Investigating The Interplay of Piezoelectricity and Synaptic Plasticity in Se-Based Photodiodes for Optically Controlled Memristors on Flexible Substrates

Neuromorphic computing, situated at the forefront of the AI revolution, offers the promise of overcoming the Von Neumann bottlenecks in terms of energy consumption while excelling in tasks such as unsupervised learning, analysis of probabilistic and fast changing data. In the era of increasing significance for computer vision, visual learning, and soft robotics, <b>the fabrication of devices on flexible substrates becomes extremely relevant.</b><br/>This work proposes a pioneering investigation of t<b>he synaptic plasticity in Selenium (Se)-based photodiodes</b> with a particular focus on the complex <b>interplay between piezoelectricity and memory</b>. These Se-based photodiodes are fabricated on flexible polyimide substrates, introducing a novel dimension to their functionality by exhibiting a substantial piezoelectric effect, leading to a remarkable variation of the open circuit voltage under different strains as previously reported.<br/>Our research methodology involves using <b>both continuous and pulsed light to explore the relationship between piezoelectricity and memory within these Se-based devices.</b> Remarkably, we observe changes in the dark current-voltage (JV) characteristics both before and after a train of illumination pulses, indicative of the device's behavior being very similar to an o<b>ptically controlled memristor</b>. It should be noted that t<b>his change in resistivity is fully reversible </b>and can be induced even at relatively low illumination power densities, as low as 5 mW.cm^-2.<br/>To further investigate the interplay between piezoelectricity and memory, we follow a classic approach of sequential light write pulses and read voltage pulses. This monitoring process involves tracking the current at specific voltage pulses, thus providing indirectly the persistent photoconductivity. This dynamic data is collected as a function of time under various bending angles of the devices. Our findings bring important insights on the role played by the interface <b>band alignment between the Se absorber and the ZnMgO window layer </b>in driving the piezoelectric effect and affecting the memory effect. Our research makes use of different illumination wavelengths, ranging from 350nm to 1400nm, which shows that the system is also <b>wavelength selective</b>.<br/>To assess the synaptic plasticity in these devices, we systematically characterize the device’s short term and long-term plasticity in terms of the Recovery Time Constant. Additionally, we utilize paired pulse facilitation ratio and rate sensitivity as figures of merit to gauge the device's performance in neuromorphic computing applications.<br/>This potentially seminal work is to our knowledge the first reported exploration of the <b>interplay between piezoelectricity and visual memory</b> within the context of Se-based photodiodes. Moreover, our research proposes a pioneering use of Selenium-based photodiodes in the frame of neuromorphic computing applications. These findings hold important promise for advancing flexible, optically controlled memristors and their diverse applications in the domains of visual AI, soft robotics, and beyond.