The causal effect of cosmic filaments on dark matter halos

The way in which the large-scale cosmic environment affects galactic properties is not yet understood. Dark matter halos, which embed galaxies, initially evolve following linear theory. Their subsequent evolution is driven by non-linear structure formation in the halo region and in its outer environment. In this work, we present the first study where we explicitly control the linear part of the evolution of the halo, thus revealing the role of non-linear effects on halo formation. We focus specifically on the effect of proximity to a large cosmological filament. We employ the splicing method to keep fixed the initial density, velocity, and potential fields where a halo will form while changing its outer environment, from an isolated state to one where the halo is near a large filament. In the regime of Milky Way-mass halos, we find that mass and virial radius of such halos are not affected by even drastic changes of environment, whereas halo spin and shape orientation with respect to a massive filament is largely impacted, with fluctuations of up to 80 % around the mean value. Our results suggest that halo orientation and shape cannot be predicted accurately from a local analysis in the initial conditions alone. This has direct consequences on the modeling of intrinsic alignment for cosmic shear surveys, like Euclid. Our results highlight that non-linear couplings to the large-scale environment may have an amplitude comparable to linear effects, and should thus be treated explicitly in analytical models of dark matter halo formation.

💡 Research Summary

This paper presents the first controlled numerical experiment that isolates the causal impact of large‑scale cosmic filaments on the properties of individual dark‑matter halos. The authors extend the “splicing” technique, originally developed to replace the density field of a halo’s Lagrangian patch, by also splicing the gravitational potential. By doing so they fix the initial density, velocity, and tidal fields inside the patch, thereby locking in all linear‑theory predictions (e.g., extended Press‑Schechter, peak‑patch, and tidal‑torque theory). Any subsequent differences in halo evolution must therefore arise from non‑linear coupling to the surrounding large‑scale environment.

The methodology proceeds as follows. Two independent Gaussian random realizations of the initial conditions are generated: one containing a target Milky‑Way‑mass halo (M≈1.7–3.5×10¹² M⊙) and another containing a massive filament. The potential field in Fourier space is constructed from a white‑noise field using the ΛCDM power spectrum, and a mask matrix M defines a spherical region Γ (radius 5 Mpc h⁻¹) that encloses the halo’s Lagrangian patch. Inside Γ the potential of the halo realization is imposed, while outside the potential of the filament realization is retained. Continuity across the boundary is enforced by solving a linear system for a correction potential φ_α using the MINRES iterative solver, with convergence criteria set to residuals <10⁻⁷. This guarantees that density, velocity, and tidal tensors are identical inside Γ for all spliced runs.

The authors then run a suite of cosmological N‑body simulations with RAMSES. First, a 50 Mpc h⁻¹ box with 512³ particles is evolved from z≈1100 (Zel’dovich ICs) to z=0, providing a catalog of halos and filaments identified with ROCKSTAR. From this catalog they select five to nine halos in the Milky‑Way mass range and trace their Lagrangian patches back to the initial redshift. For each halo they create nine spliced initial conditions in which the distance between the halo patch and the filament patch is systematically increased from essentially zero up to ~10 Mpc h⁻¹. The spliced ICs are then re‑simulated (zoom‑in where needed) to z=0, and halo properties are measured with the same pipeline (mass, virial radius, spin vector, shape tensor, and alignment angle relative to the filament spine).

The results are strikingly clear. Halo mass and virial radius show negligible dependence on the filament distance; the variations are well within statistical noise, confirming that these bulk properties are set by the local linear overdensity and are insensitive to the non‑linear large‑scale environment. In contrast, the spin magnitude and direction, as well as the major‑axis orientation, exhibit strong, systematic trends. When the halo is placed close to a filament, its spin tends to align (or anti‑align, depending on mass) with the filament spine, reproducing the well‑known “spin‑flip” mass around 10¹¹–10¹² M⊙. As the filament is moved farther away, the alignment angle decorrelates, and the spin amplitude fluctuates by up to 80 % relative to the mean of the ensemble. Shape alignment shows a similar, though slightly weaker, dependence on filament proximity. These variations far exceed the scatter expected from linear tidal‑torque theory, indicating that non‑linear torques generated by the surrounding filamentary mass distribution dominate the angular‑momentum acquisition and shape evolution of individual halos.

The authors discuss two major implications. First, the insensitivity of mass and radius validates the use of linear‑theory based mass functions for cosmological inference, even in the presence of complex large‑scale structures. Second, the strong environmental dependence of spin and shape implies that analytic models of intrinsic alignment (IA) used in weak‑lensing surveys must incorporate non‑linear coupling to filaments; otherwise, IA contamination could be mis‑estimated, jeopardizing the precision goals of missions such as Euclid, LSST, and Roman. The splicing framework itself provides a powerful “causal experiment” tool: by holding the halo’s initial patch fixed while varying only the external field, one can systematically probe the impact of any large‑scale feature (clusters, voids, sheets) across a wide mass range.

Future work suggested includes expanding the sample size to achieve statistical robustness, extending the method to hydrodynamical simulations to assess baryonic feedback, and directly linking the measured alignment statistics to observable galaxy shapes. Moreover, integrating the quantified non‑linear coupling terms into semi‑analytic models or effective field theory descriptions could improve predictions of IA and halo spin distributions, thereby tightening cosmological constraints from upcoming large‑scale structure surveys.