The Atomic World of Glass Materials

What happens inside a material that powers lasers, heals bodies, and locks away waste — when no one is looking?

Phosphate-based glass materials find wide applications in high tech areas from biomedical devices and high power lasers to waste storage and specialty coatings, but what exactly happens inside them at the atomic level and how they change the properties remains a mystery.

Unlocking this hidden world could help scientists discover valuable new glass materials with improved properties that are stronger, safer, more durable, and even more versatile than ever before.

Unlike crystalline materials, glasses do not have a regular, repeating structure. Instead, their atoms form complex, disordered networks that are hard to capture accurately by experimental methods alone.

Together, these challenges have limited scientists’ ability to reliably design and optimize them, highlighting the need for more advanced and accurate modeling approaches.

More advanced models demand greater computing power, and matching those results with real-world experimental data remains challenging, particularly for phosphorus-rich glasses. This makes computer simulations a necessity to understand their complex structures and behaviors.

Navid Marchin, a PhD candidate in Materials Science and Engineering at the University of North Texas (UNT), is the lead author on a a paper in the Journal of the American Ceramic Society from October 2025.

Marchin is trying to answer the question: How can HPC based atomistic simulations help scientists better understand the atomic structure of sodium aluminophosphate glasses?

“Improving the ability to model the atomic structure would help accelerate the design of safer, more durable, and more efficient glass materials without relying solely on costly trial-and-error experiments,” he said.

Stepping Into Supercomputing

Marchin was first introduced to supercomputing resources in 2023 in Professor Jincheng Du’s Computational Materials Science course at the Department of Materials Science and Engineering at UNT.

“Introducing our students to the world of high performance computing, or generally the computational material science methods, is critical for them to understand how these powerful tools can help to understand materials behaviors and to design new materials in addition to traditional experimental based materials research approaches,” said Du, who is a University Distinguished Research Professor and Chair of the Department of Materials Science and Engineering.

In Du’s Functional Glasses and Materials Modeling Laboratory, the students utilize atomistic simulation methods to elucidate complex structural issues of glasses, nano, and other materials for applications including communications, biomedicine, and nuclear waste disposal.

“We access TACC’s systems such as Lonestar6 or Stampede3 through projects funded by the UNT Research Computing Servies. TACC is critical to our daily and federally funded research projects,” Du said.

Now, Marchin regularly relies on TACC resources, specifically Lonestar6, for his research. “The system’s massive processing power and robust software ecosystem let me run large models and long molecular dynamics simulations that simply would not be possible on a local machine,” he said. “As a result, Lonestar6 has dramatically accelerated my work and strengthened the quality of my results.”

New Potential

In recent studies, Marchin and colleagues have developed an advanced interatomic potential that can more accurately capture the atomic structure of phosphate-based glasses. Improving the accuracy of these models is a critical step toward better predicting material properties, understanding how glasses form during processing, and designing new compositions with enhanced performance.

These models were developed based on highly accurate quantum-mechanical calculations and validated through experimental measurements, according to Marchin.

We access TACC’s systems such as Lonestar6 or Stampede3 through projects funded by the UNT Research Computing Servies. TACC is critical to our daily and federally funded research projects.

“In this work specifically, the new aluminophosphate glass model significantly reduces the gap between computer simulations and experimental observations, particularly in describing the atomic environment surrounding aluminum atoms,” he said. “TACC’s Lonestar6 CPU/GPU computing time and useful software packages, like LAMMPS, enabled us to perform large-scale molecular dynamics simulations much more efficiently.”

The researchers used supercomputers to run extensive molecular dynamics simulations of multiple glass compositions, including model development, structural relaxation, property calculations, and statistical analysis.

“Access to high performance computing allows us to simulate larger systems over longer timescales, which was essential for accurately capturing short- and medium-range atomic structures and for systematically refining and validating the new potential.”

A Clearer Picture Emerges

The new model results significantly improve agreement between simulations and experimental data, particularly for aluminum coordination and bond-angle distributions in aluminophosphate glasses. It narrows the gap between computational predictions and laboratory measurements and provides more realistic descriptions of glass structure.

“I’m pleased that technological advancements enable progress in computational materials science,” Marchin said. “Our results show that the gap between modeling and experiments can be narrowed using the latest technologies with strong potential for further improvement in future studies.”

Future work will extend this modeling approach to more complex glass compositions and apply the improved model to predict properties such as chemical durability and thermal behavior, further strengthening the link between simulations, experiments, and materials design. 

This work was funded by AGC Inc. and AGC researchers provided experimental validation of some of the simulation results. This work signifies the importance of university-industrial collaborations and integration of HPC based simulations and experiments to address fundamental material science issues.