Search for: All records

ABSTRACT In this work, we extend our recently developed multifidelity emulation technique to the simulated Lyman-α forest flux power spectrum. Multifidelity emulation allows interpolation of simulation outputs between cosmological parameters using many cheap low-fidelity simulations and a few expensive high-fidelity simulations. Using a test suite of small-box (30 Mpc h−1) simulations, we show that multifidelity emulation is able to reproduce the Lyman-α forest flux power spectrum well, achieving an average accuracy when compared to a test suite of $$0.8\, {\rm {per\ cent}}$$. We further show that it has a substantially increased accuracy over single-fidelity emulators, constructed using either the high- or low-fidelity simulations only. In particular, it allows the extension of an existing simulation suite to smaller scales and higher redshifts. 

ABSTRACT Using a set of high resolution simulations, we quantify the effect of species-specific initial transfer functions on probes of the intergalactic medium (IGM) via the Lyman-α forest. We focus on redshifts 2–6, after H i reionization. We explore the effect of these initial conditions on measures of the thermal state of the low density IGM: the curvature, Doppler width cutoff, and Doppler width distribution. We also examine the matter and flux power spectrum, and potential consequences for constraints on warm dark matter models. We find that the curvature statistic is at most affected at the $$\approx 2{{\ \rm per\ cent}}$$ level at z = 6. The Doppler width cutoff parameters are affected by $$\approx 5{{\ \rm per\ cent}}$$ for the intercept, and $$\approx 8{{\ \rm per\ cent}}$$ for the fit slope, though this is subdominant to sample variation. The Doppler width distribution shows a $$\approx 30{{\ \rm per\ cent}}$$ effect at z = 3, however the distribution is not fully converged with simulation box size and resolution. The flux power spectrum is at most affected by $$\approx 5{{\ \rm per\ cent}}$$ at high redshift and small scales. We discuss numerical convergence with simulation parameters.