AI-driven workflow targets next-generation semiconductor materials
Researchers at UNSW have developed an artificial intelligence–assisted workflow that replaces slow, iterative materials discovery with a targeted, data-driven approach.
Designing advanced materials remains largely trial and error.
Small molecular changes can radically alter performance, but with millions of possible combinations, identifying viable candidates is a major bottleneck.
Hybrid perovskites are semiconductors used in applications such as solar cells and LEDs and are built by combining inorganic layers with organic molecules.
These organic components play a critical role in determining how the material behaves, particularly how it transports electrical charge.Unlike previous trial-and-error approaches, the team’s system works backwards from a desired outcome — such as how a material should handle electrical charge — to identify molecules that could deliver it.
It then screens large numbers of candidates and filters out those unlikely to be practical to make.
Applied across millions of possibilities, the approach narrowed the field to a small set of promising candidates.
These were then checked using detailed simulations to confirm their performance.
The work tackles a long-standing expensive and time-consuming problem in the field, where researchers have typically made incremental changes to known materials rather than exploring new ones systematically.
While the candidates have not yet been tested in the lab, the researchers say the approach could help speed up the development of new materials for electronics and energy technologies by making the search process far more efficient.
