Researchers from Flinders University and Khalifa University have developed a new machine-learning platform. This system works like a smart materials discovery engine. It can greatly reduce the time needed for complex computer simulations or laboratory tests to find new materials for future semiconductors to be used in many high-tech products such as smartphones, wearable devices, communication systems, medical equipment, light-emitting diodes and solar panels.
The main problem is that there are millions of possible material combinations to test. Traditional methods of checking each one in the lab or through detailed computer models are very slow and costly. The new artificial intelligence (AI) learns the hidden chemical rules that govern the behavior of materials containing gallium. Gallium is an important mineral used in electronics, especially in high-speed computer chips and circuits that handle microwaves and infrared signals.
How the AI system finds promising new materials
The platform was trained using thousands of known semiconductor materials from global databases. It then applies Bayesian optimisation, a smart decision-making technique that helps the system focus on the most likely successful options while avoiding chemically impossible combinations. The AI checks whether proposed materials are realistic and stable before recommending them for further testing. This approach saves time and effort.
As a result, the system successfully identified several entirely new gallium-based semiconductor candidates that were not in existing databases. One important property the researchers targeted is the band gap. The band gap is the energy difference within a material that controls how it interacts with electricity and light. Different band gap sizes suit different uses. Smaller gaps work well for collecting solar energy while larger gaps are needed for powerful electronics and radiation-resistant devices.
This development offers a faster way to create better materials for future technology.
The researchers have described the methods and results of this study in a paper published in ACS Materials Letters.