Machine-Learning-Assisted Discovery of High-Efficient Oxygen Evolution Electrocatalysts

Neural Network with Optimal Neuron Activation Functions Based on Additive Gaussian Process Regression

Feed-forward neural networks (NNs) are a staple machine learning method widely used in many areas of science and technology, including physical chemistry, computational chemistry, and materials informatics. While even a single-hidden-layer NN is a universal approximator, its expressive power is limited by the use of simple neuron activation functions (such as sigmoid functions) that are typically the same for all neurons. More flexible neuron activation functions would allow the use of fewer neurons and layers and thereby save computational cost and improve expressive power. We show that additive Gaussian process regression (GPR) can be used to construct optimal neuron activation functions that are individual to each neuron. An approach is also introduced that avoids nonlinear fitting of neural network parameters by defining them with rules. The resulting method combines the advantage of robustness of a linear regression with the higher expressive power of an NN. We demonstrate the approach by fitting the potential energy surfaces of the water molecule and formaldehyde. Without requiring any nonlinear optimization, the additive-GPR-based approach outperforms a conventional NN in the high-accuracy regime, where a conventional NN suffers more from overfitting.

On the mechanistic complexity of oxygen evolution: potential-dependent switching of the mechanism at the volcano apex

The anodic four-electron oxygen evolution reaction (OER) corresponds to the limiting process in acidic or alkaline electrolyzers to produce gaseous hydrogen at the cathode of the device. In the last decade, tremendous efforts have been dedicated to the identification of active OER materials by electronic structure calculations in the density functional theory approximation. Most of these works rely on the assumption that the mononuclear mechanism, comprising the *OH, *O, and *OOH intermediates, is operative under OER conditions, and that a single elementary reaction step (most likely *OOH formation) governs the kinetics. In the present manuscript, six different OER mechanisms are analyzed, and potential-dependent volcano curves are constructed to comprehend the electrocatalytic activity of these pathways in the approximation of the descriptor G_max(U), a potential-dependent activity measure based on the notion of the free-energy span model. While the mononuclear description mainly describes the legs of the volcano plot, corresponding to electrocatalysts with low intrinsic activity, it is demonstrated that the preferred pathway at the volcano apex is a strong function of the applied electrode potential. The observed mechanistic complexity including a switch of the favored pathway with increasing overpotential sets previous investigations aiming at the identification of reaction mechanisms and limiting steps into question since the entire breadth of OER pathways was not accounted for. A prerequisite for future atomic-scale studies on highly active OER catalysts refers to the evaluation of several mechanistic pathways so that neither important mechanistic features are overlooked nor limiting steps are incorrectly determined.