New implementation of high‐level correlated methods using a general block tensor library for high‐performance electronic structure calculations

This article presents an open‐source object‐oriented C++ library of classes and routines to perform tensor algebra. The primary purpose of the library is to enable post‐Hartree–Fock electronic structure methods; however, the code is general enough to be applicable in other areas of physical and computational sciences. The library supports tensors of arbitrary order (dimensionality), size, and symmetry. Implemented data structures and algorithms operate on large tensors by splitting them into smaller blocks, storing them both in core memory and in files on disk, and applying divide‐and‐conquer‐type parallel algorithms to perform tensor algebra. The library offers a set of general tensor symmetry algorithms and a full implementation of tensor symmetries typically found in electronic structure theory: permutational, spin, and molecular point group symmetry. The Q‐Chem electronic structure software uses this library to drive coupled‐cluster, equation‐of‐motion, and algebraic‐diagrammatic construction methods. © 2013 Wiley Periodicals, Inc.     

Calculation of X‐ray scattering intensities by means of the coupled cluster singles and doubles model

Total X‐ray scattering intensity σ_ee(q) is very sensitive to electron correlation effects. In this study σ_ee(q) of N₂, CO, and N₂O have been computed by the coupled cluster singles and doubles (CCSD) method and compared with configuration interaction singles and doubles (CISD) calculations as well as experimental observations. σ_ee(q) curves by CCSD calculations are rather close to those by CISD, but although small, there still exist some discrepancies between calculated and observed values. © 2001 John Wiley & Sons, Inc. J Comput Chem 22: 1315–1320, 2001     

Low‐lying singlet excited states of isocyanogen

Recent photofragment fluorescence excitation (PHOFEX) spectroscopy experiments have observed the ùA″ singlet excited state of isocyanogen (CNCN) for the first time. The observed spectrum is not completely assigned and significant questions remain about the excited states of this system. To provide insight into the energetically accessible excited states of CNCN, optimized geometries, harmonic vibrational frequencies, and excitation energies for the first three singlet excited states are determined using equation‐of‐motion coupled‐cluster theory with singles and doubles (EOM‐CCSD) and correlation‐consistent basis sets. Additionally, excited state coupled‐cluster methods which approximate the contributions from triples (CC3) are utilized to estimate the effect of higher‐order correlation on the energy of each excited state. For the ùA″ state, our best estimate for T₀ is about 42,200 cm^?1, in agreement with the experimentally estimated upper limit for the zero‐point level of 42,523 cm^?1. © 2008 Wiley Periodicals, Inc. Int J Quantum Chem, 2008     

New algorithms for iterative matrix‐free eigensolvers in quantum chemistry

New algorithms for iterative diagonalization procedures that solve for a small set of eigen‐states of a large matrix are described. The performance of the algorithms is illustrated by calculations of low and high‐lying ionized and electronically excited states using equation‐of‐motion coupled‐cluster methods with single and double substitutions (EOM‐IP‐CCSD and EOM‐EE‐CCSD). We present two algorithms suitable for calculating excited states that are close to a specified energy shift (interior eigenvalues). One solver is based on the Davidson algorithm, a diagonalization procedure commonly used in quantum‐chemical calculations. The second is a recently developed solver, called the “Generalized Preconditioned Locally Harmonic Residual (GPLHR) method.” We also present a modification of the Davidson procedure that allows one to solve for a specific transition. The details of the algorithms, their computational scaling, and memory requirements are described. The new algorithms are implemented within the EOM‐CC suite of methods in the Q‐Chem electronic structure program. © 2014 Wiley Periodicals, Inc.     

