Quantum Computing RF

Cooling Power

Q: How much cooling power does a dilution refrigerator have at the mixing chamber?

Commercial cryogen-free dilution refrigerators typically specify cooling power at the 100 mK reference point, where modern units deliver roughly 400 to 1000 microwatts. At the more common qubit operating temperature of 10 to 20 mK, the available cooling power falls to only a few microwatts to a few tens of microwatts because the mixing-chamber cooling power scales with the square of temperature. The continuous helium-3 circulation rate sets the maximum: cooling power is approximately 84 times the molar circulation rate times temperature squared, so a fridge circulating 1 millimole per second of helium-3 provides about 8.4 microwatts at 10 mK. Larger fridges with circulation rates of several millimoles per second reach the 1 milliwatt class at 100 mK, which is what makes 100-plus qubit systems with hundreds of coaxial lines practical.

Q: Why does cooling power drop so sharply as temperature decreases?

In the dilution process, cooling comes from the enthalpy of mixing as helium-3 atoms cross the phase boundary from the concentrated phase into the dilute phase. The net cooling per mole scales with the difference of the squares of the mixing-chamber and incoming temperatures, so the available cooling power follows a temperature-squared law. Dropping from 100 mK to 10 mK reduces the temperature by a factor of 10, which reduces the ideal cooling power by a factor of 100. This is why qubit thermal management is so demanding: at 10 mK there may be only single-digit microwatts of headroom, and every attenuator, coaxial line, and circulator that conducts or dissipates heat consumes part of that tiny budget. Designers therefore stage attenuation across the 4 K, still, and cold-plate flanges so that most heat is intercepted before it reaches the mixing chamber.

Q: How is the cooling power budget split between cryostat stages?

A dilution refrigerator has several temperature stages, each with its own cooling capacity. A pulse-tube cryocooler handles the 50 K and 4 K plates, providing watts to tens of watts. The still stage near 0.7 to 1 K offers a few milliwatts to tens of milliwatts. The cold plate near 100 mK and the mixing chamber near 10 mK have the least capacity, microwatts to hundreds of microwatts. RF input lines are heat-sunk and attenuated progressively, so that the large room-temperature thermal-noise load and the Joule heating from control pulses are dissipated mostly at the warmer, higher-capacity stages. A common rule allocates the bulk of input-line attenuation at 4 K and the still, leaving only 10 to 20 dB at the coldest flanges, which keeps the mixing-chamber heat load well under the few-microwatt cooling power available there.

/KOO-ling POW-er/

In cryogenic quantum systems, the rate at which a refrigerator can remove heat at a specified stage temperature, almost always quoted as the mixing-chamber capacity of a dilution refrigerator. Because the dilution process delivers cooling proportional to the square of temperature, a fridge rated at 400 to 1000 μW at 100 mK provides only a few μW of headroom near the 10 to 20 mK qubit operating point. That tiny budget sets a hard limit on how much qubit-control RF wiring, attenuation, and amplifier dissipation a system can carry, so the heat load from every coaxial line must be intercepted at warmer stages before it reaches the mixing chamber.

Category: Quantum Computing RF

Typical (100 mK): 400 to 1000 μW

Scaling: ∝ T²

Why the Heat Budget Governs Quantum RF Wiring

Cooling power is the single most binding constraint on the input/output chain of a superconducting quantum processor. Every microwave drive line that carries control pulses from room temperature down to the qubits brings two unavoidable heat sources: blackbody and Johnson thermal-noise photons conducted from warmer stages, and ohmic dissipation when attenuators absorb that noise and the control pulses themselves. The colder the stage, the less capacity the refrigerator has to soak up that heat, so a system can only be as large as its coldest flange can stay thermalized. With only single-digit microwatts available at the mixing chamber, a 100-qubit machine routing hundreds of coaxial lines lives or dies by how carefully its thermal budget is partitioned.

The governing physics is the dilution process itself. Cooling arises from the enthalpy absorbed when helium-3 atoms cross the phase boundary into the dilute helium-3/helium-4 phase inside the mixing chamber. The net cooling per circulated mole scales with the difference of the squares of the mixing-chamber and return temperatures, which collapses to the well-known temperature-squared law for the steady-state cooling power. Halving the operating temperature therefore cuts the available cooling power by a factor of four, and this steep falloff is precisely why operating qubits below 20 mK demands aggressive heat-sinking rather than simply a bigger compressor.

