RAFIK Adam

Abstract: The design of cryogenic circuits has gained much importance in recent years, in particular for applications such as quantum computers, space exploration, and ultra-low-temperature sensing applications. Interface electronics with ultra-low levels of power consumption at temperatures as low as 4 K are required. The design of these circuits is complicated due to the lack of standard design kits for their simulation at these temperatures. In addition, the traditional design methodologies that rely on empiric transistor models often require iterative simulations, making the process time-consuming and computationally expensive.
This research project aims to facilitate the design and test of cryogenic circuits using machine learning (ML) techniques, focusing on 28 nm FD-SOI technology for the development of control and readout interfaces for silicon quantum bits. This thesis will consider the modeling, design and test of cryogenic devices and circuits, exploiting ML techniques for tasks such as: (1) the extraction of transistor models for AC, DC and noise analysis, with the models adapted for operation down to cryogenic temperatures; (2) the study of the correlations of circuit behavior with temperature that can be used for screening tests of cryogenic devices at room temperature; (3) the prediction and tuning of circuit behavior at cryogenic temperature as a function of room temperature measurements using data-driven models; (4) the analog synthesis and optimization of cryogenic circuits using reinforcement learning. In addition, generative AI techniques may be explored to propose novel circuit architectures suited for cryogenic conditions, exploiting the data-driven models extracted. The techniques developed will be demonstrated experimentally in silicon devices.