Rigorous foundations of Wilson's $\epsilon$ expansion for $O(n)$ critical exponents: large-order behavior, Borel summability, and nonperturbative validity of the matching construction

Wilson's 1972 Feynman-graph expansion computes the critical exponents of the $O(n)$-symmetric $\varphi^4$ model as explicit power series in $\epsilon=4-d$, extracted through an order-by-order matching value $u_0(\epsilon)$ of the bare coupling; the series are asymptotic (alternating, factorially growing) for finite $n$ and convergent only at $n=\infty$. The practical use of these series — resummed three-dimensional exponents validated against conformal bootstrap and Monte Carlo — is settled at the level of cross-validated physics, but its mathematical foundations are not. The open core left by Wilson's paper and fifty years of subsequent literature comprises: the exact large-order asymptotics of the exponent series at finite $n$; their Borel summability on $0\le\epsilon\le 1$ — a prove-or-disprove question for these same series — with identification of the Borel sums with the exponents of the nonperturbatively constructed continuum theory, together with the open sub-question, posed by Guida and Zinn-Justin, of whether ultraviolet renormalons obstruct summability of massive-theory quantities such as the fixed-point coupling $g^*(\epsilon)$; the nonperturbative construction of the continuum Wilson-Fisher fixed-point theory in $4-\epsilon$ dimensions at finite $n$, together with the domain of bare Hamiltonians flowing to it, within which Wilson's matching value $u_0(\epsilon)$ is recovered; and, as a consequence, rigorous error bounds for the $\epsilon=1$ extrapolation. The question is explicitly stated as open by Guida and Zinn-Justin (1998) and Kompaniets and Panzer (2017), is bounded by rigorous results only for hierarchical, fixed-dimension, and large-$N$ relatives of the model, and the renormalon side remains under active investigation.

Background

Near a continuous phase transition, thermodynamic quantities and correlation functions of widely different systems — magnets, fluids, superfluids, polymer solutions — obey power laws whose exponents, the critical exponents (for example the susceptibility exponent $\gamma$, the correlation-length exponent $\nu$, the anomalous dimension $\eta$ governing the decay of the critical pair correlation function, and the leading correction-to-scaling exponent $\omega$), are universal: they depend only on gross features such as symmetry and space dimensionality $d$, not on microscopic details. A large class of such systems is described by the $O(n)$-symmetric $\varphi^4$ (Ginzburg-Landau-Wilson) Hamiltonian, a statistical field theory of an $n$-component order-parameter field $\varphi(x)$ with a quartic interaction $(\varphi^2)^2$; $n=1$ is the Ising universality class, $n=2$ the XY class of the superfluid helium transition, $n=3$ the Heisenberg class of isotropic magnets, and $n=0$ describes long self-avoiding polymer chains.

The renormalization group (RG) — the iteration of a coarse-graining transformation acting on Hamiltonians — explains universality: long-distance critical behavior is governed by fixed points of this transformation. For $d>4$ the Gaussian (free-field) fixed point governs the transition and mean-field exponents are exact. In his 1972 paper, Wilson showed that for $d=4-\epsilon$ with small $\epsilon$ a different, interacting fixed point — the Wilson-Fisher fixed point, introduced in the companion calculation of Wilson and Fisher — controls the transition, and that the critical exponents can be computed as an explicit expansion in powers of $\epsilon$. Wilson computed the exponents for general $n$ by direct Feynman-graph (diagrammatic perturbation theory) expansion of the cutoff theory, obtaining for example $\gamma = 1 + \frac{n+2}{2(n+8)}\epsilon + \frac{(n+2)(n^2+22n+52)}{4(n+8)^3}\epsilon^2 + O(\epsilon^3)$.

Two features of Wilson's construction matter for the open question at issue. First, the exponents are extracted through a matching procedure: correlation functions are computed as series in the bare (unrenormalized) coupling $u_0$, and an RG-transient argument (the iterates of the RG carry correction terms that die out at long distance) fixes a unique value $u_0 = u_0(\epsilon)$, itself a power series in $\epsilon$, for which the series match the required critical scaling; the exponents are then read off order by order. Second, the resulting $\epsilon$ series is asymptotic, not convergent: for the Ising case $n=1$ the coefficients through eighth order (computed with Wilson's approximate recursion formula, a recursion-relation approximation to the RG transformation that is exact to order $\epsilon$ but in error at order $\epsilon^2$) are $0.167, 0.041, -0.016, 0.077, -0.2, 0.67, -2.5, 10.3$ — alternating in sign from the third order onward, with factorially growing magnitude. Wilson argued that the exact diagrammatic series is likewise asymptotic for every finite $n$ (one cannot expect a convergent series when expanding a quartic term relative to a quadratic term inside an integral over the whole field range), while for $n=\infty$ (the exactly solvable spherical-model limit) the exact $\epsilon$ series converges.

