Constrained Padé Ensembles for Thermal $\mathcal{N}{=}4$ SYM with the Exact $\mathcal O(\lambda^{5/2})$ Coefficient

By upgrading the weak-coupling expansion of thermal $\mathcal{N}=4$ SYM thermodynamics to the exact $\mathcal{O}(\lambda^{5/2})$ coefficient, this study demonstrates that the constrained log-subtracted two-point Padé ensemble collapses to a single unique curve, thereby eliminating scan uncertainty while highlighting a persistent discrepancy with the Hermite-Padé approach and identifying the need to compute the unknown $\mathcal{O}(\lambda^{-3})$ strong-coupling coefficient.

The Gist

Imagine you are trying to predict the weather for a specific city, but you only have two very different weather reports: one from a local farmer who knows the ground perfectly (Weak Coupling), and one from a satellite orbiting high above (Strong Coupling). The farmer's report is great for calm days, and the satellite's report is great for storms, but neither works well in the middle when the weather is changing.

To get a complete picture, scientists use a mathematical "bridge" called an interpolation. They try to build a smooth road connecting the farmer's data to the satellite's data. However, there are many ways to build this road, and until now, they didn't know which one was the "real" road.

This paper is about upgrading the tools used to build that bridge and seeing how much the map changes.

The Problem: Too Many Roads

In a previous study, the scientists built a bridge using the farmer's data up to a certain point (called $O(\lambda^2)$). Because the data wasn't perfect, the bridge wasn't a single line. Instead, it was a wide, fuzzy band of possible roads.

• Think of it like a foggy path where you can see three or nine different routes you could take.
• All these routes were technically "allowed" by the rules, but the scientists didn't know which one was the true path.

The Upgrade: A New, Exact Clue

Recently, a new, highly precise piece of information was discovered: an exact calculation for a specific part of the farmer's report (the $O(\lambda^{5/2})$ coefficient). It's like the farmer suddenly pulled out a GPS device that gave the exact location of a specific landmark that was previously just a guess.

The authors of this paper asked: "If we feed this new, exact GPS coordinate into our bridge-building machine, does the fog clear up? Do we get a single, clear road?"

The Experiment: The "Fog" Clears

They took their old bridge-building method (called the LSTP ensemble) and updated it with this new, exact number. Here is what happened:

1. The Collapse: Before the update, there were 9 possible routes (which looked like 3 distinct paths). After adding the new exact number, 8 of those paths were instantly rejected because they didn't fit the new data.
2. The Single Path: The remaining paths were so similar they merged into one single, unique curve. The "fog" completely vanished. The uncertainty band shrank to zero.
3. The Result: They now know exactly what the road looks like in the middle region. The "crossover" point (where the road changes from following the farmer to following the satellite) is now pinned down to a specific number: 4.79.

The Twist: Two Different Maps

Here is the interesting part. The scientists also looked at a different way of building the bridge, called the Hermite-Padé (HP) method. This is like using a different type of construction crew.

• They kept the HP method exactly as it was (without the new GPS update) to compare apples to apples.
• The Surprise: The new, single LSTP road does not match the old HP road. They cross at different points and have different shapes.
• The Analogy: Imagine the LSTP crew built a perfect, single-lane highway. The HP crew built a different highway nearby. Both look solid, but they don't connect to each other. The new GPS data told the LSTP crew exactly where to build, but it didn't tell them if their highway is the only correct one in the universe, because the HP highway is still different.

Why This Matters

This paper teaches us two big lessons:

1. Precision is Powerful: One single, exact piece of data can eliminate a huge amount of guesswork. It turned a "maybe this, maybe that" situation into a "this is definitely it" situation for the LSTP method.
2. The Mystery Remains: Even with perfect data on one side, we still have a disagreement between two different mathematical methods. This tells us that to find the absolute truth, we need more than just better data from the "farmer" (weak coupling). We need better data from the "satellite" (strong coupling) to see which of these two different roads is the real one.

The Bottom Line

The scientists have successfully narrowed down the possibilities for how heat and energy behave in a specific type of quantum system (N=4 SYM). They went from a wide, fuzzy guess to a single, precise prediction. However, because a different mathematical method gives a different answer, the final "truth" is still waiting for the next piece of the puzzle: a new calculation from the high-energy (strong coupling) side.

In short: They found the missing piece of the puzzle that made the picture clear, but they realized there are still two different pictures on the table, and they need one more clue to decide which one is real.

Technical Summary

1. Problem Statement

The thermodynamics of thermal $\mathcal{N}=4$ Supersymmetric Yang–Mills (SYM) theory in four dimensions is characterized by the Stefan-Boltzmann-normalized entropy ratio, $f(\lambda) = S/S_0$, where $\lambda = g^2 N_c$ is the 't Hooft coupling.

• The Challenge: Neither the weak-coupling perturbation series (valid for small $\lambda$) nor the strong-coupling AdS/CFT expansion (valid for large $\lambda$) converges in the intermediate "crossover" region.
• Previous Work: In a prior study (Ref. [7]), the authors used Constrained Padé Ensembles (specifically the Log-Subtracted Two-Point Padé or LSTP method) to interpolate between these limits. This resulted in an "admissible family" of curves with a significant uncertainty band, rather than a single unique solution.
• The Trigger: A recent calculation provided the exact coefficient for the $O(\lambda^{5/2})$ term in the weak-coupling expansion. The paper investigates whether incorporating this exact higher-order coefficient narrows the uncertainty band or selects a unique interpolating curve.

2. Methodology

The authors revisit the LSTP framework used in Ref. [7] but upgrade the input data.

