🔭 astrophysics

Large-scale halo velocity correlations and the impact of finite simulation volumes

This paper demonstrates that finite simulation volumes suppress large-scale halo velocity correlations due to missing long-wavelength modes, a bias that can be corrected by marginalizing over the minimum wavenumber in the analysis to recover accurate growth rate constraints, while also revealing that this suppression effect is more pronounced for more massive halos.

Original authors: Yao-Tsung Chuang, Teppei Okumura, Takahiro Nishimichi

Published 2026-02-05

📖 5 min read🧠 Deep dive

CC0 1.0

Original authors: Yao-Tsung Chuang, Teppei Okumura, Takahiro Nishimichi

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, invisible ocean. In this ocean, matter (like dark matter and galaxies) isn't spread out perfectly evenly; it's clumpy, with some areas denser than others. Just like water flows from high places to low places, this matter "flows" toward the densest clumps due to gravity. This flow is called peculiar velocity.

Scientists want to measure how fast and in what direction this cosmic ocean is flowing to understand how the universe grows and changes. To do this, they build massive computer simulations—virtual universes inside a box—to see how these flows behave.

However, the authors of this paper discovered a tricky problem with these computer boxes: they are too small.

Here is a simple breakdown of their findings using everyday analogies:

1. The "Missing Waves" Problem

Imagine you are trying to study the waves in the ocean, but you are looking at them through a small, square window in a wall.

The Reality: The real ocean has waves of all sizes, from tiny ripples to massive swells that stretch for miles.
The Simulation: Your computer box is like that small window. It can only "see" waves that fit inside the box. Any wave that is longer than the box gets cut off. It's as if the ocean suddenly stops having big waves just because your window is too small to see them.

The paper shows that because these simulations are missing the "big waves" (long-wavelength modes), the speed of the cosmic flow they measure comes out too slow on large scales. It's like trying to measure the speed of a river by only looking at a tiny puddle; you miss the powerful current coming from upstream.

2. The "Cut-Up Cake" Surprise

The researchers wanted to know if the problem was just the size of the box or something else. They did a clever experiment:

They took a huge virtual box (2 billion light-years across) that had all the big waves.
They then chopped this huge box into smaller pieces, making them the same size as the small boxes (1 billion light-years across).

The Surprise: Even though these new small pieces were cut from a "perfect" big box, the measurements inside them were still too slow.

The Analogy: Imagine baking a giant cake with perfect chocolate swirls all the way through. If you cut a small slice out of the middle, that slice still has the swirls. But if you baked a new small cake from scratch, it might not have the swirls at all.
The Lesson: The problem isn't just the size of the area you are looking at; it's about whether the "big waves" were already imprinted in the simulation when it started. If the simulation box was too small to begin with, the big waves were never there to begin with, and cutting the box up later doesn't fix it.

3. Fixing the Measurement

The scientists found a way to correct this error.

The Old Way: They assumed the simulation was perfect and tried to calculate the flow speed. This gave them the wrong answer (biased results).
The New Way: They introduced a "correction knob" (called $k_{min}$ ) in their math. This knob tells the computer, "Hey, we know we are missing the biggest waves, so let's adjust our calculation to account for that missing piece."
The Result: When they turned this knob, the calculated speed of the cosmic flow matched the true speed perfectly. However, if they ignored the missing waves and didn't use this correction, their results were significantly off.

4. Bigger Clumps, Bigger Errors

They also noticed that the size of the "clumps" (galaxy clusters vs. smaller groups) mattered.

The Analogy: Imagine trying to feel the wind. If you are a tiny leaf, you might not notice a gentle breeze. But if you are a giant tree, you feel the wind pushing against your whole trunk.
The Finding: The error caused by the missing big waves was stronger for massive galaxy clusters than for smaller groups. The bigger the object, the more it feels the "missing" big waves, and the more the simulation underestimates its speed.

The Bottom Line

The paper concludes that when scientists use computer simulations to predict how the universe is moving, they must be very careful about the size of their "virtual box." If the box is too small, it hides the biggest cosmic waves, leading to incorrect conclusions about how fast the universe is growing.

To get the right answer, they can't just look at the data; they have to mathematically "fill in the blanks" for the waves that the computer box was too small to hold. This is crucial for future studies that rely on these simulations to understand the universe's expansion and gravity.

