Fisher Forecasts for Primordial non-Gaussianity from Persistent Homology

Matteo Biagetti, Juan Calles, Lina Castiblanco, Alex Cole, Jorge Noreña

Abstract: We study the information content of summary statistics built from the multi-scale topology of large-scale structures on primordial non-Gaussianity of the local and equilateral type. We use halo catalogs generated from numerical N-body simulations of the Universe on large scales as a proxy for observed galaxies. Besides calculating the Fisher matrix for halos in real space, we also check more realistic scenarios in redshift space. Without needing to take a distant observer approximation, we place the observer on a corner of the box. We also add redshift errors mimicking spectroscopic and photometric samples. We perform several tests to assess the reliability of our Fisher matrix, including the Gaussianity of our summary statistics and convergence. We find that the marginalized 1-$\sigma$ uncertainties in redshift space are $\Delta f_{\rm NL}^{\rm loc} \sim 16$ and $\Delta f_{\rm NL}^{\rm equi} \sim 41 $ on a survey volume of $1$ $($Gpc$/h)^3$. These constraints are weakly affected by redshift errors. We close by speculating as to how this approach can be made robust against small-scale uncertainties by exploiting (non)locality.

Introduction

Cosmologists seek to decode the "fossil record" of the early universe by examining the large-scale structure (LSS) of galaxies. A primary target is Primordial non-Gaussianity (PNG), which describes subtle deviations from a Gaussian distribution in initial density fluctuations. PNG serves as a "smoking gun" for distinguishing between various inflationary models.

However, the non-linear evolution of gravity over billions of years introduces its own non-Gaussian signatures, often masking the primordial signal. Traditional statistical tools, such as the bispectrum, are powerful but can be sensitive to late-time noise and observational systematic errors. This research explores Persistent Homology (PH), a technique from Topological Data Analysis, as a robust alternative for extracting PNG signals by looking at the global connectivity of the cosmic web.

Methodology: Persistent Homology

Persistent Homology treats the distribution of galaxies as a "point cloud" and analyzes its connectivity across multiple scales. The core of PH is the filtration process. By "growing" spheres around each galaxy, the method tracks the formation and disappearance of topological features: clusters (0D), loops (1D), and voids (2D).

The filtration process: connecting points into simplicial complexes to identify topological features.

As these spheres expand, they form simplicial complexes. Features are characterized by their birth (when they first appear) and death (when they are filled in or merged). To ensure robustness against outliers, the authors utilize an $\alpha$DTM-filtration (Distance-to-Measure), which prevents isolated points from disproportionately affecting the topological summary.

Topological Summary Statistics

The results of PH are typically summarized in Persistence Diagrams, which plot the birth and death scales of every identified feature. To make these diagrams suitable for statistical analysis and Fisher forecasting, they are often converted into Persistence Images—continuous, pixelated representations of the topological data.

Persistence Diagrams and Images for 0, 1, and 2-cycles used as summary statistics for the Fisher analysis.

Fisher Forecasting and Realism

The study utilizes the Eos Dataset, consisting of high-resolution N-body simulations, to forecast the sensitivity of future galaxy surveys to two types of PNG: Local (common in multi-field inflation) and Equilateral (common in models with non-trivial interactions).

Crucially, the authors move beyond idealized "real space" simulations to include realistic observational effects, such as Redshift Space Distortions (accounting for galaxy peculiar velocities) and simulated redshift errors for both spectroscopic and photometric survey conditions.

Key Results and Insights

The Fisher matrix analysis yields promising marginalized 1-$\sigma$ uncertainties for a survey volume of $1 (Gpc/h)^3$:

  • Local PNG: $\Delta f_{\rm NL}^{\rm loc} \sim 16$
  • Equilateral PNG: $\Delta f_{\rm NL}^{\rm equi} \sim 41$

A standout finding is the topological protection of these signals. Large-scale features like cosmic voids are inherently resilient to small-scale perturbations. Even with significant photometric redshift errors, the constraints on PNG parameters only degraded by approximately 50%. This resilience suggests that PH can capture information from all higher-order correlations simultaneously, providing a cleaner window into the early universe than traditional methods.

Limitations and Future Work

While these results are encouraging, several avenues for improvement remain. The current analysis uses 1D histograms of birth/death scales; employing full Persistence Images or machine learning techniques (Simulation-Based Inference) could further enhance sensitivity. Additionally, future work must incorporate more complex galaxy-halo connections (galaxy bias) and actual survey geometries to fully validate PH for upcoming missions like Euclid or DESI.

Summary updated

Abstract

We study the information content of summary statistics built from the multi-scale topology of large-scale structures on primordial non-Gaussianity of the local and equilateral type. We use halo catalogs generated from numerical N-body simulations of the Universe on large scales as a proxy for observed galaxies. Besides calculating the Fisher matrix for halos in real space, we also check more realistic scenarios in redshift space. Without needing to take a distant observer approximation, we place the observer on a corner of the box. We also add redshift errors mimicking spectroscopic and photometric samples. We perform several tests to assess the reliability of our Fisher matrix, including the Gaussianity of our summary statistics and convergence. We find that the marginalized 1-$σ$ uncertainties in redshift space are $Δf_{\rm NL}^{\rm loc} \sim 16$ and $Δf_{\rm NL}^{\rm equi} \sim 41 $ on a survey volume of $1$ $($Gpc$/h)^3$. These constraints are weakly affected by redshift errors. We close by speculating as to how this approach can be made robust against small-scale uncertainties by exploiting (non)locality.
BibTeX
@article{biagetti2022fisher,
  author = {Matteo Biagetti and Juan Calles and Lina Castiblanco and Alex Cole and Jorge Noreña},
  title = {{Fisher Forecasts for Primordial non-Gaussianity from Persistent Homology}},
  journal = {Journal of Cosmology and Astroparticle Physics},
  year = {2022},
  doi = {10.1088/1475-7516/2022/10/002},
}
Figure