Reimagining Cryogenic Computing with Superconducting and Topological Devices

Cryogenic computing - operating at or below 4 K - has emerged as a promising solution to address the scalability issues of quantum computers, which currently rely on room temperature controller and memory units. In addition, cryogenic computing offers unique advantages for high-performance computing and deep space exploration. Various technologies, including non-superconducting, superconducting, and hybrid, have been explored. However, each of these technologies has significant limitations. Nonsuperconducting devices suffer from low speed and high power consumption, while superconducting systems based on Josephson junctions (JJs) face severe scalability issues due to flux trapping and inductive coupling. Hybrid systems, meanwhile, lack efficient interface circuitry to integrate superconducting and non-superconducting technologies.

This dissertation utilizes two unique technologies: superconducting devices (ferroelectric superconducting quantum interference devices (FE-SQUIDs) and heater cryotrons) and topological devices exploiting the quantum anomalous Hall effect, to develop a suitable cryogenic computing platform. First, we demonstrate a non-volatile FE-SQUID-based memory with voltage-controlled operation, high scalability, and separate read-write paths. Using the same device, we develop a voltage-controlled superconducting logic family that addresses the limited fanout and cascadability issues of existing current-controlled logic circuits. We further present a cryogenic in-memory computing architecture based on FE-SQUIDs to reduce data movement and cooling costs associated with von Neumann systems. Additionally, we design a ternary content addressable memory (TCAM) using FE-SQUIDs which supports exact search without any peripheral circuitry, as well as nearest-match operations through Hamming distance calculation.

We also explore the reconfigurable potential of heater cryotron devices. A single cryotron circuit is shown to execute multiple 1-, 2-, and 3-input logic operations without any structural changes, enabling logic camouflaging for enhanced hardware security.

Finally, this dissertation is the first to utilize topological devices in circuit and systemlevel applications. Leveraging the quantum anomalous Hall effect, we develop a scalable, non-volatile, and ultra-low-power memory system. This topological memory is also capable of performing both basic and complex in-memory computing tasks, further advancing the development of a practical and efficient cryogenic computing platform.