From Hartree–Fock and Heitler–London to chemical orbitals

For chemistry the theoretical representation of the forces connecting atoms in molecules was and is a central problem. The Atomic Orbital and the Molecular Orbital are basic building blocks in the Heitler–London (HL) and in the Linear Combination of Atomic Orbitals–Molecular Orbital (LCAO-MO) methods, which have lead to the construction of modern Valence Bond and Hartree–Fock methods (and related extensions). However, accurate predictions from non semi-empirical methods often require enormous amount of computer power, if applied to molecules of reasonable size and current chemical interest. We have critically re-examined the two basic methods and suggested a few extensions. Merging of the Hartree–Fock with the Heitler–London algorithms, as recently proposed in the Hartree–Fock–Heitler–London (HF–HL) method, reduces the length of the expansions needed in AO or MO ab initio models in the computation of binding energy; this simplification allows easy interpretation of the resulting wave function. The HF–HL method is exemplified with systematic computations on ground and excited state of the hydrides and homonuclear diatomic molecules with atoms of the first and second period of the periodic table. Further, we show that the HF–HL method is derivable from a wave function constructed with a new type of orbital, the Chemical orbital (CO), which embodies the characterization of MO near equilibrium, AO at dissociation and at the united atom. Preliminary computations with CO are included. The new method provides the conceptual origin of both the HF and VB approaches, thus the foundation of an 80 years effort in variational quantum chemistry.     

Parallel Fock matrix construction with distributed shared memory model for the FMO‐MO method

A parallel Fock matrix construction program for FMO‐MO method has been developed with the distributed shared memory model. To construct a large‐sized Fock matrix during FMO‐MO calculations, a distributed parallel algorithm was designed to make full use of local memory to reduce communication, and was implemented on the Global Array toolkit. A benchmark calculation for a small system indicates that the parallelization efficiency of the matrix construction portion is as high as 93% at 1,024 processors. A large FMO‐MO application on the epidermal growth factor receptor (EGFR) protein (17,246 atoms and 96,234 basis functions) was also carried out at the HF/6‐31G level of theory, with the frontier orbitals being extracted by a Sakurai‐Sugiura eigensolver. It takes 11.3 h for the FMO calculation, 49.1 h for the Fock matrix construction, and 10 min to extract 94 eigen‐components on a PC cluster system using 256 processors. © 2010 Wiley Periodicals, Inc. J Comput Chem, 2010     

Implementation of a parallel direct SCF algorithm on distributed memory computers

A parallel direct self-consistent field (SCF) algorithm for distributed memory computers is described. Key features of the algorithm are its ability to achieve a load balance dynamically, its modest memory requirements per processor, and its ability to utilize the full eightfold index permutation symmetry of the two-electron integrals despite the fact that entire copies of the Fock and density matrices are not present in each processor's local memory. The algorithm is scalable and, accordingly, has the potential to function efficiently on hundreds of processors. With the algorithm described here, a calculation employing several thousand basis functions can be carried out on a distributed memory machine with 100 or more processors each with just 4 MBytes of RAM and no disk. The Fock matrix build portion of the algorithm has been implemented on a 16-node Intel iPSC/2. Results from benchmark calculations are encouraging. The algorithm shows excellent load balance when run on 4, 8, or 16 processors and displays almost ideal speed-up in going from 4 to 16 processors. Preliminary benchmark calculations have also been carried out on an Intel Paragon. © 1995 by John Wiley & Sons, Inc.     

Attractive electron–electron interactions within robust local fitting approximations

An analysis of Dunlap's robust fitting approach reveals that the resulting two‐electron integral matrix is not manifestly positive semidefinite when local fitting domains or non‐Coulomb fitting metrics are used. We present a highly local approximate method for evaluating four‐center two‐electron integrals based on the resolution‐of‐the‐identity (RI) approximation and apply it to the construction of the Coulomb and exchange contributions to the Fock matrix. In this pair‐atomic resolution‐of‐the‐identity (PARI) approach, atomic‐orbital (AO) products are expanded in auxiliary functions centered on the two atoms associated with each product. Numerical tests indicate that in 1% or less of all Hartree–Fock and Kohn–Sham calculations, the indefinite integral matrix causes nonconvergence in the self‐consistent‐field iterations. In these cases, the two‐electron contribution to the total energy becomes negative, meaning that the electronic interaction is effectively attractive, and the total energy is dramatically lower than that obtained with exact integrals. In the vast majority of our test cases, however, the indefiniteness does not interfere with convergence. The total energy accuracy is comparable to that of the standard Coulomb‐metric RI method. The speed‐up compared with conventional algorithms is similar to the RI method for Coulomb contributions; exchange contributions are accelerated by a factor of up to eight with a triple‐zeta quality basis set. A positive semidefinite integral matrix is recovered within PARI by introducing local auxiliary basis functions spanning the full AO product space, as may be achieved by using Cholesky‐decomposition techniques. Local completion, however, slows down the algorithm to a level comparable with or below conventional calculations. © 2013 Wiley Periodicals, Inc.     

