Cryogenic Wiring
Engineering a Low-Heat-Load Microwave Harness
A superconducting quantum processor sits at the bottom of a dilution refrigerator, typically at 10 to 20 mK, yet every gate and measurement originates from room-temperature electronics. The cabling that bridges those six orders of magnitude in temperature is the cryogenic wiring harness. It must transport drive pulses, flux bias, and dispersive readout tones across roughly 1 to 12 GHz while obeying a brutal constraint: the available cooling power at the mixing chamber is only tens of microwatts. Every cable, connector, and attenuator is selected first for its thermal behavior and second for its RF performance.
The dominant noise problem is Johnson thermal radiation traveling down each input line from the 300 K source. Left unchecked, this would populate the qubit's microwave environment with many thermal photons, destroying coherence. The standard fix is staged attenuation: discrete cryogenic attenuators are bolted to the 4 K, still, 100 mK, and mixing-chamber plates so each one re-references the noise to its colder local temperature. Because an attenuator dissipates the incident power as heat, the attenuator on the coldest plate must be small in value and well heat sinked, and the bulk of the attenuation is placed higher up where cooling power is abundant.
Output and readout lines follow the opposite philosophy. They must add as little loss and noise as possible so the weak signal from the qubit survives to the first cryogenic amplifier. Here designers favor superconducting NbTi coax between the still and mixing chamber, which carries microwave signals with negligible RF loss and very low thermal conductivity, plus isolators and directional couplers to block amplifier back-action. Material choice along the line is therefore stage dependent rather than uniform.
Conductive Heat Load and Heat Sinking
The conductive heat flow through a cable segment between two stages is governed by the integrated thermal conductivity, not a single room-temperature value, because k(T) of metals drops sharply at low temperature. Each segment is clamped and often bobbin-wound at every plate so it thermalizes fully before continuing downward. A harness of 100 or more lines forces a hard budget: the summed mixing-chamber load from all cables and attenuators must stay within a few microwatts of the available cooling power.
Thermal and RF Governing Relations
Q = (A / L) × ∫TcTh k(T) dT (W)
Thermal photon occupation seen by qubit:
n̄ = 1 / (ehf / kBTeff − 1)
Effective noise temperature after attenuation:
Teff ≈ Tsrc / A + Tstage × (1 − 1/A) , A = 10(dB/10)
Where A = cross-section area, L = segment length, k(T) = temperature-dependent thermal conductivity, f = signal frequency, Teff = effective temperature of the noise at the qubit, A = linear attenuation factor. Example: a 1 m, 0.86 mm stainless coax from 4 K to 10 mK carries only a few μW; a 20 dB MXC attenuator drives Teff toward the local 10 mK plate so n̄ ≈ 0.01 photon at 6 GHz.
Coax Material Selection by Stage
| Coax Type | Conductor / Shield | Thermal Conductivity | RF Loss (6 GHz) | Typical Use |
|---|---|---|---|---|
| Stainless steel (SS-SS) | 304 SS center & shield | Very low (~1/25 of Cu) | ~8 to 12 dB/m | 300 K to 4 K input lines |
| BeCu-clad SS | BeCu center, SS shield | Low | ~4 to 7 dB/m | 4 K to still, moderate loss |
| NbTi superconducting | NbTi center & shield | Very low | < 1 dB/m (below Tc) | Still to MXC readout output |
| Silver-plated Cu | Cu with Ag plating | High | ~1 to 3 dB/m | Avoided below 4 K (heat leak) |
| NbTi flexible (semi-rigid) | NbTi/CuNi composite | Low | ~1 to 2 dB/m | Flux & fast-flux bias lines |
Frequently Asked Questions
How many attenuators do you put on a qubit drive line and at which stages?
A common XY-drive recipe distributes attenuation across the plates to kill 300 K Johnson noise: roughly 20 dB at 4 K, 10 dB at the still (~0.8 K), 6 dB at 100 mK, and 20 dB at the mixing chamber (~10 to 20 mK), for about 56 to 60 dB total. The mixing-chamber attenuator is the most thermally critical because its dissipated power lands on the coldest stage, so it is kept small and heavily heat sinked. The goal is n̄ well below one photon while still allowing fast gates.
Why use stainless steel or NbTi coax instead of copper inside a dilution refrigerator?
Copper conducts heat far too well; a copper line would overwhelm the mixing chamber's microwatt-scale cooling power. Stainless steel coax has roughly 1/25 the thermal conductivity of copper, slashing conductive heat leak while accepting higher RF loss. Between the still and mixing chamber, superconducting NbTi coax carries microwave signals with near-zero RF loss and very low thermal conductivity, making it ideal for sensitive readout output lines.
How is the heat load of a cryogenic wiring harness budgeted across stages?
Conductive flow per segment is Q = (A/L) × ∫ k(T) dT between the two stage temperatures, and every segment is clamped and heat sinked at each plate so it thermalizes before continuing. With only 10 to 20 μW of cooling power near 10 mK, a 100-line harness must hold its total mixing-chamber load to a few microwatts. That budget is the main limit on how many qubit control lines one fridge can host.