EPFL scientists have developed a unique system that can be used for detecting and analyzing molecules with very a level of high precision and without using any bulky equipment. This latest development paves the way for large-scale, image-based detection of materials assisted by artificial intelligence. The study has been reported in Science.

Organic compounds are typically detected and analyzed by infrared spectroscopy, but this method calls for complex procedures as well as huge and costly instruments which make device miniaturization quite difficult and thus limits its applications in certain medical and industrial applications and for collecting data out in the field, for example, for determining the concentrations of pollutants. Infrared spectroscopy is also essentially restricted by low sensitivities and thus needs large amounts of samples.

Conversely, a research team from EPFL's School of Engineering and Australian National University (ANU) has designed a sensitive and compact nanophotonic system that is capable of identifying the absorption characteristics of a molecule without using traditional spectrometry. The researchers have already detected organic compounds, polymers, and pesticides with their system, which can also be used with CMOS technology.

The novel system features an engineered surface which is surrounded with countless numbers of very small sensors known as metapixels. These metapixels can create a unique bar code for each molecule that the surface comes into contact, and using advanced pattern recognition and sorting technology (for example, artificial neural networks) these bar code can be extensively analyzed. This study – which turns out to be a meeting point of nanotechnology, physics, and big data – has been reported in the Science journal.