Article
The wiring and heat limits of superconducting qubits
Krinner and colleagues' 2019 EPJ Quantum Technology study presents a cryogenic setup suitable for 50 qubits at 14 millikelvin [1]. Its microwave wiring spans a large temperature gradient and adds heat near the sample.
- Published
- Reading time
- 5 min
- Cited
- 3 sources
Why do superconducting qubits run so cold?
A superconducting qubit is a circuit with discrete energy levels, controlled and measured by microwave signals. Heat can populate excited states, while thermal photons arriving through a cable can disturb the circuit and its readout. In EPJ Quantum Technology, Krinner and colleagues report typical qubit frequencies near 5 gigahertz and readout resonators near 7 gigahertz, making thermal radiation in the wiring a practical concern [1]. EPJ Quantum Technology also presents a thermally optimized setup suitable for operation of 50 qubits at 14 millikelvin [1].
A dilution refrigerator uses a helium isotope mixture to remove heat continuously from its coldest plate, called the mixing chamber. The chip is thermally anchored there, while intermediate plates intercept heat before it reaches the sample. The EPJ Quantum Technology design's 14 millikelvin operating estimate is device- and refrigerator-specific [1]. The goal is to keep thermal excitation and stray radiation low while preserving enough cooling capacity for wiring and control signals. [1]
What heat do qubit control and readout cables carry?
A drive line delivers microwave pulses to change a qubit's state. A flux line sets the magnetic flux that tunes a circuit, and an output line carries the weak readout signal to amplifiers. Each line runs from room-temperature instruments down to the chip. Heat conducted along the cable is the passive load; power dissipated in attenuators and cables during operation is the active load. Krinner et al. distinguish both contributions and also account for radiation from warmer refrigerator stages. [1]
The drive line must attenuate room-temperature noise before the signal reaches the chip. Attenuators do this by dissipating some incoming signal power at a known, thermally anchored stage, so they reduce noise but also add heat locally. In EPJ Quantum Technology, Krinner et al. explain that a 20-decibel attenuator transmits 1 percent of incident signal power and dissipates the remainder as heat [1]. The EPJ Quantum Technology noise model adds thermal radiation set by the attenuator's temperature [1]. Spreading attenuation across warmer plates reduces the noise reaching the chip without spending the entire heat budget at the coldest plate. [1]
How much cooling power is available at each stage?
The heat budget shrinks sharply toward the chip. EPJ Quantum Technology lists its nominal 50 kelvin stage at 35 kelvin, with 30 watts specified at 45 kelvin [1]. EPJ Quantum Technology lists its nominal 4 kelvin stage at 2.85 kelvin, with 1.5 watts specified at 4.2 kelvin [1]. EPJ Quantum Technology lists the Still plate at 882 millikelvin, with 40 milliwatts specified at 1.2 kelvin [1]. EPJ Quantum Technology lists the cold plate at 82 millikelvin, with 200 microwatts specified at 140 millikelvin [1]. EPJ Quantum Technology lists the mixing chamber at 6 millikelvin, with 19 microwatts specified at 20 millikelvin [1]. The two temperatures in each pair matter: the power ratings are not all measured at the stages' nominal operating temperatures. [1]
EPJ Quantum Technology measures per-drive-line passive loads of 45 milliwatts at 50 kelvin, 1.0 milliwatt at 4 kelvin, 4 microwatts at the Still, 0.4 microwatt at the cold plate, and 0.013 microwatt at the mixing chamber [1]. The table reports run-to-run errors alongside these averages. Even a small load at the mixing chamber matters because its available cooling power is measured in microwatts, and many parallel lines add their individual loads. [1]
Why do cryostats use coax, attenuators, and NbTi cable?
A cable must balance low thermal conduction against electrical performance. EPJ Quantum Technology describes stainless-steel drive coax to limit passive heat flow and niobium-titanium output coax between 4 kelvin and the mixing chamber to reduce microwave loss [1]. They measure both passive conduction and signal dissipation, showing why a line cannot be chosen by attenuation alone: a material that carries heat poorly may also attenuate the desired signal more strongly. [1]
Attenuators are clamped to intermediate plates so they thermalize the radiation they absorb instead of warming the chip. The wiring also needs thermal contact at each stage; a poorly anchored cable can conduct heat past a plate that should have intercepted it. Designers therefore budget the cable, attenuator, filters, connectors, and driven signal together. The useful limit is not simply how many channels fit through the cryostat, but how much heat each stage can absorb while the sample remains cold enough for coherent operation. [1]
Why does wiring constrain quantum-computer scale?
A larger processor needs more control, flux, and readout paths, but the coldest stages cannot absorb heat in proportion to an unconstrained number of cables. Each additional coax also needs connectors, routing room, and reliable thermal anchoring. EPJ Quantum Technology analyzes a 50-qubit setup with multiplexed readout shared among sets of 6 to 7 qubits [1]. EPJ Quantum Technology estimates that the approach could serve at least 150 qubits when disregarding space constraints [1]. That is a scaling estimate rather than a processor demonstration. [1]
The bottleneck is electrical and thermal at once. More lines can improve independent control, but they increase conductive heat leaks and the number of parts that must remain calibrated through cooldowns. Shared lines reduce the number of physical feedthroughs, but multiplexers and switches add electronics whose own heat and noise must be managed. A scalable design must therefore compare total cooling power, line count, signal quality, and chip-control needs at every temperature stage rather than optimize a single cable in isolation. [1][2]
Can multiplexers and flexible cables reduce the load?
Multiplexing combines signals so fewer physical lines can serve multiple channels. In a 2023 Nature Electronics paper, Acharya and colleagues demonstrated low-power cryogenic CMOS control electronics operating below 15 millikelvin and benchmarked the system with a superconducting qubit [2]. With appropriate thermal filtering, the Nature Electronics paper reports single-qubit gate fidelities above 99.9 percent, while noting marginal effects on coherence time [2]. This is evidence that control electronics can work at the qubit's temperature, not proof that a large processor can be wired without a heat penalty. [2]
Flexible stripline cables offer another way to route dense microwave wiring. A 2024 study published by AIP Publishing in Applied Physics Letters tested flexible stripline and semi-rigid coaxial control lines on a transmon and found no measurable difference in the qubit coherence or energy-relaxation performance in their experiment [3]. That result supports flex cables as a candidate for compact routing, but it does not establish identical heat loads or signal performance for every cable design. The engineering goal remains the same: deliver clean control and readout while keeping each refrigerator stage within its measured cooling budget. [3]
Sources
- Engineering cryogenic setups for 100-qubit scale superconducting circuit systemsEPJ Quantum Technology, 2019
- Multiplexed superconducting qubit control at millikelvin temperatures with a low-power cryo-CMOS multiplexerNature Electronics, 2023
- Equivalence of flexible stripline and coaxial cables for superconducting qubit control and readout pulsesAIP Publishing, 2024
SuperMatics: AI built for superconductors.