Computation-enabled TNT Multiplexed Sensors Fabricated by Two-Photon Stereolithography. – TNT-Sensor-IA

The TNT-Sensor-IA project aims in designing of a new generation of multiplexed array microsensors integrating a metamaterials whose porosity can be spatially programmed. Such a direct laser writing approach associates the latest advances in multiphoton stereolithography (SLA) with those in the artificial intelligence. A progressive sensor specialization processing based on deep learning methods will be performed in order to detect TNT (2,4,6-trinitrotoluene) traces in complex gas mixtures. Note that the majority of the sensors use machine learning only through a unidirectional procedure that provides a simple feed back analysis of the sensor performance. The originality of our project is to propose an iterative algorithmic strategy to reprogram the sensor after each learning cycle. This advantage stems from the computational flexibility proposed by SLA to reprogram on demand the porosity of this metamaterial over a wide space of possible configurations. Hence, beyond the strategic issue to target TNT, the principal constituent of 85 % of unexploded land mines worldwide, the great ability of our multiplexed sensors for learning make them very promising candidates for other sensing applications.