IISc researchers develop new ML model to find semiconductor properties

Researchers at the Indian Institute of Science (IISc), in collaboration with University College London, have developed machine learning-based methods to predict material properties with limited data. The research shows that this approach can aid in the discovery of semiconductors. It can also predict how quickly ions can move within electrodes in a battery, helping to build better energy storage devices. The research team, led by Sai Gautam Gopalakrishnan, Assistant Professor at the Department of Materials Engineering, developed a model based on Graph Neural Networks (GNNs) for the study.

IISc, in a release, explained that the lack of data on material properties—which is needed to train models that can predict which types of materials possess specific properties, such as electronic band gaps, formation energies, and mechanical properties—is a hindrance. This is due to expensive and time-consuming methods currently in use.

In transfer learning, researchers use a large model first pre-trained on a large dataset and then fine-tuned to adapt to a smaller target dataset. “In this method, the model first learns to do a simple task like classifying images into, say, cats and non-cats, and is then trained for a specific task, like classifying images of tissues into those containing tumors and those not containing tumors for cancer diagnosis,” Gopalakrishnan explained.

“The architecture of the GNN, such as the number of layers and how they are connected, determines how well the model can learn and recognize complex features in the data,” IISc scientists noted.

The IISc team found that their transfer learning-based model, which was first pre-trained and then fine-tuned, performed much better than models trained from scratch. The team also used a framework called Multi-property Pre-Training (MPT), in which they simultaneously pre-trained their model on seven different bulk 3D material properties. “This model was also able to predict the band gap value for 2D materials that it was not trained on,” the institute added.

“The team first determined the training data size required for predicting material properties. They also pre-trained the model by tuning only some layers while ‘freezing’ the others,” Reshma Devi, first author and PhD student at the Department of Materials Engineering, said. She added that the researchers provided data on material properties such as dielectric constant and formation energy of the material as the input, enabling the model to predict values for specific material properties, like the piezoelectric coefficient.

Gopalakrishnan believes that the GNN model can be used to make better semiconductors by predicting their tendency to form point defects, contributing to India’s push towards semiconductor manufacturing.