The study, a branch-paired reanalysis of previously published antenna-coupled YBCO bicrystal-junction data, reports no new measurements. Instead, it asks a narrower and sharper question: under what conditions does a detected maximum still represent an identifiable Shapiro-step voltage? The answer matters because a maximum-finding procedure always returns a number unless it is explicitly allowed to reject a measurement. A global maximum can be displaced by common-mode voltage offset, estimator bias, broadening, branch asymmetry, or an overlapping response component, and it can survive even after the local structure near the expected Josephson voltage has become too weak or too ambiguous to support a physical frequency estimate. Converting every maximum into a frequency can therefore yield a result that is not an identifiable Shapiro-step measurement at all.

The distinction between a measurable response and a physically meaningful voltage assignment has deep roots in the theory of driven Josephson junctions. The shape and visibility of the experimentally recorded observable depend on weak-link microstructure, spatial current inhomogeneity, damping and circuit loading, antenna matching, and quasi-optical coupling, as described in classical junction models. Nonlinear superconducting dynamics can further be shaped by Josephson coupling, microwave dissipation and disorder, and the external drive, and in YBCO junctions noise and the specific junction realization can modify the measured response. All of these effects reshape the observed differential response without altering the fundamental voltage-frequency relation itself, which remains exact whenever a genuine step is present. A measured maximum may broaden, weaken, become asymmetric, compete with neighboring structures, or disappear even while the underlying relation is unchanged.

This composite character explains why a response maximum can migrate away from the calibrated reference even while a reference-centered component persists underneath. If a broad, displaced contribution grows in amplitude as temperature rises, the sum of the contributions can peak at a voltage well removed from the Josephson value, producing an apparent detuning of several percent. The underlying primary component may remain exactly where the ac Josephson relation places it, yet an estimator that reports the location of the global maximum will faithfully return the displaced value. The error is not in the physics of the Josephson effect but in the assumption that the largest observed feature coincides with the locked response. This is precisely the scenario that a phenomenological reconstruction with a fixed, reference-centered primary component can reproduce: the composite maximum moves while the stipulated Josephson contribution stays put.

The role of noise in shaping these outcomes deserves particular attention. In high-temperature superconducting junctions, noise can round and broaden Shapiro steps, and the effective noise temperature of a junction is known to vary with its fabrication route and geometry. Noise-rounded steps were described in a ramp-type YBCO junction driven at 176 gigahertz using overdamped resistively shunted junction noise theory, and comparative studies across YBCO junction configurations have identified excess noise as a configuration-dependent figure of merit. Broadening from noise reduces the isolation of the first step from neighboring structure, making it easier for an overlapping feature or a common-mode offset to capture the global maximum. Low-frequency noise and branch asymmetry, both documented in step-edge and bicrystal devices, add further pathways by which the two polarities of the response can diverge, which is why pairing the positive and negative branches is such a powerful diagnostic: a genuine symmetric detuning of the locked response affects both branches equally, whereas voltage offsets and inter-branch drift generally do not.

Temperature acts on every one of these ingredients simultaneously. As the operating temperature changes, the critical current, normal resistance, characteristic voltage, linewidth, and coupling regime of the weak link all evolve. The characteristic frequency of the junction, set by the product of critical current and normal resistance divided by the flux quantum, determines whether the drive frequency lies well above or near the junction’s natural scale, and this ratio in turn governs the shape and amplitude of the Shapiro steps. Prior studies of related YBCO detectors have documented non-monotonic step amplitudes across this ratio, broadband response, and strong temperature evolution of the operating regime. None of these scalar quantities, however, directly answers the identifiability question, because each describes an aspect of the response while the identifiability of a step center depends on the joint condition that a paired, unique, and continuous maximum exists near the expected voltage.

The reanalysis quantified this with two devices measured on a common platform of a 330 nanometer YBCO layer and a 250 nanometer gold layer on a magnesium oxide bicrystal substrate, with the grain-boundary weak link between matching strips connected to a gold bow-tie antenna. Device A, with 5.4 micrometer-wide matching strips and a 1 millimeter hemispherical silicon lens, shows a clean progression at 155 gigahertz: branch-paired errors of 0.467 to 0.539 percent from 10 to 50 kelvin, a displacement to 3.481 percent at nominal 60 kelvin with branch widths growing from roughly 2.9 to about 12.9 and 10.8 microvolts, and complete loss of a baseline-accepted first-step center at nominal 70 kelvin even though remote maxima persist in the records. At 70 kelvin the global maximum yields a nominal error of 25.423 percent, which the continuity-aware analysis rejects rather than converts into a frequency.

