Enabling Large-Scale Correlated Electronic Structure Calculations: Scaling the RI-MP2 Method on Summit

Second-order Møller-Plesset perturbation theory using the Resolution-of-the-Identity approximation (RI-MP2) is a state-of-the-art approach to accurately estimate many-body electronic correlation effects. This is critical for predicting the physicochemical properties of complex molecular systems; however, the scale of these calculations is limited by their extremely high computational cost. In this paper, a novel many-GPU algorithm and implementation of a molecular-fragmentation-based RI-MP2 method are presented that enable correlated calculations on over 180,000 electrons and 45,000 atoms using up to the entire Summit supercomputer in 12 minutes. The implementation demonstrates remarkable speedups with respect to other current GPU and CPU codes, excellent strong scalability on Summit achieving 89.1% parallel efficiency on 4600 nodes, and shows nearly-ideal weak scaling up to 612 nodes. This work makes feasible ab initio correlated quantum chemistry calculations on significantly larger molecular scales than before on both large supercomputing systems and on commodity clusters, with a potential for major impact on progress in chemical, physical, biological and engineering sciences.

[1]  Masato Kobayashi,et al.  Two‐level hierarchical parallelization of second‐order Møller–plesset perturbation calculations in divide‐and‐conquer method , 2011, J. Comput. Chem..

[2]  Giuseppe M. J. Barca,et al.  Three- and four-electron integrals involving Gaussian geminals: Fundamental integrals, upper bounds, and recurrence relations. , 2017, The Journal of chemical physics.

[3]  Peter M. W. Gill,et al.  Molecular integrals Over Gaussian Basis Functions , 1994 .

[4]  Edward G Hohenstein,et al.  Basis set consistent revision of the S22 test set of noncovalent interaction energies. , 2010, The Journal of chemical physics.

[5]  Georg Hetzer,et al.  Multipole approximation of distant pair energies in local MP2 calculations , 1998 .

[6]  G. Kresse,et al.  Quartic scaling MP2 for solids: A highly parallelized algorithm in the plane wave basis. , 2016, The Journal of chemical physics.

[7]  Matthew L. Leininger,et al.  Is Mo/ller–Plesset perturbation theory a convergent ab initio method? , 2000 .

[8]  Kazuo Kitaura,et al.  A combined effective fragment potential-fragment molecular orbital method. II. Analytic gradient and application to the geometry optimization of solvated tetraglycine and chignolin. , 2011, The Journal of chemical physics.

[9]  Rodney J. Bartlett,et al.  Is fifth-order MBPT enough? , 1985 .

[10]  Marco Häser,et al.  Møller-Plesset (MP2) perturbation theory for large molecules , 1993 .

[11]  M. Plesset,et al.  Note on an Approximation Treatment for Many-Electron Systems , 1934 .

[12]  Alistair P. Rendell,et al.  Scaling the Hartree-Fock Matrix Build on Summit , 2020, SC20: International Conference for High Performance Computing, Networking, Storage and Analysis.

[13]  D. Bernholdt,et al.  Large-scale correlated electronic structure calculations: the RI-MP2 method on parallel computers , 1996 .

[14]  Giuseppe M. J. Barca,et al.  Many-Electron Integrals over Gaussian Basis Functions. I. Recurrence Relations for Three-Electron Integrals. , 2016, Journal of chemical theory and computation.

[15]  Kasper Kristensen,et al.  The divide–expand–consolidate MP2 scheme goes massively parallel , 2013 .

[16]  Mark S Gordon,et al.  A Hybrid Distributed/Shared Memory Model For the RI-MP2 Method in the Fragment Molecular Orbital Framework. , 2019, Journal of chemical theory and computation.

[17]  Takahito Nakajima,et al.  MPI/OpenMP Hybrid Parallel Algorithm of Resolution of Identity Second-Order Møller-Plesset Perturbation Calculation for Massively Parallel Multicore Supercomputers. , 2013, Journal of chemical theory and computation.

[18]  Hans-Joachim Werner,et al.  Scalable electron correlation methods I.: PNO-LMP2 with linear scaling in the molecular size and near-inverse-linear scaling in the number of processors. , 2015, Journal of chemical theory and computation.

[19]  Ryan P A Bettens,et al.  Energy-Based Molecular Fragmentation Methods. , 2015, Chemical reviews.

[20]  Alistair P. Rendell,et al.  High-Performance, Graphics Processing Unit-Accelerated Fock Build Algorithm. , 2020, Journal of chemical theory and computation.

[21]  Michael W. Schmidt,et al.  Recent developments in the general atomic and molecular electronic structure system. , 2020, The Journal of chemical physics.

[22]  Spencer R Pruitt,et al.  Fragmentation methods: a route to accurate calculations on large systems. , 2012, Chemical reviews.

[23]  Emily A. Carter,et al.  Pseudospectral Møller-Plesset perturbation theory through third order , 1994 .

[24]  R. Bartlett,et al.  Correlation energy in LiH, BH, and HF with many‐body perturbation theory using slater‐type atomic orbitals , 1976 .

[25]  Jan Almlöf,et al.  Elimination of energy denominators in Møller—Plesset perturbation theory by a Laplace transform approach , 1991 .

[26]  L. Radom,et al.  Deceptive convergence in Møller-Plesset perturbation energies , 1986 .

[27]  Peter M W Gill,et al.  Two-Electron Integrals over Gaussian Geminals. , 2016, Journal of chemical theory and computation.

[28]  Takahito Nakajima,et al.  Massively parallel algorithm and implementation of RI‐MP2 energy calculation for peta‐scale many‐core supercomputers , 2016, J. Comput. Chem..

[29]  F. Stillinger Mo/ller–Plesset convergence issues in computational quantum chemistry , 2000 .

[30]  Leo Radom,et al.  Why does unrestricted Mo/ller–Plesset perturbation theory converge so slowly for spin‐contaminated wave functions? , 1988 .

[31]  Peter J. Knowles,et al.  On the convergence of the Møller-Plesset perturbation series , 1985 .

[32]  Aaron Vose,et al.  Massively parallel and linear-scaling algorithm for second-order Møller-Plesset perturbation theory applied to the study of supramolecular wires , 2017, Comput. Phys. Commun..

[33]  Martin W. Feyereisen,et al.  Use of approximate integrals in ab initio theory. An application in MP2 energy calculations , 1993 .

[34]  Jeppe Olsen,et al.  Surprising cases of divergent behavior in Mo/ller–Plesset perturbation theory , 1996 .

[35]  Giuseppe M. J. Barca,et al.  Recurrence relations for four-electron integrals over Gaussian basis functions , 2017, 1703.00846.

[36]  Mark S. Gordon,et al.  Novel Computer Architectures and Quantum Chemistry. , 2020, The journal of physical chemistry. A.

[37]  V. Frecer,et al.  How accurate is the description of ligand–protein interactions by a hybrid QM/MM approach? , 2017, Journal of Molecular Modeling.

[38]  J. VandeVondele,et al.  Second-Order Møller-Plesset Perturbation Theory in the Condensed Phase: An Efficient and Massively Parallel Gaussian and Plane Waves Approach. , 2012, Journal of chemical theory and computation.

[39]  Carlos Amador-Bedolla,et al.  GPU Algorithm for the Scaled Opposite-Spin (SOS) MP2 Energy Evaluation , 2017 .

[40]  Nathaniel J. Bloomfield,et al.  Q-MP2-OS: Møller-Plesset correlation energy by quadrature. , 2020, Journal of chemical theory and computation.