Spectral Hierarchy of the Cosmic Web

arXiv:2603.15834 · astro-ph.CO, cs.CV · Submitted 2026-03-16 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.

Vera: Next we'll be talking about the paper "Spectral Hierarchy of the Cosmic Web".

Jocelyn: The paper was written by Francisco-Shu Kitaura and Francesco Sinigaglia from Instituto de Astrofísica de Canarias and Department of Astrophysics, University of La Laguna and Institute for Fundamental Physics of the Universe and SISSA - International School for Advanced Studies and INAF - Osservatorio Astronomico di Trieste and INFN – National Institute for Nuclear Physics.

Vera: Stay tuned as we take you through the paper and discuss its implications.

Summary and Methodology: Vera: We're moving into the summary section of "Spectral Hierarchy of the Cosmic Web," which explains exactly how this classification works. It’s not just a guess; it’s a very systematic approach that uses simple scale-weighting kernels.

Jocelyn: I like that phrase, "scale-weighting kernels," because it sounds like they are controlling the focus, making sure we can dial in exactly what we want to see in the structure. It’s not just random noise; it’s a controlled examination of the density field.

Subrahmanyan: The core mechanism is essentially using Fourier space to filter different levels of information, and then they run a standard eigenvalue-based web classification on those results. This allows us to map the physical environment into four defined categories: void, sheet, filament, and knot.

Vera: And I think it’s important to understand that this isn' not just classifying the density; it’s classifying the *structure* of the density contrast at different scales. We are seeing how gravity is pulling and pushing matter in a highly organized way.

Jocelyn: The paper tells us that by measuring a specific "web contrast" field and cross-correlating it with halos, we can quantify the information content itself. That’s really useful for determining which parts of our data are driven by which structural features.

Subrahmanyan: This is where the theoretical underpinning really matters—the they are using second derivatives of the filtered field to define these classification tensors. It connects the morphology directly to fundamental physics, not just visual aesthetics.

Vera: And when we look at the results, it seems like even down to a certain Nyquist limit relevant for our fast mock generation, this hierarchy retains significant information. That’s huge for us because it means we don’t have to use expensive simulations for everything.

Jocelyn: It suggests that using this approach could provide a much more efficient way of creating synthetic data that reflects the actual clustering we observe in the sky.

Subrahmanyan: The way they are bridging the gap between local geometry and large-scale physics is what makes this entire concept so powerful for future work.

Improvements and Practical Utility: Vera: Now, looking at "Spectral Hierarchy of the Cosmic Web," Kitaura and Sinigaglia really show how their framework improves traditional methods. They aren't just a replacement; they are an extension that is much more versatile.

Jocelyn: It’s great that we can see these additional higher-derivative levels, which are naturally occurring in things like bias modeling and effective field theory. That provides a direct bridge between the observed cosmic web and the complex physics of tracers.

Subrahmanyan: The way they have aligned these levels with operator families—like grad two delta, grad four delta and so on—is key to understanding how this structure relates to both long-range and short-range nonlocality.

Vera: And I think the practical implications for us are enormous, especially when we consider things like subgrid modeling. We can now condition our mock galaxy production using this hierarchy instead of just relying on simple density bins.

Jocelyn: The idea of conditioning a tracer catalogue based on these web environments is such a clean way to handle environment-dependent effects without needing dozens of extra parameters in the model. It’ keeps things physically grounded.

Subrahmanyan: This isn't just about simplifying the math; it’s about capturing the full physical story of assembly bias across different stages of structure formation, tying together short-range and long-range influences.

Vera: The way they are using this classification to create a compact information-theoretic representation is also incredibly useful. We can summarize the environment into just four types, but that summary holds all the detail we need.

Jocelyn: It’s like having a very sophisticated categorization system for our survey targets, allowing us to predict how they should behave based on where they sit in the cosmic web structure.

Subrahmanyan: The ability to see how these discrete levels map onto continuous fields of curvature or ridge measures shows that we are gaining control over the entire range of scales, ensuring the model is physically consistent.

Results and Information Content: Vera: Let's look at the results presented in "Spectral Hierarchy of the Cosmic Web" to see what we can learn about information content. The visual evidence is very compelling, showing how each level probes a different aspect of the structure.

