⚛️ general relativity

Galaxy bias renormalization: Two-loop Power Spectrum, One-loop Trispectrum and Bispectrum

This paper presents a complete, fully renormalized framework for fifth-order galaxy bias at the one- and two-loop levels, providing explicit computations for the two-loop power spectrum, one-loop bispectrum, and trispectrum while incorporating stochastic renormalization and demonstrating a pronounced scale-dependence in higher-gradient bias coefficients.

Original authors: Thomas Bakx, Mathias Garny, Henrique Rubira, Zvonimir Vlah

Published 2026-07-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Thomas Bakx, Mathias Garny, Henrique Rubira, Zvonimir Vlah

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ 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, cosmic ocean. In this ocean, galaxies aren't just floating randomly; they are like buoys or lighthouses that form in specific patterns based on the underlying currents (the dark matter). Astronomers want to map these patterns perfectly to understand the history and shape of the universe.

However, there's a problem. When scientists try to calculate how these "buoys" (galaxies) cluster together using math, the equations get messy. If they try to look at the smallest, most chaotic details (the "ultraviolet" or high-energy parts of the ocean), the math explodes and gives infinite, nonsensical answers.

This paper is like a new, highly sophisticated instruction manual for fixing those math explosions. Here is the breakdown of what the authors did, using simple analogies:

1. The Problem: The "Blurry" Telescope

Think of the universe's structure as a high-resolution photo. The "Leading Gradient" (LG) operators are like the main focus of the camera—they capture the big, clear shapes of the galaxy clusters. But to get a perfect picture, you also need to account for the tiny, fuzzy details right at the edge of the lens.

Previously, scientists had a good manual for the main focus (the LG part), but they were missing the instructions for the "fuzzy edges" (the Next-to-Leading Gradient or NLG operators). Without these, their calculations for how galaxies cluster at different scales were incomplete and prone to errors.

2. The Solution: A New "Renormalization" Recipe

The authors developed a complete "recipe" to fix these math explosions. In physics, this process is called renormalization.

The Analogy: Imagine you are baking a cake, but your measuring cups are slightly broken, so you keep adding too much flour, and the cake gets ruined.
The Fix: The authors created a new set of "counter-measures" (counterterms). These are like special adjustments you add to the recipe to cancel out the extra flour.
The Innovation: They didn't just fix the main ingredients (the flour/sugar); they also fixed the subtle spices (the gradients). They created a complete list of every possible "spice" (operator) needed to describe the universe up to a very high level of detail (fifth-order).

3. The "Two-Loop" and "One-Loop" Maps

The paper calculates three specific types of cosmic maps:

The Power Spectrum (Two-Loop): This is a map of how galaxy clusters are distributed in general. The authors fixed the math so this map is accurate even when looking at very complex, overlapping patterns (two loops of calculation).
The Bispectrum (One-Loop): This map looks at how three galaxies interact with each other.
The Trispectrum (One-Loop): This is an even more complex map looking at how four galaxies interact.

The authors showed how to calculate all three of these maps simultaneously without the math breaking down. They proved that if you use their new "counter-term" adjustments, the infinite errors disappear, leaving a clean, finite prediction.

4. The "Stochastic" Noise

Sometimes, the universe isn't perfectly smooth; there is random "noise" (like static on a radio) caused by small-scale physics we can't see directly.

The Analogy: Imagine trying to hear a conversation in a crowded room. The "deterministic" part is the conversation you can hear clearly. The "stochastic" part is the random chatter and clinking of glasses.
The Fix: The authors showed how to mathematically separate the clear conversation from the random noise, even when the noise comes from groups of 2, 3, or 4 people talking at once. They created rules to handle this "noise" so it doesn't ruin the main calculation.

5. The "Running" Coefficients (The Shape-Shifting Rules)

One of the most interesting findings is that the "rules" of the universe (the bias coefficients) aren't static; they change depending on the scale you are looking at.

The Analogy: Think of a chameleon. If you look at it from far away, it looks green. If you look close up, it looks blue. The chameleon hasn't changed its skin; your perspective has.
The Discovery: The authors found that the "spices" (bias coefficients) they added to the recipe change their value as you zoom in or out of the universe. They calculated exactly how these values change (the "Renormalization Group Equations"). They found that the "fuzzy edge" spices (NLG) have a very strong tendency to change their value as you look at different scales, which is a crucial detail for future experiments.

Summary

In short, this paper provides the complete, error-free instruction manual for predicting how galaxies cluster together in the universe. They added the missing "fine-tuning" ingredients (NLG operators) to the existing recipe, fixed the math for complex interactions (bispectra and trispectra), and showed how the rules of the game change depending on how closely you look. This allows astronomers to use upcoming, super-precise telescopes to measure the universe with much greater accuracy than before.

Technical Summary: Galaxy Bias Renormalization at Two-Loop and One-Loop Levels

Problem Statement
The next generation of galaxy surveys (e.g., DESI, Euclid, LSST) aims to constrain cosmological parameters at the percent level, necessitating theoretical models of galaxy clustering with equivalent precision. While the Effective Field Theory (EFT) of Large Scale Structure (LSS) provides a systematic framework for computing statistics of biased tracers (galaxies, halos) via an expansion in the linear density field and its derivatives, a complete treatment requires addressing ultraviolet (UV) divergences in loop integrals.

