Critical probability distributions of the order parameter from the functional renormalization group

The Gist

The Big Picture: Predicting the "Mood" of a Crowd

Imagine you are standing in a massive stadium filled with thousands of people. Each person is holding a sign that says either "Yes" or "No."

In most situations, if you ask a few people what they think, their answers are random. If you add up all the answers, the result follows a predictable pattern called a Bell Curve (or Gaussian distribution). This is the famous "Central Limit Theorem" in statistics. It's like flipping a coin a million times; you expect roughly 50% heads and 50% tails, with very few extreme deviations.

But what happens when the people start talking to each other?

If the people in the stadium are all shouting at each other, copying each other, or getting excited together, they become strongly correlated. Suddenly, the "Bell Curve" breaks down. You might see the whole stadium suddenly switch to "Yes" or "No" all at once. The rules of normal statistics no longer apply.

This paper is about figuring out exactly what that new, weird pattern looks like when a system is in this "super-connected" state, specifically at a critical tipping point (like water turning into steam).

The Problem: A Missing Map

For a long time, physicists knew that these "strongly connected" systems existed (like magnets at a specific temperature where they lose their magnetism). They knew the patterns were different from the normal Bell Curve. However, they didn't have a good mathematical map to calculate exactly what that new pattern looked like.

Previous methods were like trying to guess the shape of a cloud by looking at a single drop of water. They could get the general idea, but they couldn't calculate the precise shape of the probability distribution (the "mood" of the crowd) for every possible scenario.

The Solution: The "Functional Renormalization Group" (FRG)

The authors of this paper used a powerful mathematical tool called the Functional Renormalization Group (FRG).

Think of FRG as a smart camera with a zoom lens.

1. Zooming Out: Imagine looking at the stadium from a helicopter. You see the whole crowd as a blur.
2. Zooming In: As you zoom in, you start to see small groups of friends talking.
3. The Process: The FRG method works by gradually changing the zoom level. It starts by ignoring the tiny details (the individual people) and focuses on the big groups. Then, it slowly brings the details back in, step-by-step, calculating how the "mood" of the big groups changes as it absorbs the influence of the smaller groups.

By doing this mathematically, the authors could build a complete map of the probability distribution without needing to simulate every single person in the stadium.

The Key Discovery: A Family of Shapes

The most surprising thing the authors found is that there isn't just one shape for this "critical" pattern. There is an entire family of shapes.

They introduced a variable called $\zeta$ (zeta). You can think of $\zeta$ as the ratio between the size of the stadium and the size of the "conversation circles."

• If the stadium is huge compared to the conversation circles: The crowd acts mostly like independent groups, and the shape looks a bit like a normal Bell Curve.
• If the conversation circles are as big as the stadium: The whole crowd is one giant connected unit. The shape becomes very different, with "fat tails" (meaning extreme outcomes are much more likely than in a normal crowd).

The paper shows that by adjusting this ratio ($\zeta$), you can smoothly morph from one shape to another. They calculated the exact mathematical formula for every single shape in this family.

The "Rate Function": The Cost of Being Weird

In the paper, they talk about something called a "Rate Function."

Think of the Rate Function as a "Cost of Unusualness."

• In a normal crowd, it is very "cheap" (probable) to have a 50/50 split. It is very "expensive" (improbable) to have 90% "Yes."
• In these critical, connected systems, the "cost" changes. The paper calculates exactly how expensive it is to have a specific outcome.

They found that the "cost" of being unusual in these critical systems is different from what standard math predicts. Their calculations showed that the "tails" of the distribution (the rare, extreme events) are heavier than expected.

Did They Get It Right?

To prove their math worked, they compared their FRG "camera" results with Monte Carlo simulations.

• The Simulation: This is like running a computer program where they actually simulate millions of people in a stadium, letting them interact, and counting the results. It's the "gold standard" but takes a lot of computer power.
• The Result: The shapes predicted by their FRG math matched the computer simulations almost perfectly.

The "Paradox" Solved

The paper also solves a confusing puzzle that physicists had been arguing about for decades.

• The Puzzle: There is a famous concept in physics called a "Fixed Point" (a specific mathematical state that describes critical systems). Scientists thought this "Fixed Point" described the probability of the crowd's mood. But, the math didn't quite add up because the "Fixed Point" looked slightly different from the actual probability distribution.
• The Resolution: The authors showed that the "Fixed Point" is actually describing the system before the very last step of the zoom-in process. Their new method (FRG) takes that Fixed Point and adds the final missing piece (the "zero-momentum mode") to get the true probability distribution. It's like realizing the Fixed Point was a blueprint, and their method finished the actual building.

Summary

In short, this paper uses a sophisticated mathematical "zoom lens" (FRG) to calculate exactly how likely different outcomes are in a system where everything is connected to everything else. They discovered that there is a whole family of these probability shapes, depending on the size of the system, and they proved their math is correct by matching it with massive computer simulations. They also clarified a long-standing confusion about how these shapes relate to the fundamental laws of physics.

Technical Summary

Problem Statement
The paper addresses the challenge of calculating the probability distribution function (PDF) of the sum of strongly correlated random variables, a problem central to critical phenomena, non-equilibrium dynamics, and quantitative finance. While the Central Limit Theorem (CLT) dictates that sums of independent variables converge to a Gaussian distribution, and generalized CLTs exist for variables with diverging variance (Lévy laws), the behavior of sums of strongly correlated variables (where the correlation matrix sum diverges) remains less understood theoretically.

