image.imageformat.1286.dpr2.1236685061

Ethz.ch

Physicists at ETH Zurich led by Tilman Esslinger found that a superfluid loses the least energy flowing through a channel of intermediate width, near 4 µm (0.00016 in), and loses more energy as the channel gets wider, in a result Nature published November 18 under the title “Quantum resistance paradox of low-dimensional superfluids.” The finding cuts against the common expectation that a wider conductor resists less, and the minimum sits in a window where neither of the two standard dissipation models fits the data.

The experiment used a cloud of neutral lithium atoms, not electrons, and the paper tests no electronic hardware. The authors present the data as a benchmark for theory and say it suggests a route to minimizing dissipation in superconducting devices, a suggestion they do not test on any device.

The paper is open access, and a preprint has been on arXiv since May 26, 2026. Below: how the team built a programmable channel, how it read resistance from oscillating atom counts, and why the exact position of the minimum remains fuzzy.

Digital Micromirror Device Sets Channel Width Digitally, Run by Run

Two superfluid reservoirs hold about 270,000 lithium-6 atoms in total (2.7 ± 0.1 × 10⁵), cooled to 0.144 times the Fermi temperature, and are joined by a channel made of light. The gas sits at unitarity, the strongly interacting point where the scattering length diverges. The channel is 27 µm (0.0011 in) long inside a transport region of about 51 µm (0.0020 in), and a vertical confinement frequency of about 10.5 kHz squeezes the gas into a quasi-two-dimensional layer.

Two repulsive beams define the channel, and one of them is shaped by a digital micromirror device, a chip of tiny tilting mirrors. Larry Hornbeck of Texas Instruments invented the technology in 1987, and it became the core of DLP projectors. Here it lets the team set the channel width digitally from 0.9 µm (0.000035 in) to 19.8 µm (0.00078 in) between runs, with a width uncertainty of about 0.5 µm (0.000020 in). A third, attractive Gaussian beam tunes the local chemical potential in the channel.

The case for the platform is that the gas is defect-free. In solid-state superconductors, disorder, impurities and geometric imperfections can mask how dissipation depends on geometry, and here width is the main variable that changes. Esslinger’s group built on earlier two-reservoir work, including a 2012 Nature paper that observed the drop in resistance at the onset of superfluidity and a 2015 result on quantized conductance in a narrow channel.

RLC Circuit Model Converts Atom-Number Oscillations Into Resistance

The team starts each run with an imbalance of about 10% in atom number between the reservoirs (ΔN/N ≈ 0.1) and watches it discharge over time. The authors model the system as an RLC circuit, the same equation that describes a weight on a spring, with the resistor playing the role of friction. The imbalance oscillates at just under 10 Hz in the narrowest channels and about 18 Hz in the widest, and the damping of those oscillations carries the resistance information.

Capacitance comes from the reservoirs’ compressibility, modeled with a half-harmonic trap, and inductance comes from fitting damped sinusoids to the oscillations. The tradeoff is that the circuit picture rests on modeling choices: according to the preprint, a calibration error in capacitance would rescale the inductance and resistance values without changing the fit quality or the conclusions about mechanisms.

Phase Slips Explain Narrow Channels Over Many Orders of Magnitude

In a one-dimensional wire, residual resistance comes from phase slips: moments when the superfluid’s order parameter is briefly suppressed and its phase jumps by 2π. The four narrowest channels match the Langer-Ambegaokar-McCumber-Halperin (LAMH) theory of thermally activated phase slips, which predicts resistance falling exponentially as the channel widens. The fit holds up to a width of about 2.5 µm (0.000098 in).

The preprint’s abstract says the exponential scaling holds over more than ten orders of magnitude, while its introduction says roughly ten. The published Methods cite more than four orders of magnitude for an alternative fit that adjusts the initial current, a reminder that the dynamic range depends on how the fit is set up.

Vortex Model Was Built for Superconducting Strips, Not Atom Clouds

In a two-dimensional film, residual resistance comes instead from vortices crossing the sample, each relaxing the phase difference by 2π. The four widest channels follow a finite-size vortex model based on a 2024 theory by Nakagawa and Kogan, with resistance rising as a power law in width and a fitted exponent of 1.98 ± 0.04.

That theory was derived for thin-film superconducting strips of arbitrary width, a geometry used in superconducting single-photon detectors. In those detectors, thermally driven single-vortex crossings can produce false counts with no photon, according to separate work by Bulaevskii, Graf and Kogan. The wide-channel fit here therefore uses a model built for those strip geometries. The experiment tests no detector, and the paper makes no claim about one.

Why Does Superfluid Resistance Rise When Channels Widen?

Between roughly two and four healing lengths, neither model works. The healing length ξ, the scale over which the superfluid order parameter recovers, is about 1.2 µm (0.000047 in) here, estimated indirectly from an earlier experiment. In that window, both phase slips and vortices appear suppressed, which the authors attribute to a shift in the dominant dissipative mechanism.

