Dark photons — hypothetical ultralight vector bosons that mix kinetically with the ordinary photon — would oscillate as a coherent background field if they constitute the dark matter. The oscillation frequency corresponds to the dark photon mass. A magnetometer sensitive to oscillating magnetic fields could detect them, but the signal would be indistinguishable from ordinary magnetic noise at the same frequency. The background and the dark matter look the same to a sensor.
Bloch, Dery, Hochberg, Volansky, and collaborators (arXiv 2602.22308, February 2026) exploit geometry to break this degeneracy. Their three-axis magnetometer sits inside a conductive shielded room. The shielding attenuates external electromagnetic fields, but dark photons pass through — their coupling to ordinary matter is too weak to be shielded. However, the dark photon signal's polarization creates a specific directional pattern in the three-axis readout. One axis is geometrically defined to have zero dark-photon response — a null axis, determined by the orientation of the detector relative to the dark photon's polarization direction in the galactic frame.
The null axis becomes a noise reference. Any signal appearing on the null axis is not dark photons; it is ordinary noise. Subtracting the null-axis noise from the signal axes removes common-mode contamination without removing the dark photon signal. The noise floor drops. The sensitivity improves.
The NASDUCK' experiment (the prime denotes this upgraded analysis) constrains the kinetic mixing parameter across the mass range 4×10⁻¹² to 2×10⁻⁹ eV, improving on previous laboratory limits by up to three orders of magnitude. No dark photon signal is found. The constraints are the strongest laboratory bounds in this mass range.
The detector didn't change. The shielding didn't change. What changed was recognizing that the geometry contains a direction where the signal should be absent — and using that silence as the reference against which everything else is measured.