Assessment of the orbital‐optimized coupled‐electron pair theory for thermochemistry and kinetics: Improving on CCSD and CEPA(1)

An assessment of the orbital‐optimized coupled‐electron pair theory [or simply “optimized CEPA(0),” OCEPA(0)] [Bozkaya and Sherrill, J. Chem. Phys. 2013, 139, 054104] for thermochemistry and kinetics is presented. The OCEPA(0) method is applied to closed‐ and open‐shell reaction energies, barrier heights, and radical stabilization energies (RSEs). The performance of OCEPA(0) is compared with those of the MP2, CEPA(0), OCEPA(0), CEPA(1), coupled‐cluster singles and doubles (CCSD), and CCSD(T) methods [at complete basis set limits employing cc‐pVTZ and cc‐pVQZ basis sets]. For the most of the test sets, the OCEPA(0) method performs better than CEPA(0), CEPA(1), and CCSD, and provides accurate results. Especially, for open‐shell reaction energies and barrier heights, the OCEPA(0)–CEPA(1) and OCEPA(0)–CCSD differences become obvious. Similarly, for barrier heights and RSEs, the OCEPA(0) method improves on CEPA(0) by 1.6 and 2.3 kcal mol^?1. Our results demonstrate that the CEPA(0) method dramatically fails when the reference wave function suffers from the spin‐contamination problem. Conversely, the OCEPA(0) method can annihilate spin‐contamination in the unrestricted‐Hartree–Fock initial guess orbitals and can yield stable solutions. For overall evaluation, we conclude that the OCEPA(0) method is quite helpful not only for problematic open‐shell systems and transition‐states but also for closed‐shell molecules. Hence, one may prefer OCEPA(0) over CEPA(0), CEPA(1), and CCSD as an method, where N is the number of basis functions, for thermochemistry and kinetics. As discussed previously, the cost of the OCEPA(0) method is as much as of CCSD and CEPA(1) for energy computations. However, for analytic gradient computations, the OCEPA(0) method is two times less expensive than CCSD and CEPA(1). Further, the stationary properties of the OCEPA(0) method making it promising for excited state properties via linear response theory. © 2014 Wiley Periodicals, Inc.     

Accurate Quantum‐Chemical Prediction of Enthalpies of Formation of Small Molecules in the Gas Phase

The coupled‐cluster approach, including single and double excitations and perturbative corrections for triple excitations, is capable of predicting molecular electronic energies and enthalpies of formation of small molecules in the gas phase with very high accuracy (specifically, with error bars less than 5 kJ mol^?1), provided that the electronic wavefunction is dominated by the Hartree–Fock configuration. This capability is illustrated by calculations on molecules containing O–H and O–F bonds, namely OH, FO, H₂O, HOF, and F₂O. To achieve this very high accuracy, it is imperative to account for electron‐correlation effects in a quantitative manner, either by using explicitly correlated two‐particle basis functions (R12 functions) or by extrapolating to the limit of a complete basis. Besides taking into account harmonic zero‐point vibrational energies, it is also necessary to account for anharmonic corrections to the zero‐point vibrational energies, to include the core orbitals into the coupled‐cluster calculations, and to account for spin–orbit corrections and scalar relativistic effects. These additional corrections constitute small but significant contributions in the range of 1–4 kJ mol^?1 to the enthalpies of formation of the aforementioned molecules. The highly accurate coupled‐cluster results, obtained by employing R12 functions and by including various corrections, are compared with standard Kohn–Sham density‐functional calculations as well as with the Gaussian‐2 and complete‐basis‐set model chemistries.     

Relativistic and Non‐Relativistic Electronic Molecular‐Structure Calculations for Dimers of 4p‐, 5p‐, and 6p‐Block Elements

