A summation rule for the four-coupling beta function of four-dimensional nonplanar Euclidean scalar $\varphi^{4}$ theory under factorial coefficient bounds, and a domain on which the summed beta function is well defined

Technical Summary: A Summation Rule for the Four-Coupling Beta Function in Nonplanar $\phi^4$ Theory

Problem Statement
The paper addresses the mathematical challenge of assigning a well-defined real number to the formal beta-function series of four-dimensional nonplanar Euclidean scalar $\phi^4$ theory. The theory involves four running couplings, $c = (\lambda, \alpha, \mu, \nu) \in \mathbb{R}^4$. The beta function $B(c)$ is given as a formal power series:
$$B(c) = \sum_{p=2}^{\infty} \sum_{|m|=p} \beta(m) c^m$$
The coefficients $\beta(m)$ are assumed to satisfy Hypothesis H, a factorial bound of the form $|\beta(m)| \leq (p-1)! C^{p-1}$ for $|m|=p \geq 2$. This bound reflects the typical growth of coefficients in constructive multiscale or loop-vertex expansions.

The core difficulty is that under Hypothesis H, the series $B(c)$ diverges everywhere except at the origin (Theorem 9.1). Furthermore, the Borel transform of the series may possess a natural boundary at the circle of convergence (Theorem 9.2), rendering standard summation techniques (Borel, lateral, median, accelero-summation) undefined for the entire admissible class of coefficients. The paper seeks a summation rule that is:

1. Well-defined: Produces a finite real number on an explicit domain.
2. Faithful: Asymptotic to the formal series to all orders (Gevrey-1) as $c \to 0$.
3. Regular (in a limited sense): Linear in coefficients and smooth ($C^\infty$) in couplings.
4. Optimal: Approaches the intrinsic non-perturbative ambiguity limit.

Methodology
The author constructs an explicit summation rule $S$ based on a cut Borel-Laplace transform. The construction involves three main components:

1. The Borel Transform: Defined as $\hat{B}(z) = \sum_{p \geq 2} \frac{b_p(z)}{(p-1)!}$, where $b_p(c)$ are the homogeneous parts of degree $p$. Under Hypothesis H, $\hat{B}(z)$ is proven to be holomorphic on a sharp polydisc $\Pi_C = \{z \in \mathbb{C}^4 : \|z\|_\infty < 1/C\}$.
2. The Cut: Instead of integrating along the full ray $[0, \infty)$, which may exit the domain of holomorphy, the integration is truncated at a height $T(c)$. This cut is placed using a real-analytic, positively homogeneous gauge $\nu_k(c) = (\sum c_i^{2k})^{1/(2k)}$ (for $k \geq 1$). The cut height is defined as $T(c) = \frac{\vartheta}{C \nu_k(c)}$, where $\vartheta \in (0, 1)$ is a parameter.
3. The Summation Rule: The summed beta function is defined as:
   $$S_B(c) = \int_0^{T(c)} e^{-\tau} \hat{B}(\tau c) \frac{d\tau}{\tau} \quad (c \neq 0), \quad S_B(0) = 0.$$
   Term-by-term integration reveals that $S_B(c)$ is equivalent to a Poisson-averaged truncation of the divergent series:
   $$S_B(c) = \sum_{p \geq 2} \Pr[\text{Poi}(T(c)) \geq p] b_p(c) = \mathbb{E}\left[ \sum_{p=2}^K b_p(c) \right], \quad K \sim \text{Poi}(T(c)).$$
   This formulation ensures the series is absolutely convergent, linear in coefficients, and smooth in $c$.

Key Contributions and Results

• Existence and Regularity: The rule $S_B$ is well-defined on an explicit nonempty open domain $D = \{c : C\nu_k(c) < \vartheta\}$ (and actually on all of $\mathbb{R}^4$). The resulting function is $C^\infty$ on $\mathbb{R}^4$ and real-analytic on $\mathbb{R}^4 \setminus \{0\}$. Its Taylor expansion at the origin exactly recovers the formal series $B$ (Theorem 1.2(iv)).
• Faithfulness and Asymptotics: $S_B$ is Gevrey-1 asymptotic to $B$ to all orders. Specifically, the remainder after truncating at order $N$ is bounded by terms involving $N! x^{N+1}$ and an exponentially small term $e^{-T(c)}$. The one-loop coefficient is reproduced exactly in the asymptotic sense (Theorem 1.2(ii)).
• Optimal Truncation: At the optimal truncation order $N^*(c) \approx T(c)$, the remainder is bounded by $O(e^{-\vartheta_k / (C\|c\|_\infty)})$, where $\vartheta_k$ can be made arbitrarily close to 1 by increasing $k$ (Theorem 1.2(iii)).
• Consistency: If the classical Borel sum or the ordinary sum of a convergent series exists, $S_B$ reproduces it up to an error of order $O(e^{-(1-\sigma)T(c)})$, which is exponentially small (Theorem 1.2(v)).
• Impossibility of Exact Regularity: The paper proves that no linear summation rule can be both well-defined on the entire admissible class and "regular" (i.e., exact on convergent series). Regularity requires weights equal to 1, which leads to divergence; well-definedness requires weights decaying factorially, which introduces an error on convergent series (Theorem 9.6). Thus, the constructed rule pays the minimal necessary price: an exponentially small error on convergent inputs.

Optimality and Intrinsic Ambiguity
The paper establishes that the construction is essentially optimal relative to the information provided by Hypothesis H:

• Divergence: There exist admissible coefficient families for which the series diverges at every $c \neq 0$ (Theorem 9.1).
• Natural Boundaries: There exist admissible families where the Borel transform has a natural boundary at $|\tau| = 1/(C\|c\|_\infty)$, making any continuation-based summation impossible (Theorem 9.2).
• Intrinsic Ambiguity: Two functions satisfying the Gevrey-1 bounds can differ by exactly $(2\pi/C)e^{-1/(C\|c\|_\infty)}$ (Theorem 9.11). This represents the "renormalon" ambiguity associated with a simple pole in the Borel plane.
• Rate Optimality: No summation rule can achieve an accuracy better than $o(e^{-1/(C\|c\|_\infty)})$. The family of rules $S_{k, \vartheta}$ realizes every exponent $\eta < 1$, approaching the barrier $\eta = 1$ arbitrarily closely (Theorem 1.2(vi)).

Significance
The paper provides a rigorous, self-contained solution to the problem of summing the beta function for nonplanar $\phi^4$ theory under factorial coefficient bounds. It demonstrates that while the series is divergent and standard summation methods fail due to natural boundaries, a specific cut Borel-Laplace transform yields a well-defined, smooth, and asymptotic function.

Crucially, the work clarifies the trade-offs inherent in summation theory for such models: one cannot have a rule that is simultaneously well-defined on the whole class, regular (exact on convergent series), and free of non-perturbative ambiguity. The constructed rule achieves the best possible performance by being smooth and asymptotic while accepting an exponentially small error that matches the intrinsic non-perturbative ambiguity of the problem. The results are applied to the renormalization-group flow, showing that the summed flow is well-posed, confined within the perturbative region, and free of fixed points where the one-loop term dominates.