Accelerating resolution-of-the-identity second-order Møller-Plesset quantum chemistry calculations with graphical processing units

The modification of a general purpose code for quantum mechanical calculations of molecular properties (Q-Chem) to use a graphical processing unit (GPU) is reported. A 4.3x speedup of the resolution-of-the-identity second-order M?ller-Plesset perturbation theory (RI-MP2) execution time is observed in single point energy calculations of linear alkanes. The code modification is accomplished using the compute unified basic linear algebra subprograms (CUBLAS) library for an NVIDIA Quadro FX 5600 graphics card. Furthermore, speedups of other matrix algebra based electronic structure calculations are anticipated as a result of using a similar approach.     

Two-level hierarchical parallelization of second-order Møller-Plesset perturbation calculations in divide-and-conquer method

A two‐level hierarchical parallelization scheme including the second‐order Møller–Plesset perturbation (MP2) theory in the divide‐and‐conquer method is presented. The scheme is a combination of coarse‐grain parallelization assigning each subsystem to a group of processors, with fine‐grain parallelization, where the computational tasks for evaluating MP2 correlation energy of the assigned subsystem are distributed among processors in the group. Test calculations demonstrate that the present scheme shows high parallel efficiency and makes MP2 calculations practical for very large molecules. © 2011 Wiley Periodicals, Inc. J Comput Chem, 2011     

Tensor factorizations of local second-order Møller-Plesset theory

Efficient electronic structure methods can be built around efficient tensor representations of the wavefunction. Here we first describe a general view of tensor factorization for the compact representation of electronic wavefunctions. Next, we use this language to construct a low-complexity representation of the doubles amplitudes in local second-order M?ller-Plesset perturbation theory. We introduce two approximations--the direct orbital-specific virtual approximation and the full orbital-specific virtual approximation. In these approximations, each occupied orbital is associated with a small set of correlating virtual orbitals. Conceptually, the representation lies between the projected atomic orbital representation in Pulay-Saeb? local correlation theories and pair natural orbital correlation theories. We have tested the orbital-specific virtual approximations on a variety of systems and properties including total energies, reaction energies, and potential energy curves. Compared to the Pulay-Saeb? ansatz, we find that these approximations exhibit favorable accuracy and computational times while yielding smooth potential energy curves.     

General biorthogonal projected bases as applied to second-order Møller-Plesset perturbation theory

With low-order scaling correlated wave function theories in mind, we present second quantization formalism as well as biorthonormalization procedures for general--singular or nonsingular--bases. Of particular interest are the so-called projected atomic orbital bases, which are obtained from a set of atom-centered functions and feature a separation of occupied and virtual spaces. We demonstrate the formalism by deriving and implementing second-order M?ller-Plesset perturbation theory in it, and discuss the convergence and preconditioning of the iterative amplitude equations in detail.     

Application of second-order Møller-Plesset perturbation theory with resolution-of-identity approximation to periodic systems

Efficient periodic boundary condition (PBC) calculations by the second-order M?ller-Plesset perturbation (MP2) method based on crystal orbital formalism are developed by introducing the resolution-of-identity (RI) approximation of four-center two-electron repulsion integrals (ERIs). The formulation and implementation of the PBC RI-MP2 method are presented. In this method, the mixed auxiliary basis functions of the combination of Poisson and Gaussian type functions are used to circumvent the slow convergence of the lattice sum of the long-range ERIs. Test calculations of one-dimensional periodic trans-polyacetylene show that the PBC RI-MP2 method greatly reduces the computational times as well as memory and disk sizes, without the loss of accuracy, compared to the conventional PBC MP2 method.     

Variational second-order Møller-Plesset theory based on the Luttinger-Ward functional

In recent years there have been some rather successful applications of a new variational technique for calculating the total energies of electronic systems. The new method is based on many-body perturbation theory and uses the one-electron Green function as the basic "variable" rather than the wave function of traditional variational calculations. It is the purpose of the present work to promote the new methods within the realm of traditional theoretical chemistry by demonstrating their utility for calculating the correlation energies of a number of atoms at a level corresponding to second-order M?ller-Plesset perturbation theory. The generalization to any desired order of perturbation theory is not hard to accomplish.     

An efficient atomic orbital based second-order Møller-Plesset gradient program

Based on the orbital-invariant atomic orbital formulation of the MP2 (M?ller-Plesset second-order perturbation theory) energy and gradient [P. Pulay and S. Saeb?, Theor. Chim. Acta 69, 357 (1986)], we have derived and programmed detailed working equations for closed-shell MP2 gradients. The orbital-invariant form avoids the difficulties of other formulations with frozen orbitals, and allows the use of arbitrary occupied orbitals, an important consideration for local correlation theories, although the present program uses canonical molecular orbitals. The atomic orbital formulation offers savings both in storage and computer time. Test calculations on systems containing up to approximately 100 atoms and approximately 1000 basis functions, performed on a single personal computer, are reported. Parallelization of the code is underway.     