Previous work (Paper I) established a systematic renormalization scheme for leading-gradient (LG) bias operators up to fifth perturbative order. However, two-loop predictions for the galaxy power spectrum, as well as one-loop predictions for the bispectrum and trispectrum, require the inclusion of next-to-leading-gradient (NLG) operators—those carrying two additional spatial derivatives. These operators are necessary to absorb UV-sensitive contributions that mix with LG operators at higher loop orders and to encode physical non-locality effects of order $(kR_L)^2$ . Furthermore, a consistent joint modeling of these statistics requires a unified treatment of stochastic renormalization for products of two, three, and four operators at coincidence.

Methodology
The authors generalize the renormalization scheme introduced in [15, 16] and extended in Paper I to incorporate the full basis of NLG bias operators at second gradient order. The methodology proceeds through several key steps:

Operator Basis Construction: The authors construct a non-redundant basis of scalar NLG bias operators at second gradient order and up to third perturbative order. This involves classifying operators by their perturbative order in the linear density contrast $\delta_L$ and the number of spatial gradients. The basis includes:
- "Standard" operators with two overall gradients ( $\nabla^2$ ).
- Operators without overall gradients (specific to biased tracers).
- "Backreaction" operators arising from the interaction of counterterms with the tidal tensor, necessitating an extended set of building blocks beyond the standard $\Pi^{[n]}_{ij}$ tensors. Specifically, the authors introduce a working hypothesis involving the building block $\nabla_k \nabla_l \nabla^{-2} \Pi^{[n]}_{kl}$ to ensure the basis is sufficient for renormalizing the deterministic sector of the one-loop bi- and trispectra and the two-loop power spectrum.
Deterministic Renormalization: A generalized renormalization condition is imposed, requiring that renormalized bias parameters match tree-level predictions not only in the large-scale limit ( $k \to 0$ ) but also in their quadratic dependence on external wavenumbers. This fixes the full renormalization mixing matrix $Z_{AB}$ , which couples LG and NLG operators. The authors derive the complete one-loop and two-loop renormalization matrices, establishing new commutation relations between single-hard and double-hard limits of bias kernels.
Stochastic Renormalization: The framework is extended to the operator product expansion (OPE) for stochastic terms. The authors derive renormalization conditions for products of two, three, and four operators at coincident points. These conditions absorb UV-sensitive mode-coupling loops into contact terms, defining stochastic counterterms for the power spectrum, bispectrum, and trispectrum.
Correlator Construction: The renormalized operator expressions are translated into explicit prescriptions for the renormalized two-loop power spectrum, one-loop bispectrum, and one-loop trispectrum. The authors identify which specific NLG operators yield distinct, non-degenerate contributions at each loop order and for each statistic.
Renormalization Group (RG) Analysis: The authors derive the coupled RG equations governing the running of both LG and NLG bare bias coefficients with respect to the UV cutoff $\Lambda$ . They solve these equations at one- and two-loop orders to analyze the scale dependence of higher-gradient bias coefficients.

Key Contributions and Results

Complete Operator Basis: The paper provides the complete non-redundant basis of 20 second-gradient operators (17 "standard" plus 3 "backreaction" operators) required to renormalize the two-loop power spectrum and one-loop bispectrum/trispectrum for general biased tracers.
Renormalization Matrices: The authors compute the full block structure of the one-loop and two-loop renormalization matrices ( $Z_{AB}$ ), including the mixing between LG and NLG sectors. They derive the associated anomalous dimension matrices and establish commutation-of-limit relations that connect double-hard limits to products of single-hard limits.
Stochastic Counterterms: A systematic treatment of stochastic renormalization is provided for products of up to four operators. The authors derive the necessary contact-term coefficients ( $Z_{abc}$ , $Z_{abcd}$ , etc.) and demonstrate how they absorb UV divergences in the mode-coupling loops of the bispectrum and trispectrum.
Operator Counting: The paper provides a detailed count of the independent free parameters required for various combinations of N-point functions. For the full set $\{P^{2L}, B^{1L}, T^{1L}\}$ , the analysis identifies 24 deterministic LG operators, 20 deterministic NLG operators, and 7 stochastic parameters at leading gradient order.
RG Solutions and Scale Dependence: Solving the RG equations reveals a pronounced scale dependence of higher-gradient bias coefficients. The authors find that the running of NLG coefficients is significantly influenced by mixing with LG coefficients and among NLG coefficients themselves. They demonstrate that the one-loop RG evolution captures most of the UV sensitivity, while intrinsic two-loop contributions provide subdominant corrections.
Matter vs. Biased Tracers: The analysis highlights that while the matter density field (where $b_\delta=1$ and other LG biases vanish) has a closed system for counterterm running, general biased tracers exhibit mixing between LG and NLG sectors, leading to more complex running behavior.

Significance
The paper claims to provide a "complete treatment" of fifth-order renormalized galaxy bias in real space, extending the EFT framework to include next-to-leading gradient corrections. By performing renormalization manifestly at the operator level, the results ensure that UV divergences are consistently absorbed across all N-point functions, allowing for a joint analysis of the power spectrum, bispectrum, and trispectrum with identical renormalized bias parameters.

The authors emphasize that their results can be easily extended to higher N-point functions, higher loop orders, and field-level analyses. The derivation of the RG equations serves as a non-trivial consistency check of the bias renormalization approach and offers a tool to verify the theoretical "running" of EFT coefficients obtained from data fits. The work establishes a systematic foundation for comparing fifth-order galaxy bias expansions directly to observational data from current and upcoming surveys, addressing the need for percent-level theoretical precision.