Technical Summary: Large-scale halo velocity correlations and the impact of finite simulation volumes

Problem Statement
Velocity correlation functions measured from the peculiar velocity fields of dark matter and halos in cosmological N-body simulations exhibit a systematic suppression in amplitude on large scales compared to theoretical linear predictions. This discrepancy is particularly problematic for precision cosmology, where velocity statistics are used to constrain the growth rate parameter $f(z)\sigma_8(z)$ and test gravity. The suppression is attributed to the finite volume of simulation boxes, which imposes a hard cutoff on Fourier modes, eliminating long-wavelength perturbations that are naturally present in the real Universe. While this effect is known to alter large-scale correlations, its specific impact on velocity statistics, its dependence on halo mass, and its consequences for parameter inference in simulation-based modeling (such as emulators) have not been fully quantified.

Methodology
The authors investigate this issue using suites of N-body simulations from the DARK QUEST project, specifically comparing two box sizes:

L1G: $1 \, h^{-1}$ Gpc side length.
L2G: $2 \, h^{-1}$ Gpc side length.

To disentangle the effects of box size from the imprint of long-wavelength modes, the authors perform a controlled sub-volume analysis:

Subdivision: The larger L2G boxes are subdivided into smaller volumes with side lengths of $1 \, h^{-1}$ Gpc ( $L\times0.5_{2G}$ ) and $0.3 \, h^{-1}$ Gpc ( $L\times0.15_{2G}$ ).
Comparison: They compare velocity correlations measured in these sub-volumes against those from independent L1G simulations. Crucially, the sub-volumes retain the long-wavelength modes of the parent L2G box, whereas the independent L1G simulations do not.
Modeling: The theoretical velocity autocorrelation function is modeled using a Hankel transform of the matter power spectrum $P_m(k)$ . The authors introduce a free parameter, $k_{min}$ , representing the lower limit of the integral to mimic the missing long-wavelength modes.
Analysis: They fit the model to simulation data to constrain $f\sigma_8$ , testing two scenarios: (a) marginalizing over $k_{min}$ as a free parameter, and (b) fixing $k_{min} \approx 0$ (ignoring the finite volume effect). They also analyze the dependence of these effects on halo mass by splitting halos into group and cluster mass ranges.

Key Results

Origin of Suppression: The suppression of velocity correlation amplitudes is driven by the absence of long-wavelength modes in the simulation volume, not merely by the physical size of the analyzed region. Sub-volumes carved from the larger L2G box ( $L\times0.5_{2G}$ ) exhibit the same level of suppression as the original L2G box, despite having the same volume as the independent L1G box. This confirms that the "memory" of the larger box's modes persists in the sub-volumes.
Mass Dependence: The suppression is mass-dependent. More massive halos (cluster-sized, $M_h > 10^{14} h^{-1} M_\odot$ ) exhibit a systematically stronger suppression of velocity correlation amplitude compared to less massive halos (group-sized, $M_h \sim 10^{13} h^{-1} M_\odot$ ).
Parameter Inference ( $f\sigma_8$ ):
- Marginalizing $k_{min}$ : When $k_{min}$ is treated as a free parameter and marginalized over, the recovered $f\sigma_8$ is consistent with the true input value of the simulations ( $f\sigma_8 = 0.4852$ ), regardless of the box size.
- Ignoring the Effect: If $k_{min}$ is fixed to zero (ignoring the finite volume effect), the constraint on $f\sigma_8$ becomes significantly biased, particularly when the full covariance matrix is not utilized (i.e., when correlations between separation bins are ignored). Even with the full covariance matrix, a shift in the best-fit value persists, though it remains within $1\sigma$ .
Covariance Importance: The bias in $f\sigma_8$ is exacerbated when the off-diagonal elements of the covariance matrix are neglected.

Significance and Claims
The paper argues that finite-volume effects in N-body simulations constitute a critical, often underappreciated systematic for simulation-based modeling of velocity statistics.

Modeling Implications: For simulation-based approaches like emulators, simply using the raw output of finite boxes without correcting for missing long-wavelength modes leads to biased cosmological constraints. The authors propose that introducing an effective large-scale cutoff ( $k_{min}$ ) as a free parameter is a necessary step to recover unbiased constraints on $f\sigma_8$ .
Distinction from Observations: The authors highlight a fundamental difference between simulation artifacts and observational window functions. While observational window functions can be explicitly modeled and deconvolved, the hard cutoff in simulations is imprinted in the initial conditions and persists through nonlinear evolution, making it a more rigid limitation.
Physical Interpretation: The mass-dependent suppression implies that finite-volume effects and velocity bias cannot be cleanly separated when analyzing biased tracers. This complicates tests of the Einstein equivalence principle using simulation-calibrated models, as unaccounted finite-volume effects could mimic or obscure genuine physical signals.

The work concludes that future analyses of peculiar velocity and kinetic Sunyaev–Zel'dovich (kSZ) data must consistently account for missing long-wavelength modes when using simulations as theoretical templates to ensure robust cosmological interpretations.