Scalar-Magnetometer Search for Ultralight Dark Photon Dark Matter with a Single-Site, Two-Sensor Array: A 6-Channel DTFT Likelihood Analysis with Scalar Optically Pumped Magnetometers

Abstract

We report on a laboratory search for ultralight dark photon dark matter using a single-site, two-sensor scalar magnetometer array. The experiment employs two scalar optically pumped magnetometers (OPMs) operated in a differential configuration to suppress common-mode noise and enhance sensitivity to spatially coherent dark photon fields. We analyze 10.5 hours of continuous data with a six-channel complex data vector evaluated at the three physical frequencies of the expected dark photon signal triplet. Assuming Gaussian noise, we develop a likelihood framework to set robust, frequency-resolved upper limits on the kinetic-mixing parameter $\varepsilon$, which governs the coupling between Standard Model photons and dark photons. Within the mass range $4\times10^{-15}\,\mathrm{eV} \leq m_{A'} \leq 3\times10^{-14}\,\mathrm{eV}$, we obtain the most stringent direct laboratory limits to date on $\varepsilon$, complementing existing astrophysical bounds including those inferred from observations of the Leo-T dwarf galaxy.

Figures (6)

• Figure 1: Photograph of a single Twinleaf PPM sensor head (the black rectangular unit) placed in its custom-designed, non-magnetic alignment mount. The mount, constructed from 3D-printed material, allows for precise adjustment of the sensor's orientation. The hand-drawn markings on the base serve as a reference for aligning the sensor's primary axis with the direction of the local geomagnetic field, $\hat{\mathbf{B}}_b$, a critical step for minimizing heading errors.
• Figure 2: The raw time-domain data from one of the optically pumped magnetometer sensors, showing the full 10.5h measurement period used for this analysis. The signal is dominated by the large, quasi-static geomagnetic field (approximately 49150nT). A clear diurnal variation with an amplitude of several nT is visible, upon which high-frequency instrumental and environmental noise is superimposed. The expected dark photon signal is many orders of magnitude smaller and is therefore completely buried within the noise fluctuations.
• Figure 3: The Linear Spectral Density (LSD) of the data from a single representative sensor, calculated over the entire 10.5-hour observation period. The plot reveals the instrument's noise floor as a function of frequency, which consists of a broadband continuum and numerous discrete spectral lines. The continuum shows a characteristic $1/f$ behavior at low frequencies, while at higher frequencies, the narrow peaks correspond to a mixture of instrumental artifacts, which are largely unique to this specific sensor, and environmental interference, which is observed in both. This complete spectrum represents the background against which the search for a narrow dark photon signal triplet is performed.
• Figure 4: Our 95% C.L. upper limit on the kinetic mixing parameter $\varepsilon$ as a function of the dark photon mass $m_{A'}$ (solid cyan line). Excluded regions from other searches are shown for comparison, including terrestrial constraints from SNIPE Hunt SNIPE_Hunt_Placeholder, the SuperMAG network (1-min FedderkeSuperMAG and 1-sec FedderkeSuperMAG_1sec datasets), and AMAILS AMAILS_Placeholder, as well as astrophysical and cosmological constraints from FIRAS FIRAS_Placeholder and Leo T LeoT_Placeholder. Our result establishes the most stringent direct laboratory limits to date in the mass range from approximately $4 \times 10^{-15}\,\text{eV}$ to $3 \times 10^{-14}\,\text{eV}$.
• Figure 5: Quantile-Quantile (Q-Q) plots testing the Gaussianity of the whitened noise data, aggregated from the frequency sidebands around a representative central frequency. The left panel shows the real part and the right panel shows the imaginary part. The empirical quantiles of the data (blue dots) closely follow the theoretical quantiles of a standard normal distribution (red line), providing strong visual support for the complex Gaussian noise assumption central to our likelihood framework.