🔬 condensed matter

Seven- and eight-loop critical exponents of the three-dimensional Ising model

This paper presents precise estimates of the critical exponents $\eta$ , $\nu$ , and $\omega$ for the three-dimensional Ising model by resumming seven- and eight-loop renormalization-group series, revealing that while the error for $\eta$ decreases rapidly with loop order, the slow convergence of all exponents leads to a systematic tension with current conformal bootstrap benchmarks.

Original authors: D. Shapoval, Yu. Honchar, B. Delamotte, M. Dudka, Yu. Holovatch

Published 2026-07-10

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: D. Shapoval, Yu. Honchar, B. Delamotte, M. Dudka, Yu. Holovatch

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, chaotic dance floor. At a very specific temperature, the dancers (atoms) suddenly stop wiggling randomly and lock into a perfect, rigid formation. This moment is called a "phase transition," and the rules that govern how they lock together are described by mysterious numbers called critical exponents. For decades, physicists have been trying to calculate these numbers for the simplest dance floor of all: the 3D Ising model (a theoretical grid of spins that can be up or down).

For a long time, the best way to guess these numbers was to use a mathematical tool called perturbation theory. Think of this like trying to predict the path of a rollercoaster by looking at the track one tiny inch at a time. You take a step, make a guess, take another step, and refine your guess. The more steps (or "loops") you take, the closer you should get to the truth.

The New High-Stakes Map

In this paper, the authors decided to take the rollercoaster ride to the absolute limit. They used the most recent, incredibly complex calculations available—seven and eight loops (which is like calculating the track for millions of tiny inches)—to update their map of the 3D Ising model. They focused on three specific numbers:

$\eta$ (eta): How fast the dancers' influence fades away.
$\nu$ (nu): How the "dance floor" stretches as it gets ready to lock up.
$\omega$ (omega): How quickly the dancers settle into their final rhythm.

The Magic Trick: Resummation

Here's the catch: when you add up too many of these tiny steps, the math usually explodes into nonsense. The numbers get huge and wild. To fix this, the authors used a clever "magic trick" called resummation.

Imagine you are trying to listen to a song, but the speaker is crackling and distorting the sound. You can't just turn up the volume; you need a special filter to clean up the noise. The authors used a combination of conformal mapping (which is like stretching a crumpled piece of paper flat so you can read the writing) and a homographic transformation (a mathematical twist that moves the "bad spots" in the calculation out of the way). They then optimized their filter using two rules:

Minimal Sensitivity: Make sure the answer doesn't wiggle if you tweak the filter slightly.
Fastest Convergence: Make sure the answer settles down quickly as you add more loops.

The Big Surprise: A Slight Tension

The authors found something fascinating and a little bit frustrating.

For the exponent $\eta$ , the magic worked beautifully. As they added more loops (going from 7 to 8), the error bar (the "maybe" range) shrank rapidly. It was like zooming in with a camera and suddenly getting a crystal-clear picture. Their final estimate for $\eta$ is 0.03617(10).

However, for $\nu$ and $\omega$ , the picture didn't get as sharp. Even with all that extra math, the error bars didn't shrink much. The values they found were:

$\nu$ : roughly 0.6299 (with an uncertainty of about 0.0001)
$\omega$ : roughly 0.820 (with an uncertainty of about 0.007)

Here is the twist: The authors compared their results to the "gold standard" of the field, a method called the conformal bootstrap (which uses pure logic and symmetry to find the answer without doing the messy loop calculations). The conformal bootstrap gives a very precise value for $\nu$ of 0.629971(4).

The authors' results are very close, but not quite identical. There is a slight, systematic tension. It's like two expert watchmakers building the same clock; one says it ticks at 12:00:00.00, and the other says 12:00:00.01. They are both incredibly accurate, but they don't match perfectly.

What They Ruled Out (and What They Suspect)

The paper explicitly argues against the idea that adding more loops will automatically make the perturbative method converge perfectly to the "exact" answer. For a long time, physicists thought, "If we just keep adding loops, the answer will just get better and better until it matches the conformal bootstrap."

This paper suggests that this is not happening for $\nu$ and $\omega$ . The convergence is slowing down.

Why? The authors suggest a geometric reason. Imagine the path the math takes to find the answer is a river.

The "exact" path is a wide, slow-moving Great River flowing through a valley.
The perturbative method (the loop calculations) is like a small stream trying to follow that river.
The problem is that the small stream doesn't flow exactly along the center of the Great River; it has to zigzag a bit because the river bends in directions the small stream can't easily see.
This "zigzagging" creates non-analytic terms—mathematical glitches that standard loop calculations can't see. These glitches get in the way, preventing the answer from settling down perfectly, even with 8 loops.

How Sure Are They?

The authors are very sure that their numbers are accurate within their own method. They have calculated the uncertainty carefully, and their error bars are small.

They are confident that the error on $\eta$ is shrinking fast.
They are confident that the error on $\nu$ and $\omega$ is not shrinking as fast as hoped.
They are suggesting (not proving) that the reason for the mismatch with the conformal bootstrap is these hidden, non-analytic glitches in the math.

They do not claim to have solved the mystery of why the two methods disagree. Instead, they highlight that the "standard" way of doing these calculations might have a hidden limit that we haven't fully understood yet. The paper is a high-precision measurement that says, "We did our best with the most advanced tools we have, and we found a tiny, stubborn gap between our best guess and the gold standard."

The Takeaway

This paper is a masterclass in pushing math to its breaking point. The authors took the most complex calculations available, cleaned them up with the best filters, and found that while they are incredibly precise, the old-school method of "just adding more loops" might have hit a wall for some of these numbers. The gap between their result and the conformal bootstrap is small, but it's real, and it hints that the universe's dance floor might be a little more complex than our current maps can fully capture.

