PNG-pmwd Suite
The PNG-pmwd Suite is a collection of \(22{,}410\) halo catalogs developed to support data-driven analyses of large-scale structure formation and the impact of primordial non-Gaussianity (PNG).
The suite includes:
6 Latin hypercube sets evaluated at \(z=0.5030475235459835\).
All datasets incorporate non-Gaussian initial conditions.
Two additional datasets, one corresponding to a fiducial flat \(\Lambda{\rm CDM}\) cosmology and 1P design that varies one parameter at a time.
\(22{,}421\) CPU hours used.
The dark matter snapshots were generated using the Differentiable Cosmological Simulation with the Adjoint Method, a fast N-body solver fully differentiable implemented in JAX and freely available at pmwd code.
Initial conditions are set using second-order Lagrangian displacement (2LPT) at redshift \(z=9\), with a linear power spectrum based on the the transfer functions of Eisenstein & Hu. The linear matter power spectrum for the fiducial cosmology can be found here.
Each simulation evolves \(512^3\) dark matter particles in a periodic box of size \((1 Gpc/h)^3\) over 20 time steps, reaching the final redshift of \(z=0.5030475235459835\).
To implement primordial non-Gaussian initial conditions, we developed a modified version of the code based on the prescription by Scoccimarro, Hui, Manera, and Chan.
> The updated codebase with local and equilateral templates can be found at: pmwd PNG version.
Halos are identified using a Friends-of-Friends (FoF) algorithm (Davis, Efstathiou, Frenk, and White) with a linking length of 0.2. We use the public nbodykit library and store halos containing at least 20 particles. Dark matter particle snapshots are not stored.
Citation
If you use data from the PNG-pmwd Suite, please cite:
Primordial non-Gaussianity – Fast simulations and persistent summary statistics (2025)
Note
Powered@NLHPC: This research was partially supported by the supercomputing infrastructure of the NLHPC (CCSS210001).
PNG-pmwd Suite