Galaxies in the simulated cosmic web: I. Filament identification and their properties

TL;DR

This study develops physically motivated cosmic filament catalogs by applying the DisPerSE algorithm to the dark matter density field in the EAGLE and IllustrisTNG100 simulations at $z = 0$ and $z = 2$, carefully calibrating smoothing, persistence thresholds, and masking to recover filaments that host well-resolved galaxies. The authors quantify filament widths, lengths, and central overdensities, and compare DM-based filaments to those traced by galaxies, revealing a heterogeneous filament population with generally thin, substructure-dominated cores and varying gas properties across simulations. They also analyze gas temperature and density structures around filaments, highlighting substantial differences in gas morphology between EAGLE and TNG100 due to different feedback implementations, and demonstrate that galaxy-based filaments, while capturing the thickest DM filaments, can misrepresent thinner filaments and introduce artefacts. The paper concludes that filaments are complex, multi-scale environments whose DM structure is robust across simulations, but baryonic properties require cross-simulation validation, setting the stage for Part II to study galaxy–filament co-evolution in detail.

Abstract

As the environment harbouring the majority of galaxies, filaments are thought to play a key role in the co-evolution of galaxies and the cosmic web. In this first part of a series to understand the link between galaxies and filaments through cosmological simulations, we address two major current obstacles on this path: the difficulty of meaningful filament identification, and their poorly constrained properties and internal structure. We use the public EAGLE and TNG100 simulations to build physically motivated filament catalogues with the DisPerSE algorithm, based on the dark matter (DM) field at redshift z = 0 and z = 2, explicitly accounting for the multi-scale nature of filaments and with careful validation of results. Filament widths, lengths, and densities vary by factors ~5-100 in both simulations, highlighting the heterogeneous nature of filaments as a cosmic environment. All filaments are relatively thin, with overdensity profiles of galaxies, DM, and gas dropping to the cosmic mean within <3 Mpc from their spines. Contrary to groups and clusters, filament cores are highly substructure dominated, by as much as ~80 per cent. Filament gas maps reveal rich temperature and density structures that limit the applicability of simple cylindrically symmetric models. EAGLE and TNG100 agree that z = 2 filament spines are traced by overdense cool gas in pressure equilibrium with a >10x hotter envelope. However, significant differences in detail between their predicted gas property maps imply that individual simulations cannot yet describe the baryon structure of filaments with certainty. Finally, we compare our fiducial filament network to one constructed from galaxies. The two differ in many aspects, but the distance of a galaxy to its nearest galaxy-based filament still serves as a statistical proxy for its true environment.

Figures (24)

• Figure 1: Close-up view of one filament in the EAGLE-Ref100 simulation. Shown is the dark matter (top) and gas (bottom) surface density in a $6 \times 6 \times 6$ Mpc cube at $z = 0$. In both cases, the colour map changes halfway between the cosmic average ($\overline{\Sigma}$) and the 99.99 percentile, to highlight the low- and high-overdensity regions. Galaxies with $M_\mathrm{star} > 10^9\,\mathrm{M}_\odot$ are depicted as open green circles. Dark matter is concentrated into clumps with a density $> 10 \times$ above the smooth filament background. The gas is more smoothly distributed and shows a continuous filament structure, but still has noticeable peaks up to $\approx\,$5 $\times$ the smooth background, due to the presence of galaxies and feedback-driven haloes. Neither corresponds to the smooth density field implicitly assumed by topological filament finders, which complicates the filament identification.
• Figure 2: Filament network at $z = 0$ of EAGLE (left) and TNG100 (right) compared to the underlying DM density field. Orange lines represent filaments in a 15 Mpc thick slice. Background images show the DM density projected in the same slice. Galaxies with $M_\mathrm{star} > 10^9\,\mathrm{M}_\odot$ are shown as indigo circles. The filaments identified by DisPerSE trace well the main filamentary features of the DM density field that harbour resolved galaxies.
• Figure 3: Fraction of haloes that mark the endpoint of filaments. Shown is the fraction of FoF haloes, in 0.1 dex bins of $M_\mathrm{200c}$, that are within 500 kpc of a cosmic web node as identified by DisPerSE (blue: all nodes, red: only those at the end of a joined filament; see text). Shaded bands represent the binomial 1$\sigma$ uncertainty on the fraction. There is no single mass threshold above which haloes become filament endpoints, but for the joined filaments the transition occurs broadly at the scale of massive galaxies ($M_\mathrm{200c} \approx 2\cdot10^{12}\,\mathrm{M}_\odot$).
• Figure 4: Sensitivity of the filament network to parameter choices. Each panel shows a full projection through the EAGLE-Ref25 simulation, with filaments in orange, galaxies ($M_\mathrm{star} > 10^9 \mathrm{M}_\odot$) in indigo, and the projected DM density as greyscale image in the background. Deviations from our fiducial parameters are highlighted in red. In the top row, the middle panel (green frame) uses our fiducial values, while the left- and right-hand panels adopt a $10\times$ lower and higher DisPerSE persistence threshold $T$, respectively. In the bottom row, the left-hand and middle panels instead show the effect of smoothing the DM field with a smaller or larger kernel $S$ than our fiducial choice. Finally, the bottom-right panel shows the filaments without smoothing the spines. Of the six networks shown here, only our fiducial choice traces the DM filaments adequately.
• Figure 5: Comparison of raw and joined filaments. Different coloured lines correspond to five individual ('raw') DisPerSE filaments that we join to one filament for our analysis. The background image shows the projected DM density in a 16 $\times$ 16 $\times$ 4 Mpc slab, oriented along the best-fit plane of the joined filament. Blue circles mark 1 and 3 $r_\mathrm{200c}$ of the two most massive halo in the field, with masses just under $10^{13}$ and $10^{12}\,\mathrm{M}_\odot$, respectively. Linking these two haloes, the joined filament is more physically meaningful than the raw ones.