Description
Primordial non-Gaussianity (PNG) represents a window into the nature of inflation, and large-scale structure (LSS) surveys can promisingly sharpen its constraints. To fully exploit this potential, cosmological simulations play a crucial role, allowing us to study the signatures and test the detectability of primordial features from LSS. I will present GENGARS, a framework to generate non-Gaussian initial conditions for N-body simulations from an arbitrary separable PNG shape. Building on the reduced-bispectrum-kernel formulation, we employ a Schwinger parameterization that improves efficiency by order of magnitudes, while preserving accuracy. I will outline the method and validation against 2LPT-PNG on the standard local, equilateral and orthogonal cases, emphasizing control of the induced large-scale primordial power spectrum correction, a critical requirement for accurate PNG simulations. I will then discuss how the same pipeline naturally accommodates non-standard templates, such as oscillatory PNG, and show their imprint on late-time statistics.