An asymptotic series $\sum_k f_k \epsilon^k$ with $f_k \sim c(-a)^k k!$ carries, in favorable cases, a unique resummation: the Borel transform (dividing coefficients by $k!$) converges in a disk, extends analytically to the positive real axis, and the Laplace integral of the continuation defines the Borel sum, which may reproduce the physical quantity; the series is then called Borel summable. The obstruction is a singularity of the Borel transform on the positive axis, produced in $\varphi^4$ theories either by instantons (nontrivial saddle points of the Euclidean action that dominate high orders of perturbation theory, following Lipatov) or, in massless minimal-subtraction renormalization (a scheme in which divergences are removed by subtracting only the pole parts in $\epsilon$), possibly by ultraviolet renormalons (ambiguities of renormalized perturbation theory originating at large momentum). For the coupling series of the massive (nonzero-mass, away-from-criticality) theory at fixed integer dimensions $d=2$ and $d=3$, Borel summability is a theorem of constructive field theory (the rigorous construction of the Euclidean $\varphi^4$ measure and its perturbation series); for the $\epsilon$ expansion of the critical exponents no such theorem exists.

Setting $\epsilon=1$ (three dimensions) therefore requires summing a divergent series far from its asymptotic regime. Since the 1970s the standard tool has been the Borel-Leroy transformation (a Borel transform shifted to exploit the known leading coefficient growth) combined with a conformal mapping of the Borel plane (mapping the known singularity domain to a disk to accelerate the analytic continuation), with Padé, order-dependent-mapping, hypergeometric-Meijer, and variational alternatives. These resummations now use six- and seven-loop series and agree with conformal-bootstrap determinations (a numerical method imposing the consistency of operator product expansions and crossing symmetry on the critical correlation functions) and with Monte Carlo simulation at the $10^{-3}$–$10^{-2}$ level. Yet the mathematical status of the series itself — rigorous summability, exact large-order constants, nonperturbative uniqueness of the matching value $u_0(\epsilon)$, the domain of bare Hamiltonians flowing to the fixed point, and rigorous error bounds at $\epsilon=1$ — remains unsettled, and this is the open problem at issue.

Problem Statement

The problem is to give Wilson's $\epsilon$ expansion for the critical exponents of the $O(n)$-symmetric $\varphi^4$ model at finite $n$ — the series being those for general $n$, their coefficients rational functions of $n$, with the physically emphasized cases $n=0,1,2,3$ — a rigorous mathematical foundation. Its components are scientifically linked: the first three are successive layers of the same analytic structure of the expansion, and the fourth is their operational consequence, corresponding to the comparison Wilson's paper itself attempted (his Table I compares the $\epsilon$-series values at $\epsilon=1$ with the best three-dimensional estimates of his time):

(1) Exact large-order behavior. Determine the precise large-order asymptotics of the $\epsilon$ series of the critical exponents, and of the renormalization-group functions generating them, at finite $n$: the factorial growth constants, the sign pattern, and the subleading corrections.

(2) Borel summability. Prove, or disprove, that these series are Borel summable on $0\le\epsilon\le 1$ and that their Borel sums reproduce the critical exponents of the nonperturbatively constructed continuum $O(n)$ theory of component (3) at $\epsilon=1$ ($d=3$); whether the analytic continuation and the Laplace integral can be controlled jointly in $\epsilon$ over the interval, rather than separately at each fixed $\epsilon$, is a known obstacle rather than an assumed requirement. As an explicitly labelled open sub-question in the sense of Guida and Zinn-Justin, settle whether ultraviolet renormalons obstruct summability of quantities defined only in the massive renormalized theory (the theory away from criticality, with a nonzero mass scale), such as the fixed-point coupling $g^{*}(\epsilon)$ — in their formulation, the value of the renormalized coupling at the Wilson-Fisher fixed point defined through massive normalization conditions, to be distinguished from the minimal-subtraction fixed-point coupling of Kompaniets and Panzer, who use $D=4-2\varepsilon$ so that $d=3$ corresponds to $\varepsilon=1/2$. The massless exponent series remain the main target — Guida and Zinn-Justin conjecture them renormalon-free — but the $g^{*}(\epsilon)$ renormalon sub-question is also required: component (2) is settled only when both the exponent-series result and the $g^{*}(\epsilon)$ sub-question are resolved. Resolving the $g^{*}(\epsilon)$ sub-question alone does not complete this component for the exponent series.

