A tenth of a millimeter is 100 micrometers, and 100 micrometers set against a 100-nanometer wire is a factor of a thousand [4]. The phys.org account of the Optica paper calls it more than 100 times wider [3], which is true only in the loose sense. Anyone sizing an optical system off that sentence should do the division themselves.
The reason the old width was fixed had nothing to do with collecting light. A photon has to break superconductivity across the full width of the wire to produce a pulse, and NIST postdoc Eli Mueller puts the difficulty plainly: getting a device into a regime where one photon can manage that across 100 microns is very hard, so wires stayed in the hundreds of nanometers [7]. Narrow wires carry less current, and current is what sets sensitivity. At lower current a low-energy photon makes a smaller disturbance and the readout blip is fainter [8]. The cap on current was not the material. It was the edges, where fabrication defects let current pile up in eddies, and where that pileup fires false counts of its own [6][9].
The fix therefore addresses the edges rather than the photon. Rails of superconducting material border the central wire and carry current in the same direction; their magnetic field meets the wire's own and redistributes the flow, so nothing accumulates at the boundary [11]. With the edges no longer the weak point, the wire can sit closer to the transition between the superconducting and normal states, which Mueller identifies as where sensitivity to very low-energy photons lives, and a hot spot can then be raised across an arbitrarily wide wire [10]. Width becomes a consequence of the current you can hold rather than a limit on it [12].
What the account does not carry is the scoreboard. No efficiency, no dark count rate and no timing jitter are given for the rail-stabilized devices [16]. The 98% is attached to the conventional nanoscale generation [1], and until the wide wires are measured on the same axes, the comparison that matters has not been made. Polarization insensitivity is asserted for them [13], which is worth real money on any channel where the orientation of the photon's field is not under the experimenter's control, and that includes the deep-space links and quantum networks NIST offers as motivation [14].
Read narrowly, this is a fabrication argument rather than a detection one. Specialized nanometer-scale patterning is the entry fee for SNSPDs today [5], and a stripe a tenth of a millimeter across is a different class of part to make. Whether that translates into detectors built outside the few labs with the process to do it depends on numbers that live in the paper and not in the summary of it.
