Cryogenic CMOS control electronics for qubits

Short answer: Today's qubits are controlled by electronics at room temperature, connected by many cables into a dilution refrigerator. Researchers are testing chips built with standard CMOS technology that operate inside the cryostat, to cut the number and length of lines. Published experiments from 2020 and 2023 show such chips driving or multiplexing qubits at millikelvin temperatures, with measured power use and effects on qubit performance. These are laboratory demonstrations, not deployed systems.

The wiring problem

A 2024 review on arXiv (2410.15895) says fault-tolerant quantum computing is expected to need the integration of several million qubits, and that the interconnection complexity and latency this implies are, in its words, currently unsolvable with state-of-the-art room-temperature control and readout electronics. It describes cryogenic integrated circuits, including CMOS radio-frequency chips and rapid-single-flux-quantum-logic chips, as potential alternatives. Placed inside the cryostat, they reduce the number and length of transmission lines, and operating cold can suppress electronic noise and improve qubit control fidelity. The review adds that CMOS design at these temperatures is held back by a lack of reliable cryogenic transistor models, plus noise and power dissipation problems.

A 2020 chip for generating control signals

A 2020 Nature Electronics paper starts from the objection that CMOS is generally discounted because its power dissipation could heat fragile qubits. The authors report a CMOS platform that provides several electrical control signals for qubits at 100 mK. The chip is configured by digital input signals from room temperature and uses switched-capacitor circuit cells to generate static and dynamic voltages. They use it to bias a quantum dot device and to switch its conductance with voltage pulses generated on the chip. From six cells they report an average power dissipation of 18 nW per cell for 100 mV pulses, and estimate that a system with a thousand cells could be cooled by a commercially available dilution refrigerator.

A 2023 multiplexer below 15 mK

A 2023 Nature Electronics paper reports a low-power radio-frequency multiplexing cryo-electronics system operating below 15 mK with minimal cross-coupling. The authors connect it to a superconducting qubit and observe that the qubit's relaxation times are unaffected and the coherence times only marginally affected, in both static and dynamic operation. They report that single-qubit gate fidelities above 99.9% can be achieved with appropriate thermal filtering, which they note is above the threshold for surface-code error correction. They also demonstrate time-division multiplexing by windowing calibrated control pulses.

What these results do not show

Each experiment uses a small number of qubits or cells, and the thousand-cell cooling figure is an estimate. Whether such chips can be scaled to the qubit numbers the review describes, within a real cooling budget, is not settled by these papers.

FAQ

Why not keep control electronics at room temperature?

The 2024 review says the interconnect complexity and latency become unmanageable at the scale of millions of qubits.

Does a cold chip heat the qubits?

The 2020 paper measured 18 nW per cell, and the 2023 paper observed relaxation times unaffected and coherence times marginally affected.

Which fidelities were reported?

The 2023 paper reports single-qubit gate fidelities above 99.9% with appropriate thermal filtering.

What other cryogenic chip types exist?

The review names rapid-single-flux-quantum-logic chips alongside CMOS, and focuses mainly on CMOS as the more mature.

Is this in use in quantum computers today?

These papers describe laboratory demonstrations and make no such claim.