PTChallenge Simulations in Cosmology

• PTChallenge simulations are a suite of large-volume cosmological N-body realizations designed to rigorously test one-loop EFT bispectrum predictions.
• They employ high resolution (3072^3 particles) across 10 independent realizations to suppress sample variance on scales k ≲ 0.2 h/Mpc.
• COBRA factorization accelerates computations, enabling rapid, precise parameter inference with sub-permille accuracy in bispectrum analysis.

Searching arXiv for PTChallenge and related large-volume simulation suite papers. PTChallenge simulations denote a large-volume suite of cosmological $N$-body realizations used to validate and stress-test one-loop effective-field-theory predictions for the redshift-space galaxy bispectrum multipoles. In the usage established by "One-Loop Galaxy Bispectrum: Consistent Theory, Efficient Analysis with COBRA, and Implications for Cosmological Parameters," the suite consists of 10 independent realizations, each with side-length $3.84$ Gpc$/h$, total volume $566\,(\mathrm{Gpc}/h)^3$, and $3072^3$ particles evolved to $z=0.61$; halos are populated through an HOD tuned to BOSS CMASS-2 LRGs, so that number density and bias closely match existing redshift surveys (Bakx et al., 29 Jul 2025). Its defining role is methodological: the enormous volume suppresses sample variance below the percent level on scales $k\lesssim0.2\,h/\mathrm{Mpc}$, making the suite a stringent empirical test-bed for one-loop bispectrum theory and for fast inference pipelines such as COBRA.

1. Simulation suite and observational targeting

The PTChallenge suite is specified in the literature as 10 independent $N$-body realizations with side-length $3.84$ Gpc$/h$, totaling $3.84$0, and using $3.84$1 particles evolved to $3.84$2 (Bakx et al., 29 Jul 2025). The halo population is generated with an HOD tuned to BOSS CMASS-2 LRGs, and the resulting number density and bias are reported to closely match existing redshift surveys.

These design choices determine the suite’s scientific utility. The quoted volume suppresses sample variance below the percent level on scales $3.84$3, which is why the simulations are described as the most powerful test-bed to date for one-loop bispectrum theory. This suggests that PTChallenge occupies a niche distinct from smaller validation ensembles: it is not merely a mock catalog generator, but an environment in which residual theoretical error, stochastic terms, discreteness systematics, and scale-cut choices can be isolated with unusually high statistical leverage.

Property	Specification	Role
Number of realizations	10 independent $3.84$4-body realizations	Ensemble validation
Box size and volume	$3.84$5 Gpc$3.84$6 per side; total $3.84$7	Sample-variance suppression
Mass resolution and epoch	$3.84$8 particles evolved to $3.84$9	Large-scale structure calibration

2. Place within one-loop redshift-space bispectrum theory

The PTChallenge simulations are analyzed within Eulerian perturbation theory in redshift space, where the galaxy density field is expanded as

$/h$0

At tree level the bispectrum is

$/h$1

At one loop, the construction adds four loop diagrams,

$/h$2

each given as a 3-D integral over products of $/h$3–$/h$4 kernels and three factors of the linear power spectrum $/h$5. In the EFT treatment, these SPT loops are supplemented by higher-derivative counterterms $/h$6, pure stochastic terms $/h$7, and mixed stochastic–deterministic terms $/h$8 and $/h$9, all stated to be required by power counting in a $566\,(\mathrm{Gpc}/h)^3$0CDM universe with effective slope $566\,(\mathrm{Gpc}/h)^3$1 at $566\,(\mathrm{Gpc}/h)^3$2 (Bakx et al., 29 Jul 2025).

The full one-loop bispectrum is therefore written as

$566\,(\mathrm{Gpc}/h)^3$3

with 45 bias and EFT parameters in total at one loop in redshift space. Multipole decomposition in the plane-parallel approximation is defined by

$566\,(\mathrm{Gpc}/h)^3$4

where $566\,(\mathrm{Gpc}/h)^3$5 and $566\,(\mathrm{Gpc}/h)^3$6 is the azimuth around the line of sight.