Application of Gaussian-type geminals in local second-order Møller-Plesset perturbation theory

In this work Gaussian-type Geminals (GTGs) are applied in local second-order Moller-Plesset perturbation theory to improve the basis set convergence. Our implementation is based on the weak orthogonality functional of Szalewicz et al., [Chem. Phys. Lett. 91, 169 (1982); J. Chem. Phys. 78, 1420 (1983)] and a newly developed program for calculating the necessary many-electron integrals. The local approximations together with GTGs in the treatment of the correlation energy are introduced and tested. First results for correlation energies of H(2)O, CH(4), CO, C(2)H(2), C(2)H(4), H(2)CO, and N(2)H(4) as well as some reaction and activation energies are presented. More than 97% of the valence-shell correlation energy is recovered using aug-cc-pVDZ basis sets and six GTGs per electron pair. The results are compared with conventional calculations using correlation-consistent basis sets as well as with MP2-R12 results.     

A local second-order Møller-Plesset method with localized orbitals: a parallelized efficient electron correlation method

Using orthogonal localized occupied orbitals we have developed and implemented a parallelized local second-order M?ller-Plesset (MP2) method based on the idea developed by Head-Gordon and co-workers. A subset of nonorthogonal correlation functions (the orbital domain) was assigned to each of the localized occupied orbitals using a distance criterion and excitations from localized occupied orbitals that were arranged into subsets. The correlation energy was estimated using a partial diagonalization and an iterative efficient method for solving large-scale linear equations. Some illustrative calculations are provided for molecules with up to 1484 Cartesian basis sets. The orbital domain sizes were found to be independent of the molecular size, and the present local MP2 method covered about 98%-99% of the correlation energy of the conventional canonical MP2 method.     

Explicitly correlated second-order Møller-Plesset perturbation theory employing pseudospectral numerical quadratures

We implemented explicitly correlated second-order M?ller-Plesset perturbation theory with numerical quadratures using pseudospectral construction of grids. Introduction of pseudospectral approach for the calculation of many-electron integrals gives a possibility to use coarse grids without significant loss of precision in correlation energies, while the number of points in the grid is reduced about nine times. The use of complementary auxiliary basis sets as the sets of dealiasing functions is justified at both theoretical and computational levels. Benchmark calculations for a set of 16 molecules have shown the possibility to keep an error of second-order correlation energies within 1 milihartree (mH) with respect to MP2-F12 method with dense grids. Numerical tests for a set of 13 isogyric reactions are also performed.     

A kinetic energy fitting metric for resolution of the identity second-order Møller-Plesset perturbation theory

A kinetic-energy-based fitting metric for application in the context of resolution of the identity second-order M?ller-Plesset perturbation theory is presented, which is derived from the Poisson equation. Preliminary tests of the applicability include the evaluation of the error in the correlation energy, compared to standard M?ller-Plesset perturbation theory, with respect to the auxiliary basis set employed. We comment on the potential merits of this fitting metric, compared to standard resolution of the identity second-order M?ller-Plesset perturbation theory, and discuss its scaling behavior in the limit of large molecules.     

Local explicitly correlated second-order Møller-Plesset perturbation theory with pair natural orbitals

We explore using a pair natural orbital analysis of approximate first-order pair functions as means to truncate the space of both virtual and complementary auxiliary orbitals in the context of explicitly correlated F12 methods using localised occupied orbitals. We demonstrate that this offers an attractive procedure and that only 10-40 virtual orbitals per significant pair are required to obtain second-order valence correlation energies to within 1-2% of the basis set limit. Moreover, for this level of virtual truncation, only 10-40 complementary auxiliary orbitals per pair are required for an accurate resolution of the identity in the computation of the three- and four-electron integrals that arise in explicitly correlated methods.     

An atomic orbital-based reformulation of energy gradients in second-order Møller-Plesset perturbation theory

A fully atomic orbital (AO)-based reformulation of second-order M?ller-Plesset perturbation theory (MP2) energy gradients is introduced, which provides the basis for reducing the computational scaling with the molecular size from the fifth power to linear. Our formulation avoids any transformation between the AO and the molecular orbital (MO) basis and employs pseudodensity matrices similar to the AO-MP2 energy expressions within the Laplace scheme for energies. The explicit computation of perturbed one-particle density matrices emerging in the new AO-based gradient expression is avoided by reformulating the Z-vector method of Handy and Schaefer [J. Chem. Phys. 81, 5031 (1984)] within a density matrix-based scheme.     

