- You are here: Home > DL Talk Abstract - The Quest for Superconducting Supercomputers
You are here
DL Talk Abstract - The Quest for Superconducting Supercomputers
Superconducting Electronics (SE) is around since quite a long time, covering in meantime the whole area from ultrasensitive sensors to extremely fast switching elements. In scientific applications, it has become quite a success story: Superconducting Quantum Interference Devices (SQUIDs) are used in applications that require very high sensitivity to magnetic flux or anything, magnetic flux can be derived from; Transition Edge Sensors (TES) begin to dominate the field of detector arrays in astronomical telescopes, and the digital operation of Single Flux Quantum (SFQ) circuits has been demonstrated up to a frequency of about 800 GHz. Nevertheless, the number and financial volume of commercial applications of SE is small and nearly not growing. But there could be a significant change coming: future supercomputers are so power-hungry that in a few generations each supercomputer would need an own electrical power station.
Superconducting digital circuits could be a way out of this dilemma. By using superconducting circuits the energy consumption can be reduced by 3 to 5 orders of magnitude (including the cooling) compared to equivalent semiconductor supercomputers. This has been recognized by one of the largest users of supercomputers in the USA, and recently a major project has been started to investigate ways how to realize such a computer. The main obstacle is posed by the need for a scalable and low-power superconducting memory. New ideas are needed since the classical SQUID-like memory cell is far too big for a large-scale integration. Magnetic Josephson junctions could be the way out, but they need to be scalable to nanometer dimensions, they need to be low-power and they need to be comparable in speed to semiconductor memories. In the area of interfacing to room-temperature electronics also new elements are needed, preferable three-terminal devices with power gain, being fast enough to overcome the need for parallelization.
The need of new superconducting memory elements and three-terminal devices requests the interaction with fundamental and device oriented research groups. Without having such devices available, the quest for a superconducting supercomputer will be a failure.
In this talk, I will shortly introduce Superconducting Electronics, give an overview over the supercomputer project and report on ongoing work for superconducting memory elements.