PTChallenge’s relevance follows directly from this theoretical structure. Because one-loop redshift-space bispectra involve many loop contributions and an expanded stochastic sector, a simulation suite with sub-percent sample variance is unusually effective for determining whether the EFT bookkeeping is complete. The literature emphasizes that the first comprehensive treatment of stochastic EFT contributions significantly improves the match to data, and PTChallenge is the environment in which that claim is tested most stringently (Bakx et al., 29 Jul 2025).

3. COBRA factorization and computational acceleration

A central feature of PTChallenge analyses is the use of the COmpressed BAse-space Recovery Approach (COBRA). Rather than decomposing the linear power spectrum $566\,(\mathrm{Gpc}/h)^3$7 into $566\,(\mathrm{Gpc}/h)^3$8 FFTLog power laws, COBRA constructs an optimal numerical basis $566\,(\mathrm{Gpc}/h)^3$9 with $3072^3$0 by applying an SVD to a template bank of $3072^3$1 (Bakx et al., 29 Jul 2025). The power spectrum is approximated as

$3072^3$2

Because each one-loop diagram is polynomial in bias and growth rate and factorizes in $3072^3$3, one can precompute rank-3 tensors

$3072^3$4

where $3072^3$5 labels the basis monomial in bias and $3072^3$6. At run time the complete one-loop bispectrum is assembled by the tensor contraction

$3072^3$7

This reduces the cost of computing the one-loop EFT bispectrum to around a second per cosmology on a single core, with negligible loss of accuracy. The cosmology dependence is reported to be captured to sub-permille accuracy with just eight templates, and the practical $3072^3$8 FFTLog step introduces additional error below $3072^3$9 (Bakx et al., 29 Jul 2025). Direct comparison against brute-force FFTLog integration across 10 cosmologies yields below $z=0.61$0 error on $z=0.61$1 and below $z=0.61$2 on PTChallenge errors; IR resummation is also reproduced with sub-per-mille accuracy using eight COBRA modes.

In the PTChallenge context, this computational compression has methodological consequences beyond speed. It makes one-loop redshift-space bispectrum fitting operationally comparable to lower-order pipelines, enabling broad parameter scans and systematic scale-cut studies on a simulation suite whose statistical precision would otherwise expose even very small numerical artifacts.

4. Data vector, priors, covariance, and scale cuts

The PTChallenge fitting configuration uses a joint data vector consisting of $z=0.61$3 with $z=0.61$4 up to $z=0.61$5, $z=0.61$6 as a real-space proxy on $z=0.61$7, and $z=0.61$8 with $z=0.61$9 (Bakx et al., 29 Jul 2025). Scale cuts are deliberately asymmetric across multipoles: for the one-loop monopole $k\lesssim0.2\,h/\mathrm{Mpc}$0, $k\lesssim0.2\,h/\mathrm{Mpc}$1; for the one-loop quadrupole and hexadecapole, $k\lesssim0.2\,h/\mathrm{Mpc}$2 and $k\lesssim0.2\,h/\mathrm{Mpc}$3, $k\lesssim0.2\,h/\mathrm{Mpc}$4 in order to avoid discreteness bias. By contrast, the tree-level monopole analysis is limited to $k\lesssim0.2\,h/\mathrm{Mpc}$5.

The prior structure is broad. Shot-noise amplitudes are assigned $k\lesssim0.2\,h/\mathrm{Mpc}$6 Gaussian priors; EFT counterterms and higher-derivative terms have widths $k\lesssim0.2\,h/\mathrm{Mpc}$7; $k\lesssim0.2\,h/\mathrm{Mpc}$8; $k\lesssim0.2\,h/\mathrm{Mpc}$9; $N$0; stochastic-expansion coefficients $N$1; and all cosmological parameters are flat uninformative. The covariance is Gaussian analytic with volume $N$2, excludes cross-covariance of power-spectrum and bispectrum multipoles, and applies discreteness weights to $N$3 and $N$4.

