Speaker
Description
Primordial non-Gaussianity (PNG) provides a unique window into the nature of inflation, and large-scale structure (LSS) surveys offer a promising route to sharpen its constraints. While local PNG leaves a characteristic imprint on large-scale 2-point clustering statistics, many well-motivated non-local PNG models mainly induce higher-order correlations, making their signatures more difficult to model and detect. Cosmological simulations are therefore essential to connect primordial physics with late-time observables in the nonlinear Universe.
I will present $\texttt{GENGARS}$, a framework to generate non-Gaussian initial conditions for N-body simulations from arbitrary separable primordial bispectra. I will outline the main idea behind the method, emphasizing how the kernel choice controls the variance and accuracy of the generated correlation functions, and describe the prescription adopted in $\texttt{GENGARS}$ to make the calculation computationally feasible while preserving the target primordial bispectrum. I will then discuss how this pipeline can be applied beyond standard templates, including oscillatory feature models, to assess how primordial correlations propagate into late-time matter and halo statistics. Finally, I will briefly comment on how this framework can provide a basis for future field-level and simulation-based inference applications.