Predicting Redox Potentials with Machine Learning
Author Information
Author(s): Jinnouchi Ryosuke, Karsai Ferenc, Kresse Georg
Primary Institution: Toyota Central R&D Labs., Inc.
Hypothesis
Can machine learning improve the accuracy of predicting the absolute standard hydrogen electrode potential and redox potentials of various molecules and atoms?
Conclusion
The study demonstrates that a hybrid functional with machine learning can accurately predict redox potentials with an average error of 140 mV.
Supporting Evidence
- The hybrid functional can predict redox potentials across a wide range of potentials.
- The method achieved an average error of 140 mV in predicting redox potentials.
- Machine learning models were trained on semi-local exchange-correlation functionals.
Takeaway
This study shows how computers can help scientists predict important chemical properties, like how easily substances can gain or lose electrons, using smart algorithms.
Methodology
The study used machine learning models combined with first-principles calculations to predict redox potentials through thermodynamic integration.
Limitations
The computational cost remains high for accurate predictions, and the method may introduce errors from approximations.
Statistical Information
Confidence Interval
−4.52 ± 0.09 V for ASHEP
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website