HIPSTER: HIgh-k Power SpecTrum and bispectrum EstimatoR

Overview

HIPSTER is a code to quickly compute small-scale power spectra and bispectra for galaxy surveys and N-body simulations, based on the work of Philcox & Eisenstein (2019, accepted by MNRAS, arXiv) and Philcox (2020, in prep.). This computes the Legendre multipoles of the power spectrum, \(P_\ell(k)\) or bispectrum \(B_\ell(k_1,k_2)\) in configuration space, by computing weighted pair counts over the survey or simulation box truncated at some maximum radius. This does not include shot-noise or aliasing, fully accounts for window function effects and is optimized for small-scale power spectrum and bispectrum computation. Note that the bispectrum code algorithm has the same complexity as the power spectrum algorithm, thanks to spherical harmonic decompositions. By combining HIPSTER with conventional FFT-based methods, spectra can be measured efficiency across a vast range of wavenumbers.

The code can be run either in aperiodic or periodic mode, for galaxy surveys or N-body simulations respectively. The periodic mode contains various optimizations relating to the periodic geometry, as detailed in the second paper. For the bispectrum, only periodic mode is supported.

The source code is publicly available on Github, and contains many modules modified from the RascalC covariance matrix code.

To compute a periodic matter power spectrum up to \(\ell=L\) from a particles in a simulation box (data.dat), with pair counts truncated at radius R0 with \(k\)-space binning file binning.csv on 4 CPU-cores, we simply run:

./hipster_wrapper_periodic.sh --dat data.dat --l_max L -R0 R0 -k_bin binning.csv --nthreads 4

For a galaxy power spectrum with a non-trivial survey geometry, we also need random particle files (randoms.dat) and the usage is simply:

./hipster_wrapper.sh --dat data.dat --ran_DR randoms.dat --ran_RR randoms.dat -l_max L -R0 R0 -k_bin binning.csv --nthreads 4

To compute a bispectrum up to \(\ell=L\) from a particles in a simulation box (data.dat), with pair counts truncated at radius R0 with \(k\)-space binning file binning.csv on 4 CPU-cores, we simply run:

./hipster_wrapper_bispectrum.sh --dat data.dat --l_max L -R0 R0 -k_bin binning.csv --nthreads 4

This is described in detail in the accompanying pages.

For any queries regarding the code please contact Oliver Philcox.