The goal of so-called neuromorphic computing is to go beyond software-based artificial neural networks and create new hardware whose architecture more closely resembles the functioning of the biological nervous system than that of digital computers. Such an approach promises extremely energy-efficient information processing solutions. One leading technology in this area is the development of hardware networks from so-called memristive devices, or resistive switching memories. In recent years, numerous neuromorphic hardware devices capable of complex information processing have been introduced, but their operation has typically been limited to the bottom end of the MHz frequency range, or in many cases to kHz frequencies. So far, only individual memristive devices have been demonstrated to operate at frequencies above GHz, but complex neuromorphic circuits capable of information processing at such high frequencies are not yet available. Recently, the collaboration of ETH Zürich and BME delivered the world record for non-volatile and volatile memristive devices with an ultra-short switching times of ~15 ps [1,2]. Building on these achievements, the goal of the doctoral research project is to go beyond ultrafast device characterization and develop neuromorphic information-processing circuits in the GHz frequency range. To this end, the applicant will systematically test potential building blocks for high frequency neuromorphic applications, covering various memristive material systems and operation mechanisms. It will be investigated, how these memristive building blocks can be assembled into low-power analog information processing circuits. The target goals include efficient radio frequency information processing in edge devices, optical information processing by optocoupled memristive circuits, and high frequency information processing with coupled oscillator networks.

[1] M. Csontos, M., Horst, Y., Olalla, N. J., Koch, U., Shorubalko, I., Halbritter, A. & Leuthold, J. Picosecond Time‐Scale Resistive Switching Monitored in Real‐Time. Adv Elect Materials 9, https://doi.org/10.1002/aelm.202201104 (2023). 

[2] Schmid, S. W., Pósa, L., Török, T. N., Sánta, B., Pollner, Z., Molnár, G., Horst, Y., Volk, J., Leuthold, J., Halbritter, A. & Csontos, M. Picosecond Femtojoule Resistive Switching in Nanoscale VO₂ Memristors. ACS Nano 18, 21966–21974 https://doi.org/10.1021/acsnano.4c03840 (2024).