Accurate oxidation and reduction potentials are central to electrolyte design, photoredox catalysis, and electrochemical stability screening, yet measuring them experimentally for every candidate of interest is slow and resource-intensive. Computational prediction offers a scalable alternative, provided the method is accurate enough to be chemically useful.
The benchmark
To validate our redox-potential workflow, we turned to the dataset compiled by Roth, Romero, and Nicewicz (Synlett 2016), one of the most comprehensive experimental references available: over 180 organic molecules with measured half-peak potentials spanning 17 functional-group classes. We ran 155 of these molecules through Compular Lab automated DFT pipeline from molecular structure to predicted redox potential producing 155 redox entries (some molecules have both an oxidation and a reduction potential) with all calculations completed in under an hour.
Key results
Across a 5 V experimental window the calculated potentials track experiment with an R² of 0.95, a mean absolute error (MAE) of 0.23 V, and an RMSE of 0.30 V, after applying a small uniform calibration offset of +0.23 V that removes a systematic underestimation common to DFT-based redox predictions. A concordance index of 0.91, meaning the method correctly ranks over 90 % of all molecule pairs, confirms that the protocol captures relative ordering, not just absolute values. The best-performing classes — amines, phenols, and imines — achieve sub-0.15 V MAE, with ethers, aldehydes, and aromatic hydrocarbons also falling below 0.2 V. Alkenes, ketones, and aromatic heterocycles come in around 0.2–0.3 V. Thiols and sulfides, alkyl halides, and carboxylic acids cluster around 0.3 V, while acyl/sulfonyl chlorides remain the most challenging class at roughly 0.5 V. The parity plot below summarizes the comparison across all molecules and functional-group classes.
The full picture
The full figure maps every molecule in the benchmark onto a single potential scale, grouped by functional class. It gives an at-a-glance view of which chemistries are easiest to oxidize or reduce, and where your candidate sits relative to the rest.
What this means for screening
With an R² of 0.95, a concordance index above 0.90, and sub-0.15 V accuracy for several key classes, the method is accurate enough to rank candidates, flag outliers, and narrow a large library before committing to experiment.
This benchmark shows what Compular’s platform delivers in practice: you enter your molecules, hit run, and get back predicted redox potentials for hundreds of candidates in a single batch, in under an hour. No setup, no queue management, no post-processing. Whether you are screening electrolyte additives for stability windows, selecting photoredox catalysts, or ranking intermediates in an electrochemical synthesis, Compular handles the computation so you can focus on the chemistry.
In a future post we will show how to move beyond isolated molecules and predict the electrochemical stability of windows of full formulations — and the individual components within them. Stay tuned!
To learn more about how Compular can accelerate your redox screening workflows, reach out to
