Going for gold in thermal conductivity
Mechanical Engineering Professor Alan McGaughey coordinated the “Phonon Olympics,” bringing together developers and expert users to benchmark three leading open-source thermal conductivity calculation packages.
Although there’s no medal at the end of the Phonon Olympics, for Alan McGaughey, the collaboration required to evaluate the accuracy of three widely used open-source thermal conductivity calculation packages was worth more than gold.
For the last decade, researchers seeking to understand the properties of new materials have turned to open-source packages to perform thermal conductivity calculations. These packages enable a broader community to study thermal transport, but until now users had no way of knowing whether or not each package would produce consistent and accurate results.
Coordinated by McGaughey, six teams participated in the “Phonon Olympics” to test each of the three most cited packages: ALAMODE, phono3py, and ShengBTE. For each package, one team was made up of the developers while the other consisted of expert users. Each team conducted benchmark calculations for four materials: geranium, rubidium bromide, monolayer molybdenum diselenide, and aluminum nitride. As published in the Journal of Applied Physics, the thermal conductivities calculated by the teams fell within 15% of their mean values for each of the four materials.
The team conducted benchmark calculations for four materials using three of the most widely used open-source thermal conductivity calculation packages.
“This is a good sign that when used properly, these codes give us a consistent prediction of thermal conductivity,” said McGaughey, professor of mechanical engineering. “This outcome is better than what we expected.”
The team also documented best practices for how to perform the calculations. They hope this will help novice users recognize that how they use these open-source packages impacts their results.
With this type of program, users can’t just download a code and hit “go.” For example, they must determine how to model the atomic interactions, build a digital sample of the material, and weigh accuracy vs. computational cost.
A lot of people write code and put it on the internet, but it’s important that they are rigorously checked to make sure they give trustworthy results.
Alan McGaughey, Professor, Mechanical Engineering
“Having more researchers in this space will lead to more advancement, but only if their findings are well grounded. A lot of people write code and put it on the internet, but it’s important that they are rigorously checked to make sure they give trustworthy results,” he said.
The dedication from every team member involved in this four-year project is a testament to just how important validating these open-source codes is.
“Through pandemics and personal changes, 17 people pulled together to get this job done,” McGaughey said. “Being part of a team effort like this is really special, both personally and for the greater research community.”