Two of the most hot button topics at the minute are artificial intelligence (AI) and net zero energy - therefore, it was only a matter of time before we started to see overlap between the two sectors.
Stemming from research at the University of Toronto, new methods are being explored towards utilising AI to discover new and more efficient materials for net zero energy technology; with machine learning proving to significantly speed up the amount of time needed to find new materials with desired properties.
In the pursuit of trying to find better materials than those we already have for net zero energy applications (e.g. lithium-ion batteries), AI could be utilised to find entirely new materials for battery applications, or using materials that are already known but have never been considered for use in net zero energy applications.
Given the current limitations with materials currently used with green technologies, such as cost, (in)efficiency, or the limits of their capabilities having already been reached, it is hoped that implementing machine learning will significantly speed up the time taken to create new and better materials for clean energy application by combining elements of existing ones.
The basis of the machine learning model used in this application relies on data found in the Materials Project, an open-source database of more than 140,000 know materials developed over the past decade. This database also includes information about the components of known materials, including crystal structure, molecular composition, density, energy conductivity and stability.
To use lithium-ion batteries as an example, and how they may be improved, AI may be utilized to figure out the stability of new materials and how much energy they can store.
The challenge currently is that the calculations required to do this kind of research work do not scale very well. For example, more complex materials such as an alloy require twice as many atoms to encode, making their properties four times slower to calculate using conventional methods; with these types of calculations currently relying on a quantum chemistry approach, proving to be slow and require a lot of computing power.
In contrast, the newly developed method incorporating the aforementioned AI model can do such calculations 1,000 times faster.
Previous models in this field were able to reproduce the stabilities of known materials, but could not predict the stabilities for materials with unknown crystal structures (the way atoms, ions and molecules are arranged in a material—an essential factor in determining its physical properties). By training the new machine learning model on what has been named 'distorted structures', insights are provided into how new materials will perform under strain, allowing the model to relax a crystal structure to its more stable configuration.
As such, knowing the precise crystal geometry allows accurate predictions of what the properties of new materials will look like and how they will ultimately perform; with this new AI implemented method of assessing materials significantly speeding up this process and potentially opening up a lot of possibilities.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
