What is it about?
In some solid materials, ions can move and thereby conduct a current. Such materials are interesting for energy applications such as batteries, but also for sensors and new computing architectures. By computer simulations, researchers search for very conductive materials, but such simulations do not produce an exact result. Instead, there is a statistical scatter in the outcome. Some ways of analysing the simulation results can lead to a smoother appearance of the data. We showed that this smoothing gives just the illusion of a more precise result, without a real benefit. Further, we demonstrated that, contrary to previous beliefs, running a simulation for a longer time or on a larger system does not necessarily lead to a more precise result. We thus point towards possible pitfalls for wasting computational resources and formulate rules of thumb to keep the simulations resource-efficient.
Featured Image
Photo by BoliviaInteligente on Unsplash
Why is it important?
Computer simulations are becoming increasingly important in materials science because they allow for a faster screening of different materials than would be possible by experiments. Still, such simulations are very demanding and are typically carried out on high-performance computing infrastructure. An efficient use of simulation resources can accelerate materials research and reduce the energy consumption associated with the simulations.
Perspectives
Our analysis involved the mathematics of a random walk, which describe something (or someone) taking steps in random directions. There are textbook formulas that describe how far such a random walker will likely have moved from its starting point, after a given amount of time. But in our case, the "thing" undergoing a random walk is actually a "nothing": a vacant place in the crystal structure that can swap places with neighbouring ions. Therefore, the ions also move randomly, but their movement follows different mathematical rules. I was amazed to see how drastically such systems would in some cases differ from ordinary random-walk statistics, while being in close alignment with them in other cases.
Adrian Usler
Read the Original
This page is a summary of: A general expression for the statistical error in a diffusion coefficient obtained from a solid‐state molecular‐dynamics simulation, Journal of Computational Chemistry, February 2023, Wiley,
DOI: 10.1002/jcc.27090.
You can read the full text:
Contributors
The following have contributed to this page