Past the minimum, resistance climbs again. The paper’s abstract states the consequence plainly: widening a channel increases its resistance. The authors call this deviation from expected monotonic behavior the quantum resistance paradox and list it among their three main findings.

Independent theory offers context but not an explanation. A 2021 analysis of superconducting strips near zero temperature by König, Protopopov, Levchenko, Gornyi and Mirlin examined resistance generated by vortex configurations. The lithium gas sits at a finite temperature, T/T_F = 0.144, a different regime, so this experiment does not directly test that analysis.

Energy Analysis, Not Direct Resistance Fit, Locates Minimum With Fuzzy Position

The minimum is not a direct resistance reading. The paper infers it from a model-free energy analysis: the fraction of the initial energy still present at the extremal points of the oscillation, at times beyond about 0.15 seconds, peaks near 4 µm (0.00016 in), described as roughly 3ξ. More energy retained means less dissipation. The authors report that these energy-ratio estimates carry a typical positive bias of 0.5% with a standard deviation below 2%.

The paper’s own numbers are not perfectly consistent. With ξ near 1.2 µm (0.000047 in), 3ξ is about 3.6 µm (0.00014 in), not 4 µm, and the paper elsewhere puts 4ξ at about 5 µm (0.00020 in), so the minimum’s exact position carries some uncertainty. The team tested for systematic effects from capacitance, width, temperature and atom number. A width blur of about 0.5 µm (0.000020 in), for example, shifts resistance by a constant factor that the fit absorbs without changing the slope.

The authors also state that the oscillations are not Josephson oscillations. The relative phase between reservoirs was not controlled and sampled values from 0 to 2π, as measured in a slightly different geometry.

Single Temperature, Fixed Interaction Strength and No Microscopic Theory Bound Result

The sample sits at a single temperature, T/T_F = 0.144. Temperature enters the LAMH model, but the paper presents no temperature series, so how the minimum moves with temperature is untested. The same goes for interaction strength, held at unitarity.

The authors state that in the crossover window neither model provides a microscopically satisfactory description. The vortex model also underestimates oscillation amplitudes near that window, so extending it to finite widths remains open. The paper offers no quantitative prediction for the depth or location of the minimum. The LAMH fit also shows a relatively high chi-squared, which the authors attribute possibly to the box-potential approximation or to local heating by the attractive beam.

Swiss National Science Foundation Grants Fund Experiment Amid National Quantum Push

The authors acknowledge four Swiss National Science Foundation grants, alongside an ETH Zurich grant. Esslinger is also listed as a signatory of a 2022 letter from the Swiss quantum community urging a national initiative covering quantum computing, simulation, materials, sensing and communication, along with infrastructure and education. Switzerland’s Federal Council adopted the Swiss Quantum Initiative in May 2022.

Benchmark for Theorists, Open Question for Device Designers: Wider Is Not Always Lower Loss

For theorists, the dataset is a clean target: a defect-free, strongly interacting system with controllable geometry and a clear anomaly the standard models cannot explain, which the authors present explicitly as a benchmark for theories. For device physicists, the hint is that making a strip wider may not always reduce losses, and that an intermediate width may suppress both topological dissipation channels. Whether that carries over to solid-state superconductors, with their disorder, different carriers and different dimensional constraints, is the question the paper leaves to follow-up work.

The obvious next experiments are temperature and interaction-strength scans, which would show whether the minimum is robust or fine-tuned. A successful theory would need to reproduce both the location and the depth of the dip. The authors say the data are in Supplementary Information, and fitting code is available from the corresponding author on request.

Frequently Asked QuestionsWhat are phase slips and vortices, and why do they cause resistance in a superfluid?

A phase slip is a brief event in a narrow wire where the superfluid’s order parameter is suppressed and its phase jumps by 2π, releasing energy. A vortex is a quantized whirlpool-like defect that, in a two-dimensional film, can cross the sample and relax the phase difference by the same 2π. In both cases the relaxation dissipates energy, which shows up as residual resistance, according to the arXiv preprint’s introduction.

Could this result reduce energy loss in superconducting devices?

Not yet. The authors say the measurements suggest a route to minimizing dissipation in superconducting devices, but the experiment used neutral lithium atoms and no electronic hardware. The vortex model behind the wide-channel fit was built for thin superconducting strips, so strip width is a natural place to test the idea, but the paper reports no such test.

What is a unitary Fermi gas, and why did the team use one?

A unitary Fermi gas is a gas of fermionic atoms with strong interactions tuned so the scattering length diverges. The team used lithium-6 at unitarity because it forms a clean superfluid whose geometry can be programmed, free of the defects that blur solid-state experiments.

Is the resistance minimum a direct measurement?

No. It is inferred from a model-free energy analysis of how much of the initial energy remains at the oscillation extrema, with a typical positive bias of 0.5% and a standard deviation below 2%. The preprint also notes the vortex model underestimates oscillation amplitudes near the crossover.