Exploring the potential of memristors for quantum computing and artificial intelligence – News

Jan Billiard and Dominic Drouin, teachers to me Department of Electrical and Computer Engineering from the Faculty of Engineering and members of the Quantum Institute at UdeS.

Photo: Michel Caron – Audi

Although the term biomimicry made its way to Le Robert in 1975, this process has been in use for a much longer period. Leonardo da Vinci said: “Learn from nature, you will find the future there.”

The future, for many researchers at the Quantum Institute (IQ), is a quantum computer. In addition to its architecture, algorithms and debugging, this research tool with huge potential will also require the participation of engineers who will ensure its control. This is a formidable challenge, and there is still a lot of work to be done, because maintaining quantum states requires extremely cold temperatures. It is therefore necessary to develop electronics that can operate in these extreme temperatures.

To create a quantum computer, electronics capable of operating at extremely cold temperatures must be developed.

To do this, a research team from the Institute of Quantum (IQ) and the Interdisciplinary Institute for Technological Innovation (3IT) is interested in memristors (contraction memory Based on resistant). This electronic component, which Professor Leon Chua of Berkeley theoretically predicted in 1971, did not materialize until forty years later. It is a nanoscopic electronic component whose resistance can be changed at will. This property makes memristors very promising candidates for the realization of synthetic synapses in the framework of artificial intelligence (AI) enhanced circuits.

What we are trying to do requires a combination of the expertise found in 3IT and IQ. We direct our research projects at the interface of artificial intelligence, emerging nanoelectronics, and quantum science; You need all three at the same time.

Jan Billiard, Professor of Engineering and IQ Member

“At 3IT, we develop resistive memories, also called ‘memristors,'” continues the researcher. These are new nano-components that make it possible to develop high-performance electronic circuits specifically for artificial intelligence. Then, three years ago, an idea arose: why not apply These technologies are on the scale of a quantum computer? Concretely, it is about contributing to the automatic control of qubits using artificial intelligence, whether one chooses quantum dots on silicon or even superconducting circuits. Classical electronics is needed to control quantum chips in a cryostat. If we want to expand If quantum technology scales in the thousands or even millions of qubits, we will have to automate processes using artificial intelligence and using very powerful classical control electronics.

The research team defines AI as a promising way to automate certain control actions of quantum systems, from reading qubits to tomography, including the state of qubits and controlling quantum devices on silicon.

It takes high-performance computers to run the AI, to avoid heating the cryostat. You also need improved electronics for everything to work efficiently.

“Our team is collaborating with Roger Melko, a professor at the University of Waterloo and researcher at the Perimeter Institute, and Stephanie تشيižek, PhD in the application of artificial intelligence to quantum, explains Dominic Drouin, a professor at the College of Engineering. This research project is funded, among other things, through an Invitation Intelligence for projects.Because we needed empirical data to train neural networks on quantum dots on silicon, we relied on the research work done by Sophie Rochet and Julian Cameran-Lamer when they were studying for their PhD in Professor Michel Piero Ladriere’s group.In addition to working at the interface of several disciplines and taking advantage of From 3IT and IQ resources, we relied on a collaborative approach. By combining all of these together, we were able to demonstrate self-calibration of quantum points through machine learning, which is a step closer to automating certain actions.”

Upcoming business

Yann Beilliard reminds us of the temperature challenge: “We need cryogenic resistance memories, specifically adapted to the operating constraints of quantum systems so that we can implement high-performance AI-based control methods directly in the cryostat. On the other hand, all the memories that arose even Now they are operated at room temperature.The next step is therefore to design memristors that are specially adapted for cryogenic conditions to open all applications.So it will be necessary to work on both materials and component architecture, in particular with superconducting materials, a first in the field of memristors. »

Here nature may have its limits and the resources of 3IT, IQ and collaborators will come alone.

Leave a Comment