Expanded access: DOE’s nuclear waste glass database

Scientists and engineers around the world have worked for decades developing technologies to safely and permanently dispose of nuclear waste. Though these wastes originate from a multitude of processes, they all are highly toxic and extremely harmful to life in any form.

In 1982, borosilicate glass was deemed the preferred waste form for the disposal of high-level nuclear waste.¹ Since then, researchers confirmed glass is an effective waste form for low-activity waste as well.²

In the years leading up to and following that 1982 decision, an enormous amount of data has been generated by U.S. national labs, universities, and international institutions relating to the properties of nuclear waste glasses. Until recently, most of this data could only be accessed from isolated reports or institutional databases maintained by scientists who performed the original experiments.

In 2019, researchers from Savannah River and Pacific Northwest National Laboratories joined forces to create the first online, open-access repository of nuclear waste glass data. This effort, which now also includes experts from Clemson and Alfred Universities, is sponsored by the U.S. Department of Energy’s Office of Environmental Management.

Credit: Images provided by Savannah River and Pacific Northwest National Laboratories and Clemson and Alfred Universities. Compiled by Cyndy Griffith, ACerS.

The primary goal of the database development team is to consolidate the data from these decades of research into a single source that can be used by glass scientists from around the world to study vitrified radioactive waste forms. They have worked to incorporate data from multiple institutions; identify the Nuclear Quality Assurance level for each study; develop a forward-facing “landing page” for the front end of the database; and implement filtering methods to target specific glass properties such as composition, viscosity, and durability.

In 2024, the development team began work to bootstrap the former “SciGlass” database,³ which contains information on more than 400,000 glasses, to the website. Along with this bootstrapping, which allowed for greater compositional breadth, the developers incorporated machine learning-based predictive models to enhance the database’s utility.

The result is the new Virtual Center for Nuclear Waste Glass Science, which currently includes data for more than 6,000 glass compositions with various properties related to processability and product quality. In contrast to other glass databases, the Virtual Center uses statistical mechanics-based estimates to approximate various properties, adding physically meaningful results to the software implementation. These estimates are made possible through using the new Stat Mech Glass Python module.⁴

The Virtual Center for Nuclear Waste Glass Science is undergoing its final rounds of beta testing before being officially launched to the general public. The website can be found at https://srnl.mcdc.cecas.clemson.edu/database.

Industry Perspectives is a regular column discussing trends in ceramics and glass markets.