Q2MM¶
Q2MM fits molecular mechanics force fields to quantum-mechanical reference data. A force field is an inexpensive energy model whose parameters describe bonds, angles, and other interactions. Use Q2MM when standard parameters do not cover your chemistry and you need to control which parameters are fitted and which stay frozen.
Start without your own QM calculation¶
The first-run guide uses a small CH3F molecule with an approved, precomputed geometry and Hessian distributed inside q2mm. A Hessian describes the energy curvature as atoms move. There is no publication archive to recover and no GPU or QM installation to configure.
From the documented source installation, with JAX_PLATFORMS=cpu set:
This runs genuine JAX molecular mechanics on CPU and a budget-limited SciPy
optimizer, then explicitly saves ch3f.frcmod and its provenance manifest in
your chosen directory. It is a first execution success, not a convergence
or force-field quality claim.
The separate --bounded-ci mode is only a software-path proof: zero energy,
an echoed reference Hessian/geometry, and no parameter update. Do not use it as
native-backend or scientific validation.
Use this source revision or its local wheel
This breaking alpha API/CLI is unreleased. An arbitrary PyPI version may not contain it. Examples are source-only; approved small-system inputs are installed package data. Use Python 3.15 final and the installation instructions.
Then choose your fitting problem¶
The tutorial progresses from the executable example to your own inputs: one FCHK, XYZ plus an explicit Hessian, or several structures sharing a complete force field and an OPT/custom parameter subset. Stationary-point kind (minimum or transition state), force-field form, and runtime components are explicit choices, never inferred from filenames.
There is one public lifecycle: prepare → evaluate → optimize → assemble →
save. Settings construction loads no engines; prepare starts no hidden QM
job. optimize returns a numerical result, while
OptimizationRun.from_result builds the final force field and provenance.
Read the canonical lifecycle
for the full API and breaking migration notes.
Continue¶
Getting started covers installation, outputs, and a local wheel. The tutorial explains inputs and settings in Python. Publication coverage gives the current source-linked availability and blockers; it does not claim exact reproduction. Theory and methods explains Hessian-informed initialization and transition-state handling.