Parallelization of four‐component calculations. I. Integral generation,SCF, and four‐index transformation in the Dirac–Fock package MOLFDIR

The treatment of relativity and electron correlation on an equal footing is essential for the computation of systems containing heavy elements. Correlation treatments that are based on four‐component Dirac–Hartree–Fock calculations presently provide the most accurate, albeit costly, way of taking relativity into account. The requirement of having two expansion basis sets for the molecular wave function puts a high demand on computer resources. The treatment of larger systems is thereby often prohibited by the very large run times and files that arise in a conventional Dirac–Hartree–Fock approach. A possible solution for this bottleneck is a parallel approach that not only reduces the turnaround time but also spreads out the large files over a number of local disks. Here, we present a distributed‐memory parallelization of the program package MOLFDIR for the integral generation, Dirac–Hartree–Fock and four‐index MS transformation steps. This implementation scales best for large AO spaces and moderately sized active spaces. © 2000 John Wiley & Sons, Inc. J Comput Chem 21: 1176–1186, 2000     

Molecular-orbital-free algorithm for the excited-state force in time-dependent density functional theory

Starting from the equation of motion in the density matrix formulation, we reformulate the analytical gradient of the excited-state energy at the time-dependent density functional theory level in the nonorthogonal Gaussian atom-centered orbital (AO) basis. Analogous to the analytical first derivative in molecular-orbital (MO) basis, a Z-vector equation has been derived with respect to the reduced one-electronic density matrix in AO basis, which provides a potential possibility to exploit quantum locality of the density matrix and avoids the matrix transformation between the AO and the MO basis. Numerical tests are finished for the excited-state geometry optimization and adiabatic excitation energy calculation of a series of small molecules. The results demonstrate the computational efficiency and accuracy of the current AO-based energy gradient expression in comparison with the MO-based scheme.     

A novel parallel algorithm for large-scale Fock matrix construction with small locally distributed memory architectures: RT parallel algorithm

We developed a novel parallel algorithm for large-scale Fock matrix calculation with small locally distributed memory architectures, and named it the "RT parallel algorithm." The RT parallel algorithm actively involves the concept of integral screening, which is indispensable for reduction of computing times with large-scale biological molecules. The primary characteristic of this algorithm is parallel efficiency, which is achieved by well-balanced reduction of both communicating and computing volume. Only the density matrix data necessary for Fock matrix calculations are communicated, and the data once communicated are reutilized for calculations as many times as possible. The RT parallel algorithm is a scalable method because required memory volume does not depend on the number of basis functions. This algorithm automatically includes a partial summing technique that is indispensable for maintaining computing accuracy, and can also include some conventional methods to reduce calculation times. In our analysis, the RT parallel algorithm had better performance than other methods for massively parallel processors. The RT parallel algorithm is most suitable for massively parallel and distributed Fock matrix calculations for large-scale biological molecules with more than thousands of basis functions.     

Density matrix structure for saturated molecules

Conclusion The existence of a common Hamiltonian matrix structure for saturated systems results in common structural properties of the density matrices for the whole class of molecules, such as the zero occupation of AO in the first approximation, the density matrix perturbations due to a heteroatom, etc. This fact can be taken as a quantum-mechanical foundation for viewing saturated molecules as a separate class of compounds. The endowment of this system with the transferability of electronic structure properties, relative to atoms and bonds, to high accuracy, within the framework of the effective Hamiltonian method follows from an analysis of the general expressions for the density matrix elements. The transferability of the saturated system Hamiltonian matrix elements requisite for this is supported by a comparison among the self-consistent Fock matrix elements of various hydrocarbons in a localized orbital basis [9]. Independently of the detailed structure of the actual molecules, the influence of a heteroatom on the electron density distribution in saturated systems dies off quickly with distance from the heteroatom. From an analysis of expressions for the nondiagonal elements of the density matrix corresponding to nonneighboring AO we establish a connection between the degree of electron localization in saturated systems and the size or certain Hamiltonian matrix elements.There is a consequent analogy between saturated and alternatively conjugated hydrocarbons, which, starting from the common structure of the Hamiltonians, also leads to common properties of the density matrices [14]. However, the study of the influence of heteroatoms on the density matrices in these systems by means of perturbation theory is complicated by the dependence of the matrix N(4) on the molecular structure, which makes it necessary to introduce highly simplified approaches for the solution of Eq. (2) [8]. Therefore, for alternating hydrocarbons we have succeeded in establishing only the sign of the orbital — orbital polarizability (alternating polarity theorem [14]), while, as for saturated systems, the equality N=1 permits an analytic expression for the polarizability.V. Kapsukas Vilnius State University. Translated from Zhurnal Strukturnoi Khimii, Vol. 29, No. 5, pp. 3–8, September–October, 1988.     

