Infrared In-Sensor Computing Based on Flexible Photothermoelectric Tellurium Nanomesh Arrays.

The inherent limitations of traditional von Neumann architectures hinder the rapid development of internet of things technologies. Beyond conventional, complementary metal-oxide-semiconductor technologies, imaging sensors integrated with near- or in-sensor computing architectures emerge as a promising solution. In this study, the multi-scale van der Waals (vdWs) interactions in 1D tellurium (Te) atomic chains are explored, leading to the deposition of a photothermoelectric (PTE) Te nanomesh on a polymeric polyimide substrate. The self-welding process enables the lateral vapor growth of a well-connected Te nanomesh with robust electrical and mechanical properties, including a PTE responsivity of ≈120 V W-1 in the infrared light regime. Leveraging the PTE operation, the thermal-coupled bi-directional photoresponse is investigated to demonstrate a proof-of-principle in-sensor convolutional network for edge computing. This work presents a scalable approach for assembling functional vdWs Te nanomesh and highlights its potential applications in PTE image sensing and convolutional processing.