Researchers have built a levitating superconducting oscillator that loses energy extremely slowly. A lead-plated sphere of about 6.3 milligrams was suspended in a magnetic trap at millikelvin temperatures and left to ring down for more than 110 hours. That long decay time corresponds to a resonance linewidth smaller than 0.8 microhertz — a record low for a superconducting object in a magnetic field.

The device works by running opposing currents through two superconducting coils to make a nearly flat levitation potential in the horizontal plane. The team used a 2 mm diameter lead sphere as the levitated object and avoided so-called flux pinning problems by choosing a type I superconductor (lead). Two extra pairs of lateral coils let them nudge the trapped sphere and tune its natural frequency. The whole setup sits inside a dilution refrigerator at a base temperature around 5 millikelvin. To watch the motion they measured radio‑frequency signals between small coils and calibrated that signal against optical measurements and simulations.

In the cold vacuum run that produced the longest result, the horizontal natural frequency was about 2.7 hertz and the ring-down time was (114 ± 3) hours, equivalent to a coherence time of roughly 4.1×10^5 seconds. When they repeated experiments with the cell filled with liquid helium, the measured decay times changed with temperature as expected from the fluid’s behaviour. The authors were able to detect drag caused by trace amounts of helium‑3 impurities in superfluid helium‑4 at the level of roughly 10^−8 of the fluid, and they report a drag force of order 3×10^−15 newtons (about 3 femtonewtons) for one measured case.

This level of isolation and long coherence matters because mechanical oscillators used as force sensors are limited by how fast they lose energy. A slower loss means less thermal noise and better sensitivity. The authors give a figure of merit T/τ (bath temperature divided by coherence time) of about 10^−8 kelvin per second for their device, and they say the system is compatible with adiabatic nuclear demagnetisation — a cooling step that could push temperatures below one millikelvin and lower T/τ toward 10^−10 kelvin per second. Such sensors could help probe tiny forces relevant to precision tests of the boundary between quantum and classical physics.