Empa Develops First Atomic-Scale Model of Hydrogen-Containing Amorphous Alumina to Advance Energy and Materials Research

Empa researchers led by Simon Gramatte (front) and Vladyslav Turlo have succeeded for the first time in simulating amorphous aluminum oxide with hydrogen inclusions with atomic precision. Image: Empa

(IN BRIEF) Empa researchers have created the first atomically precise model of amorphous aluminum oxide thin films with hydrogen inclusions, using a combination of experiments, advanced simulations, and machine learning. The work, published in npj Computational Materials, provides detailed insights into how hydrogen interacts with oxygen in the material, influencing density and chemical states. This understanding paves the way for improved thin-film coatings, membranes, and particularly hydrogen separation materials for green hydrogen production. The machine learning approach reduces simulation times from theoretical billions of years to just one day and could be applied to other amorphous materials, enabling targeted optimization of their properties for industrial and energy applications.

(PRESS RELEASE) DÜBENDORF, 14-Aug-2025 — /EuropaWire/ — Empa scientists have achieved a major milestone in materials research by successfully creating the first highly accurate atomic-scale model of amorphous aluminum oxide (Al₂O₃) thin films, including hydrogen inclusions. This breakthrough, made possible through the integration of advanced experiments, high-performance simulations, and machine learning, provides unprecedented insight into the disordered atomic structure of a material widely used for protective coatings, membranes, and passivation layers.

Aluminum oxide is one of the most extensively studied compounds in materials science, known for its versatility in electronics, industrial applications, and technical ceramics. While its various crystalline forms are well understood, the amorphous state—lacking a regular, repeating atomic structure—has remained elusive at the atomic level. This irregularity makes such materials far more difficult to examine and model than crystals. As Empa researcher Vladyslav Turlo explains, simulating an amorphous alumina thin film from scratch at atomic precision would normally take longer than the age of the universe.

Clarity out of chaos: In amorphous alumina, aluminum atoms (grey) and oxygen atoms (red) do not arrange in an ordered crystalline structure. The model also visualizes hydrogen atoms (blue) closely binding to neighboring oxygen atoms, which alters the material’s properties. Image: Empa

To overcome this challenge, researchers from three Empa laboratories collaborated closely. Thin films of amorphous aluminum oxide were fabricated using atomic layer deposition and characterized with advanced analytical techniques, including hard X-ray photoelectron spectroscopy (HAXPES)—a method available in Switzerland exclusively at Empa. These measurements revealed not only the positions of aluminum and oxygen atoms but also the distribution of hydrogen atoms, which vary depending on manufacturing methods and have a significant impact on the material’s density and chemical behavior.

The resulting machine learning-driven model, detailed in npj Computational Materials, is the first to represent amorphous alumina with hydrogen at true atomic precision. It shows that above certain concentrations, hydrogen binds to oxygen atoms, altering the chemical states of neighboring elements and making the material less dense. This new understanding is particularly important for applications such as hydrogen separation membranes in green hydrogen production, where controlled permeability is critical.

According to Turlo, amorphous alumina could become a leading material for hydrogen filtration in water-splitting processes powered by renewable energy or direct sunlight. The model offers a powerful tool for optimizing the mechanical, optical, and permeability properties of amorphous alumina, enabling more targeted design for specific applications. Furthermore, the approach could be adapted to model other amorphous materials, dramatically accelerating research timelines—from billions of years of hypothetical computation down to about a day.

By bringing order to the atomic chaos of amorphous materials, Empa’s interdisciplinary team has opened the door to innovations in coatings, membranes, and sustainable energy technologies, demonstrating the transformative power of combining experimental science with machine learning.