Reconfigurable Superconducting and Photonic Technologies of the Future

Objectives and Purpose
The main objective of the research program is to introduce a new and advanced computational paradigm using the concepts of neuromorphic computation.
Computation based on the conventional  Von Neumann computational architecture, where a Central Processing Unit (CPU)  executes computation and exchanges data with the memory units through a communications bus, suffers from a number of shortages and bottlenecks. 
The processing units based on the current CMOS technology have approached its limits of reliable computation, mainly because of the minimum achievable feature sizes, facing the adverse effects of quantum physics, such as tunneling, and the need to dissipate the heat generated in very densely populated  Si chips.
In addition, perhaps more importantly, the bus that is needed to facilitate high data rate communication in between the CPU and memory, constitute a major bottleneck in reaching higher speeds in computation.
The neuromorphic computational paradigm, on the other hand, is an emerging approach that emulates some of the assumed operational characteristics of the neurons constituting the brain.  
In neuromorphic computing, instead of CMOS transistors, the basic units of computation are neuron-like entities, emulating the neurons in the brain. There are various means of realizing electronic or electro-optical neurons.
Following are their most prominent advantages: low energy consumption; scalability, capability to build large scale systems; fault-tolerance due its distributed construction; capability to learn from new data and experience makes it suited for applications like image and speech recognition; fast response time and low latency which are important features for real-time applications such as robotics; unified computation and memory functions where both computation and data storage take place within the neuron, eliminating the computational bottleneck between the CPU and memory to a large extent.

Major contribution to the global scientific challenges:
One way of implementing neuromorphic computing is nanophotonic computing with ultrafast  photodetection.  The entire electro-optical chip need to be implemented in a single chip. In our particular application, superconducting switches realize neurons, while the synaptic weights are implemented using phase change memory (PCM) circuit elements.
The technology demonstration will combine vision and cognition on a single chip, designed as an in-sensor neuromorphic computing with atto-joule switching power consumption, sub-nanosecond latency, and high compactness (3000 neurons and >100K synapses). This has not been possible with the current technology.
It is expected that the technology willl have applications in a number of crucial commercial applications. We will attempt to demonstrate three specific cases:  

1. online medical image processing of PET scans,
2. quantum ghost imaging,
3. optical fiber dispersion compensation in high-speed telecommunication links. 

Contribution to the professional skills and capabilities and technology objectives:

1. Integrating artificial vision and recognition units on the same chip. In this implementation, superconducting detectors act as a retina, superconducting switches as neurons, and phase change memory elements as their synapses. 
2. Generating cryogenic memristor models and implement innovative design strategies to  mitigate drift/loss issues. 
3. Integrate on-chip circuitry for the synaptic weight updates allowing for high-speed and efficient learning and adaptability. The entire electro-optical training and weight update circuitry will be reconfigurable, allowing for optimization and test benching different training algorithms. 
4. Provide three prototyped technology demonstrators mentioned above to bridge the gap between the laboratory and industrial/clinical environments.

Contribution to the Human Resources Development:
During the 48-month project period, it is expected to train graduate students and TUBITAK engineers in related fields. 

Project Partners

1. SINGLE QUANTUM, Netherlands
2. TÜBİTAK, Türkiye
3. TuDELFT, Netherlands
4. THE UNIVERSITY OF OXFORD, UK
5. UNIVERSITEIT GENT, Belgium
6. UNIVERSITEIT GRONINGEN, Netherlands
7. FRAUNHOFER, Germany