⚛️ high-energy theory

Two-loop renormalization and running of galaxy bias

This paper systematically extends the framework of galaxy bias renormalization to two-loop order by deriving explicit renormalization results for a minimal basis of 29 deterministic operators, incorporating stochasticity, and establishing two-loop renormalization group equations that reveal enhanced UV sensitivity and suggest potential applications of quantum field theory resummation techniques.

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

Published 2026-07-10

📖 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 soup. In this soup, invisible matter swirls around, and occasionally, it clumps together to form galaxies. Astronomers are like chefs trying to taste the soup to figure out the recipe, but there's a catch: the galaxies aren't just floating around randomly; they are "biased" tracers. This means they don't perfectly trace the underlying matter; they prefer certain spots, like how a specific type of fish might only swim near the surface while others dive deep.

To understand the universe's recipe, scientists use a mathematical tool called "bias expansion." Think of this as a recipe book where they list ingredients (mathematical operators) to describe how galaxies form. For a long time, this book only had recipes up to a certain complexity. But as our telescopes get sharper, we need more complex recipes to match the data.

The Big Discovery: A New, Massive Recipe Book
The main finding of this paper is that the authors have successfully written out the next two chapters of this recipe book, extending it all the way to the "fifth order." In plain English, they have figured out how to describe the galaxy soup with incredible precision, accounting for interactions that happen at two different "loops" (or layers of complexity) simultaneously.

They found that to do this, they needed a specific set of 29 unique ingredients (mathematical operators). It's like realizing that to bake the perfect cosmic cake, you don't just need flour and sugar; you need exactly 29 specific spices, and they have now listed every single one of them.

What They Ruled Out: The "Redundant" Ingredients
Here is where it gets interesting. When they tried to use all 29 ingredients for their specific two-loop calculation, they discovered that some of them were actually "fake" or redundant.

Imagine you are building a tower with blocks. You think you need 29 blocks, but when you try to stack them for this specific tower, you realize that 12 of those blocks are actually just copies of each other or don't fit the shape of the tower at all. The paper explicitly rules out using these 12 specific operators for the two-loop power spectrum. They found that certain complex combinations of ingredients (specifically those built from five or four instances of a basic building block) cancel each other out perfectly. So, for this specific calculation, you don't need them; they are unnecessary clutter.

The "Running" Bias: A Shifting Goalpost
The paper also tackles a tricky problem: the "smoothing scale." Imagine you are looking at the galaxy soup through a foggy window. If you change the fog (the smoothing scale), the picture changes. Usually, scientists want their answers to be independent of how foggy the window is.

The authors show that the "bias coefficients" (the numbers that tell us how galaxies cluster) aren't fixed constants. They are more like a shifting goalpost. As you change the scale at which you look at the universe (the foggy window), these numbers "run" or change.

They calculated exactly how these numbers change using something called the Renormalization Group Equations (RGE). Think of this as a map that tells you how the bias numbers morph as you zoom in or out. They found that:

Third-order bias parameters (the more complex ingredients) do change as you zoom.
There is a specific "growing mode" in this change. It's like a snowball rolling down a hill; as you look at larger and larger scales, one specific combination of bias parameters grows larger and larger, while others settle down.

The Quantum Connection: Borrowing from Particle Physics
One of the most playful parts of the paper is the analogy they draw with Quantum Field Theory (QFT), the physics of tiny particles. In particle physics, scientists use a trick called "resummation" to handle huge numbers that appear in their equations (called "large logarithms").

The authors suggest that the same trick might work for galaxies. They propose that the "large logarithm" in the universe isn't a log of numbers, but rather a ratio related to how the universe's density changes with scale. They found that for certain types of universes (specifically those with a "power-law" spectrum where the spectral index $n$ is close to $-3$), this ratio becomes very large.

They suggest that by using their new "running" bias numbers, we can "resum" (or sum up) the messy, high-energy parts of the universe's behavior that usually break our calculations. It's like having a magic filter that cleans up the noise in the soup, allowing us to see the true flavor even when the pot is boiling.

How Sure Are They?
The authors are very confident about the math they have derived. They have explicitly calculated the coefficients for all 29 operators and proved that 12 of them are redundant for this specific calculation. They have derived the equations for how the bias "runs" and shown that the math works out consistently with previous one-loop results.

However, regarding the "resummation" idea (using the bias running to clean up the noise), they are suggesting it as a powerful new tool. They have shown that the math allows for this interpretation and that the numbers look promising (the "noise" is suppressed by a factor of about 0.05 to 0.1 in their models), but they note that applying this to real-world data is a future step. They haven't "solved" the problem of galaxy clustering yet, but they have built a much stronger, more precise ladder to climb it.

The Bottom Line
This paper is a massive upgrade to the theoretical toolkit for studying the universe. It gives us a complete list of the 29 ingredients needed for high-precision galaxy clustering, tells us which 12 to throw away for specific calculations, and provides a new map for how the "flavor" of the universe changes as we zoom in and out. It suggests that by using these new maps, we might finally be able to taste the universe's recipe without getting a mouthful of mathematical noise.

