Australian researchers build new AI that could solve challenge of cheaper solar power

If movies are to be believed, artificial intelligence is a one-way ticket to a dystopian future, with films like The Terminator, Blade Runner and The Matrix pointing to a bleak future for humanity – but new Australian research suggests AI could actually play a key role in avoiding the climate crisis.

Australian researchers have unveiled a new artificial intelligence platform that has the potential to accelerate the development of cheaper and higher performance next generation solar cells, with the ability to discover new materials that do not exist yet.

Researchers from the ARC Centre of Excellence in Exciton Science in Melbourne, have successfully demonstrated a new type of machine learning model that is able to predict the energy conversion efficiency of new materials, including those used in next generation organic solar cells.

The model, developed by researchers based at RMIT University and Monash University, allows scientists to model ‘virtual materials’ that do not yet exist, allowing progress towards the development of cheaper and higher performance solar cells to be fast-tracked.

According to new research published in the journal Computational Materials, the new artificial intelligence platform is significantly faster than other machine learning programs, and its source code has been released freely for use by other researchers.

The researchers believe the new model could help speed up the development of cheap and efficient organic solar cells, seen as a potentially cheaper alternative to traditional silicon based solar cells, but which have yet to achieve large-scale commercial deployment.

There is a wide range of potential materials that could be suitable for use in organic solar cells, and identifying the optimal materials has been the focus of a significant amount of solar energy research. The use of artificial intelligence algorithms has the potential to fast track the assessment of prospective materials, as well as materials that have yet to be created, by modelling virtual versions of the materials using computers.

The researchers say that the use of artificial intelligence would also provide potentially more consistent results that lab-based experiments on prospective materials, allowing more consistent grounds for comparison.

“Our aim is to demonstrate that simple, interpretable molecular descriptors and machine learning methods can model and predict important organic photovoltaic properties,” the research paper says.

“While it is clearly ideal to model experimentally measured properties directly, there are many variables that can affect the organic photovoltaic performance metrics, for example, the device design; processing conditions; dopants, dyes, solvents, and other additives; and others. Thus, measured organic photovoltaic properties can vary from experiment to experiment and between labs.”

Previous machine learning algorithms have been computationally intensive and expensive, involving calculations at the quantum level, slowing the rate at which new materials could be evaluated and requiring the use of a large amount of computational resources.

Dr Nastaran Meftahi of RMIT University and the ARC Centre of Excellence in Exciton Science. Credit: Exciton Science

However, researchers at the ARC Centre of Excellence in Exciton Science developed a new model that uses ‘chemically interpretable signature descriptors’ of the materials being analysed, which greatly reduces the amount of computational resources required, while still providing sufficient insight into the behaviour of materials in applications like solar cells.

“The majority of the other models use electronic descriptors which are complicated and computationally expensive, and they’re not chemically interpretable,” lead researcher Dr Nastaran Meftahi said.

“It means that the experimental chemist or scientist can’t get ideas from those models to design and synthesise materials in the lab. If they look at my models, because I used simple, chemically interpretable descriptors, they can see the important fragments.”

The model has been developed through a partnership with other Australian research intuitions, including CSIRO’s Data 61, Monash University, La Trobe University, and the UK’s University of Nottingham.