A Sub-kHz Mechanical Resonator Passively Cooled to 6 mK

Highly coherent mechanical resonators are invaluable to ultrasensitive detection techniques by enabling detection of small forces. Studying mechanical resonators in a thermal equilibrium state at millikelvin temperatures provides a promising path to increase their coherence time. Here, we passively cool a 700 Hz massive (1.5 ng) mechanical cantilever down to 6.1(4) mK by means of nuclear demagnetization, as confirmed by detecting its thermal motion via a lock-in based detection scheme. At the lowest temperatures the thermal motion of the resonator is still clearly distinguishable from the background noise. Our data analysis confirms that at these temperatures the motion is still thermally distributed. These results pave the way for passive cooling low-frequency resonators to the sub-millikelvin regime, which would enable new tests of quantum mechanics and advances in ultrasensitive force detection.

💡 Research Summary

The authors present a pioneering demonstration of passive cooling of a sub‑kilohertz mechanical resonator to the sub‑millikelvin regime using nuclear demagnetization. A silicon cantilever, equipped with a 7.3 µm Nd₂Fe₁₄B magnetic tip, has a resonance frequency of approximately 700 Hz, a spring constant of 26 µN·m⁻¹, and an effective mass of 1.5 ng. The experimental platform combines a cryogen‑free dilution refrigerator (base temperature ≈ 20 mK) with a PrNi₅ nuclear demagnetization stage. A high‑purity silver wire provides a thermally conductive but mechanically isolated link between the cantilever, a superconducting pickup loop, and the demagnetization stage. This configuration allows the cantilever and its read‑out chip to reach temperatures well below the refrigerator’s base temperature while preserving the vibration isolation afforded by a mass‑spring suspension system.

Thermal motion of the cantilever is sensed via magnetic flux coupling into a two‑stage SQUID current sensor. A calibration coil placed between the pickup loop and the SQUID input coil generates a known test flux, enabling precise determination of the flux‑to‑voltage conversion factor κ. The detection scheme is essentially non‑invasive; dissipation is limited to the weak magnetic coupling between resonator and detector. To monitor the resonator’s energy in real time, the authors employ a digital lock‑in amplifier referenced to the measured resonance frequency (≈ 700 Hz) with a 1 Hz bandwidth, roughly twice the resonator’s thermal noise bandwidth. The resonator’s energy correlation time τ = 2Q/ω₀ is about 7 s, so a two‑hour acquisition yields roughly 10²⁸ statistically independent energy samples.

The authors construct histograms of the measured energy and compare them to the Boltzmann distribution exp(−E/kBT). Both the mean energy and the slope of the exponential tail give consistent temperature estimates. In the first cooling run (run A) the cantilever temperature saturates at 6.1 ± 0.4 mK, while in a second run (run B) it saturates at 7.7 ± 0.4 mK. Simultaneous measurements with a magnetic‑flux‑fluctuation thermometer (MFFT) attached to the same silver wire report bath temperatures of 3.4 mK (run A) and 3.1 mK (run B). The cantilever temperature is systematically higher than the MFFT temperature, indicating a residual thermal link limitation rather than excess external vibration, which would produce diurnal temperature fluctuations that are not observed.

A linear fit of the form T_cantilever = c · T_MFFT yields c ≈ 1.07 for run A (close to unity) and c ≈ 1.33 for run B. The authors attribute the discrepancy in run B to calibration uncertainties, possibly arising from unaccounted electrostatic forces on the cantilever tip. Their calibration accounts only for magnetic forces; any residual charge could introduce an additional drive of unknown sign, affecting the inferred displacement amplitude.

The work demonstrates that nuclear demagnetization can cool low‑frequency, relatively massive mechanical resonators to temperatures where their thermal motion remains observable yet thermally limited. This passive cooling preserves the resonator’s equilibrium state, unlike active sideband or feedback cooling schemes that inject external drive and decoherence. Achieving sub‑10 mK temperatures for a 700 Hz resonator opens the path to dramatically higher mechanical quality factors by suppressing intrinsic loss channels (e.g., two‑level systems, surface losses). Higher Q and lower force noise directly benefit a range of precision experiments, including nanoscale magnetic resonance imaging, solid‑state spin detection, and tabletop tests of gravity at short ranges.

Looking forward, the authors suggest that further reduction of the thermal resistance between the silver wire and the demagnetization stage, as well as refined displacement calibration that includes electrostatic contributions, could push the resonator temperature below 1 mK. At such temperatures, the resonator’s quantum ground‑state occupation would become appreciable even without active cooling, enabling new tests of quantum‑mechanical collapse models such as CSL and providing an ultra‑low‑noise platform for force sensing at the zeptonewton level.