For RF component designers, this reframes familiar specifications. A cryogenic attenuator is rated not only by its insertion loss and return loss but by how much power it can dissipate at a given flange and how well it thermalizes. Likewise, an isolator, circulator, or coaxial cable is judged by its thermal conductance from the warmer stage above it. The art of cryogenic RF design is delivering clean, low-noise signals to the qubit while spending as little of the mixing-chamber cooling power as possible.

Cooling Power and Heat-Load Equations

Mixing-Chamber Cooling Power (general):
Q̇_mc ≈ ṅ₃ × (95 T_mc² − 11 T_in²) [W]

Ideal-Exchanger Maximum (T_in → T_mc):
Q̇_mc,max ≈ 84 × ṅ₃ × T_mc² [W] (84 = 95 − 11)

Temperature-Squared Scaling:
Q̇(T₂) / Q̇(T₁) ≈ (T₂ / T₁)²

Stage Heat Balance (margin):
Q̇_cool(T) > P_conduction + P_{RF dissipated} + P_radiative

Where ṅ₃ = helium-3 molar circulation rate (mol/s), T_mc = mixing-chamber temperature (K), T_in = incoming helium-3 temperature (K). The 95 and 11 J·K⁻²·mol⁻¹ coefficients are the dilute- and concentrated-phase enthalpy terms; with an ideal continuous heat exchanger T_in approaches T_mc and the net reduces to 84 T_mc². Example: ṅ₃ = 1 mmol/s at T_mc = 10 mK → Q̇_mc,max ≈ 8.4 μW.

Cooling Power by Cryostat Stage

Stage	Typical Temperature	Cooling Power	Cooling Mechanism	RF Heat Sources Absorbed
1st PT stage	40 to 50 K	20 to 40 W	Pulse-tube cryocooler	Cable conduction, radiation shield
2nd PT stage	3 to 4 K	0.5 to 2 W	Pulse-tube cryocooler	Bulk of input-line attenuation, HEMT amps
Still	0.7 to 1 K	2 to 30 mW	Helium-3 evaporation	Mid-stage attenuators, line thermalization
Cold plate	~100 mK	400 to 1000 μW	Dilution unit (cold ³He return / heat exchanger)	Final-stage attenuation, isolators
Mixing chamber	10 to 20 mK	2 to 30 μW	Dilution (phase boundary)	Last 10 to 20 dB, qubit package, circulators

Common Questions

Frequently Asked Questions

How much cooling power does a dilution refrigerator have at the mixing chamber?

Commercial cryogen-free units specify cooling power at 100 mK, where modern fridges deliver about 400 to 1000 μW. At the 10 to 20 mK qubit operating point it falls to a few μW up to a few tens of μW, because cooling power scales as T². The continuous helium-3 circulation rate sets the ceiling: Q̇_mc ≈ 84 × ṅ₃ × T², so 1 mmol/s gives roughly 8.4 μW at 10 mK. Several-mmol/s fridges reach the 1 mW class at 100 mK, enabling 100-plus qubit systems with hundreds of coaxial lines.

Why does cooling power drop so sharply as temperature decreases?

Dilution cooling comes from the enthalpy of mixing as helium-3 crosses the phase boundary into the dilute phase, and the net cooling scales with the difference of the squares of the mixing-chamber and incoming temperatures. Going from 100 mK to 10 mK lowers temperature by 10x but cuts ideal cooling power by 100x. That is why every attenuator, coaxial line, and circulator that conducts or dissipates heat consumes part of a single-digit-μW budget, so attenuation is staged across the 4 K, still, and cold-plate flanges.

How is the cooling power budget split between cryostat stages?

A pulse-tube cryocooler handles the 50 K and 4 K plates with watts to tens of watts. The still near 0.7 to 1 K offers a few mW to tens of mW. The cold plate near 100 mK and the mixing chamber near 10 mK have the least, μW to hundreds of μW. RF input lines are heat-sunk and attenuated progressively so the large room-temperature noise load and pulse Joule heating dissipate mostly at warmer stages, typically leaving only 10 to 20 dB at the coldest flanges to keep the mixing-chamber load well under its few-μW capacity.

Related Terms

← Cool-Down Time Cooper Pair →