The Python Simulations of Chemistry Framework: 10 years of an open-source quantum chemistry project

Abstract

Over the past decade, the Python-based Simulations of Chemistry Framework (PySCF) has developed into a widely used open-source platform for electronic structure theory and quantum chemical method development. This article reviews the major advances since the previous overview in 2020, covering new modules and methodology, infrastructure changes, and performance benchmarks.

Figures (9)

• Figure 1: Wall times of ground-state HF (upper), and CCSD and LNO-CCSD(T) correlation energy (lower) calculations for the diamond primitive cell using the GTH-cc-pVDZye2022correlation Gaussian basis set with various density fitting schemes: RSDF (blue), FFTDF (orange), and ISDF (purple), as well as a plane wave basis (green). A plane wave energy cutoff of 60 a.u. and the GTH-HF-revstein2020double pseudopotential were used throughout. The calculations were performed on 32 Intel Ice Lake 8352Y CPU cores [for LNO-CCSD(T) and ISDF] and on 32 AMD EPYC 7742 CPU cores (for all other methods).
• Figure 2: Average per-iteration wall times as a function of the number of threads for RCCSDT, UCCSDT, and RCCSDTQ calculations. All calculations were performed for the hydrogen thioperoxide (H2OS) molecule within the frozen-core approximation, using the cc-pVTZ basis for RCCSDT and UCCSDT ($N_{\text{occ}} = 7$, $N_{\text{vir}} = 79$) and the cc-pVDZ basis for RCCSDTQ ($N_{\text{occ}} = 7$, $N_{\text{vir}} = 29$). "Full" refers to implementations that explicitly store the complete $T_3$ (RCCSDT and UCCSDT) or $T_4$ (RCCSDTQ) amplitude tensors, while "Compact" refers to implementations employing compact tensor storage and contraction by utilizing index-permutation symmetry. All benchmarks were run on a single 96-core AMD Genoa node. Tensor contractions were performed using pytblis,pytblisrepo a Python wrapper for the TBLIS library.tblisrepo
• Figure 3: IR spectra of the protonated water cluster. The first panel shows the experimental gas-phase H2-predissociation spectrum of H^+(H2O)6.H2. The second and third panels show the experimental IR$^2$MS$^2$ spectra of H^+(H2O)6.H2, probing the transitions at 3159 cm$^{-1}$ and 3715 cm$^{-1}$ respectively (indicated by the red and blue arrows). (The experimental spectra were reprinted with permission from Ref. heine2013isomer. Copyright 2013 American Chemical Society). The last two panels show the computed IR spectra for the Zundel-like and Eigen-like conformers at the LNO-CCSD(T)/cc-pVTZ level of theory. The intensity under 2000 cm$^{-1}$ in the computed spectra is multiplied by 3 for clarity, and the spectra are convoluted using a Gaussian kernel with a width of 1 cm$^{-1}$. Reprinted with permission from Ref. lnoccad. Copyright 2024 AIP Publishing.
• Figure 4: (a) The left graph shows the error in energies of molecules in the HEAT dataset tajti2004heat for various methods relative to CCSDTQP results. AFQMC with a CISD trial state is more accurate than CCSD(T) for all molecules and it shows significant improvement over the results when a HF trial state is used. (b) The right graph shows the walltime of performing calculations on trans-polyacetylene of increasing size with AFQMC/CISD (GPU-accelerated) and CCSD(T) (CPU-based) using the aug-cc-pVDZ basis. Both restricted and unrestricted formalisms are shown for increasing system lengths. AFQMC sampling was scaled linearly with size to maintain a constant stochastic error of $\sim$1 m$E_\text{h}$. The inset displays the relative walltime ratios between the two methods. (Copyright 2025 ACS Publishing.)
• Figure 5: Capabilities of the TDDFT derivative couplings in PySCF. (a) The geometry of the $S_1/S_2$ minimum energy crossing point of furan using TDA (pink atoms) and TDA-ris (red atoms). (b) and (c) Potential energy surfaces within the branching plane of the $S_1/S_2$ minimum energy determined using the TDA and the TDA-ris methods, respectively. Figures are adapted with permission from Ref. pu2026. Copyright © 2026, American Chemical Society.