Quantum Computing and Quantum RF Cryogenic Microwave Engineering Informational

What is the purpose of infrared filtering in the microwave lines of a quantum computer?

Q: Where in the cryostat should IR filters be placed?

At the mixing chamber stage (10-20 mK) for maximum effectiveness, as close to the qubit sample as practically possible. This ensures any thermal radiation from the filter body is at millikelvin temperature. Some labs place additional Eccosorb filters at the 4K stage to block the high IR flux from the upper stages, reducing the thermal load on the mixing chamber filters. On output lines, place the IR filter between the first circulator and the quantum-limited amplifier.

Q: How much insertion loss is acceptable?

For input (qubit drive) lines: 3-5 dB of IR filter insertion loss at the qubit frequency (4-8 GHz) is acceptable because the signal level is adjusted by the room-temperature electronics to compensate. For output (readout) lines: every dB of insertion loss before the first amplifier degrades the system noise temperature by 26% (at T << hf/k), directly reducing readout fidelity. Target <2 dB total IR filter loss on the output line. Eccosorb filters achieve 1-2 dB insertion loss at 6 GHz in a well-designed 5 cm section; copper powder filters achieve 1-3 dB depending on length and powder loading.

Q: Can commercial low-pass filters replace custom IR filters?

Partially. Commercial microwave low-pass filters (e.g., Mini-Circuits VLF series) provide excellent rejection above their cutoff frequency but may have limited attenuation at very high frequencies (>50 GHz) due to package leakage and resonant modes. They are useful as part of a multi-stage filtering approach: a commercial LPF for sharp cutoff at 12-15 GHz combined with an Eccosorb or copper powder filter for broadband absorption above 20 GHz. Commercial filters also need to be verified for cryogenic compatibility (solder joints, substrate materials, and magnetic components may fail or change properties at millikelvin temperatures).

Infrared (IR) filtering in quantum computer microwave lines blocks stray thermal photons at frequencies above the superconducting gap frequency (~100 GHz for aluminum) from reaching the qubit chip, preventing quasiparticle generation that degrades qubit coherence. Superconducting qubits (transmons, flux qubits) rely on the superconducting energy gap (2*Delta/h ≈ 100 GHz for Al) to maintain coherence. Photons with energy above 2*Delta can break Cooper pairs, creating quasiparticles that cause energy relaxation (T1 decay), dephasing, and charge parity fluctuations. Even at 20 mK, room-temperature blackbody radiation (peak at ~18 THz) leaking down the coaxial cables through the cryogenic stages produces enough IR photons to poison the qubit. Standard microwave attenuators provide attenuation at microwave frequencies (1-20 GHz) but are transparent to IR radiation above 100 GHz. IR filters are specifically designed to pass the qubit operating band (4-8 GHz) while blocking everything above approximately 15-20 GHz. Common filter types: Eccosorb CR-110/CR-124 loaded coaxial filters (lossy ferrite-filled epoxy surrounding a center conductor, providing 40-60 dB per cm above 20 GHz), copper powder filters (copper powder mixed with Stycast epoxy in a coaxial geometry, providing distributed absorption), and custom low-pass filters with cutoff frequencies around 10-12 GHz. Filters are thermally anchored at the mixing chamber or cold plate stage.

Category: Quantum Computing and Quantum RF

Updated: April 2026

Product Tie-In: Cryogenic Components, Attenuators, Circulators, Cables

IR Filtering for Superconducting Qubits

Quasiparticle poisoning from stray infrared radiation is one of the primary mechanisms limiting superconducting qubit coherence in modern quantum processors. Without adequate IR filtering, qubit T1 times can be reduced by 10-100× from their intrinsic limits.

Quasiparticle Generation Mechanism

When a photon with energy hf > 2*Delta (the superconducting gap energy) is absorbed by the superconducting thin film of a qubit, it breaks a Cooper pair into two quasiparticles. For aluminum (Tc = 1.2K, Delta = 180 μeV): the gap frequency is 2*Delta/h ≈ 87 GHz. For niobium (Tc = 9.2K, Delta = 1.5 meV): gap frequency ≈ 725 GHz. Quasiparticles in the qubit junction cause stochastic tunneling events that exchange energy between the qubit and the environment (T1 relaxation pathway) and cause random phase shifts (T2 dephasing). The quasiparticle density in equilibrium at 20 mK should be exponentially small (~10^-50 per Cooper pair), but non-equilibrium quasiparticles from stray radiation maintain densities of 10^-8 to 10^-6 per Cooper pair, far above equilibrium. Reducing this non-equilibrium population through IR filtering directly improves qubit T1.