• Weak-Coupling Expansion:
  The entropy ratio is expanded as:
  $$f(\lambda) = 1 - \frac{3}{2\pi^2}\lambda + \frac{3+\sqrt{2}}{\pi^3}\lambda^{3/2} + A_{2\log}\lambda^2 \log\left(\frac{\lambda}{\pi^2}\right) + A_{20}\left(\frac{\lambda}{\pi^2}\right)^2 + A_{5/2}\frac{\lambda^{5/2}}{\pi^5} + \dots$$
  The study compares two scenarios:

  1. "Before" Scan: Truncated at $O(\lambda^2)$ (matching Ref. [7]).
  2. "After" Scan: Includes the exact $O(\lambda^{5/2})$ coefficient ($A_{5/2} \simeq +0.7537$) and shifts the weak-coupling matching points to intermediate values ($\lambda \in \{10^{-4}, 10^{-3}, 0.1, 0.5, 1.0\}$) where the new term is numerically significant.

• Interpolation Ansatz (LSTP):

  • A log-subtracted function $g(\lambda)$ is defined to remove the $\lambda^2 \log \lambda$ singularity.
  • $g(\lambda)$ is approximated by a rational function $R(z)$ in a conformal variable $z = \sqrt{\lambda}/(1+\alpha\sqrt{\lambda})$.
  • The approximant is a near-diagonal $[4/4]$ Padé rational function with 9 unknown coefficients.
  • Constraints: The coefficients are fixed by matching 5 weak-side points and 4 strong-side points ($\lambda \in \{10, 30, 100, 300\}$).
  • Admissibility Filters: Solutions are accepted only if they satisfy:
    1. $0.75 \leq f(\lambda) \leq 1$ on the evaluation grid.
    2. Monotonicity ($df/d(\log \lambda) \leq 0$).
    3. No poles on the positive real $\lambda$ axis.

• Comparison Route: The Hermite-Padé (HP) central curve from the previous work is retained unchanged to compare the robustness of different interpolation routes.

3. Key Contributions & Results

A. Collapse of the Admissible Ensemble

The most significant finding is the drastic reduction in the solution space:

• Old Scan ($O(\lambda^2)$): Produced 9 nominal survivors (3 distinct curves) with a broad crossover range $\lambda_c \in [2.95, 6.73]$ and a maximum band width of $0.047$.
• New Scan ($O(\lambda^{5/2})$): The inclusion of the exact $A_{5/2}$ coefficient causes the admissible set to collapse to a single distinct curve.
  • Only 3 parameter combinations $(\alpha, \Lambda_0)$ survived the filters, all sharing $\Lambda_0 = 4.0$.
  • Due to numerical degeneracy in $\alpha$, these 3 survivors are identical, resulting in 1 unique LSTP curve.
  • The pointwise band width drops to $< 10^{-15}$ (effectively zero).
  • New Crossover Value: $\lambda_c \simeq 4.7863$, with $f(\lambda_c) \simeq 0.8371$.

B. Route Dependence and Discrepancy

While the LSTP uncertainty is resolved, a discrepancy between interpolation methods remains:

• The unique LSTP curve does not coincide with the HP central curve.
• Crossover Shift: The LSTP crossover ($\lambda_c \approx 4.79$) is significantly higher than the HP crossover ($\lambda_c \approx 3.52$) and the Müller Padé value ($\lambda_c \approx 3.14$).
• Cause: The positive sign of the exact $A_{5/2}$ coefficient raises the weak-coupling series at intermediate $\lambda$, pushing the inflection point to larger values. The HP route was not updated with this coefficient, highlighting a "route dependence" that weak-coupling data alone cannot resolve.

C. Strong-Coupling Residual Estimator

The authors define a finite-coupling estimator $\hat{S}_3(\lambda^*)$ to probe the unknown $O(\lambda^{-3})$ strong-coupling coefficient:
$$\hat{S}_3(\lambda^*) = \frac{4}{3}\lambda^{*3} \left[ f(\lambda^*) - \frac{3}{4}\left(1 + \frac{15\zeta(3)}{8}\lambda^{-3/2}_*\right) \right]$$
Evaluated at $\lambda^* = 20$:

• LSTP Route: $\hat{S}_3 \approx +2.14$
• HP Route: $\hat{S}_3 \approx -21.86$
• Significance: The two routes disagree not just in magnitude but in sign. This indicates that the current weak-coupling constraints are insufficient to determine the true sign of the asymptotic strong-coupling coefficient $S_3$.

4. Significance and Conclusion

• From Prediction to Selection: The exact $O(\lambda^{5/2})$ coefficient transforms the problem from a "prediction" (estimating the next term) to a "selection" problem. It proves that the LSTP family is far more rigid than previously thought; a single exact coefficient can eliminate the entire uncertainty band.
• Limitations of Weak-Coupling Data: The study demonstrates that even with exact higher-order weak-coupling terms, interpolation ambiguity persists if the interpolation ansatz (the mathematical structure of the method) differs between routes (LSTP vs. HP).
• Future Directions:
  1. The primary remaining uncertainty is the unknown $O(\lambda^{-3})$ strong-coupling coefficient. Computing this is the necessary next step to resolve the sign disagreement between the LSTP and HP routes.
  2. The HP ansatz itself requires rebuilding (extending the rational structure) to accommodate the $O(\lambda^{5/2})$ term without generating spurious artifacts, which would allow for a symmetric comparison between the two methods.

In summary, the paper establishes that while exact weak-coupling coefficients can uniquely determine a specific interpolation family (LSTP), they cannot yet resolve the structural differences between competing interpolation methods, necessitating further strong-coupling calculations.