Within the framework of the project „Bio-inspired circuits with energy-efficient superconducting microelectronics“, which is embedded in the Ilmenau School of Green Electronics, new approaches for a computer architecture optimally adapted to extremely energy-efficient superconducting microelectronic components are being explored in cooperation with colleagues from the Faculty of Computer Science and Automation, Department of Distributed Systems and Operating Systems (Prof. Dr. Boris Koldehofe, Dr. Peter Amthor).
The research activities are motivated by the fact that modern IT applications are characterized by a drastic increase in energy consumption, particularly due to the rapid spread of mobile and cloud computing, big data processing, smart cities and smart infrastructure, high-performance computing, and the Internet of Things in society. As transistor scaling is expected to reach its limits in the near future, another approach to reducing the energy required for computing operations is seen in abandoning inefficient von Neumann-based hardware architectures that rely on digitally encoded signals based on voltage levels. With a bio-inspired design, neuromorphic, spike-encoded circuits can overcome this problem by representing analog values and enabling their highly efficient encoding for today’s energy-critical computing tasks.
Superconducting microelectronics, in particular, is considered a promising option, as signals are generated as voltage pulses by nonlinear oscillators. Signal propagation in these circuits resembles known processes in biological nerves. Therefore, they form the basis for a truly generic bio-inspired concept. Moreover, this is the only known solid-state technology that enables both high speed and energy efficiency, with an energy consumption of approximately 10⁻¹⁹ joules per bit operation.
However, there is currently no optimal architectural design or corresponding computational model for the use of these circuits.
The research objective is the implementation of neuromorphic principles at the circuit level using superconducting microelectronics, in particular through the manipulation of individual magnetic flux quanta as carriers of information. Integrated circuits are to be developed and applied to relevant use cases, demonstrating superior energy efficiency.
This project is strongly oriented toward computer science and addresses a critical research gap in the development, demonstration, and evaluation of spike-encoded dendritic circuits and in applying their physical properties to a computer architecture and a corresponding computational model for real-world applications.
These research activities began on February 1, 2025.
Link to the webpage
Contact: Prof. Hannes Töpfer
Back