Tautomerism of xanthine: The second-order MØller-Plesset study

Four 9H and four 7H tautomers of DNA base xanthine were studied by the ab initio LCAO-MO method at the MP2/6-311G^**//HF/6-31G^** and MP2/6-31G^**//HF/6-31G^** approximations. All calculated structures are minima at the HF/6-31G^** potential energy surface with the dioxo 7H tautomer (A1) being the global minimum. The second most stable tautomer, dioxo-9H (B1) is by 9 kcal/mol less stable. For the A1 B1 transition the calculated MP2 energy gap corresponds to the equilibrium constant of 2 × 10^–7. Therefore, only the major tautomeric form A1 is predicted to be detectable in the gas phase. The 7H and 9H groups of tautomers are discussed separately. Within both groups, the dioxo form (A1-7H, B1-9H) is the most stable one and is succeeded by the 2-dihydroxy (A2, B2) form. However, while the energy difference between A1 and A2 is 10 kcal/mol, the energy difference between B1 a B2 is only 2 kcal/mol. The effect of polar environment was estimated by the SCRF method, using a spherical cavity, at the HF/6-31G^** level. These calculations did not change the gas phase stability order of the tautomers. However, the energy difference between A1 and B1 decreased from 9 kcal/mol at the HF/6-31G^** level to 4 kcal/mol at the SCRF HF/6-31G^** level.     

New implementation of second-order Møller-Plesset perturbation theory with an analytic Slater-type geminal

The author introduces a new method for the exchange commutator integrals in explicitly correlated M?ller-Plesset second order perturbation theory. The method is well suited with an analytic Slater-type geminal correlation factor. He also explains the scheme for auxiliary integrals needed for the correlation factor. Based on different Ans?tze, he analyzes the performance of the method on correlation energies and reaction enthalpies in detail.     

A hybrid scheme for the resolution-of-the-identity approximation in second-order Møller-Plesset linear-r(12) perturbation theory

In the framework of second-order M?ller-Plesset linear-r(12) (MP2-R12) perturbation theory, a method is developed and implemented that uses an auxiliary basis set for the resolution-of-the-identity (RI) approximation for the three- and four-electron integrals. In contrast to previous work, the two-electron integrals that must be evaluated never involve more than one auxiliary basis function. The new method therefore scales linearly with the number of auxiliary basis functions and is much more efficient than the previous one, which scaled quadratically. A general formulation of MP2-R12 theory is presented for various ansatze, approximations, and orbitals (canonical or localized). The new method is assessed by computations of the valence-shell second-order M?ller-Plesset correlation energy of a few small closed-shell systems. The preliminary calculations indicate that the difference between the new and previous methods is about one order of magnitude smaller than the errors that occur due to basis-set truncations and RI approximations and under the assumptions of generalized and extended Brillouin conditions.     

Analytical energy gradients for local second-order Møller-Plesset perturbation theory using density fitting approximations

An efficient method to compute analytical energy derivatives for local second-order M?ller-Plesset perturbation energy is presented. Density fitting approximations are employed for all 4-index integrals and their derivatives. Using local fitting approximations, quadratic scaling with molecular size and cubic scaling with basis set size for a given molecule is achieved. The density fitting approximations have a negligible effect on the accuracy of optimized equilibrium structures or computed energy differences. The method can be applied to much larger molecules and basis sets than any previous second-order M?ller-Plesset gradient program. The efficiency and accuracy of the method is demonstrated for a number of organic molecules as well as for molecular clusters. Examples of geometry optimizations for molecules with 100 atoms and over 2000 basis functions without symmetry are presented.     

Ab initio second-order Møller-Plesset calculation of the vibrational spectra of C4 clusters

Second-order Møller-Plesset calculations on the C₄ potential surface yielded three isomers, a linear triplet and rhombic and tetrahedral singlets. A large discrepancy between observed and calculated frequencies of the Σ_u⁺ vibration of linear ¹²C₄ is outside the expected range of error. It is tentatively suggested that an observed absorption at 1544 cm⁻¹ previously assigned to C₅ might belong instead to C₄.     

Singularities of Møller-Plesset energy functions

The convergence behavior of M?ller-Plesset (MP) perturbation series is governed by the singularity structure of the energy, with the energy treated as a function of the perturbation parameter. Singularity locations, determined from quadratic approximant analysis of high-order series, are presented for a variety of atoms and small molecules. These results can be used as benchmarks for understanding the convergence of low-order methods such as MP4 and for developing and testing summation methods that model the singularity structure. The positions and types of singularities confirm previous qualitative predictions based on functional analysis of the Schrodinger equation.