SIMULATION OF CRYOGENIC CMOS-OXRAM CIRCUITS FOR AI-DRIVEN SPIN QUBIT CONTROL

Project Description

Context: The latest major breakthrough in quantum computing (QC) has been the demonstration of quantum systems with more than 50 superconducting qubits allowing quantum supremacy for the first time. Other very promising qubit technologies include spin qubits based on solid-state quantum dots (QDs). They leverage the great maturity of CMOS technologies to offer low cost and highly scalable quantum devices. Major research centers like CEA, QuTech and Intel have started to report high quality spin qubits based on advanced CMOS technologies. However, the tuning and control of QDs are still performed mostly by hand with bulky classical electronics located outside the cryostat. The absence of fully integrated cryo-electronics capable of automatically tuning the QDs makes it currently impossible to build a large-scale quantum computer due to the “wiring bottleneck” between the quantum devices and the control electronics. In that scope, the 3IT-1QBit consortium is starting an ambitious research program to solve this problem using artificial intelligence and cryogenic neuromorphic circuits. We propose a 6-month internship focused on the simulation of cryogenic-compatible circuits based on TiO2 memristors (i.e. OxRAM) and CMOS devices. These CMOS-OxRAM circuits will allow to implement in-memory computing for neural networks specifically designed for low-power QD auto-tuning. 

Research project: The intern will be in charge of optimising the performance and energy consumption of CMOS-OxRAM circuits specifically designed to run neural network algorithms dedicated to QD auto-tuning. The impact of temperature and hardware constrains/non-idealities will be considered thanks to experimental measurements of CMOS-OxRAM circuits fabricated in the scope of the program. This project will build upon the work of the INPAQT group at 3IT on TiO2-based memristors and neural networks for QD auto-tuning:
• Fully CMOS-compatible passive TiO2-based memristor crossbars for in-memory computing - ScienceDirect
• Miniaturizing neural networks for charge state autotuning in quantum dots - IOPscience

The student will have to (i) learn an already developed software architecture allowing the simulations of neural networks and OxRAM-based analog circuits specialized to recognize specific patterns in stability diagrams of QDs, (ii) implement at the system-level auto-tuning algorithms of single and double QDs, then benchmark the performance of CMOS-OxRAM circuit designed for the same task, (iii) investigate the impact of circuit non-idealities on the accuracy of the auto-tuning process and develop mitigation methods based on hardware-aware training, (iv) Develop cryogenic models of CMOS and OxRAM devices to implement hardware-aware retraining methods allowing the neural network to perform correctly at 4 Kelvin.

Supervision & work environment: The project will be realized under the direction of Pr. Dominique Drouin within the IRL-LN2, an International Research Laboratory of the French CNRS based in Sherbrooke (QC, Canada). Pr. Fabien Alibart and Pr. Yann Beilliard will also participate in the supervision. The work will be carried out mainly at the Interdisciplinary Institute for Technological Innovation (3IT) and at the Quantum Institute (IQ) of UdeS, in close collaboration with the company 1QBit. 3IT is a unique institute in Canada, specializing in the research and development of innovative technologies for energy, electronics, robotics and health. The IQ is a state-of-the-art institute whose mission is to invent the quantum technologies of tomorrow and transfer them to the industry. 1QBit is a Canadian leader in QC, AI and high-performance computing. Its multidisciplinary team designs control systems, compilers and service architectures for exotic and next generation computing platforms. The student will thus benefit from an exceptional research environment that combines students, professionals, professors and industrialists working hand-in-hand to develop the future technologies for AI and QC.

Researched profile: 
• Specialization in electrical and computer engineering
• Strengths: knowledge in analog circuit design (Cadence, LTSpice), neural networks, Python/PyTorch
• Excellent adaptability, autonomy, teamwork, problem solving skills, strong taste for interdisciplinary R&D
• Fluent in French or English
Contacts: jobnano@usherbrooke.ca
Documents to provide: CV, transcripts of the past two years and references

Funding offered

To be discussed

Partner(s)

1QBit

The last update was on 13 March 2024. The University reserves the right to modify its projects without notice.