A TU Wien-led team combines supercomputing and automation to predict high-performance thermoelectric materials—shifting the search from trial-and-error to intelligent design.
New technologies demand new materials—but discovering them can take decades. Now, a research team led by Andrej Pustogow at TU Wien has created an automated computer protocol that predicts thermoelectric materials capable of converting waste heat into usable electricity. Their approach slashes the time and cost of discovery, offering a faster route to green energy materials.
Instead of relying on textbook wisdom or laborious lab experiments, the team used supercomputers to simulate how metals like nickel, cobalt, and iron interact with every other element in the periodic table. Starting with transition metals, they scanned thousands of possible combinations. Among the results, nickel-germanium (Ni₃Ge) stood out for its remarkable thermoelectric performance—confirmed later in lab tests.
While previous research by the team hinted at this (including work on nickel-gold alloys), this is the first time a systematic, AI-assisted method has delivered accurate predictions. Despite the raw power of supercomputers like Austria’s Scientific Cluster, brute force isn’t practical—evaluating all possible element combinations could take longer than the universe has existed. So, the team narrowed the search intelligently using physical principles and automation.
The selected candidate, Ni₃Ge, not only performed well but proved that even simple binary compounds can deliver. If two-element systems show promise, imagine what we’ll find in more complex combinations. The approach is already gaining traction beyond academia. Industry giants like Google and Microsoft are building AI tools trained on decades of experimental and theoretical materials data. The best results still come from human intuition—amplified by computing power. This marks a strategic shift in materials science. For almost a century, the focus has been on semiconductors.