Jocelyn: It’s clear from Figure two that as you move from the large-scale tidal web at i=-two to the higher derivative levels, you are systematically zooming in on smaller and smaller details. The visual evidence is very persuasive.

Subrahmanyan: And I find it particularly interesting that while the classical i=-two level loses predictive power quickly at high wavenumbers, those higher-order levels are much more robust there is important information to be found.

Vera: The cross-power spectrum analysis in Figure four shows this trend quantifying the persistence of information, which is something we really need for our observational work. It confirms that the hierarchy retains predictive power up to the Nyquist scale.

Jocelyn: Seeing that i=zero two and levels are dominating at higher k means that our small-scale measurements will be particularly well-represented by this model, which is a massive win for us.

Subrahmanyan: It suggests that the information about local curvature and short-range dynamics is often much more significant in the nonlinear regime than what simple potential models can capture.

Vera: The fact that we are seeing this behavior both in the high-resolution FastPM simulation and then on a coarse mesh with the Abacus halo data reinforces how robust this method is across different scales.

Jocelyn: It tells us that our chosen analysis framework can handle both the massive, large-volume simulations and the smaller, coarser meshes typical of quick mock generation without losing vital environmental information.

Subrahmanyan: The mathematical consistency between these different scale tests proves that the structure has a deep, underlying hierarchical organization that is independent of how we initially view it.

Conclusion and Final Wrap-Up: Vera: We are coming to the end of our discussion on "Spectral Hierarchy of the Cosmic Web," and I think we can all agree this is a remarkably robust piece of work. It’ offers us a truly comprehensive way to characterize the environment.

Jocelyn: It feels like this paper has finally provided a practical, interpretable tool for us, unifying everything from large-scale gravitational pull to local curvature sensitivity. We have something powerful here for our future surveys.

Subrahmanyan: I think the ability to map these discrete web categories onto the continuous language of bias operators is the most significant theoretical contribution of this work. It ties together all those disparate concepts neatly into one coherent framework.

Vera: The entire concept of "Spectral Hierarchy of the Cosmic Web" provides a powerful, scale-ordered ladder that handles everything from infrared tidal structure to local short-range dynamics.

Jocelyn: It’s exciting to know that we can now use this hierarchy not just for theoretical modeling but also as a practical basis for designing subgrid models in our simulations.

Subrahmanyan: We must remember the core idea—that the web classification is not merely an aesthetic choice, it encodes the same tensor invariants that drive large-scale nonlocality and bias.

Vera: The results show us that this hierarchy keeps information relevant all the way up to our mesh Nyquist limit, which is a huge practical win for mock generation.

Jocelyn: It’s a powerful tool for us, providing a compact signature of the environment that works across different scales and gives us confidence in our analysis.

Subrahmanyan: We can't wait to see how this framework is used to explore other phenomena, like its potential applications in studying galaxy assembly bias and further structure.

Vera: Thank you all for joining us; we’re really looking forward to the next paper on arXiv!

Jocelyn: Goodbye everyone, and I hope our listeners are excited about this "Spectral Hierarchy of the Cosmic Web."

Subrahmanyan: Keep an eye on this work, it' a truly fundamental change in how we view cosmic structure.

Instituto de Astrofísica de Canarias · Department of Astrophysics, University of La Laguna · Institute for Fundamental Physics of the Universe · SISSA - International School for Advanced Studies · INAF - Osservatorio Astronomico di Trieste · INFN – National Institute for Nuclear Physics

astro-ph.CO, cs.CV

Submitted: 2026-03-16

Updated: 2026-09-03

Comments: 33 pages, 7 figures, 1 table, revised version

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 92/100

The gist: The cosmic web, defined by its anisotropic structure of voids, sheets, filaments, and knots, is a fundamental concept in large-scale structure analysis.

Key concepts

Spectral Hierarchy of the Cosmic Web
This is a systematic approach that classifies the physical structure of matter into four defined categories. The method uses scale-weighting kernels and Fourier space filtering, moving beyond simple density to classify the structure of density contrast at different scales.
Cosmic Web Structure Types
The hierarchy maps the physical environment into four distinct categories based on how matter is organized. These types are void, sheet, filament, and knot. This classification allows researchers to categorize the density field based on its geometric properties.
Subgrid Modeling Utility
The framework provides a practical tool for creating mock galaxy production and subgrid models. Instead of using simple density bins, researchers can condition their simulations on these defined web environments, capturing environment-dependent effects efficiently.