We report results of non‐relativistic and two‐component relativistic single‐reference coupled‐cluster with single and double and perturbative triple excitations [CCSD(T)] treatments for the 4p‐block dimers Ga₂ to Br₂, the 5p‐block dimers In₂ to I₂, and their atoms. Extended basis sets up to pentuple zeta are employed and energies extrapolated to the complete basis‐set limit. Relativistic and non‐relativistic results for the dissociation energy D_e are in close agreement with each other and previously published data, provided non‐relativistic or scalar‐relativistic results are corrected for spin–orbit contributions taken from the literature. An exception is Te₂ where theoretical results scatter by 0.085 eV. By virtue of this agreement it is unexpected that comparison with the experimental D₀ or D_e dissociation energies (zero‐point vibrational effects are negligible in this context) reveal errors larger than 0.1 eV for Ga₂, Ge₂, and Sb₂. Only relativistic treatments are presented for the 6p‐block cases Tl₂ to At₂. Sufficient agreement with experimental data is found only for Pb₂ and Bi₂, the deviation of the computed and experimental D₀ values for Po₂ is again larger than 0.1 eV. Deviations of 0.1 eV between the computed and experimental D₀ values are a major reason for concern and call for additional investigations in both fields to clarify the situation.     

Spin–Orbit Contributions in High‐Spin Nitrenes/Carbenes: A Hybrid CASSCF/MRMP2 Study of Zero‐Field Splitting Tensors

Zero‐field splitting (ZFS) tensors ( D tensors) of organic high‐spin oligonitrenes/oligocarbenes up to spin‐septet are quantitatively determined on the basis of quantum chemical calculations. The spin–orbit contributions, D ^SO tensors are calculated in terms of a hybrid CASSCF/MRMP2 approach, which was recently proposed by us. The spin–spin counterparts, D ^SS tensors are computed based on McWeeny–Mizuno’s equation in conjunction with the RODFT spin densities. The present calculations show that more than 10 % of ZFS arises from spin–orbit interactions in the high‐spin nitrenes under study. Contributions of spin‐bearing site–site interactions are estimated with the aid of a semi‐empirical model for the D tensors and found to be ca. 5 % of the D ^SO tensor. The analysis of intermediate states reveal that the largest contributions to the calculated D ^SO tensors are attributed to intra‐site spin flip excitations and delocalized π and π* orbitals play an important role in the inter‐site spin–orbit interactions.     

Gibbs energy of activation for thermal isomerization of (1Z)‐acetaldehyde hydrazone and (1Z)‐acetaldehyde N,N‐dimethylhydrazone by Gaussian‐4 theory and CCSD(T)/CBS computations

In this article, we examined the Gibbs energy of activation for the Z/E thermal isomerization reaction of (1Z)‐acetaldehyde hydrazone and (1Z)‐acetaldehyde N,N‐dimethylhydrazone, at 298.15 K in the solvent of cyclohexane. We carried out computations employing both the Gaussian‐4 (G4) theory and the coupled cluster method using both single and double substitutions and triple excitations noniteratively, CCSD(T). The CCSD(T) energy is extrapolated to the complete basis set (CBS). We compared the calculated results to the available experimental observation. It appeared that both G4 and CCSD(T)/CBS computations overestimated the experimental value by as much as about 6 and 12 kcal/mol in the present two cases. We discussed possible sources of error and proposed the experimental kinetic data could be questionable. © 2009 Wiley Periodicals, Inc. J Comput Chem, 2009     

Linear‐scaling self‐consistent field calculations based on divide‐and‐conquer method using resolution‐of‐identity approximation on graphical processing units

Graphical processing units (GPUs) are emerging in computational chemistry to include Hartree?Fock (HF) methods and electron‐correlation theories. However, ab initio calculations of large molecules face technical difficulties such as slow memory access between central processing unit and GPU and other shortfalls of GPU memory. The divide‐and‐conquer (DC) method, which is a linear‐scaling scheme that divides a total system into several fragments, could avoid these bottlenecks by separately solving local equations in individual fragments. In addition, the resolution‐of‐the‐identity (RI) approximation enables an effective reduction in computational cost with respect to the GPU memory. The present study implemented the DC‐RI‐HF code on GPUs using math libraries, which guarantee compatibility with future development of the GPU architecture. Numerical applications confirmed that the present code using GPUs significantly accelerated the HF calculations while maintaining accuracy. © 2014 Wiley Periodicals, Inc.     