Technical Summary: Two-loop renormalization and running of galaxy bias

Problem Statement
Precise measurements of the large-scale structure (LSS) of the Universe require robust theoretical modeling of galaxy clustering. A central component of this modeling is the bias expansion, which maps the observed galaxy density contrast to the underlying matter density field. While one-loop renormalization and running of bias parameters up to third order have been established, the demand for higher-precision data from next-generation surveys necessitates extending these frameworks to two-loop order and including bias operators up to fifth order in perturbation theory. Furthermore, a systematic understanding of the renormalization group (RG) evolution of these higher-order operators and their connection to the resummation of large logarithms (analogous to Quantum Field Theory) remains to be fully developed.

Methodology
The authors employ a systematic extension of the bias renormalization framework to two-loop order and fifth-order perturbation theory. The methodology involves three primary pillars:

Bias Operator Basis: The authors review and explicitly construct the minimal complete basis of 29 deterministic bias operators up to fifth order in the density field, utilizing the building blocks $\Pi^{[m]}_{ij}$ and enforcing symmetries (spatial rotations) and the Cayley-Hamilton relation to eliminate redundancies.
Kernel Limits and Renormalization: To handle Ultraviolet (UV) divergences arising in loop integrals, the authors compute the single-hard ( $p \gg k$ $p ≫ k$ ) and double-hard ( $p, q \gg k$ $p, q ≫ k$ ) limits of the bias kernels.
- For deterministic terms, they derive the limits where loop momenta become large, showing that the double-hard limit of all fifth-order kernels depends on a single universal function $g(p/q)$ of the ratio of loop momenta.
- For stochastic terms, they analyze the contact terms arising from the renormalization of composite operators (products of bias operators), deriving the necessary counter-terms for the power spectrum.
Renormalization Group Equations (RGE): Instead of relying solely on the Wilson-Polchinski shell-approach, the authors derive RGEs by enforcing the independence of physical observables (N-point functions) from the smoothing scale $\Lambda$ . They compute the one-loop and two-loop $\beta$ -functions for bias coefficients, generalizing previous results to include fifth-order operators and two-loop running.

Key Contributions and Results

Two-Loop Renormalization Conditions: The authors derive explicit expressions for renormalized bias operators $[O_a]$ at two-loop order. They demonstrate that the renormalization of deterministic bias at two loops requires subtracting contributions from both single-hard and double-hard limits. Crucially, they show that the double-hard limit contribution is governed by a universal function $g(p/q)$ , defined in Eq. (3.14), which depends only on the ratio of loop momenta and is independent of the specific bias operator.
Stochastic Bias Renormalization: The paper extends bias renormalization to include stochastic contributions via composite operator renormalization. They prove that for the power spectrum, stochastic renormalization at any loop order simply amounts to subtracting the $k=0$ limit of the power spectrum, ensuring the renormalized power spectrum vanishes as $k \to 0$ .
Redundancy in the Two-Loop Power Spectrum: A significant finding is the identification of operator redundancies specific to the two-loop power spectrum. The authors show that 12 of the 29 leading-gradient bias operators (specifically those constructed from five instances of $\Pi^{[1]}_{ij}$ , four instances of $\Pi^{[1]}_{ij}$ , and certain mixed terms) do not contribute to the renormalized two-loop power spectrum. This reduces the effective number of operators needed for numerical evaluation.
Explicit Two-Loop Power Spectrum: The authors provide a manageable, explicit expression for the fully renormalized two-loop power spectrum of biased tracers (Eq. 4.40), suitable for numerical evaluation. This expression incorporates both deterministic and stochastic renormalization.
Renormalization Group Running: The authors derive the one-loop and two-loop RGEs for bias coefficients.
- They identify a linear combination of bias operators with enhanced UV sensitivity, associated with a positive eigenvalue of the one-loop RG matrix. This implies that certain bias operators grow in the UV, preventing the system from converging to a simple fixed point when running from the IR to the UV.
- They provide numerical solutions for the running of linear, second, and third-order bias parameters, showing that including the running of third-order parameters and higher-order operators significantly alters the evolution of lower-order biases.
RG-Improved Perturbation Theory: Drawing an analogy with QFT, the authors propose a systematic resummation of higher-loop terms via the RG. They identify a quantity analogous to the "large logarithm" in QFT. For power-law spectra, this is $1/(n+3)$ (where $n$ is the spectral index), and for $\Lambda$ CDM, it is related to $\ln(\Lambda/k_{eq})$ . They demonstrate that the two-loop contribution is parametrically suppressed relative to the "one-loop squared" contribution by this factor, suggesting that solving the one-loop RG equations captures the dominant UV-sensitive parts of higher loops.

Significance
The paper claims to provide a systematic extension of galaxy bias renormalization to two-loop order and fifth-order perturbation theory, filling a gap required for next-generation LSS analyses. By explicitly working out the renormalization conditions and RGEs, the authors offer a framework that:

Controls the dependence on the smoothing scale $\Lambda$ for both deterministic and stochastic bias terms.
Provides a concrete, reduced basis of operators for calculating the two-loop power spectrum.
Establishes a connection between the bias RG and the resummation of large logarithms, suggesting that RG-improved perturbation theory can be used to systematically organize and improve the convergence of the perturbative expansion in LSS, similar to techniques used in QFT.

The work serves as a theoretical foundation for future applications in galaxy clustering analyses, particularly for field-level analyses where the smoothing scale is an explicit parameter, and for matching field-level bias measurements to N-point function analyses.

More like this