xTB Optimizer

A MoleditPy plugin for semiempirical geometry optimization using the xTB method family (GFN2-xTB, GFN1-xTB) via the tblite package and the ASE LBFGS optimizer.

Optimization runs on a background thread — the UI stays fully responsive during calculation.

Features

  • GFN2-xTB and GFN1-xTB methods
  • Live per-step energy and force table
  • Scrolling log output
  • Cancel button (stops the optimizer mid-run cleanly)
  • Undo-compatible (coordinates committed via push_undo_checkpoint)

Installation

1. Install dependencies

Windows: use conda-forge, not pip. tblite has no pre-built pip wheel for Windows — pip install tblite will attempt to compile from source and fail unless LAPACK is present.

mamba install -c conda-forge tblite-python ase
# or
conda install -c conda-forge tblite-python ase

2. Install the plugin

Copy the XTB_Optimizer/ folder (or just xtb_optimizer.py) to your MoleditPy plugin directory:

  • Windows: C:\Users\<YourName>\.moleditpy\plugins\
  • Linux / macOS: ~/.moleditpy/plugins/

Then restart MoleditPy or use Plugins › Reload All Plugins.

Usage

  1. Load a molecule and generate 3D coordinates.
  2. Open 3D Edit › xTB Optimizer…
  3. Choose method, force threshold, and max steps.
  4. Click ▶ Run. Monitor progress in the table and log.
  5. Close the dialog when done — optimized coordinates are already applied.

Settings

Setting Default Description
Method GFN2-xTB Hamiltonian: GFN2-xTB (more accurate) or GFN1-xTB (faster)
Force threshold 0.05 eV/Å LBFGS convergence criterion
Max steps 500 Maximum optimizer iterations

Notes

  • Dummy atoms (*) are not supported and will be rejected before the run starts.
  • Molecules without 3D coordinates must be converted first (3D Edit › Generate 3D Coordinates).
  • The plugin uses the ASE LBFGS optimizer internally.