Natural bond orbital‐based energy density analysis for correlated methods: Second‐order Møller–Plesset perturbation and coupled‐cluster singles and doubles

Natural bond orbital‐based energy density analysis (NBO‐EDA), which split energies into atomic and bonding contributions, is proposed for correlated methods such as coupled‐cluster singles and doubles (CCSD) and second‐order Møller–Plesset (MP2) perturbation. Applying NBO‐EDA for CCSD and MP2 to ethylene and the Diels–Alder reaction, we are successful in obtaining useful knowledge regarding electron correlation of π‐ and σ‐type orbitals, and clarifying the difference of the reaction barriers and heat of reaction calculated by CCSD and MP2. © 2008 Wiley Periodicals, Inc. Int J Quantum Chem, 2008     

DLPNO‐CCSD(T) scaled methods for the accurate treatment of large supramolecular complexes

In this work, we present scaled variants of the DLPNO‐CCSD(T) method, dubbed as (LS)DLPNO‐CCSD(T) and (NS)DLPNO‐CCSD(T), to obtain accurate interaction energies in supramolecular complexes governed by noncovalent interactions. The novel scaled schemes are based on the linear combination of the DLPNO‐CCSD(T) correlation energies calculated with the standard (LoosePNO and NormalPNO) and modified (Loose2PNO and Normal2PNO) DLPNO‐CCSD(T) accuracy levels. The scaled DLPNO‐CCSD(T) variants provide nearly TightPNO accuracy, which is essential for the quantification of weak noncovalent interactions, with a noticeable saving in computational cost. Importantly, the accuracy of the proposed schemes is preserved irrespective of the nature and strength of the supramolecular interaction. The (LS)DLPNO‐CCSD(T) and (NS)DLPNO‐CCSD(T) protocols have been used to study in depth the role of the CH–π versus π–π interactions in the supramolecular complex formed by the electron‐donor truxene‐tetrathiafulvalene (truxTTF) and the electron‐acceptor hemifullerene (C₃₀H₁₂). (NS)DLPNO‐CCSD(T)/CBS calculations clearly reveal the higher stability of staggered (dominated by CH–π interactions) versus bowl‐in‐bowl (dominated by π–π interactions) arrangements in the truxTTF•C₃₀H₁₂ heterodimer. Hemifullerene and similar carbon‐based buckybowls are therefore expected to self‐assemble with donor compounds in a richer way other than the typical concave–convex π–π arrangement found in fullerene‐based aggregates. © 2017 Wiley Periodicals, Inc.     

Computerized implementation of higher‐order electron‐correlation methods and their linear‐scaling divide‐and‐conquer extensions

We have implemented a linear‐scaling divide‐and‐conquer (DC)‐based higher‐order coupled‐cluster (CC) and Møller–Plesset perturbation theories (MPPT) as well as their combinations automatically by means of the tensor contraction engine, which is a computerized symbolic algebra system. The DC‐based energy expressions of the standard CC and MPPT methods and the CC methods augmented with a perturbation correction were proposed for up to high excitation orders [e.g., CCSDTQ, MP4, and CCSD(2)_TQ]. The numerical assessment for hydrogen halide chains, polyene chains, and first coordination sphere (C1) model of photoactive yellow protein has revealed that the DC‐based correlation methods provide reliable correlation energies with significantly less computational cost than that of the conventional implementations. © 2017 Wiley Periodicals, Inc.     

