Continuous Heat Exchanger
How Counterflow Recuperation Sets the Base Temperature
In a dilution refrigerator the cooling power comes from forcing helium-3 across the phase boundary in the mixing chamber, where it crosses from the concentrated phase into the dilute helium-3/helium-4 phase. To keep that circulation going, warm helium-3 returning from the still (near 0.6 to 0.7 K) must be cooled to within a few millikelvin of the mixing chamber before it arrives, otherwise its enthalpy lands directly on the coldest stage and raises the base temperature. The continuous heat exchanger performs this duty by running the incoming concentrated stream and the outgoing dilute stream in opposite directions along a common wall, so each fluid element gives up heat to a slightly colder neighbor over the full length of the device.
The exchanger is described as continuous because it is a single uninterrupted counterflow channel, in contrast to the chain of thermally isolated step exchangers that follow it. A typical implementation is a tube-in-tube or annular geometry several meters long, often coiled, with the concentrated flow in the inner bore and the dilute return in the surrounding annulus. The wall is thin enough to pass heat radially yet long enough that axial conduction along the metal does not short-circuit the temperature gradient. Performance is rated by its effectiveness, the fraction of the maximum possible enthalpy transfer that the device actually achieves, and by the number of transfer units (NTU) that scales with area divided by the minimum heat-capacity flow rate.
Below about 30 mK the continuous design runs out of headroom. The Kapitza thermal boundary resistance between liquid helium-3 and the metal wall climbs as the inverse cube of temperature, so the wetted area needed to pass even microwatts of heat becomes enormous. That is the point where designers hand off to discrete sintered-silver step exchangers, each packing square meters of submicron-powder surface into a small volume. The continuous exchanger and the step chain together form the recuperative backbone that lets the system deliver clean cooling to a cryogenic RF cold plate or qubit package.
Effectiveness, NTU, and Kapitza Resistance
ε = [1 − e−NTU(1−Cr)] / [1 − Cr × e−NTU(1−Cr)]
Number of transfer units:
NTU = U × A / (ṁ × cp)min , Cr = (ṁcp)min / (ṁcp)max
Kapitza boundary resistance:
RK = a / (A × T3) ≈ 0.02 to 0.05 K4·m2/W ÷ (A × T3)
Where U = overall heat-transfer coefficient, A = wetted area, ṁ = mass flow, cp = specific heat, Cr = heat-capacity-rate ratio, T = absolute temperature. Example: NTU ≈ 8, Cr ≈ 0.6 → ε ≈ 0.98 (effectiveness saturates fast, so NTU above ~15 buys little). Note RK × 1000 for every 10× drop in T.
Continuous vs. Step Exchanger Selection
| Property | Continuous (counterflow) | Step (sintered silver) | Continuous-flow cryostat HX |
|---|---|---|---|
| Geometry | Tube-in-tube / annulus, 1 to 5 m long | Discrete sponge cells in series | Coiled capillary in vapor stream |
| Useful T span | ~0.6 K to 30 mK | ~30 mK to 5 mK | 300 K to 4 K (gas precool) |
| Surface area | 0.01 to 0.1 m² (smooth wall) | 100 to 3,000 m² (sinter) | 0.01 to 0.05 m² |
| Dominant loss | Axial conduction, pressure drop | Kapitza resistance | Longitudinal conduction |
| Target effectiveness | ε ≈ 0.95 to 0.99 | Per-stage ΔT limited | ε ≈ 0.85 to 0.95 |
| Typical count | 1 per fridge | 3 to 6 in series | 1 per cold finger |
Frequently Asked Questions
Why does a dilution refrigerator need both continuous and discrete (step) heat exchangers?
The continuous exchanger handles the upper span, roughly the still at 0.6 to 0.7 K down to about 30 to 50 mK, where helium-3 viscosity and conduction are manageable and a single long counterflow channel works well. Below ~30 mK the Kapitza resistance rises as 1/T3, so a continuous unit would need impractical area; designers then switch to a series of sintered-silver step exchangers, each thermally isolated, packing thousands of m² into a few cm³. A typical fridge uses one continuous plus three to six step exchangers to reach 5 to 10 mK.
How much surface area is needed to overcome Kapitza resistance at millikelvin temperatures?
RK = a / (A × T3), with a near 0.02 to 0.05 K4·m²/W for sintered silver in liquid helium-3. Because resistance scales as 1/T3, the area needed for a given heat flow grows 1000× for every 10× drop in temperature. To carry tens of microwatts at 10 mK the exchanger must present hundreds to thousands of m², which is why submicron silver powder (about 700 angstrom particles) is sintered into the channels, giving 1 to 3 m² per gram.
What limits the effectiveness of a continuous counterflow heat exchanger?
Three competing losses: axial conduction along the walls that short-circuits heat from warm to cold end, viscous pressure drop in the narrow helium-3 channels that adds dissipation, and Kapitza resistance that blocks full equilibrium across the wall. A practical unit targets ε of 0.95 to 0.99, which for a heat-capacity-rate ratio near 0.6 corresponds to NTU of roughly 5 to 15 (counterflow effectiveness saturates quickly with NTU). Pushing effectiveness higher needs more length and area, which raises both conduction and pressure drop, so the optimum balances recuperation against parasitic load. Residual unrecovered enthalpy directly raises base temperature and cuts cooling power.