Spec Libraries

SPEC libraries are hosted on GitHub at the URL https://github.com/spec-org

The following libraries are available

The following libraries are scheduled to be released in FY21 (see below for more information):

  • August 2021: ASCI-MCSCF
  • September 2021: DMRG

GFCC

The Green’s Function Coupled Cluster Library (GFCC) is designed for the Green's function calculation of a molecular system at the Coupled-Cluster level. The library is developed in the C++ programming language with a framework that enables scalability via the scalable design of the algorithm, efficiency via multi-layer parallelism, memory management, scheduling of tensor operations, and flexibility by providing a long-term sustainable platform for the development of the class of methods on machines ranging from workstations to modern supercomputers.

ChangeLog

version 1.0 (released 09-27-2019)

  • Concurrent computation of frequency-dependent Green's function matrix elements and spectral function in the CCSD/GFCCSD level (enabled via MPI process groups)
  • Support for multidimensional real/complex hybrid tensor contractions and slicing on both CPU and GPU
  • On-the-fly Cholesky decomposition for atomic-orbital based two-electron integrals
  • Direct inversion of the iterative subspace (DIIS) is customized and implemented as the default complex linear solver
  • Gram-Schmidt orthogonalization for multidimensional complex tensors
  • Model-order-reduction (MOR) procedure for complex linear systems
  • Automatic resource throttling for various inexpensive operations
  • Checkpointing (or restarting) calculation employing parallel IO operations for reading (writing) tensors from (to) disk

version 1.1 (released 06-04-2020)

  • The default linear solver has been changed from DIIS to Generalized Minimal REsidual Method (GMRES)
  • Load balancing across concurrent computation of different orbitals for a given frequency
  • TAMM (Tensor Algebra for Many-body Methods): Improved distributed tensor contraction performance on GPUs

B. Peng, A. Panyala, K. Kowalski, S. Krishnamoorthy, “GFCCLib: Scalable and Efficient Coupled-Cluster Green's Function Library for Accurately Tackling Many Body Electronic Structure Problems”, Computer Physics Communications 265, pp. 108000 (2021); DOI:10.1016/j.cpc.2021.108000

B. Peng, K. Kowalski, A. Panyala, S. Krishnamoorthy, "Green’s function coupled cluster simulation of the near-valence ionizations of DNA-fragments", The Journal of Chemical Physics 152, pp 011101 (2020); https://dx.doi.org/10.1063/1.5138658

MC-MPn-Direct

version 1.0 (released 06-09-2020)

The MC-MP methods are the Monte Carlo version of MP2, MP3, and MP4. Similarly, the MC-GF methods are the Monte Carlo version of second- and third-order green's function theory. F12 methods perform explicitly correlated MP2 and GF2 calculations. The acronyms correspond to:

  • MC-MP2: second-order Monte Carlo perturbation theory
  • MC-MP3: third-order Monte Carlo perturbation theory
  • MC-MP4: fourth-order Monte Carlo perturbation theory
  • MC-GF2: second-order Monte Carlo Green's function theory
  • MC-GF3: third-order Monte Carlo Green's function theory
  • MC-MP2-F12: explicitly correlated second-order Monte Carlo perturbation theory
  • MC-GF2-F12: explicitly correlated second-order Monte Carlo Green's Function theory

A. E. Doran and So Hirata,  "Monte Carlo Second- and Third-Order Many-Body Green’s Function Methods with Frequency-Dependent, Nondiagonal Self-Energy", Journal of Chemical Theory and Computation 15, pp. 6097-6110 (2019); https://dx.doi.org/10.1021/acs.jctc.9b00693

X2Chem

(version 1.0, released 05/03/2021)

An eXact 2-Component relativistic quantum chemistry library. This library provides a convenient tool that wraps X2C transformation modules with an interface to user-defined integral engines. The output of this library is the X2C transformed relativistic two-component Hamiltonian.

SCHEDULED FOR RELEASE in FY22 / FY23

ASCI-MCSCF: Adaptive Sampling Configuration Interaction Multi-Configuration Self-Consistent Field library in a determinant basis that will be able to handle hundreds of millions of determinants

Scalable DMRG: Massively parallel quantum chemical density matrix renormalization group library.