Terminology

Summary

The cosmic web, defined by its anisotropic structure of voids, sheets, filaments, and knots, is a fundamental concept in large-scale structure analysis. However, existing methods are often method-dependent, failing to provide a unified framework for modeling environment-dependent phenomena. This paper introduces the spectral hierarchy of the cosmic web—a systematic classification scheme that unifies traditional potential/tidal web definitions with new curvature and higher-derivative levels. This approach provides an explicit bridge between cosmic-web environments and long- and shortrange nonlocal bias ingredients, offering a practical, scalable basis for modeling galaxy formation in both high-fidelity simulations and fast mock production.

How the Spectral Hierarchy is Constructed

The hierarchy is built upon applying simple scale-weighting kernels to the density field, delta(k), organized by integer even powers of k. This family of filtered fields is defined as:

delta(i)(k) -k i (k), i in 2 Z

where i indexes the degree of derivative and nonlocality. By applying the generalized Hessian tensor, H ab = d a d b delta(i)(x), this construction creates a spectral ladder that organizes information content across scales in a controlled manner. This structure allows researchers to interpret web information as a compact summary of the field content that enters bias expansions and effective stress-tensor descriptions.

Physical Interpretation of Hierarchy Levels

The hierarchy levels are not merely mathematical constructs; they correspond to distinct physical regimes, ranging from ultra-long-range effects to highly localized nonlinear dynamics:

  • ** i = -4 (IR/Gauge):** Corresponds to relativistic/gauge-screening sensitivity, encoding extremely long-range modes and coupling between super-horizon perturbations.

  • ** i = -2 (Tidal/Potential):** Represents the standard Newtonian gravitational potential, capturing the anisotropic collapse and long-range tidal environments responsible for generating halo angular momentum. This is the classical tidal web.

  • ** i = 0 (Curvature):** Defines a curvature web, emphasizing peaks, ridges, and saddle morphology through the density Hessian, providing a short-range nonlocal correction.

  • ** i = +2 and i = +4 (UV/Local)::** These higher-derivative levels are increasingly sensitive to small-scale structure. They encode local relaxation and are relevant near the ultraviolet cutoff of a numerical mesh, aligning with higher-order gradient corrections in effective field theory.

Quantifying Information Content

To test the predictive power of this hierarchy, the paper utilizes two distinct simulation sets: a high-resolution FastPM particle-mesh simulation and the coarse Abacus dark matter catalogue. The method involves three steps: (i) constructing generalized web classifications from the Hessian tensors, (ii) compressing these into a four-value field (w(x)) representing knots, filaments, sheets, or voids, and (iii) measuring the cross-correlation between this compressed web field and the halo overdensity field.

Results: Information Retention Across Scales

The results demonstrate that the spectral hierarchy retains significant tracer-relevant information from very large scales down to the mesh Nyquist limit. The analysis shows a clear scale-dependent hierarchy of information content:

  • At intermediate nonlinear scales (k about 1 h Mpc-1), all levels are highly correlated.

  • As k increases toward the ultraviolet end, the more local (curvature/higher-derivative) levels dominating, such as i=2 and i=4, provide substantially more information than the classical tidal-tensor level (i=-2).

Implications for Mock Generation

The spectral hierarchy offers a powerful tool for practical applications. It provides a practical, interpretable conditioning basis for fast mock galaxy production, allowing researchers to bypass the pitfalls of truncated explicit bias expansions. By subdividing a tracer catalogue into subsets based on these physically meaningful web environments, local positive-definite bias models can be applied in each subset, naturally accounting for environment-dependent physics without introducing excessive complexity or negative predicted densities.

Improvements for AI systems

The following improvements leverage the structured, multi-scale nature of the Spectral Hierarchy to enhance AI systems used in large-scale structure modeling, mock generation, and bias estimation. These changes are specific, technical, and designed for high-fidelity applications where traditional methods fail to capture environmental dependence efficiently.