Technical Summary: Seven- and Eight-Loop Critical Exponents of the Three-Dimensional Ising Model

Problem Statement
The determination of critical exponents for the three-dimensional (3D) Ising universality class serves as a primary benchmark for evaluating perturbative Renormalization Group (RG) methods. While non-perturbative approaches such as the Conformal Bootstrap (CB), Monte Carlo (MC) simulations, and the Non-Perturbative Renormalization Group (NPRG) have achieved high-precision results, the reliability and convergence properties of high-order perturbative $\epsilon$ -expansions ( $\epsilon = 4-d$ ) remain a subject of methodological scrutiny. Specifically, there is an open question regarding whether the slow convergence of resummed perturbative series at high loop orders (seven and eight) introduces systematic biases or underestimates error bars, potentially leading to tensions with the "exact" values provided by the Conformal Bootstrap. This study addresses these issues by analyzing the recently computed seven- and eight-loop RG series for the 3D Ising model.

Methodology
The authors employ the field-theoretical RG framework within the $\epsilon$ -expansion, utilizing the perturbative series for the beta function $\beta(g)$ and anomalous dimensions $\gamma_\phi(g)$ and $\gamma_{\phi^2}(g)$ up to the eighth loop order. The core of the methodology involves a sophisticated resummation procedure designed to handle the factorial divergence of the series:

Borel-Conformal Mapping (BCM): The divergent series are first transformed via a Borel-Leroy transformation to suppress factorial growth. The resulting Borel transform is then analytically continued using a conformal mapping that relocates the dominant singularities to the boundary of the convergence domain, accelerating the convergence of the truncated series.
Homographic Transformation: To further improve convergence and stability, the expansion parameter $\epsilon$ is subjected to a homographic (Möbius) transformation, introducing an additional tunable parameter $q$ . This transformation helps shift singularities in the Borel plane away from the physical integration path.
Optimization Criteria: The resummation involves auxiliary parameters ( $a, b, \alpha, q$ $a, b, α, q$ ). To determine optimal values and quantify uncertainties, the authors implement two complementary strategies:
- Method I (Global Min-Max): Based on the Principle of Minimal Sensitivity (PMS) and the Principle of Fastest Apparent Convergence (PFAC). This approach minimizes a total error functional $E_{total}$ that combines the sensitivity of the observable to resummation parameters with the discrepancy between successive loop orders.
- Method II (Local RMS Stability): A complementary approach using Root-Mean-Square (RMS) variations to assess local stability within parameter neighborhoods, combined with PFAC consistency checks. This method aims to resolve the structure of stability plateaus more finely than the global min-max approach.
Error Estimation: Uncertainties are estimated by analyzing the spread of results within stability windows and the discrepancy between loop orders, employing a weighted mean over the stable region and a global minimum approach to define the final values and error bars.

Key Contributions

Extension to Eight Loops: The paper presents the first application of the BCM resummation technique combined with homographic transformations to the complete seven-loop and substantial eight-loop RG series for the 3D Ising model, utilizing the recent computations by Schnetz.
Refined Error Analysis: The authors introduce and compare two distinct optimization algorithms (Method I and Method II) for multi-parameter resummation. They demonstrate that while Method I provides a conservative global estimate, Method II offers a more detailed view of local stability, confirming the robustness of the results across different parameter landscapes.
Systematic Tension Identification: The study rigorously quantifies the convergence behavior of the critical exponents $\eta$ , $\nu$ , and $\omega$ . It highlights that while the error bar for $\eta$ decreases rapidly with increasing loop order, the convergence for $\nu$ and $\omega$ is notably slower.

Results
The authors report precise estimates for the critical exponents $\eta$ , $\nu$ , and the correction-to-scaling exponent $\omega$ at seven- and eight-loop orders.

Convergence Patterns: The exponent $\eta$ shows rapid convergence, with the error bar shrinking significantly from the six-loop to the eight-loop order. In contrast, $\nu$ and $\omega$ exhibit slow convergence; the estimated values remain accurate in absolute terms but do not show a proportional reduction in uncertainty with added loop orders.
Comparison with Benchmarks: The eight-loop estimates for $\eta$ , $\nu$ , and $\omega$ are found to be in good absolute agreement with high-precision Conformal Bootstrap results. However, the authors observe a "slight but systematic tension" between the perturbative estimates and the CB benchmarks. This tension arises because the slow convergence of the perturbative series prevents the central values from settling exactly onto the CB values even at eight loops, despite the absolute accuracy of the estimates.
Consistency: The results obtained via Method I and Method II are mutually consistent and align well with previous six-loop studies (e.g., Kompaniets & Panzer), validating the stability of the resummation framework.

Significance and Claims
The paper claims that its primary significance lies in clarifying the behavior of high-order perturbative RG series in the regime where straightforward convergence to "exact" results ceases to be evident. The authors argue that the observed slow convergence and systematic tension with Conformal Bootstrap estimates are likely intrinsic features of the perturbative expansion, potentially linked to non-analytic contributions near the fixed point (as suggested by Sokal's geometric interpretation of RG flow).

The work does not claim to resolve the tension or prove the superiority of one method over another; rather, it provides a rigorous quantitative assessment of the limitations of perturbative resummation at very high orders. The authors conclude that while perturbative RG remains a powerful tool, the slow convergence of $\nu$ and $\omega$ suggests that non-analytic effects or the specific structure of the RG flow may limit the precision achievable through standard resummation techniques alone, necessitating continued cross-validation with non-perturbative methods like the Conformal Bootstrap.