A submatrix algorithm for the matrix-vector multiplication of very large matrices

In self-consistent field (SCF) calculations the construction of the Fock matrix is most time-consuming step. The Fock matrix construction may formally be seen as a matrix-vector multiplication, where the matrix is the supermatrix,??_ijkl, and the vector is the first-order density matrix, γ_ij. This formalism should be optimal for vector machines. This is not, however, fully utilized in most programs running on computers with small core memory. The size of the ?? matrix, typically in the order of 10⁶–10⁸ elements, has forced programmers to implement other nonvectorizable methods. We will present a submatrixbased algorithm which will partition the supermatrix so that vectorizable methods can be employed. The method will also reduce the input/output.     

Corrected small basis set Hartree‐Fock method for large systems

A quantum chemical method based on a Hartree‐Fock calculation with a small Gaussian AO basis set is presented. Its main area of application is the computation of structures, vibrational frequencies, and noncovalent interaction energies in huge molecular systems. The method is suggested as a partial replacement of semiempirical approaches or density functional theory (DFT) in particular when self‐interaction errors are acute. In order to get accurate results three physically plausible atom pair‐wise correction terms are applied for London dispersion interactions (D3 scheme), basis set superposition error (gCP scheme), and short‐ranged basis set incompleteness effects. In total nine global empirical parameters are used. This so‐called Hartee‐Fock‐3c (HF‐3c) method is tested for geometries of small organic molecules, interaction energies and geometries of noncovalently bound complexes, for supramolecular systems, and protein structures. In the majority of realistic test cases good results approaching large basis set DFT quality are obtained at a tiny fraction of computational cost. © 2013 Wiley Periodicals, Inc.     

Hartree-Fock calculations with linearly scaling memory usage

We present an implementation of a set of algorithms for performing Hartree-Fock calculations with resource requirements in terms of both time and memory directly proportional to the system size. In particular, a way of directly computing the Hartree-Fock exchange matrix in sparse form is described which gives only small addressing overhead. Linear scaling in both time and memory is demonstrated in benchmark calculations for system sizes up to 11 650 atoms and 67 204 Gaussian basis functions on a single computer with 32 Gbytes of memory. The sparsity of overlap, Fock, and density matrices as well as band gaps are also shown for a wide range of system sizes, for both linear and three-dimensional systems.     

Direct inversion of the iterative subspace with contracted planewave basis functions

Ways to reduce the computational cost of periodic electronic structure calculations by using basis functions corresponding to linear combinations of planewaves have been examined recently. These contracted planewave (CPW) basis functions correspond to Fourier series representations of atom‐centered basis functions, and thus provide access to some beneficial properties of planewave (PW) and localized basis functions. This study reports the development and assessment of a direct inversion of the iterative subspace (DIIS) method that employs unique properties of CPW basis functions to efficiently converge electronic wavefunctions. This method relies on access to a PW‐based representation of the electronic structure to provide a means of efficiently evaluating matrix–vector products involving the application of the Fock matrix to the occupied molecular orbitals. These matrix–vector products are transformed into a form permitting the use of direct diagonalization techniques and DIIS methods typically employed with atom‐centered basis sets. The abilities of this method are assessed through periodic Hartree–Fock calculations of a range of molecules and solid‐state systems. The results show that the method reported in this study is approximately five times faster than CPW‐based calculations in which the entire Fock matrix is calculated. This method is also found to be weakly dependent upon the size of the basis set, thus permitting the use of larger CPW basis sets to increase variational flexibility with a minor impact on computational performance. © 2018 Wiley Periodicals, Inc.     