Goodness of fit is reported as $N$5 up to $N$6, while pulls appear at $N$7 for the monopole or for $N$8. This establishes PTChallenge not only as a high-volume validation set, but as a calibrated threshold detector for the breakdown scale of the one-loop description.

5. Parameter recovery and cosmological information content

The principal scientific result extracted from PTChallenge is that the one-loop prediction provides an excellent match to the bispectrum data up to $N$9, as shown by the precise recovery of $3.84$0, $3.84$1, $3.84$2, and the amplitude of equilateral primordial non-Gaussianity $3.84$3 (Bakx et al., 29 Jul 2025). In the reported baseline comparison, combining the power spectrum with tree-level $3.84$4 at $3.84$5 gives

$3.84$6

Extending the monopole to one loop and increasing its scale reach to $3.84$7 improves the errors to

$3.84$8

Adding $3.84$9 and $/h$0 at $/h$1 further tightens the constraints to

$/h$2

and also reduces the error on $/h$3 by approximately $/h$4.

These staged improvements are consistent with the aggregate summary that, relative to the tree-level bispectrum monopole, COBRA-based one-loop bispectrum multipoles shrink the posteriors on $/h$5, $/h$6, and $/h$7 by $/h$8, $/h$9, and $3.84$00, respectively (Bakx et al., 29 Jul 2025). A plausible implication is that the main gain comes from extending theoretical control in scale and angular structure rather than from introducing qualitatively new late-time parameters.

6. Primordial non-Gaussianity, theory breakdown, and survey implications

PTChallenge is also used to assess sensitivity to equilateral primordial non-Gaussianity under fixed background cosmology. The tree-level monopole with $3.84$01 yields

$3.84$02

whereas the one-loop monopole with $3.84$03 yields

$3.84$04

corresponding to an approximately $3.84$05 tighter constraint (Bakx et al., 29 Jul 2025).

The same simulation suite also delineates the regime where the model ceases to be reliable. Beyond $3.84$06, theory bias grows above $3.84$07, signaling breakdown of the one-loop description. Two-loop corrections become comparable to PTChallenge errors at $3.84$08; discreteness weights for $3.84$09 and $3.84$10 require further study; and the covariance treatment neglects non-Gaussian contributions and power-spectrum–bispectrum cross terms, though these are described as small on non-squeezed triangles.

For ongoing surveys, the reported implication is pragmatic. Even after rescaling the covariance to DESI-Y5 volume, $3.84$11, theory biases at $3.84$12 remain below $3.84$13, so one-loop $3.84$14 to $3.84$15 is presented as robust; one-loop multipoles with $3.84$16 to $3.84$17 can be included with approximately $3.84$18–$3.84$19 improvement in $3.84$20 and $3.84$21 (Bakx et al., 29 Jul 2025). This supports the view that PTChallenge functions as a bridge between formal EFT consistency and operational survey analysis.

7. Terminological scope and a distinct usage in pntcc tooling

Within current large-scale-structure usage, “PTChallenge simulations” refers to the cosmological $3.84$22-body suite described above (Bakx et al., 29 Jul 2025). A separate usage appears in work on probabilistic ntcc tooling, where “PTChallenge” labels a pntcc model or workflow executed in Ntccrt and VerificationPntccM, rather than a cosmological simulation ensemble (Toro, 2018). In that setting, the system is a C++-based meta-interpreter with a front-end parser and AST generator, a simulation machine term

$3.84$23

and a verification machine term

$3.84$24

compiled onto a finite-domain Gecode back-end (Toro, 2018).

That usage is technically unrelated to the galaxy-bispectrum benchmark suite. The distinction matters because the acronym can suggest a common framework where none is established in the cited literature. In the cosmological context, PTChallenge denotes a giant-volume $3.84$25-body test-bed for one-loop EFT bispectrum analysis; in the pntcc context, it denotes a model execution and verification scenario in a discrete-time probabilistic process calculus. The shared label is therefore best understood as terminological overlap rather than scientific continuity.