AI characterizes the interface between zinc oxide and water

Metal oxide particles in contact with water are important for diverse fields, ranging from medicine to water-splitting photocatalysts for renewable energy production. At the interface between the particles and water, where neither the atoms of the oxide nor the water molecules are in their “natural” environment, intriguing surface chemistry often occurs.

Chemical identities of molecules near the interface can be determined using vibrational spectroscopy, but how the water is structured at the interface can be difficult to interpret from these measurements.

As researchers report in The Journal of Chemical Physics, they are instead applying computational techniques to characterize the structure of the interface and calculate the vibrational fingerprints of each chemical species, which can be directly compared to experimental data.

The authors characterized the interface between zinc oxide (ZnO) and water via a machine learning technique based on artificial neural networks. It predicts the water structure on the surface as well as its vibrational spectra. A one-dimensional vibrational model further allowed them to describe all hydroxide bonds by including anharmonic effects.

“Some water molecules near the ZnO surface spontaneously dissociate — resulting in a mixture of hydroxide ions and water at the interface,” said co-author Vanessa Quaranta. “Vibrational frequencies of a molecule depend sensitively upon its chemical environment and, in this case, the extent to which hydroxide ions and water molecules surround themselves with other hydroxide ions, water molecules, and metal oxide atoms. Hydroxide ions often vibrate at higher frequencies than water molecules, but we found exceptions because of specific interactions with the metal oxide surface.”

Using a machine learning technique contributed to a better understanding of experimentally recorded vibrational spectra, according to co-author Matti Hellström. And such techniques can now be applied to other systems.

Source: “Maximally resolved anharmonic OH vibrational spectrum of the water/ZnO$(10\overline{1}0)$ interface from a high-dimensional neural network potential,” by Vanessa Quaranta, Matti Hellström, Jörg Behler, Jolla Kullgren, Pavlin D. Mitev, and Kersti Hermansson, The Journal of Chemical Physics (2018). The article can be accessed at https://doi.org/10.1063/1.5012980 .