The second device, fabricated on the same film stack and antenna platform but with 4.6 micrometer-wide matching strips and no silicon lens in the matched records, remains identifiable at all fifteen overlapping temperature-frequency coordinates spanning 50, 60, and 70 kelvin and 150 to 170 gigahertz. Its mean absolute paired error stays near 0.6 percent at all three temperatures, with a maximum of 1.071 percent. Notably, it achieves this despite a lower effective characteristic voltage and a higher ratio of drive frequency to characteristic frequency at elevated temperature, quantities that a naive figure-of-merit analysis might have flagged as unfavorable. The lesson is that identifiability is an emergent property of the whole measurement chain, including junction realization, electromagnetic coupling, thermal stability, and readout, rather than a deterministic function of any single device parameter.

The continuity requirement in the analysis framework adds a dimension that single-point maximum finding lacks. A physically meaningful step assignment should evolve smoothly as temperature or frequency is swept, because the Josephson voltage tracks the drive frequency exactly and the response shape changes gradually with operating conditions. A maximum that appears far from the trajectory traced by neighboring measurements, or one that cannot be matched between the two voltage polarities, fails this continuity test and is rejected. This explicit rejection capability is what separates the framework from conventional peak extraction, which by construction always returns a voltage. Allowing the analysis to declare a measurement non-identifiable converts a silent failure mode into a visible diagnostic, which is essential for any application that relies on the intrinsic calibration of the Josephson relation rather than on external calibration.

The retrospective nature of the study imposes honest limits on what can be concluded. The archived data lack independent bidirectional sweeps at the critical conditions, contemporaneous traceable voltage calibration, and synchronized temperature logging, so the observed displacement at nominal 60 kelvin cannot be definitively attributed to junction physics rather than to instrumental gain drift. A constant gain error of roughly 3.481 percent could in principle account for the paired detuning of about 11.158 microvolts at 155 gigahertz, and the archival resistor check, which demonstrates within-scan linearity only up to about 100.46 microvolts, below the first-step voltage of roughly 320.514 microvolts, cannot exclude such an error. Across the five matched frequencies the paired errors span 3.481 to 5.444 percent, and the best constant percentage-gain counterfactual of 4.401 percent leaves a residual root-mean-square of 0.670 percentage points, while more flexible counterfactual models require parameter combinations that are unstable when individual frequencies are excluded. These ambiguities do not undermine the central methodological contribution, but they do illustrate the evidentiary standard that any claim of physical detuning must meet.

For researchers developing high-temperature superconducting detectors, the practical implications are concrete. Spectroscopic and metrological applications that read a frequency directly from a step voltage should incorporate branch pairing, continuity tracking, and an explicit non-identifiable state into their analysis pipelines, rather than relying on linewidth, responsivity, or characteristic-frequency proxies alone. Prospective experiments designed to separate intrinsic response evolution from instrumental drift would combine controlled changes of radio-frequency coupling on a single device with traceable calibration, stabilized temperature, and repeated bidirectional sweeps. Such measurements would also help distinguish among the candidate mechanisms for displaced composite maxima, including spatially nonuniform weak-link behavior, changing antenna coupling, distributed nonlinear transport, local heating, and flux-related structure. Until then, the reanalysis stands as a caution and a tool: the ac Josephson relation remains exact, but the number reported by a maximum-finding algorithm is a hypothesis about the data that must be tested, not a calibrated frequency that can be taken on faith.

Subject of Research: Shapiro-step voltage identifiability in antenna-coupled YBCO Josephson weak links

Article Title: Shapiro-step identifiability in YBCO Josephson weak links: A branch-paired reanalysis

Article References: Holdengreber, E. (2026). Shapiro-step identifiability in YBCO Josephson weak links: A branch-paired reanalysis. Results in Physics, 90, Article 108771. https://doi.org/10.1016/j.rinp.2026.108771

Image Credits: AI Generated

DOI: 10.1016/j.rinp.2026.108771

Keywords: Shapiro steps, YBCO, Josephson junctions, weak links, superconducting detectors, voltage metrology, branch-paired analysis, identifiability, high-temperature superconductors, bicrystal junctions, differential conductance, terahertz sensing