Acceleration of the GAMESS‐UK electronic structure package on graphical processing units

The approach used to calculate the two‐electron integral by many electronic structure packages including generalized atomic and molecular electronic structure system‐UK has been designed for CPU‐based compute units. We redesigned the two‐electron compute algorithm for acceleration on a graphical processing unit (GPU). We report the acceleration strategy and illustrate it on the (ss|ss) type integrals. This strategy is general for Fortran‐based codes and uses the Accelerator compiler from Portland Group International and GPU‐based accelerators from Nvidia. The evaluation of (ss|ss) type integrals within calculations using Hartree Fock ab initio methods and density functional theory are accelerated by single and quad GPU hardware systems by factors of 43 and 153, respectively. The overall speedup for a single self consistent field cycle is at least a factor of eight times faster on a single GPU compared with that of a single CPU. © 2011 Wiley Periodicals, Inc. J Comput Chem, 2011     

Theoretical determination of lowest structures of all‐metal aromatic clusters M4L2 (M = Al,Ga, In,Tl; L = Li,Na, K,Rb, Cs)

The researches of all‐metal aromatic clusters have been a thermic theme in inorganic aromaticity domain both experimentally and theoretically since the Al₄L^? (L = Li, Na, Cu) clusters were created by laser vaporization. In systemic determination of the lowest structures of 20 gaseous all‐metal aromatic clusters M₄L₂ (M = Al, Ga, In, Tl; L = Li, Na, K, Rb, Cs), the isomer energy differences of four low‐lying structures of each cluster were evaluated at high‐quality quantum chemistry levels. Single point calculations at the coupled cluster level were performed at geometries optimized at the MP2, B3LYP, and B3PW91 levels, and harmonic frequency calculations and zero point energy corrections were implemented following optimizations at the B3LYP and B3PW91 levels. In addition to Li‐ and Na‐containing species, theoretical investigations came down to those new clusters including K, Rb, and Cs. For many clusters, the most convincing theoretical evidences indicate that the lowest structures are a square bipyramidal isomer rather than an edge‐caped square pyramidal species. A few discrepancies were addressed at the MP2, B3LYP, and B3PW91 levels in comparison with the coupled cluster results. These findings are significant because some clusters were generated by laser vaporization and served as theoretical prototypes to test the new means for assessing aromaticity. © 2009 Wiley Periodicals, Inc. Int J Quantum Chem, 2010     

Reduced‐cost sparsity‐exploiting algorithm for solving coupled‐cluster equations

We present an algorithm for reducing the computational work involved in coupled‐cluster (CC) calculations by sparsifying the amplitude correction within a CC amplitude update procedure. We provide a theoretical justification for this approach, which is based on the convergence theory of inexact Newton iterations. We demonstrate by numerical examples that, in the simplest case of the CCD equations, we can sparsify the amplitude correction by setting, on average, roughly 90% nonzero elements to zeros without a major effect on the convergence of the inexact Newton iterations.     

The Evolution of Bonding and Thermodynamic Properties of Boron‐Doped Small Carbon Clusters: An Ab Initio Study

Theoretical studies on BC_n (n=1–6) clusters are carried out using density functional theory, Møller–Plesset second‐order perturbation theory (MP2), coupled‐cluster calculations including up to triple excitations (CCSD(T)), and higher‐level approaches. All possible isomers depending on the positions of the boron atom are generated and the lowest‐energy isomers are determined for doublet and quartet electronic states. The three potential evolution paths of the clusters are determined as a function of their size. The energetic and electronic consequences for the increased size of structures differ significantly, which leads to representatives of the ground electronic state from different structural groups. The ab initio calculated thermal functions allow enhancements to the available atomization energies and improve the agreement between the calculated and experimental heat content.     

A new approach for second‐order perturbation theory

A new second‐order perturbation theory (MP2) approach is presented for closed shell energy evaluations. The new algorithm has a significantly lower memory footprint, a lower FLOP (floating point operations) count, and a lower time to solution compared to previously implemented parallel MP2 methods in the GAMESS software package. Additionally, this algorithm features an adaptive approach for the disk/distributed memory storage of the MP2 amplitudes. The algorithm works well on a single workstation, small cluster, and large Cray cluster, and it allows one to perform large calculations with thousands of basis functions in a matter of hours on a single workstation. The same algorithm has been adapted for graphical processing unit (GPU) architecture. The performance of the new GPU algorithm is also discussed. © 2016 Wiley Periodicals, Inc.