Specifically, in the context of the three-dimensional Ising model at criticality, the fluctuations of the order parameter (total spin) diverge, rendering the standard saddle-point approximation and Gaussian CLT inapplicable. Furthermore, there exists a paradox in the Renormalization Group (RG) framework: while the PDF is an observable and thus RG-scheme independent, RG fixed points are scheme-dependent. Additionally, while there is a single RG fixed point, there is an infinite family of critical PDFs (or rate functions) indexed by the ratio $\zeta = L/\xi_\infty$ (system size over correlation length). Previous theoretical efforts have largely remained at a conceptual level or relied on ad hoc methods for isolated cases, lacking a systematic technique to compute these PDFs for strongly correlated systems.

Methodology
The authors employ the Functional Renormalization Group (FRG) in its modern version to solve these issues. The methodology involves:

1. Formulation via Constrained Partition Function: The PDF of the normalized total spin $\hat{s}$ is interpreted as the partition function $Z_M$ of a system with a modified Hamiltonian $H_M[\hat{\phi}] = H[\hat{\phi}] + \frac{M^2}{2}(\int \hat{\phi} - s)^2$, where $M \to \infty$ enforces the constraint.
2. FRG Flow Equations: A scale-dependent effective action $\Gamma_{M,k}[\phi]$ is introduced, which interpolates between the microscopic Hamiltonian and the full effective action. The flow is governed by the Wetterich equation with a regulator $R_{M,k}$ that freezes low-wavenumber modes.
3. Rate Function Definition: By taking the limit $M \to \infty$, the authors define a scale-dependent rate function $I_k(s) = L^{-d}\check{\Gamma}_k[\phi=s]$, where $\check{\Gamma}_k$ is the limit of the effective action. The PDF is recovered as $P(s) \propto \lim_{k\to 0} \exp(-L^d I_k(s))$.
4. Approximation Scheme: The authors utilize the Local Potential Approximation (LPA), the simplest functional approximation, where the effective action is approximated by a local potential term. In this approximation, the anomalous dimension $\eta$ vanishes, though the authors argue this does not significantly compromise the shape of the rate function for the 3D Ising model.
5. Numerical Implementation: The flow equations are solved numerically for various values of $\zeta$ (ranging from $-4$ to $4$) to generate the family of scaling functions $I_\zeta(\tilde{s})$, where $\tilde{s}$ is the rescaled order parameter.

Key Contributions

• Resolution of RG Paradoxes: The paper demonstrates that the FRG framework naturally resolves the apparent contradictions between the scheme-independence of observables and the scheme-dependence of fixed points. It clarifies that the rate function $I_{\zeta=0}$ is distinct from, though closely related to, the fixed-point effective potential $\tilde{U}^*$. The fixed point describes the system excluding the zero-momentum mode, whereas the rate function includes it (frozen by the $M \to \infty$ term), allowing for non-convex shapes that the true effective potential cannot possess.
• Computation of the Full Family of PDFs: The authors successfully compute the entire family of universal scaling functions for the order parameter PDF, indexed by $\zeta$, rather than just a single critical distribution.
• Quantitative Agreement with Simulations: The FRG results are compared against Monte-Carlo (MC) simulations of the 3D Ising model on a cubic lattice. The study finds very accurate agreement across the entire range of $\zeta \in [-4, 4]$, validating the FRG approach for calculating PDFs of strongly correlated variables.

Results

• Scaling Functions: The computed rate functions $I_\zeta(\tilde{s})$ exhibit the expected scaling behavior. For $\zeta \ll 1$ (near criticality), the functions are non-convex with a double-well structure reminiscent of the fixed-point potential but distinct due to the inclusion of the zero mode. For $\zeta \gg 1$ (disordered phase, $L \gg \xi_\infty$), the functions become strictly convex and Gaussian-like at small $\tilde{s}$, recovering CLT behavior, while retaining non-Gaussian tails.
• Comparison with MC: The FRG results (solid lines) align closely with MC simulation data (symbols) for the normalized rate functions. A minor rescaling of the parameter $\zeta$ ($\zeta_{MC} \approx 0.9 \zeta_{LPA}$) is required to achieve collapse, which the authors attribute to inaccuracies in the correlation length calculation within the LPA approximation.
• Regulator Independence: The study confirms that the resulting scaling functions are largely independent of the specific choice of regulator function, a necessary condition for universality.
• Convexity: The rate functions are found to become strictly convex for $\zeta \gtrsim 2.2$, whereas the fixed-point potential remains non-convex.

Significance
The paper claims that the Functional Renormalization Group provides the correct framework to quantitatively calculate the PDFs of strongly correlated random variables, a task previously hindered by the lack of systematic techniques beyond conceptual connections to RG. By resolving the paradoxes regarding the relationship between fixed points and observables, the work elucidates the deep connection between RG theory and probability theory. The authors note that while the current results rely on the LPA, the method is generalizable to other pure statistical systems at thermal equilibrium and potentially to disordered or non-equilibrium systems, provided the formalism is adapted. The paper concludes that the LPA captures the essential shape of the rate functions, though systematic improvements (e.g., derivative expansions) may be necessary for regions with strong non-triviality, such as the coexistence region at low temperatures.