Much like a tongue freezes to a frigid metal pole, ice can cause speed up the adsorption, or stickiness, of molecules. An icy surface can also cause molecules to degrade in the presence of light, releasing trace gases. Before researchers can measure these reactions and incorporate their impacts in global atmospheric models, researchers first need to understand the structure of ice itself.
To that end, a recent study from Lawrence Livermore National Laboratory (LLNL) used a combination of spectroscopy, simulation and machine learning to examine the surface of ice.
In the inner bulk of ice, the team found that the protons are disordered: while the oxygen atoms are fixed in a distinct pattern, the hydrogen atoms are randomly oriented. In contrast, at the surface of ice, they discovered that the protons are ordered: both the oxygen and hydrogen atoms are fixed in place.
The scientists obtained data by simulating vibrational sum-frequency generation (SFG) spectroscopy, which probes the vibrational properties of asymmetric regions of materials such as surfaces or interfaces. This technique is well established, but experimentally interpreting results can be challenging due to the lack of molecular information. By developing a neural network and deploying simulations, the researchers were able to assign spectra peaks to specific water molecule configurations.
"These machine learning models enabled an efficient exploration of various proton arrangements at the ice surface and significantly improved our ability to interpret experimental measurements," said Margaret Berrens, LLNL physicist in the Quantum Simulations Group and first author of the study.
The publication demonstrates an efficient way to simulate and compute surface spectra and proves the utility of SFG spectroscopy as a tool for exploring ice interfaces.
"Our findings and methodology will enhance understanding of the intricate chemical processes that occur in unique and critical atmospheric conditions," said Anh Pham, LLNL scientist and principal investigator of the project.
Looking forward, the team aims to use a similar workflow to examine solid-liquid interfaces.
The research appeared on the cover of ACS JACS Au and has been selected to be featured as an ACS Editors' Choice. Davide Donadio (UC Davis) acted as Principal Investigator for this work. Other co-authors include Marcos Calegari Andrade (LLNL), John Fourkas (University of Maryland). The research is funded by the Center for Enhanced Nanofluidic Transport, an Energy Frontier Research Center funded by the Department of Energy, Office of Science, Basic Energy Sciences, the LLNL Grand Challenge Program, and the National Science Foundation under Grant No. 2305164.
