Suppressing variance in 21 cm signal simulations during reionization
Current best limits on the 21 cm signal during reionization are provided at large scales (≳100 Mpc). To model these scales, enormous simulation volumes are required, which are computationally expensive. We find that the primary source of uncertainty at these large scales is sample variance, which de...
Gespeichert in:
Veröffentlicht in: | Astronomy and astrophysics (Berlin) 2023-01, Vol.669, p.A6 |
---|---|
Hauptverfasser: | , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Current best limits on the 21 cm signal during reionization are provided at large scales (≳100 Mpc). To model these scales, enormous simulation volumes are required, which are computationally expensive. We find that the primary source of uncertainty at these large scales is sample variance, which determines the minimum size of simulations required to analyse current and upcoming observations. In large-scale structure simulations, the method of ‘fixing’ the initial conditions (ICs) to exactly follow the initial power spectrum and ‘pairing’ two simulations with exactly out-of-phase ICs has been shown to significantly reduce sample variance. Here we apply this ‘fixing and pairing’ (F&P) approach to reionization simulations whose clustering signal originates from both density fluctuations and reionization bubbles. Using a semi-numerical code, we show that with the traditional method, simulation boxes of
L
≃ 500 (300) Mpc are required to model the large-scale clustering signal at
k
= 0.1 Mpc
−1
with a precision of 5 (10)%. Using F&P, the simulation boxes can be reduced by a factor of 2 to obtain the same precision level. We conclude that the computing costs can be reduced by at least a factor of 4 when using the F&P approach. |
---|---|
ISSN: | 0004-6361 1432-0746 |
DOI: | 10.1051/0004-6361/202244986 |