTATIONAL CHEMISTRY
(2023)
Article
Chemistry, Multidisciplinary
Frank Neese
Summary: This paper describes the SHARK integral generation and digestion engine, which is based on the McMurchie/Davidson approach and utilizes an efficient BLAS algorithm. SHARK can handle various types of basis function integrals and features programming constructs that simplify workflows and avoid code duplication.
JOURNAL OF COMPUTATIONAL CHEMISTRY
(2023)
Article
Chemistry, Physical
Robert Izsak
Summary: A second quantized formulation for evaluating spin-dependent properties in the UHF basis is introduced in this article. The practical usage of this formulation is illustrated using the configuration interaction singles ansatz, and the results are compared to spin-restricted cases of the same ansatz. On a more tentative basis, the notion of quasi-spin-adaptation is also discussed, which ensures that the ansatz becomes spin-adapted as the unrestricted orbitals approach the restricted ones.
Editorial Material
Chemistry, Physical
Conner Masteran, Ashutosh Kumar, Nakul Teke, Bimal Gaudel, Takeshi Yanai, Edward F. Valeev
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Karl Pierce, Edward F. Valeev
Summary: This paper addresses the problem of constructing a canonical polyadic (CP) decomposition for a tensor network rather than a single tensor. It demonstrates how leveraging the structure of the network during CP factor optimization can reduce the complexity of constructing an approximate CP representation. The utility of this technique is shown for approximating the order-4 Coulomb interaction tensor with two order-3 tensors using an approximate generalized square-root (SQ) factorization.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Andreas Erbs Hillers-Bendtsen, Dmytro Bykov, Ashleigh Barnes, Dmitry Liakh, Hector H. Corzo, Jeppe Olsen, Poul Jorgensen, Kurt V. Mikkelsen
Summary: We have developed a massively parallel implementation of the CPS(D-3) excitation energy model based on cluster perturbation theory. The extended algorithm efficiently scales with increasing computational resources and allows for low time-to-solution calculations on large molecular systems. The CPS(D-3) excitation energies are shown to be a computationally efficient alternative to those obtained from the coupled-cluster singles and doubles model.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Yang Guo, Fabijan Pavosevic, Kantharuban Sivalingam, Ute Becker, Edward F. F. Valeev, Frank Neese
Summary: In this work, a linear scaling explicitly correlated N-electron valence state perturbation theory (NEVPT2-F12) is developed to reduce the computational scaling of the conventional NEVPT2-F12 to near-linear scaling using the idea of a domain-based local pair natural orbital (DLPNO). The excitation energies predicted by DLPNO-NEVPT2-F12 are as accurate as the exact NEVPT2-F12 results for low-lying excited states of organic molecules. The new algorithm is applied to study cluster models of rhodopsin and can handle systems with large basis functions and active spaces.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
David B. Williams-Young, Andrey Asadchev, Doru Thom Popovici, David Clark, Jonathan Waldrop, Theresa L. Windus, Edward F. Valeev, Wibe A. de Jong
Summary: With the increasing reliance on accelerator-based architecture in modern supercomputers, the development and optimization of electronic structure methods to utilize massively parallel resources has become a recent priority. This work presents distributed memory algorithms for evaluating the Coulomb and exact exchange matrices for hybrid Kohn-Sham DFT with Gaussian basis sets using direct density-fitted and seminumerical methods. The performance and scalability of these methods are demonstrated on systems ranging from a few hundred to over one thousand atoms using up to 128 NVIDIA A100 GPUs on the Perlmutter supercomputer.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Andrey Asadchev, Edward F. Valeev
Summary: In order to improve the efficiency of Gaussian integral evaluation on modern accelerated architectures, FLOP-efficient Obara-Saika-based recursive evaluation schemes are optimized to reduce memory usage. The use of multiquantal recurrences is shown to save significant memory for evaluating 3-center 2-particle integrals. Other innovations include leveraging register memory and compile-time features of modern C++/CUDA. The performance of the proposed schemes is demonstrated for various types of integrals and the implementation is available in the open-source LibintX library.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Robert Izsak, Aleksei V. Ivanov, Nick S. Blunt, Nicole Holzmann, Frank Neese
Summary: In this article, the different measures of electron correlation in wave function theory, density functional theory, and quantum information theory are briefly reviewed. The focus is then placed on a more traditional metric based on dominant weights in the full configuration solution and its behavior regarding the choice of N-electron and one-electron basis. The impact of symmetry is discussed, and the usefulness of distinguishing between determinants, configuration state functions, and configurations as reference functions is emphasized.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Yuqi Wang, Yang Guo, Frank Neese, Edward F. Valeev, Wei Li, Shuhua Li
Summary: In this article, a series of explicitly correlated local correlation methods developed under the cluster-in-molecule (CIM) framework are presented. These methods allow F12 calculations of large molecules on a single node. The authors used these methods to investigate the relative stability between extended and folded alkane, the stability of different structures of polyglycine, and the binding energies of host-guest complexes. The results demonstrate the promising potential of combining CIM with F12 methods for studying large molecules.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Edward F. Valeev, Robert J. Harrison, Adam A. Holmes, Charles C. Peterson, Deborah A. Penchoff
Summary: Optimizing nearly exact orthonormal orbitals in real space using a multiresolution spectral element basis allows for evaluation of energy of arbitrary states of atoms and molecules. This method provides more accurate electronic energies compared to Gaussian AO bases, and is suitable for high-end models of correlated electronic states. The approach can optimize over 100 fully correlated numerical orbitals on a single computer node, with significant room for improvement.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Multidisciplinary
Ingolf Harden, Frank Neese, Giovanni Bistoni
Summary: This study investigates the aggregation process of chiral organocatalysts derived from phosphoric acids using high-level quantum mechanical calculations. The results show that the formation of catalyst dimers is possible and the aggregation process can significantly affect the stereo-controlling factors, reaction kinetics, and selectivity of the transformations.
Article
Chemistry, Multidisciplinary
Ingolf Harden, Frank Neese, Giovanni Bistoni
Summary: The formation of Bronsted acid aggregates in asymmetric organocatalytic reactions has a significant impact on the stereo-controlling factors of the transformations. This study uses high-level quantum mechanical calculations to investigate the influence of catalyst structure and reaction conditions on the spontaneity of the aggregation process, shedding light on its importance.