Microlensing constraint on Primordial Black Hole abundance with Subaru Hyper Suprime-Cam observations of Andromeda

TL;DR

This study constrains sub-solar PBH dark matter via high-cadence microlensing toward M31 with Subaru/HSC, incorporating finite-source effects and a robust multi-stage selection pipeline. By combining 2014, 2017, and 2020 data with a hierarchical Bayesian framework that integrates light-curve information and Poisson event counts, it derives upper limits on PBH abundance and, under PBH-origin hypotheses, posterior regions for the PBH mass function. The results indicate a characteristic PBH mass scale around $M_{ m PBH}\\sim 10^{-7}$–$10^{-6}\,M_$ with $f_{ m PBH}\sim \mathcal{O}(10^{-2}-10^{-1})$ for the full candidate set, while analyses restricted to secure events yield weaker constraints. The work demonstrates the continued potential of M31 microlensing with HSC as a probe of PBHs and highlights the need for multi-band, longer-baseline surveys (e.g., Rubin/LSST, Roman) to mitigate systematics and improve sensitivity to short-timescale events.

Abstract

We present updated microlensing analysis results based on high-cadence ($\sim$2~min) Subaru Hyper Suprime-Cam (HSC) observations of the Andromeda Galaxy (M31) in 2014, 2017, and 2020, yielding a total of 39.3 hours of data. We use a point-lens finite-source model for the microlensing light curve model and employ multi-stage selection procedures to identify microlensing candidates. From more than 25,000 variable candidates detected across all nights, we identify 12 microlensing candidates with light-curve timescales shorter than 5~hours, and among them, 4 secure candidates with high-significance detections. We estimate detection efficiencies using light-curve-level simulations that account for observational conditions and finite-source effects. Using a hierarchical Bayesian framework that combines the light-curve fitting information for each candidate with the Poisson statistics of the number of candidates, we derive constraints on parameters that characterize the abundance and mass scale of primordial black hole (PBH) dark matter. First, we derive upper limits on the PBH abundance under the null hypothesis that all events are assumed to be false detections. Next, employing the PBH hypothesis in which all (or only secure) candidates are assumed to be due to PBH microlensing, we derive the allowed region of the PBH parameters; the inferred mass scale is $M_{\rm PBH}\sim10^{-7}$--$10^{-6}M_\odot$, and the PBH abundance to the total dark matter is $f_{\rm PBH}\sim \mathcal{O}(10^{-2}{\rm -}10^{-1})$. Our results demonstrate that HSC-M31 monitoring remains a uniquely powerful probe of PBHs, and highlight the need for further studies for example, using Rubin Observatory LSST observations of the Large Magellanic Cloud.

Figures (13)

• Figure 1: In each panel, the solid line indicates the PSF size (seeing) as a function of the observation time from the beginning of observation on each night, and the square points indicate the 10 images with the best seeing on each night, which are therefore used to generate the reference images for image subtraction, as described in Section \ref{['ssec:image-subtraction']}. The PSF size of the 2014-11-24 data is overplotted in gray in each panel for comparison. Note that the 2020-11-12 data have relatively large seeing compared to other observation nights, and therefore we do not generate a reference image for this night. We also note that, although the 2020-11-14 data are under good conditions throughout the night, the reference image was generated from images with relatively large seeing compared to those from the 2014-11-24 and 2017-09-20 data, which are the best data in our dataset.
• Figure 2: An example of the image for the M31 region, where the reference image obtained under the best seeing conditions during a given night is subtracted from a target image. Most stars in the spiral disk regions of M31 are well subtracted. Point sources remaining in the difference image are candidates of variable stars. Cyan star symbols denote the locations of the 12 microlensing candidates that passed all the selection criteria.
• Figure 3: The detection efficiency of the microlensing events as a function of the microlensing light-curve timescale $\hat{t}\equiv t_{\rm E}\sqrt{u_T^2(\rho)-u_0^2}$. Each panel shows the detection efficiency for each observation data, as indicated in the title of panel. Different colors show the efficiency for source stars with different intrinsic brightness, i.e., $(22,23,24,25,26)$ mag, respectively. For each combination of observation date, source brightness, and HSC patch, we simulated 10,000 realizations of microlensing light curves by randomly drawing model parameters from the priors listed in Table \ref{['tab:simulation-prior']} and adding Gaussian random noise whose amplitude is estimated from the corresponding data in each HSC patch and on each observation date. The black curve in each panel indicates the effective efficiency, obtained by integrating the brightness-dependent efficiencies weighted by the brightness distribution of source stars in M31, as defined in Eq. (\ref{['eq:effective-efficiency']}). The shaded regions in each panel indicate the regions of microlensing timescales that are difficult to detect in each dataset: for each observation data, the upper timescale cut corresponds to $T_{\rm eff}$ listed in Table \ref{['tab:obs-summary']}.
• Figure 4: The rightmost panel shows the profiles of the light-curve likelihoods for each of the 12 microlensing candidates, computed using Eq. (\ref{['eq:lc-like']}). Here we adopt the monochromatic PBH mass function defined in Eq. (\ref{['eq:mono-mass-func']}). The colored lines are the likelihoods for individual candidates $\mathcal{L}_{\rm LC}(\bm{d}_j^n|M_{\rm PBH})$, each normalized by their maximum values. The bold lines show the likelihoods for the 4 secure candidates (see Table \ref{['tab:candidate-catalog']}). The joint likelihood $\mathcal{L}_{\rm LC}(\{\bm{d}_j^n\}|M_{\rm PBH})$ for all the candidates or for the 4 secure candidates is shown by the thin gray or bold black line, respectively, and each profile is normalized such that its peak is twice as high as those of the individual likelihoods. The left and middle panels show the numerator and denominator of the light-curve likelihoods. Note that, since the candidates are obtained only on 2014-11-24 and 2017-09-20, the middle panel contains only two lines.
• Figure 5: The 95% C.L. exclusion region of the PBH abundance parameter $f_{\rm PBH}$ for each mass scale of $M_{\rm PBH}$ (green region). Here, $f_{\rm PBH}$ is the PBH mass fraction to the total dark matter in the Milky Way and M31 halo regions. We obtained the region assuming a monochromatic mass function under the null hypothesis, namely that all our candidates are false positives.