PyTC

Python TransCorrelation package

User Guide

• Installation
  • Requirements
  • Dependencies
  • Install from PyPI
  • Install from source
  • GPU support
• Quick Start
  • VMC-based Jastrow Optimization
  • Design Note: Functional, Immutable Data Structures
  • ISDF-accelerated xTC Integrals
  • Examples
• GPU Memory Management
  • How auto-sizing works
  • Environment-variable overrides
  • When to use the overrides

API Reference

• API Reference
  • pytc

Development

• Contributing

Features

• JAX-based automatic differentiation for Jastrow gradients and Laplacians via folx — single-pass forward Laplacian, JIT-compiled to fused GPU kernels, avoiding the cost of finite differences entirely.

• Multi-GPU acceleration via JAX sharding for both VMC sampling and integral evaluation — near-linear strong scaling across devices with no user-side code changes.

• Transcorrelated integrals: K1, K2, K3 two-body and xTC approximated three-body integrals — fully GPU-resident contractions, never materialising the 3-body tensor explicitly.

• Interpolative Separable Density Fitting (ISDF) for the xTC kernel — empirical \(T \propto n_{\mathrm{orb}}^{1.76}\) scaling (see plot above) pushes routine calculations past 1200 orbitals on a single B200, which is beyond what standard quantum chemistry solvers can handle.

• Modular Jastrow factors: Boys-Handy, Nuclear Cusp, Neural Network (EE/EN/EEN), REXP, Polynomial, and Composite — share the same JIT-compiled evaluation path, so adding a new factor adds zero per-step overhead.

• VMC-based Jastrow optimization with second-order Newton and first-order ML optimizers (Adam) — second-order steps reach the same accuracy in an order of magnitude fewer iterations than SGD.

• PySCF interoperability with JAX-native CCSD: Builds on PySCF mean-field objects and molecular data, then runs xTC-CCSD with PyTC’s in-house JAX solver