Galaxy clustering bias was strongest in the outskirts of cosmic-web nodes in a simulation analysis, where the mean bias reached 4.1. That was up to approximately four times the Tinker et al. (2010) fit used as a reference.
The work used DisPerSE to identify the cosmic web in the TNG300 IllustrisTNG volume and an object-by-object estimator for large-scale galaxy bias. Its catalog contained 203,844 simulated central galaxies selected under the study's stated stellar- and halo-mass thresholds.
Here, bias is the study's measure of large-scale clustering strength. An effective bias for each population was obtained by taking the mean of its individual bias estimates, so the reported values compare averages across groups.
Five settings, one simulated catalog
The final environmental catalog used five settings: voids and walls, nodes, node outskirts, filaments and filament outskirts. They contained 161,801, 701, 4,859, 17,883 and 18,600 galaxies, respectively.
The study compared broad web environments with filament-specific features, including length, average galaxy density and position along the filament spine. These comparisons used the same individual-bias framework.
Among lower-mass host halos, red galaxies had about 1.5 times the bias of blue galaxies in filaments and about 1.6 times the bias in filament outskirts. After the low-mass red-galaxy sample was filtered, the original-to-filtered bias ratio was still up to about 1.5 in filaments and 1.4 in filament outskirts.
These figures are comparisons between populations in the simulation. They do not show that a particular environment causes bias to change.
Filament length tracked a broad decline
The filament-focused analysis used 8,513 unique filaments associated with at least one galaxy. Mean filament bias decreased with filament length, from 1.4 for short filaments to around -0.5 for long filaments.
That length relation persisted after stratification by halo mass and galaxy colour. At fixed average filament galaxy density, short filaments typically had about 0.8 to 1.5 higher mean bias than long filaments.
An additional analysis selected 1,346 filaments and tracked normalized bias along their spines. Short filaments had approximately uniform longitudinal profiles. Long filaments showed normalized bias increasing from about 0.85 near the saddle point to about 1.06 near the node.
What the simulation cannot settle
Taken together, the results describe associations between the study's bias measure and cosmic-web environment, filament length and position along a filament. They do not establish that filament length or environment causes the differences. The evidence comes from simulated central galaxies in the TNG300 IllustrisTNG volume, not from a direct observational sample.
One measurement limit is built into the estimator. It is intrinsically isotropic, so it cannot separate directional components associated with cosmic-web geometry. In practical terms, the analysis cannot distinguish bias along a filament from bias across it.
The front matter identifies the manuscript as arXiv:2608.25060v1, dated 25 August 2026. The acknowledgements list institutional, scholarship, hospitality, programme and outreach support from Universidad Técnica Federico Santa María, Pontificia Universidad Católica de Valparaíso, ICTP's Regular Associates Programme and the AstroGainz science outreach initiative.
Paper data and sources
Original title: Tracing Galaxy Bias Through the Cosmic Web: The Role of Filaments
Authors: Constanza A. Soto-Suárez, Antonio D. Montero-Dorta, Daniela Galárraga-Espinosa et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text