gkx

Official Resources

• Source Repository: https://github.com/sirmarcel/gkx
• License: See repository

Overview

gkx is a JAX-based workflow for computing thermal conductivity using Green–Kubo relations, designed to integrate with modern machine-learned interatomic potentials and to support efficient anharmonic thermal transport calculations from molecular dynamics trajectories.

Scientific domain: Thermal transport, Green–Kubo, anharmonicity, ML potentials
Target user community: Researchers computing thermal conductivity from MD using modern JAX/ML potential stacks

Theoretical Methods

• Green–Kubo (fluctuation–dissipation) relations for thermal conductivity
• Equilibrium molecular dynamics (EMD) workflow components

Capabilities (CRITICAL)

• Thermal conductivity from EMD trajectories
• Workflow tooling for Green–Kubo correlation functions (as implemented)
• JAX-based pipelines enabling integration with differentiable/ML workflows

Inputs & Outputs

• Input formats: Trajectory/heat-flux or on-the-fly heat flux workflow inputs (see repository)
• Output data types: Correlation functions and thermal conductivity estimates; tabulated outputs

Interfaces & Ecosystem

• JAX ecosystem; intended to be combined with ML potential tooling (see repository)

Limitations & Known Constraints

• Requires careful statistical convergence and validation against reference calculations.

Verification & Sources

Primary sources:

1. Source repository: https://github.com/sirmarcel/gkx

Confidence: VERIFIED

Verification status: ✅ VERIFIED

• Source code: PUBLIC