Psi4 1.4: Open-source software for high-throughput quantum chemistry

Abstract

PSI4 is a free and open-source ab initio electronic structure program providing implementations of Hartree-Fock, density functional theory, many-body perturbation theory, configuration interaction, density cumulant theory, symmetry-adapted perturbation theory, and coupled-cluster theory. Most of the methods are quite efficient, thanks to density fitting and multi-core parallelism. The program is a hybrid of C++ and Python, and calculations may be run with very simple text files or using the Python API, facilitating post-processing and complex workflows; method developers also have access to most of PSI4's core functionalities via Python. Job specification may be passed using The Molecular Sciences Software Institute (MolSSI) QCSCHEMA data format, facilitating interoperability. A rewrite of our top-level computation driver, and concomitant adoption of the MolSSI QCARCHIVE INFRASTRUCTURE project, makes the latest version of PSI4 well suited to distributed computation of large numbers of independent tasks. The project has fostered the development of independent software components that may be reused in other quantum chemistry programs.

FIG. 1.

Input modes for PSI4. A coupled-cluster calculation is run equivalently through its preprocessed text input language (PSIthon; left), through the Python API (PSIAPI; middle), and through structured JSON input (QCSCHEMA; right).

FIG. 2.

Structure of the distributed driver: see the final paragraph in Sec. IV for details. In brief, a user request (a) for a multi-molecule, multi-model-chemistry, or non-analytic derivative passes into planning functions (b) defined in procedure tiles (z) that generate a pool of QCSCHEMA for single-molecule, single-model-chemistry, analytic derivative inputs. These can run in several modes (c), depending on desired parallelism and recoverability. Completed QCSCHEMA passes through assembly functions (d) defined in procedure tiles (z) and denoted “ASM” that reconstitute (e) into the requested energy (“E”), gradient (“G”), or Hessian (“H”).

FIG. 3.

Input file illustrating a CBS and many-body gradient run through the distributed driver in the continuous mode [white-background lines; Fig. 2(c.ii)], distributed mode with FractalSnowflake [Fig. 2(c.iii); additional blue-background lines], and distributed mode with the full storage and queuing power of QCFRACTAL [Fig. 2(c.iv); additional red-background lines]. The lower example is “free” when using QCFRACTAL since the components required for BSSE corrections have already been computed during the upper VMFC. While this example exposes the returned QCSCHEMA AtomicResult, the traditional syntax of grad = psi4.gradient(“HF/cc-pV[DT]Z,” bsse_type=“vmfc”) runs in the mode as in Fig. 2(c.ii) and is identical to the upper example.

FIG. 4.

Wall-time comparison for the interaction energy of the adenine·thymine stacked dimer from the S22 database with various versions of PSI4 using 1 (darker green) to 16 (brown) threads, in multiples of two. PSI4 v1.4 data are obtained with the robust grid pruning algorithm.

FIG. 5.

F-SAPT0-DD3M(0)/jun-cc-pVDZ analysis of 459 atoms (5163 orbitals and 22 961 auxiliary basis functions) from the β₁AR–salbutamol co-crystal (PDB: 6H7M). (Left) Geometry of ligands (wide sticks) and residues (thin sticks) within 7 Å. (Right) Order-2 F-SAPT difference analysis of an active vs an inactive complex, with functional groups colored by contribution to ΔΔE_int (red: more attractive in the activated state; blue: more attractive in the inactive state; color saturation at ±10 kcal mol⁻¹).

FIG. 6.

Example Python implementation of TDDFT oscillator strengths.

FIG. 7.

UV-Vis spectrum of rhodamine 6G at the PBE0/aug-pcseg-2 level of theory. The spectra computed using full TDDFT (RPA) and the Tamm–Dancoff approximation (TDA) are reported in blue and orange, respectively.