AI Identifies Molecules from Light Patterns

Every molecule has a unique fingerprint made of light. When molecules spin and vibrate, they emit specific patterns of light waves that act like a barcode with no two molecules producing exactly the same pattern. Artificial neural networks (ANN) are good at learning patterns, just like their natural archetype—the human brain. Argonne scientists have taught artificial intelligence to read these light barcodes and instantly identify what molecule they’re looking at.

Traditional methods require scientists to manually compare thousands of light patterns, a process that can take weeks or months. The AI systems learns to recognize molecular fingerprints automatically, identifying unknown substances in minutes instead of months. This speed breakthrough is crucial when analyzing samples from space missions, developing new medicines, or investigating unknown materials. Datasets with up to 100 million molecules are being created by other researchers to enumerate all synthesizable molecules.

The research team is now training AI on databases containing up to 100 million different molecules. Each molecule’s light pattern gets added to the AI’s knowledge base, creating the world’s most comprehensive molecular identification system. The technology has already earned two U.S. patents and is being used by researchers worldwide to speed up chemical discovery and analysis.