Search for Axion-Like Particles in High-Magnetic-Field Pulsars with NICER

Axion-like particles (ALPs) can couple to photons in strong magnetic fields, producing characteristic fluctuations in X-ray spectra. Using data from NASA’s Neutron Star Interior Composition Explorer (NICER), we analyzed three pulsars: PSR J2229+6114, PSR J1849-0001, and PSR B0531+21, to search for such features. Each spectrum was modeled with a sliding-window power-law fitting method to identify local deviations from the smooth continuum. From these analyses, we derived constraints on the axion-photon coupling constant $g_{aγγ}$ assuming dipole and quadrupole magnetic field distributions. We obtain upper limits in the range $10^{-10}$–$10^{-13},\mathrm{GeV}^{-1}$ for ALP masses in the $0.8$–$10~\mathrm{keV}$ region.

💡 Research Summary

The authors present a search for axion‑like particles (ALPs) that couple to photons in the intense magnetic fields of young rotation‑powered pulsars, using observations from NASA’s Neutron Star Interior Composition Explorer (NICER). Three bright, high‑magnetic‑field pulsars—PSR J2229+6114, PSR J1849‑0001, and the Crab pulsar (PSR B0531+21)—are analyzed in the soft X‑ray band (approximately 0.8–10 keV). The central idea is that photon‑ALP conversion in the magnetosphere would imprint narrow, localized deviations (“wiggles”) on an otherwise smooth power‑law spectrum.

Data reduction and spectral fitting
NICER data are processed with HEASoft 6.35.2 and NICERDAS 11, employing the standard “3c50” empirical background model. For each pulsar, the authors extract spectra over a selected energy range and fit an absorbed power‑law continuum, $F(E)=AE^{-\Gamma}$, as a baseline. To search for local anomalies, they adopt a sliding‑window technique: a moving window of 4 bins (below 2 keV) or 10 bins (higher energies) is fit independently across the spectrum. Because windows overlap, each energy bin is associated with several fits; the fit with the lowest reduced χ² is taken as the representative local model. This approach is intended to increase sensitivity to narrow features that could be washed out in a global fit.

Statistical significance is quantified via a “pull” variable $z_i = (y_{\rm data,i} - y_{\rm model,i})/(s_i\sigma_i)$, where $s_i$ is a scaling factor from the local fit. Two‑tailed Gaussian probabilities $p_{\rm two-tailed,i}=2