Photonics in Quantum and Neuromorphic Computation


Cluster ML4Q of the Excellence Initiative of the German Federal and State Governments

BMBF Innovation Cluster NeuroSys

  Artistic visualization of future articificial intelligence Copyright: © AMO GmbH Artistic visualization of future articificial intelligence, based on innovative hardware

Photonics is playing an increasing role in emerging computing platforms. In the cluster “Matter and Light for Quantum Computing”, we cooperate with the Department of Physics and with the Forschungszentrum Jülich to optically interconnect spin-qubit computing platforms, in particular creating high coupling-efficiency optical interfaces between optically active gate defined quantum dots and single-mode fibers utilizing photonic crystals to constrain and engineer the available optical density of states.

In a different application field, as part of our involvement in the NeuroSys cluster, we are working on increasing the scalability of optical-electrical-optical (OEO) artificial neural networks (ANNs) to hundreds of neurons, to enable ultra-fast processing of complex equalization tasks in fiber-optical networks. In OEO ANNs, light is transmitted between neurons optically, but signal regeneration and typically also part of the nonlinear signal processing (activation function) are implemented with electronics. By offloading other operations such as signal weighting and summation in the optical domain, latencies of feed-forward signal propagation can be substantially reduced and signal processing accelerated. Current approaches however rely on narrowband resonant devices or wavelength division multiplexers (WDM) that both need to be carefully tuned and temperature controlled. The complexity associated with exercising such tight control over hundreds or even thousands of devices can be near insurmountable. We are developing a system architecture that enables massively parallel links and optical processing without requiring such devices, opening the door to a much larger scaling of the network. As ANNs are increasingly being applied to compensate for nonlinear data distortion in optical networks, we anticipate such hardware accelerators to find an important role in broadband optical communications.


Faculty of Mathematics, Computer Science and Natural Science, RWTH Aachen University

Forschungszentrum Jülich

University of Bonn