Cryogenic Systems

Cryogenic Attenuator (Detail)

Q: How much attenuation should I put on each stage of a qubit drive line?

A common distribution for a 6 GHz superconducting qubit drive line is 20 dB at the 4 K stage, 10 dB at the still (around 0.8 K), 6 dB at the cold plate (around 0.1 K), and 20 dB at the mixing chamber (around 10 mK), totaling roughly 56 to 60 dB. The goal is to add attenuation where there is cooling power to absorb the dissipated heat and to ensure the photons arriving at the qubit are thermalized to a temperature low enough that residual thermal-photon population does not dephase the qubit. The mixing-chamber attenuator dominates the final noise temperature, so it is placed last and bolted directly to the coldest plate.

Q: Why not use the same attenuator design at room temperature and at 10 mK?

Standard thin-film attenuators are built on alumina or FR4 substrates with epoxy bonds and brass bodies that contract differently from the stainless or beryllium-copper coax around them. At 10 mK those mismatches crack solder joints and degrade return loss. Cryogenic attenuators use materials with matched contraction and high thermal conductivity, such as beryllium-copper bodies, gold-plated contacts, and substrates that stay conductive of heat at millikelvin temperatures, so the resistive film actually reaches the plate temperature. They are also rated for the very low signal powers (often below -80 dBm at the qubit) where self-heating of the film is the limiting factor, not average power.

Q: How does adding a cold attenuator improve the noise temperature of a line?

An attenuator at physical temperature T contributes noise as if it were a matched load at T scaled by its loss. When you place attenuation at a colder stage, the noise added downstream is referenced to that colder physical temperature rather than to 300 K. The output noise temperature is approximately T_out = T_in / L + T_phys (1 - 1/L), where L is the linear loss factor. A 20 dB attenuator (L = 100) sitting at 10 mK presents a noise temperature near 10 mK to anything downstream, which is why the final-stage attenuator essentially sets the thermal photon floor seen by the qubit.

/kry-uh-JEN-ik uh-TEN-yoo-ay-tur/

Bolted directly to a cold plate inside a dilution refrigerator, a cryogenic attenuator is a fixed-loss coaxial component whose primary job is thermal, not just electrical. It strips the 300 K blackbody noise riding down a qubit drive line and re-emits a far weaker noise spectrum referenced to the cold plate's physical temperature, dumping the absorbed power into the cryostat heat sink. Because the residual noise temperature presented to the device equals the attenuator's physical temperature, placing the final stage at the mixing chamber (around 10 mK) sets the thermal-photon floor for the experiment. These parts feed superconducting qubits and ultra-sensitive receivers ahead of a cold low-noise amplifier.

Category: Cryogenic Systems

Typical Stages: 4 K / 0.8 K / 0.1 K / 10 mK

Total Drive Loss: ~56 to 60 dB

Why Cold Attenuation Sets the Qubit Noise Floor

In a quantum-computing or cryogenic-receiver wiring stack, signal attenuation is deliberately added at multiple temperature stages rather than all at once. The reason is thermodynamic: an attenuator behaves like a matched resistive load at its own physical temperature, so when it absorbs an incoming signal it also re-radiates Johnson noise corresponding to that temperature. A 20 dB pad sitting at room temperature would inject 300 K worth of thermal photons downstream, swamping a qubit that needs to see an effective temperature in the tens of millikelvin. By distributing the loss and anchoring each pad to a progressively colder plate, the line discards both the unwanted signal power and the warm noise that traveled with it.

The placement of the last attenuator matters most. Whatever noise temperature it presents propagates straight to the device, so the mixing-chamber pad is mounted as close to the qubit as the wiring allows and bolted to the 10 mK plate with a high-conductivity thermal path. Earlier-stage pads at 4 K and the still plate carry the bulk of the dissipated power because those stages have far more cooling capacity. A typical 6 GHz drive line splits roughly 20 dB / 10 dB / 6 dB / 20 dB across the 4 K, still, cold-plate, and mixing-chamber stages.

Thermalization is the second design pillar. The resistive film must actually reach plate temperature, which is non-trivial when the only heat path is through millimeter-scale contacts at millikelvin. Cryogenic-grade pads use beryllium-copper bodies, gold-plated interfaces, and substrates chosen for thermal conductivity that survives down to 10 mK, paired with stainless or NbTi coax that limits heat leak between stages. The same hardware appears in front of cold receivers inside a dilution refrigerator used for deep-space and dark-matter instrumentation.

The Cold-Attenuator Noise Equation

Output noise temperature of an attenuator at physical temperature T_phys:
T_out = (T_in / L) + T_phys × (1 − 1/L)

Linear loss factor from decibels:
L = 10^{(A / 10)} (A = attenuation in dB)

Mean thermal photon occupancy at frequency f:
n̄ = 1 / (e^{(hf / k_BT)} − 1)

Example: a 20 dB pad (L = 100) at T_phys = 10 mK presents T_out ≈ 10 mK to the qubit, regardless of the warm noise at its input. At 6 GHz, hf / k_B ≈ 0.29 K, so n̄ ≈ 4 × 10⁻¹³, far below one photon.

Per-Stage Attenuation Budget (6 GHz Drive Line)

Stage	Plate Temp	Attenuation	Cooling Power	Role
50 K shield	~50 K	0 to 3 dB	~1 W	Coax heat break, optional pad
4 K plate	~4 K	20 dB	~1 W	Bulk noise dump, high power budget
Still	~0.8 K	10 dB	~10 mW	Intermediate thermalization
Cold plate	~0.1 K	6 dB	~400 μW	Pre-final attenuation
Mixing chamber	~10 mK	20 dB	~20 μW	Sets final noise floor at qubit

Common Questions

Frequently Asked Questions

How much attenuation should I put on each stage of a qubit drive line?

A common split for a 6 GHz drive line is 20 dB at 4 K, 10 dB at the still (~0.8 K), 6 dB at the cold plate (~0.1 K), and 20 dB at the mixing chamber (~10 mK), totaling about 56 to 60 dB. Loss is added where there is cooling power to absorb the dissipated heat, and the final pad is bolted directly to the coldest plate because it dominates the noise temperature delivered to the qubit.

Why not use the same attenuator design at room temperature and at 10 mK?

Standard thin-film pads on alumina or FR4 with brass bodies contract differently from the stainless or beryllium-copper coax around them, cracking joints and degrading return loss at millikelvin. Cryogenic pads use matched-contraction beryllium-copper bodies, gold-plated contacts, and substrates that stay thermally conductive at 10 mK, and they are characterized at the very low powers (often below −80 dBm) where film self-heating is the limiting factor.

How does adding a cold attenuator improve the noise temperature of a line?

An attenuator radiates noise as a matched load at its own physical temperature, so T_out ≈ T_in/L + T_phys(1 − 1/L), where L is the linear loss factor. A 20 dB pad (L = 100) at 10 mK presents roughly 10 mK to anything downstream regardless of the warm noise at its input, which is why the final-stage attenuator essentially sets the thermal-photon floor seen by the qubit.

Related Terms

← Cryogenic Attenuator Cryogenic Cable →