Primordial Physics in the Nonlinear Universe: mapping cosmological collider models to weak-lensing observables

Abstract

Primordial non-Gaussianities (PNGs) are features in the initial density field that provide a window into the nonlinear dynamics of particles during the inflationary epoch. Among them, a distinctive set of signatures from "cosmological collider physics" originates through interactions of the inflaton with heavy particles active at high energies. The amplitude and form of these signatures depend on the strength and nature of the interactions. The corresponding features in large-scale structure have been studied predominantly through the use of perturbation theory, restricted to the linear regime of the density field. In this work, we implement a method for running cosmological simulations with arbitrary bispectra signals in their initial density field, and produce a simulation suite of over thirty PNG-generating templates, resolving the corresponding collider signatures in the strongly nonlinear regime of the density field. We detail the signals in a variety of late-time measurements -- the matter power spectra, matter bispectra, the halo abundance, and halo bias. We then forecast the potential constraints on the signal amplitudes using weak lensing measurements from the Year-10 dataset of the Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST). The second and third moments of the lensing convergence field produce constraints that are competitive and complementary to those from the Cosmic Microwave Background. The data products are publicly released as part of the Ulagam simulation suite. Our initial conditions generator is also publicly available at https://github.com/DhayaaAnbajagane/Aarambam.

Figures (12)

• Figure 1: Feynman diagrams for the classes of interactions considered in this work. The black (colored) lines represent inflaton fluctuations (massive particles). Panels (i), (ii), and (iii) correspond to single-, double-, and triple-exchange of massive particles, respectively. Case (i) includes the SI, SII, HSC, and EC models; (ii) represents the MSC scenario; and (iii) corresponds to the QSF scenario.
• Figure 2: The different templates considered in this work, shown in three specific limits, alongside the approximated versions using our basis functions (black lines). In all cases, the templates are adequately reproduced by our approximations. There are some oscillatory residuals in the approximated template, given our basis functions are Legendre polynomials in linear $k$. These can be suppressed by adding more modes and allowing for fine cancellations between basis functions, but such fine-tuning degrades the numerical stability of our basis set (see Section \ref{['sec:sims:ICs']} and Appendix \ref{['appx:ICs:TechDetails']}). For this reason, we retain some oscillatory deviations in the approximated template but note that in all cases the residuals are smaller than the overall variation of the bispectrum (in a given limit) across $k$.
• Figure 3: An example of the mode decomposition for a few templates shown in Figure \ref{['fig:Template']}. The rows show the shape function, $S$, in the squeezed, equilateral, and folded limits. The columns show different models, following the same nomenclature as Figure \ref{['fig:Template']}. The analytic model (red) is shown alongside the approximation (dashed black line). The individual modes are shown in the thin, translucent lines and are colored by the mode index. Larger indices generally correspond to higher-order modes. The local template is reproduced exactly using just a single mode (hidden behind the red line). The rest of the templates require some cancellations between different basis functions in order to reproduce the target template. We use a symmetric log axis to highlight the presence of a few large-amplitude modes that are canceled somewhat finely to construct the target template.
• Figure 4: Constraints on the amplitude $f_{\rm NL}$ for different models/templates and parameter spaces. We compare our forecasted constraints for LSST Y10 lensing against the CMB-derived constraints from Sohn:2024:Colliders. The former are within 30% of the latter for most models studied here, highlighting the primordial information in the lensing field. For the Equilateral Collider (EC), Sohn:2024:Colliders present the constraints at the maximum likelihood point for the phase $\phi$, whereas we show our forecasts for two choices of $\phi$. This template is also normalized differently (relative to the other templates), which results in higher values of $f_{\rm NL}$ (see Section \ref{['sec:sims:Models']}). For the Multi-Speed Collider (MSC), we can only show CMB constraints at a single point that overlaps with our exact setup, and this is indicated as a gray star. The lensing constraints do not marginalize over additional parameters, where such marginalization is expected to degrade constraints by 20-30%. See Section \ref{['sec:results:lensing']} for details.
• Figure 5: The correlation in the Fisher posterior for pairs of models. Values close to 1 and 0 indicate strong degeneracy and orthogonality, respectively, between the late-time predictions of two models. The upper (lower) triangle shows results for an LSST Y10 survey (cosmic variance, or CV, limited survey). The correlations in LSST Y10 are similar to those from a CV-limited survey. The templates from scalar exchange (QSF, SI, SII) are highly correlated with each other, while those from particle spins and sound speed variations are more orthogonal to each other and to these scalar templates. The diagonal is shaded black for visualization. The black boxes denote different classes of templates.