On August 25, NTT announced an interpretable autonomous deposition system that combines AI, robotics and automated measurement to explore β-Ga₂O₃, a potential next-generation power semiconductor material.
How does the AI-driven lab work?
The system deposits films, moves substrates, performs optical measurements and analyzes data automatically, then chooses the next experimental conditions. NTT says the deposition and evaluation cycle is about three times faster than a researcher-led process.
Researchers can therefore spend more time defining questions, interpreting results and transferring processes instead of repeating the same manual steps.
It explains why an answer works
Machine-learning optimization can find effective parameters without explaining why they work. When materials, equipment or conditions change, the original optimum may not transfer directly.
NTT analyzes temperature, sputtering power, argon flow and oxygen flow, maps their effects and interactions, and turns the data into deposition rules that people can understand.
What did the experiment achieve?
NTT says the system produced a high-quality single-crystal β-Ga₂O₃ film using sputtering, which it describes as a world first for the method. AI Watch reported that the system reached the high-quality condition through 56 deposition experiments.
The research was published in Nature Communications on August 13. This remains a research result and does not mean the material has entered commercial mass production.
What it means for AI for Science
The result shows AI in science moving from running parameter searches for researchers toward organizing knowledge that can be transferred. In materials, chemistry and semiconductor research, an explainable and reusable result is often more valuable than a single best score.
