Quantum Computing and Quantum RF Cryogenic Microwave Engineering Informational

How do I characterize the microwave loss of materials at millikelvin temperatures?

Characterizing microwave loss at millikelvin temperatures is essential for developing low-loss materials for superconducting qubits, where even parts-per-million (ppm) level dielectric loss directly limits qubit coherence. The primary measurement technique uses superconducting resonators: a CPW or lumped-element resonator is fabricated from the material under test (or on a substrate of the material) and its quality factor Q is measured at the operating temperature (10-20 mK). The internal quality factor Q_i = 1/tan(delta_eff) captures all loss mechanisms: Q_i = (1/Q_loaded - 1/Q_coupling)^-1, where Q_loaded is measured from the resonance linewidth and Q_coupling is the designed coupling Q. For high-purity sapphire substrates: Q_i > 10^7 (tan delta < 10^-7). For silicon: Q_i = 10^6-10^7 depending on processing. For amorphous SiO2 (which must be avoided for qubits): Q_i ≈ 10^3-10^4 (tan delta ≈ 10^-3-10^-4). Measurement requirements: (1) Single-photon power level: loss in dielectrics at mK temperatures is power-dependent due to two-level system (TLS) saturation. At high power, TLS defects saturate and Q_i increases; at single-photon power (the qubit operating regime), Q_i is at its minimum. Measure at average photon number n_avg ≈ 1 in the resonator, corresponding to drive power of approximately -140 to -130 dBm at the resonator input. (2) Multiple temperatures: measure Q_i vs temperature to distinguish TLS loss (constant below ~100 mK, decreasing above as TLS begin to thermally depopulate) from quasiparticle loss (exponentially decreasing below Tc).

Category: Quantum Computing and Quantum RF

Updated: April 2026

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

Millikelvin Microwave Loss Measurement

Understanding and minimizing microwave loss at millikelvin temperatures is one of the most active research frontiers in superconducting quantum computing. The loss tangent of materials at these temperatures and power levels cannot be predicted from room-temperature measurements and must be measured directly.

Technical Considerations

The dominant loss mechanism in superconducting circuits at mK temperatures is coupling to two-level system (TLS) defects in amorphous dielectric materials. TLS are atomic-scale structural defects (tunneling atoms, dangling bonds, adsorbates) in amorphous oxides, substrate surfaces, and metal-substrate interfaces. Each TLS has a resonance frequency and couples to the microwave electric field, absorbing energy. The aggregate effect of billions of TLS at different frequencies creates a frequency-independent loss tangent: tan delta_TLS = pi × P_0 × p^2 / (3 × epsilon_0 × epsilon_r), where P_0 is the TLS density of states and p is the TLS dipole moment. At single-photon power, TLS are unsaturated and provide maximum loss. At high power (>10^4 photons), TLS saturate and the effective loss decreases, making high-power measurements not representative of qubit operating conditions. The power dependence follows: Q_i(n) = Q_i(n→∞) × (1 + n/n_crit)^(1/2), where n_crit is the critical photon number for TLS saturation (typically 10-1000 depending on the TLS-resonator coupling).

Performance Analysis

The standard technique is the "hanger" or "notch" resonator measurement: a resonator is weakly coupled to a feedline, and the S21 transmission spectrum shows a dip at the resonance frequency. Fitting the complex S21 to a resonator model extracts three parameters: resonance frequency f_0 (determines the resonator properties), coupling Q (Q_c, set by the coupling capacitor design), and internal Q (Q_i, the quantity of interest). The measurement chain: room-temperature VNA or signal generator → attenuated coaxial line to mK stage → on-chip resonator → amplified output chain (JPA/TWPA + HEMT + room-temperature amplifier) → VNA or digitizer. Key measurement details: (1) Calibrate the power at the resonator input by measuring the total chain attenuation (using a through-line calibration standard on the same chip). (2) Measure at multiple power levels spanning the single-photon to saturated regime. (3) Measure at multiple temperatures (10 mK to 1K) to separate TLS from quasiparticle contributions. (4) Fabricate multiple resonators at different frequencies (4-8 GHz) on the same chip to check for frequency dependence.

Performance verification: confirm specifications against the application requirements before finalizing the design
Environmental factors: temperature range, humidity, and vibration affect long-term reliability and parameter drift
Cost vs. performance: evaluate whether the application demands premium components or standard commercial grades

Design Guidelines

Best reported Q_i values at single-photon power and 20 mK: High-resistivity silicon substrate (>10 kΩ-cm): Q_i = 2-5 × 10^6. Sapphire (C-plane, EFG-grown): Q_i = 5-10 × 10^6. Niobium on sapphire: Q_i = 1-5 × 10^6 (limited by Nb surface oxide). Aluminum on silicon (transmon typical): Q_i = 0.5-3 × 10^6. Tantalum on sapphire (recent breakthrough): Q_i = 5-30 × 10^6 (reduced TLS density at the metal-substrate interface). TiN on silicon: Q_i = 1-10 × 10^6. These Q_i values translate directly to qubit T1 limits: T1_max = Q_i / (2*pi*f) = Q_i/(2*pi*5e9). For Q_i = 10^6 at 5 GHz: T1_max = 32 μs. For Q_i = 10^7: T1_max = 320 μs. This explains why materials research is critical for pushing qubit coherence beyond current limits of ~100-300 μs.

Common Questions

Frequently Asked Questions

Why is single-photon measurement power important?

Superconducting qubits operate at the single-photon energy level (E = hf ≈ 3.3 × 10^-24 J at 5 GHz). Dielectric loss at this power level is dominated by unsaturated TLS defects that strongly absorb single photons. At higher power levels (>1000 photons), TLS saturate and the apparent loss decreases by 2-10×, giving a misleadingly optimistic Q_i. Many early qubit loss studies reported high-power Q values that did not predict actual qubit T1 times. Modern loss characterization always reports Q_i at the single-photon level, which directly predicts qubit performance.

What materials should be avoided in qubit circuits?

Amorphous materials with high TLS density are the worst offenders: SiO2 (tan delta ≈ 10^-3): avoid as substrate, interlayer dielectric, or anywhere near the qubit. Si3N4 (tan delta ≈ 10^-4): acceptable for some applications but not in high-field regions. SiNx (PECVD): highly variable, generally 10^-3 to 10^-4, avoid. Photoresist residues: must be completely removed; Even sub-monolayer residues contribute measurable loss. Native metal oxides (AlOx, NbOx): unavoidable on exposed surfaces but minimized through in-situ processing, surface treatments (HF dip for Si, buffered oxide etch for metals), and encapsulation. Best materials: crystalline substrates (silicon, sapphire), and epitaxial or columnar-growth metal films (Al, Ta, Nb, TiN) with minimal grain boundary and interface oxide.

How do I distinguish TLS loss from other loss mechanisms?

Three diagnostic signatures: (1) Power dependence: TLS loss decreases with increasing photon number (saturating behavior). Conductor loss and radiation loss are power-independent. (2) Temperature dependence: TLS loss is approximately constant below 100 mK and decreases above 100 mK as tanh(hf/2kT). Quasiparticle loss exhibits an exponential decrease below Tc. (3) Capacitive vs inductive participation: TLS loss primarily occurs in regions of high electric field (in dielectrics and at interfaces). Moving to lumped-element designs that minimize the electric field in lossy regions (parallel-plate capacitors with vacuum gap) can isolate the TLS contribution. By combining power and temperature sweeps: if the low-power low-temperature Q is limited and independent of temperature, TLS dominates. If Q improves exponentially with cooling, quasiparticle loss dominates.

Keep Reading