Quantum Computing RF

Copper Powder Filter

Q: How does a copper powder filter attenuate signals without using lumped inductors or capacitors?

A copper powder filter relies on distributed skin-effect loss rather than reactive elements. The signal wire meanders through a compacted mixture of micron-scale copper grains and epoxy. At low frequency the wire simply behaves like a resistor of about 1 to 10 ohms, so DC and control signals pass freely. As frequency rises, eddy currents induced in the individual copper grains become confined to a skin depth that shrinks as 1 over the square root of frequency, dramatically increasing the effective series resistance and dielectric loss seen by the wire. The result is monotonic absorptive attenuation that grows roughly as the square root of frequency, reaching 40 to 100 dB across 1 to 20 GHz with no sharp resonances or reflections to leak power back toward the qubit.

Q: What grain size and packing fraction give the best high-frequency attenuation?

Smaller grains and higher metal packing fraction both increase loss. Copper powder in the 1 to 50 micron range is typical; grains around 10 micron or smaller push the onset of strong attenuation down toward 0.3 GHz because the skin depth in copper equals the grain radius at a lower frequency. Packing fractions of 50 to 70 percent by volume, achieved by compacting the powder around a meander wire in stainless or copper tubing and back-filling with low-outgassing epoxy such as Stycast 2850FT, maximize grain-to-grain proximity and dielectric loss. The trade-off is DC resistance: very fine, lightly packed powder raises series resistance and self-heating, which matters on bias lines carrying milliamp currents.

Q: Why are copper powder filters mounted at the mixing chamber stage of a dilution refrigerator?

They serve a dual role as both an electrical filter and a thermalization element, so they are bolted to the coldest plate, the mixing chamber, typically at 10 to 20 millikelvin. The long meandered wire plus the large metal mass provides strong thermal anchoring that cools the inner conductor and removes heat carried down from warmer stages. Electrically, placing the filter at the final stage strips residual GHz blackbody and Johnson noise just before the line reaches the qubit, where even attowatt-level photon noise would degrade coherence. The filter body itself emits only Johnson noise at the local stage temperature, so a 10 millikelvin mount contributes negligible added noise compared with a filter sitting at 4 kelvin.

/kop-er pow-der fil-ter/

Embedding a meandered signal wire in a compacted matrix of micron-scale copper grains creates one of the workhorse absorptive components of cryogenic wiring. The filter passes DC and slow control signals with only 1 to 10 Ω of series resistance, yet relies on distributed skin-effect loss inside the conductive grains to absorb microwave noise, reaching 40 to 100 dB of attenuation from roughly 1 to 20 GHz. Because attenuation is dissipative rather than reflective, the device adds no resonant peaks and presents only thermal noise at its mount temperature. Bolted to the mixing-chamber plate of a dilution refrigerator near 10 mK, it strips residual gigahertz photon noise just before the line reaches a superconducting qubit, preserving coherence.

Category: Quantum Computing RF

Cutoff f_c: 0.1 to 1 GHz

Attenuation: 40 to 100 dB (1 to 20 GHz)

Why Quantum Hardware Needs Lossy Wiring

Superconducting qubits, single-photon detectors, and other quantum devices operate at energies so small that a single stray microwave photon arriving down a control line can excite an unwanted transition or break a Cooper pair. The coaxial and twisted-pair wiring that carries DC bias and slow control signals from room temperature into a dilution refrigerator also acts as a near-perfect waveguide for blackbody and Johnson noise from the warmer stages. Conventional reflective low-pass filters help, but their out-of-band stopband is full of reflections and resonances that can resonantly trap or re-radiate noise. A copper powder filter solves this by making the wiring itself lossy: instead of reflecting unwanted energy, it absorbs and thermalizes it inside a dense conductive medium.

The construction is deliberately simple. A length of resistive or normal-metal wire, often 0.5 to 2 m, is wound into a tight meander or spiral and packed into a copper or stainless tube filled with a mixture of fine copper powder and epoxy. Each copper grain is electrically isolated from its neighbors by the thin oxide and epoxy skin, so the matrix does not short the wire to ground, but the grains are close enough that the wire's high-frequency fields couple strongly into them. At microwave frequencies the fields induce eddy currents confined to an ever-shrinking skin depth within each grain, and that dissipation appears as a rapidly rising distributed series loss along the wire.

