Rowan Docs

Redox Potential Prediction

Python API · stjames models · API example · API example

How it works

Estimate one-electron oxidation and reduction potentials in acetonitrile. Rowan optimizes the starting molecule and each requested oxidized or reduced form, then converts their energy differences into potentials in volts relative to the saturated calomel electrode (SCE).

Settings

  • Redox type: Select "Reduction" to add an electron, "Oxidation" to remove one, or both. The web form defaults to oxidation only; at least one must be selected.
  • Mode: Sets the geometry optimization and single-point energy methods below. The web form defaults to "Rapid."
  • Solvent: Acetonitrile (MeCN) is fixed; there is no solvent selector. Standard modes optimize geometries in the gas phase and include implicit acetonitrile solvation in the single-point energies.

Modes

ModeGeometry optimizationSingle-point energy in MeCN
RecklessGFN-FFGFN2-xTB
RapidGFN2-xTBr²SCAN-3c
Carefulr²SCAN-3cωB97X-3c
Meticulousr²SCAN-3c → ωB97X-3cωB97M-D3BJ/def2-TZVPPD

More expensive modes refine the electronic energies and geometries, but solvent-model error can limit the improvement in predicted potentials. Chan's study discusses this limitation and the performance of ωB97X-3c.

Notes

At the same reference electrode, a more positive reduction potential means easier reduction; a lower oxidation potential means easier oxidation. The sign alone does not identify which process was calculated.

Results default to SCE. The results selector can display "SHE," "Fc/Fc+," or "Ag/AgCl" using conversion constants from Pavlishchuk and Addison. Changing this display scale does not change the solvent or rerun the calculation. Match the solvent and electrode scale when comparing with experiment.

The energy-to-potential conversion includes empirical reference corrections from Neugebauer et al., with an additional correction for GFN2-xTB energies. Review the optimized structures and any warnings: fragmentation or a very large predicted potential can indicate unreliable results. The workflow estimates a one-electron process; it does not model coupled proton transfer or subsequent chemical reactions.

Submission video

Benchmarks and validation

Accuracy

Rowan's published GFN2-xTB/CPCM-X protocol achieved a mean absolute error of 0.32 V on the OROP benchmark set, excluding values that did not start from a neutral species or were not in acetonitrile. This result does not validate the current rapid-mode protocol shown above.

Explore OROP and OMROP results for OMol25-trained models. These model benchmarks use separate protocols from this workflow: reduction-potential benchmarks.

Further reading