Investigate The Potential of Memristors For Quantum Computing and Artificial Intelligence

“Learn from nature, you will find the future” said Leonardo da Vinci. While the term biomimicry appeared in 1982, the process has been in use for much longer.

The future for many researchers at the Institut quantique is the quantum computer. In addition to the architecture, algorithms, and error correction, this incredibly promising research tool will also require the work of engineers to control it. This is a tremendous challenge as several of the architectures require cryogenic temperatures to maintain their quantum states. This in turn necessitates developing electronics that can operate and withstand in exceptionally low temperatures.

A research team from the Institut quantique (IQ) and the Institut Interdisciplinaire d’Innovation Technologique (3IT) is working on memristors (a contraction between the words memory and resistor). This electronic component, predicted in theory by Professor Leon Chua of Berkley in 1971, was only developed some 40 years later. It is a nanoscopic electronic component, whose resistance can be adjusted at will. This property makes memristors very promising candidates when creating artificial synapses in circuits optimized for artificial intelligence (AI).

“What we are trying to do requires both expertise from 3IT and at IQ. We focus our research projects where artificial intelligence, emerging nano-electronics and quantum science intersect. You need all three at the same time. At 3IT, we are developing resistive memories, also called memristors. These are new nanocomponents that make the development of high-performance electronic circuits specifically for AI possible. Three years ago, the idea began to emerge. Why not apply these technologies to scale up quantum computing? In practical terms, this means contributing to the automatic control of qubits with the help of AI, whether one chooses quantum boxes on silicon or superconducting circuits. Classical electronics are required to control quantum chips in the cryostat. If we want to scale up quantum technology with thousands or even millions of qubits, we will have to automate these processes using AI and use classical control electronics, which are very robust” explains Yann Beilliard, associate professor at the Faculty of Engineering.

The research team believes that AI is a promising avenue for automating certain control procedures for quantum systems, from qubit reading to tomography, qubit state and control of quantum devices on silicon.

You need high-performance computer hardware to run the artificial intelligence to avoid heating up the cryostat. You also need optimized electronics to make everything work efficiently.

“Our team is working with Roger Melko a professor at the University of Waterloo and a researcher at the Perimeter Institute and Dr. Stefanie Czischek, a postdoctoral researcher in his group, who specializes in the application of AI to quantum. This research project is partly funded by the IQ’s call for projects. And, since we needed experimental data to train neural networks for quantum boxes on silicon, we based our work on research conducted by Sophie Rochette and Julien Camirand Lemyre while they were PhD students in Professor Pioro-Ladrière’s group. In addition to working at the interface of several disciplines and taking advantage of the resources of 3IT and IQ, we relied on a collaborative approach. By bringing all these elements together, we were able to demonstrate the self-calibration of a quantum box by machine learning, which is another step towards automating certain procedures,” explains Prof. Dominique Drouin, from the Faculty of Engineering.

The Next Steps

Yann Beilliard reminds us of the temperature-related challenge. “We need cryogenic resistive memories, specifically adapted to the operating constraints of quantum systems to implement high-performance AI-based control methods directly in the cryostat. On the other hand, all the memories developed so far are made to be operated at room temperature. The next step is to design memristors specifically adapted to cryogenic conditions to unlock all applications. This will require work on both the architecture and component materials, including the use of superconducting materials, a first in the memristor field.

This is where nature may have its limits and where the combined resources of 3IT, IQ and their collaborators will make a real difference.