Because the mechanism is absorptive and broadband, a single compact filter can cover the entire problematic band from below 1 GHz to above 18 GHz without the periodic passband re-entries that plague distributed reflective designs. The same dense metal that provides the loss also gives excellent thermal contact, so the filter doubles as a thermalization anchor that cools the inner conductor of the line, a function ordinary low-pass filters cannot perform.

Governing Loss and Cutoff Equations

Skin depth in a copper grain:
δ = √(2ρ / (ωμ)) ≈ 2.1 μm at 1 GHz, 0.66 μm at 10 GHz

Distributed insertion loss (empirical):
A(f) ≈ A₀ × √(f / f₀) dB (for f > f_c)

RC corner (3 dB point):
f_c = 1 / (2π R C_eff) ≈ tens of MHz

Where ρ = copper resistivity, μ = permeability, ω = 2πf, A₀ = reference loss, f₀ = reference frequency, R = wire series resistance, C_eff = distributed wire-to-matrix capacitance. The lumped RC corner sets only the gentle 3 dB roll-off point; strong absorptive attenuation (the practical onset quoted as 0.1 to 1 GHz) sets in higher up as the √f skin-effect loss accumulates. Example: a 1 m meander with R ≈ 5 Ω and C_eff ≈ 0.5 nF gives f_c ≈ 64 MHz, then roughly 30 dB at 1 GHz scaling to > 90 dB at 10 GHz.

Copper Powder vs. Other Cryogenic Noise Filters

Filter Type	Mechanism	Passband R	Stopband Atten	Useful Band	Best Use
Copper powder	Skin-effect absorption	1 to 10 Ω	40 to 100 dB	DC to ~20 GHz	DC bias and slow control lines
Eccosorb (lossy dielectric)	Magnetic and dielectric loss	< 1 Ω	20 to 60 dB	~1 to 40 GHz	Inline microwave drive lines
Lumped LC low-pass	Reflective reactance	< 0.5 Ω	40 to 80 dB (in band)	DC to f_c only	Sharp cutoff, narrow stopband
Thin-film distributed RC	Resistive line loss	~50 Ω	20 to 50 dB	DC to ~10 GHz	On-chip integration
Stainless powder	Skin-effect + higher ρ	10 to 100 Ω	50 to 110 dB	DC to ~20 GHz	Extreme attenuation, low current

Common Questions

Frequently Asked Questions

How does a copper powder filter attenuate signals without using lumped inductors or capacitors?

It relies on distributed skin-effect loss rather than reactive elements. A meander wire threads a compacted mix of micron-scale copper grains; at DC the wire is just a 1 to 10 Ω resistor, so control signals pass. As frequency rises, eddy currents in each grain crowd into a skin depth shrinking as 1/√f, raising effective series loss so attenuation grows roughly as √f, reaching 40 to 100 dB over 1 to 20 GHz with no reflective resonances.

What grain size and packing fraction give the best high-frequency attenuation?

Smaller grains and higher metal packing both raise loss. Copper powder of 1 to 50 μm is typical; grains near 10 μm push strong attenuation onset toward 0.3 GHz. Packing fractions of 50 to 70 percent by volume, compacted around the wire and back-filled with low-outgassing epoxy such as Stycast 2850FT, maximize grain proximity. The trade-off is DC resistance and self-heating on bias lines carrying milliamp currents.

Why are copper powder filters mounted at the mixing chamber stage of a dilution refrigerator?

They act as both filter and thermalization anchor, so they bolt to the coldest plate near 10 to 20 mK. The long meander plus large metal mass cools the inner conductor, while placing the filter at the final stage strips residual GHz blackbody and Johnson noise just before the line reaches the qubit. The body emits only Johnson noise at its local 10 mK temperature, so it adds negligible noise versus a 4 K mount.

Related Terms

← Copper Pour Copper Weight →