Filter Technologies

Eccosorb filters: a length of coaxial cable filled with Eccosorb CR-110 (iron-loaded epoxy) or Eccosorb MF-124 absorbing material. The material's complex permeability provides frequency-dependent loss: low loss at 1-10 GHz (2-5 dB insertion loss in the passband), rapidly increasing loss above 20 GHz (>20 dB/cm at 100 GHz). Typical implementation: 10-15 cm of Eccosorb-filled coaxial section providing >60 dB attenuation above 50 GHz with <3 dB loss at 6 GHz. Copper powder filters: fine copper powder (1-10 μm particle size) suspended in Stycast 1266 epoxy, cast around a meander-line inner conductor in a copper housing. The skin effect in individual copper grains creates distributed loss that increases with frequency. Performance: 1-3 dB insertion loss at 5 GHz, >70 dB attenuation above 50 GHz. First demonstrated by Fukushima et al., now standard in many quantum computing labs. Custom low-pass filters: discrete-element or distributed filters with sharp cutoff at 12-15 GHz, providing >50 dB rejection above the cutoff. These are more precisely engineered but have narrower passband and potential for resonant behavior.

Installation Best Practices

IR filters should be installed at the coldest stage possible (mixing chamber or cold plate) to ensure that any thermal noise generated by the filter itself is at the lowest temperature. The filter must be in good thermal contact with the stage (bolted to a gold-plated copper bracket with indium foil or Apiezon N grease). Place IR filters after the last attenuator and before the qubit chip. On output lines, place IR filters after the first isolator/circulator. Multiple IR filter technologies can be combined in series (e.g., Eccosorb filter + copper powder filter) for maximum rejection. Total insertion loss in the qubit band should be minimized to preserve signal integrity: target <5 dB total for all IR filtering on the input line, <2 dB on the output line (where every dB of loss degrades readout SNR).

Quasiparticle Physics

Gap Frequency: f_gap = 2Δ/h
Al: Δ = 180 μeV, f_gap ≈ 87 GHz
Nb: Δ = 1.5 meV, f_gap ≈ 725 GHz
Quasiparticle Rate: Γ_qp ∝ x_qp × Δ/ℏ
where x_qp = quasiparticle density per Cooper pair

Common Questions

Frequently Asked Questions

Where in the cryostat should IR filters be placed?

At the mixing chamber stage (10-20 mK) for maximum effectiveness, as close to the qubit sample as practically possible. This ensures any thermal radiation from the filter body is at millikelvin temperature. Some labs place additional Eccosorb filters at the 4K stage to block the high IR flux from the upper stages, reducing the thermal load on the mixing chamber filters. On output lines, place the IR filter between the first circulator and the quantum-limited amplifier.

How much insertion loss is acceptable?

For input (qubit drive) lines: 3-5 dB of IR filter insertion loss at the qubit frequency (4-8 GHz) is acceptable because the signal level is adjusted by the room-temperature electronics to compensate. For output (readout) lines: every dB of insertion loss before the first amplifier degrades the system noise temperature by 26% (at T << hf/k), directly reducing readout fidelity. Target <2 dB total IR filter loss on the output line. Eccosorb filters achieve 1-2 dB insertion loss at 6 GHz in a well-designed 5 cm section; copper powder filters achieve 1-3 dB depending on length and powder loading.

Can commercial low-pass filters replace custom IR filters?

Partially. Commercial microwave low-pass filters (e.g., Mini-Circuits VLF series) provide excellent rejection above their cutoff frequency but may have limited attenuation at very high frequencies (>50 GHz) due to package leakage and resonant modes. They are useful as part of a multi-stage filtering approach: a commercial LPF for sharp cutoff at 12-15 GHz combined with an Eccosorb or copper powder filter for broadband absorption above 20 GHz. Commercial filters also need to be verified for cryogenic compatibility (solder joints, substrate materials, and magnetic components may fail or change properties at millikelvin temperatures).

Keep Reading