Constraining Inflationary Particle Production with CMB Polarization

Abstract

Following Philcox et al. (2025), we investigate a scenario with a massive partner to the inflaton ($O(100)$ times the inflationary Hubble scale), in which particles are produced during a narrow time period, leaving characteristic hot- or cold-spots in the cosmic microwave background (CMB). Using tools developed for thermal Sunyaev-Zel'dovich cluster-finding, we search component-separated Planck PR4 $E$-mode maps for these hotspots, and compare to analogous results in $T$. Our analysis pipeline is validated on simulated observations and gives unbiased constraints for sufficiently large and bright hotspots. At Planck sensitivities, the temperature data are more sensitive to small hotspots, but for sufficiently large hotspots the polarization data are more sensitive. We improve upon earlier work by building a full Poissonian likelihood for the hotspot abundance. We find no strong evidence for primordial hotspots and thereby place novel bounds on the couplings between the inflaton and massive scalars during inflation, probing physics at energies many orders of magnitude above any feasible terrestrial collider. The bounds derived from our new likelihood improve upon those of Philcox et al. (2025) by more than an order of magnitude for sufficiently light particles ($M_0\lesssim100H_I$). We also forecast the inferred bounds on inflationary physics for a search using Atacama Cosmology Telescope (ACT) data, and from an optimistic cosmic-variance-limited experiment (CV), for which $E$-mode data provide stronger constraints than $T$ on nearly all scales. ACT should improve on the Planck constraints by $\gtrsim10\%$, nearing the CV limit allowed by its sky coverage. Finally, we compare the constraining power of localized searches to that of a power spectrum analysis, and demonstrate that for sufficiently few produced particles the localized search performed herein is dominant.

Figures (15)

• Figure 1: The mean number of particle-production hotspots $N_{{\mathrm{HS}}}(g,M_0)$ as a function of the coupling $g$ and minimal mass $M_0$, where we take characteristic values for $\eta_*$ and $\Delta\eta$ equal to 100 Mpc to give a sense of the nontrivial dependence of particle production on the physical parameters in the model. We highlight the $N_{\rm HS} = 1$ contour, which we will find is the parameter space in which our single-hotspot sensitivities derived in Sec. \ref{['sec:forecasts']} and the population-level bounds in Sec. \ref{['sec:bounds']} should be equivalent
• Figure 2: Sketch of the relevant geometry, where $\vb{x}_0$ represents our position today, $\vb{x}_{\mathrm{HS}}$ is the hotspot position, $\chi_{\mathrm{rec}}=\eta_0-\eta_{\mathrm{rec}}$ is the comoving distance to the surface of last scattering, $\chi_{\mathrm{HS}}=\eta_0-\eta_{\mathrm{HS}}$ is the comoving distance to the hotspot, and $\eta_*$ is the comoving horizon size at particle production. Note that to causally impact the CMB the comoving distance between the hotspot and the surface of last scattering must be less than the horizon size (indicated by the gray band) at particle production, apart from small ISW contributions in $T$ for hotspots with $\eta_{\rm HS} - \eta_{\rm rec} > \eta_*$.
• Figure 3: Temperature and polarization hotspot templates for $g=1$, demonstrating the variability of the polarization hotspot profiles with respect to both $\eta_*$ (encoding the hotspot formation time during inflation) and $\eta_{{\mathrm{HS}}}$ (parametrizing the hotspot distance). (A): We fix $\eta_{{\mathrm{HS}}}=\eta_{\text{rec}}$ and show profiles for $\eta_*=30,100,300\,\,\text{Mpc}$ in blue, orange, and green, respectively. (B): We fix $\eta_*=100~\text{Mpc}$ and show the variability of the profile with respect to $\eta_{{\mathrm{HS}}}$. Each hotspot has a significant "cold ring" at $\theta \simeq 0.5$--$0.6^\circ$ from the center. (C): We compare the profiles for temperature (dashed) and polarization (solid). We multiply the polarization profiles by a factor of 10 for visual clarity. Notice that the temperature amplitude is characteristically around an order of magnitude larger than that in polarization.
• Figure 4: Central amplitude of the hotspots. (Left): We show a comparison between the central amplitudes of hotspot profiles in temperature (blue) and in $E$-mode polarization (orange), where we have fixed $\eta_{{\mathrm{HS}}}=\eta_{\text{rec}}$. Negative values are plotted with dashed curves. (Right): We show the central amplitude for polarization computed for various $\eta_{{\mathrm{HS}}}$ values, where we have chosen $100$ Mpc as a characteristic scale for the variation of $\eta_{\mathrm{HS}}$ away from $\eta_{\mathrm{rec}}$, while preserving causality. Note that the central amplitude of the anisotropy is strongest for hotspots located close to the surface of last scattering ($\eta_{{\mathrm{HS}}}=\eta_{\rm rec}$).
• Figure 5: Exemplar simulated hotspots with $g=30$, $\eta_*=129\ \mathrm{Mpc}$, and $\eta_{\mathrm{HS}}=\eta_{\mathrm{rec}}$. Left panels show signal-only maps; right panels show the same signals added to npipe simulations. The $T$ and $E$ hotspots are centered on the same point, and are shown in $5^\circ\times5^\circ$ maps with 1.5 arcminute pixels in the Gnomonic projection.