Instead of feeding the AI system a single density contrast (delta(x)), the input feature vector must be extended to include the local invariants derived from multiple levels of the spectral hierarchy.

Implementation: For every grid cell x, calculate and feed:

  • The Level-Specific Hessians: The 4 times 4 = 16 classification tensors, H(i)(x) = d i d j delta(i)(x), for the chosen levels (e.g., i=-2, i=0, and i=2).

  • The Eigenvalue Invariants: The three corresponding scalar invariants (I 1, I 2, I 3) derived from the eigenvalues (lambda 1, lambda 2, lambda 3) of each level's Hessian.

  • ** The Input Vector V(x):** Concatenate these features into a structured vector:

V(x) = [Inv 1(-2), Inv 2(-2), Inv 3(-2),, Inv 3(0),, lambda 1(i), lambda 2(i) for i=-4 to i=+4].

The AI system will utilize the discrete classification provided by the hierarchy as a categorical conditioning variable, moving beyond simple high density vs. low density thresholds.

Training Logic: The training objective is modified to ensure that the predicted bias b is a weighted sum of local models:

b = sum i w(x) times B sub-region(x)

where w(x) is the categorical indicator function for each sub-region, and B sub-region is a local parametric model (e.g., a small polynomial or Gaussian fit) calibrated to the high-resolution data.

The loss function must be adapted to penalize predictions that fail specifically in the transition zones between large-scale tidal dominance and small-scale curvature dominance, ensuring that the model learns when to trust H(-2) versus H(0).

The improved system will be capable of executing highly specific tasks that are impossible with standard density-based classifiers:

  1. Predictive Accuracy in Non-Linear Regimes: The system can predict the halo abundance and clustering power spectrum P(k) with high fidelity even at scales where traditional linear theory fails, because it has learned to distinguish between a feature caused by large-scale tidal strain (i=-2) and a feature caused by local density curvature (i=0), which is crucial for accurate non-linear mock generation.

  2. Efficient Nonlocal Bias Modeling: It can implement effective bias models that are physically grounded without requiring explicit, computationally expensive non-local operators in the simulation code. By using the 256 discrete environmental regions as conditioning variables, it replaces complex operator expansions with a compact lookup table of local parametric fits.

  3. Scale-Dependent Tracer Characterization: The system can identify and classify a tracer population (e.g, simulated galaxies) not just by its density, but by its environmental footprint—its unique response to tidal forces (H-2) and its ability to reside in high-curvature regions (H 0). This allows for the creation of highly nuanced, multi-scale subgrid models.

  4. Automated Subgrid Model Transfer: The system can automatically translate the hierarchical information content (e.g, This halo is in a knot defined by i=2 ) into an action for a fast mock generator, allowing researchers to transfer knowledge gained from high-resolution simulations to coarse-mesh forward models with minimal manual intervention.

Abstract

We introduce a spectral hierarchy of cosmic-web classifications obtained by applying simple scale-weighting kernels to the density field before performing a standard eigenvalue-based web classification. This unifies and extends several widely used web definitions within a single framework: the familiar potential/tidal web (large-scale, nonlocal), a curvature-based web (more local, peak- and ridge-sensitive), and additional higher-derivative levels that progressively emphasize smaller-scale structure. Because the classification is built from second derivatives of the filtered field, successive hierarchy levels align naturally with operator families that appear in renormalised bias and effective descriptions of large-scale structure, providing an explicit bridge between cosmic-web environments and long- and short-range nonlocal bias ingredients. We quantify the information content of the hierarchy with a compact statistic: we map each cell to one of four ordered web types (void, sheet, filament, knot), construct a corresponding ``web contrast'' field, and measure its cross-correlation with halos from the AbacusSummit simulation suite on a coarse mesh with ΔL 5.5,h-1 Mpc. We find that the hierarchy retains significant tracer-relevant information from very large scales down to the mesh Nyquist limit, with the more local (curvature/higher-derivative) levels dominating toward nonlinear scales. This makes the spectral hierarchy a practical, interpretable conditioning basis for fast mock-galaxy production and field-level modelling, and a flexible tool for studying environment-dependent clustering and assembly bias.

Sources

Related papers