Cryogenic Circulator
Ferrite Physics in the Millikelvin Regime
The non-reciprocal behavior of any circulator comes from the gyromagnetic precession of unpaired electron spins in a biased ferrite. A static bias field B0 sets the spins precessing at the Larmor frequency, and a microwave signal that enters the ferrite junction couples to this precession differently depending on its direction of rotation. The result is a 120° rotation of the field pattern inside a symmetric junction, steering energy to the next port clockwise while starving the port behind it. Cryogenic circulators exploit the same Y-junction or stripline geometry as room-temperature parts, but every material choice is reconsidered for operation at 4 K, where helium cooling power is scarce and superconducting circuits nearby are exquisitely sensitive to stray magnetic fields.
Cooling helps more than it hurts. As the ferrite cools, its ferromagnetic resonance linewidth ΔH narrows and its saturation magnetization 4πMs rises modestly, both of which sharpen the non-reciprocal response and can improve isolation by several dB relative to 300 K. Conductor losses also drop because the resistivity of the gold or silver plating falls with temperature. The dominant engineering challenge is therefore not RF performance but thermal and magnetic integration: the device must add negligible heat load, hold its bias field stable across the cooldown, and be shielded so its fringing flux does not couple into adjacent SQUID amplifiers or Josephson parametric amplifiers.
Because an electromagnet would dump unacceptable ohmic heat onto a dilution-refrigerator stage, the bias field is supplied by a permanent magnet, usually samarium cobalt for its low temperature coefficient. The magnet is characterized cold and the junction is matched at the operating temperature, since the bias field strengthens slightly on cooldown. Two or three circulators are frequently cascaded in a single readout line to stack isolation up to 40 to 60 dB, protecting the quantum device from the broadband noise that the first amplifier radiates back toward it.
Governing Relations
fL = γ × B0, γ ≈ 28 GHz/T
Ideal scattering matrix (ports 1→2→3):
S21 = S32 = S13 ≈ 1, S12 = S23 = S31 ≈ 0
Isolation and insertion loss:
Isolation (dB) = −20·log10|S12|, IL (dB) = −20·log10|S21|
Added noise of a lossy cold component:
Tadd ≈ (L − 1) × Tphys
Where γ = gyromagnetic ratio, B0 = bias field, L = linear loss factor, Tphys = physical temperature. Example: 0.2 dB loss (L ≈ 1.047) at 4 K adds only Tadd ≈ 0.19 K of noise.
Cryogenic vs. Room-Temperature Circulators
| Parameter | Cryogenic (4 K) | Room temperature (300 K) | Why it differs |
|---|---|---|---|
| Operating temp | 4 K to 10 mK | 300 K | Mounted on cold stages |
| Insertion loss | 0.1 to 0.5 dB | 0.3 to 0.8 dB | Lower conductor resistance cold |
| Isolation | 18 to 25 dB | 15 to 20 dB | Narrower FMR linewidth cold |
| Bias source | SmCo permanent magnet | SmCo or NdFeB | No static heat dissipation allowed |
| Added noise | < 0.2 K | Set by 290 K loss | (L − 1)·Tphys tiny when cold |
| Magnetic shield | Mandatory (mu-metal) | Optional | Protects nearby SQUIDs/qubits |
Frequently Asked Questions
Why do cryogenic circulators use a permanent magnet instead of an electromagnet?
The bias field, typically 0.05 to 0.3 T, must come from a source that dissipates no static power, because a dilution-refrigerator mixing-chamber stage offers only tens of microwatts of cooling at 10 mK. A permanent magnet such as samarium cobalt supplies the field for free. SmCo has a small reversible coefficient near −0.03 percent per kelvin, so the field strengthens slightly on cooldown and the junction is tuned at the operating temperature, not at 300 K.
How much insertion loss and isolation does a 4 K circulator provide?
In the 4 to 12 GHz band a good device shows 0.1 to 0.5 dB insertion loss and 18 to 25 dB isolation over a 30 to 50 percent bandwidth, with return loss better than 15 dB. Cooling from 300 K narrows the ferromagnetic resonance and lowers conductor resistance, so it usually gains margin versus its warm performance. Terminating the third port in a cold 50 Ω load turns it into an isolator.
Why is a circulator placed between a qubit and the first amplifier in a quantum readout chain?
The readout signal is microwatts or less, so it feeds a near-quantum-limited amplifier such as a Josephson parametric amplifier, which reflects part of its input and radiates broadband noise backward. The circulator passes the forward signal to the amplifier and diverts the reflected noise to a cold 50 Ω load, keeping it away from the qubit. Cascading two or three reaches 40 to 60 dB of total reverse isolation.
What limits how small or how broadband a cryogenic circulator can be?
Junction diameter scales inversely with frequency, so a 4 to 8 GHz part is physically larger than a 30 GHz one, and mandatory magnetic shielding adds bulk. Bandwidth is set by the ferrite saturation magnetization and the matching network; pushing past about an octave trades isolation for bandwidth. These constraints drive active research into superconducting on-chip circulators based on parametric modulation.