Improvements on the direct SCF method

Three improvements on the direct self-consistent field method are proposed and tested which together increase CPU-efficiency by about 50%: (i) selective storage of costly integral batches; (ii) improved integral bond for prescreening; (iii) decomposition of the current density matrix into a linear combination of previous density matrices—for which the two-electron contributions to the Fock matrix are available—and a remainder ΔD, which is minimized; construction of the current Fock matrix only requires processing of the small ΔD which enhances prescreening.     

Hartree–Fock High-Precision Analytical Functions of Open-Shell Atoms

The algorithm of high-precision optimization of basis functions suggested previously for calculating the analytical Hartree–Fock orbitals of closed-shell atoms is generalized to open-shell systems described by the Roothaan method (1960). Expressions for the first (free gradient) and second (Hesse matrix) derivatives of the system's energy with respect to the nonlinear parameters (orbital exponents) of the basis functions are derived in terms of density matrices for the filled and open shells. An algorithm is proposed for high-precision optimization of the nonlinear parameters using these equations based on Murtagh–Sargent and Newton minimization procedures. To illustrate the application of this algorithm, we give optimization of the basis sets of Slater type functions for atoms from the second row, as well as for Al, Si, P, K, Sc, and Fe atoms. The analytical Hartree–Fock orbitals giving nearly Hartree–Fock energies are calculated with a high degree of accuracy.     

Is large-scale ab initio Hartree-Fock calculation chemically accurate? Toward improved calculation of biological molecule properties

Numerical errors in total energy values in large-scale Hartree–Fock calculations are discussed. To obtain total energy values within chemical accuracy, 0.01 kcal/mol, stricter numerical accuracy is required as basis size increases. In molecules with 10,000 basis sizes, such as proteins, numerical accuracy for total energy values must be retained to at least 11 digits (i.e., to the order of 1.0D-10) to keep accumulation of numerical errors less than the chemical accuracy (0.01 kcal/mol). With this criterion, we examined the sensitivity analysis of numerical accuracy in Hartree–Fock calculation by uniformly replacing the last bit of the mantissa part of a double-precision real number by zero in the Fock matrix construction step, the total energy calculation step, and the Fock matrix diagonalization step. Using a partial summation technique in the Fock matrix generation step, the numerical error for total energy value of molecules with basis size greater than 10,000 was within chemical accuracy (0.01 kcal/mol), whereas with the conventional method the numerical error with several thousand basis sets was larger than chemical accuracy. Computation of one Fock matrix element with parallel machines can include the partial summation technique automatically, so that parallel calculation yields not only high-performance computing but also more precise numerical solutions than the conventional sequential algorithm. We also found that the numerical error of the Householder-QR diagonalization routine is equal to or less than chemical accuracy, even with a matrix size of 10,000. ©1999 John Wiley & Sons, Inc. J Comput Chem 20: 443–454, 1999     

Parallel MP2-energy evaluation: Simulated shared memory approach on distributed memory parallel machines

A parallel algorithm for four-index transformation and MP2 energy evaluation, for distributed memory parallel (MIMD) machines is presented. The underlying serial algorithm for the present parallel effort is the four-index transform. The scheme works through parallelization over AO integrals and, therefore, spreads the O(n³) memory requirement across the processors, reducing it to O(n²). In this sense, the scheme superimposes a shared memory architecture onto the distributed memory setup. A detailed analysis of the algorithm is presented for networks with 4, 6, 8, 10, and 12 processors employing a smaller test case of 86 contractions. Model direct MP2 calculations for systems of sizes ranging from 160 to 238 basis functions are reported for 11- and 22-processor networks. A gain of at least 40% and above is observed for the larger systems. © 1997 by John Wiley & Sons, Inc.     

On the implicit bond-dependency origins of bridge interactions

The indirect (through-bridge) components of chemical interactions between atomic orbitals (AO) are shown to originate from the indirect dependencies between AO due to the orbital intermediaries in the bond system of the molecule. They are expressed in terms of the bridge-coupling elements of the density matrix via the chain rule transformation of the implicit derivatives between the indirectly bonded AO in the molecular bond system. The elements of the charge-and-bond-order (CBO) matrix are interpreted as the canonical derivatives between the AO-projections onto the bond subspace combining the occupied Molecular Orbitals (MO). The chain-rule manipulations are then used to express the scattering amplitudes via AO intermediaries in terms of the relevant elements of the CBO matrix. The squares of such amplitudes are related to the Wiberg-type indirect bond components, which complement the familiar direct Wiberg bond-order contributions. The interference implications of the probability scatterings via the multiple cascades involving all basis functions are examined. These probability propagations are shown to preserve the stationary conditional probabilities of the underlying molecular communication channel in AO resolution.