(3) Nonperturbative construction. Construct nonperturbatively the continuum critical (Wilson-Fisher) theory in $d=4-\epsilon$ at finite $n$ whose perturbative expansion generates the $\epsilon$ series, and characterize the domain of bare ('initial') Hamiltonians that flow to this fixed point with the corresponding exponents, the task Wilson assigned to the renormalization-group approach. Wilson's matching value $u_0(\epsilon)$ — the bare coupling of the cutoff theory fixed order by order by his RG-transient argument — is recovered within this construction; a nonperturbative existence-and-uniqueness statement about $u_0(\epsilon)$ is a consequence to be formulated inside the constructed theory, not an independent formulation asserted to be equivalent to it.

(4) Controlled extrapolation. As a consequence of components (1)–(3), place the extraction of the three-dimensional critical exponents ($\epsilon=1$) on rigorous error bounds, replacing the current heuristic apparent-convergence estimates.

The objective is resolved when components (1)–(3) are settled; component (4) then follows as their corollary.

A rigorous negative result concerning these same target objects settles the corresponding component: for example, a proof that the $\epsilon$ series of the critical exponents, or of the renormalization-group functions generating them, fail to be Borel summable — with the resulting nonperturbative ambiguity quantified — equally settles component (2), whose question is posed as prove-or-disprove for these same series. A negative result about an adjacent observable outside the target series (for example about $g^{*}(\epsilon)$ alone) does not by itself complete the objective for the exponent series.

References

1. Kenneth G. Wilson, 'Feynman-Graph Expansion for Critical Exponents', Physical Review Letters 28, 548 (1972). DOI: 10.1103/PhysRevLett.28.548, https://doi.org/10.1103/PhysRevLett.28.548 (the source paper; the full text is paywalled)
2. Kenneth G. Wilson and Michael E. Fisher, 'Critical Exponents in 3.99 Dimensions', Physical Review Letters 28, 240 (1972). DOI: 10.1103/PhysRevLett.28.240, https://doi.org/10.1103/PhysRevLett.28.240
3. E. Brézin, J. C. Le Guillou and J. Zinn-Justin, 'Perturbation theory at large order. I. The phi^{2N} interaction', Physical Review D 15, 1544 (1977). DOI: 10.1103/PhysRevD.15.1544, https://doi.org/10.1103/PhysRevD.15.1544
4. E. Brézin, J. C. Le Guillou, J. Zinn-Justin and B. G. Nickel, 'Higher order contributions to critical exponents', Physics Letters A 44, 227 (1973). DOI: 10.1016/0375-9601(73)90894-3, https://doi.org/10.1016/0375-9601(73)90894-3
5. George A. Baker, Bernie G. Nickel, Melville S. Green and Daniel I. Meiron, 'Ising-Model Critical Indices in Three Dimensions from the Callan-Symanzik Equation', Physical Review Letters 36, 1351 (1976). DOI: 10.1103/PhysRevLett.36.1351, https://doi.org/10.1103/PhysRevLett.36.1351
6. J. C. Le Guillou and J. Zinn-Justin, 'Critical Exponents for the n-Vector Model in Three Dimensions from Field Theory', Physical Review Letters 39, 95 (1977). DOI: 10.1103/PhysRevLett.39.95, https://doi.org/10.1103/PhysRevLett.39.95
7. J. C. Le Guillou and J. Zinn-Justin, 'Critical exponents from field theory', Physical Review B 21, 3976 (1980). DOI: 10.1103/PhysRevB.21.3976, https://doi.org/10.1103/PhysRevB.21.3976
8. A. J. McKane, D. J. Wallace and O. F. de Alcantara Bonfim, 'Non-perturbative renormalisation using dimensional regularisation: applications to the epsilon expansion', Journal of Physics A: Mathematical and General 17, 1861 (1984). DOI: 10.1088/0305-4470/17/9/021, https://doi.org/10.1088/0305-4470/17/9/021
9. J. Magnen and R. Sénéor, 'Phase space cell expansion and Borel summability for the Euclidean phi^4_3 theory', Communications in Mathematical Physics 56, 237 (1977). DOI: 10.1007/BF01614211, https://doi.org/10.1007/BF01614211
10. P. Collet, J.-P. Eckmann and B. Hirsbrunner, 'A numerical test of Borel summability in the epsilon-expansion of the hierarchical model', Physics Letters B 71, 385 (1977). DOI: 10.1016/0370-2693(77)90246-5, https://doi.org/10.1016/0370-2693(77)90246-5
11. K. Gawędzki, A. Kupiainen and B. Tirozzi, 'Borel summability of the perturbation series in a hierarchical phi^4 model', Journal of Statistical Physics 36, 145 (1984). DOI: 10.1007/BF01015730, https://doi.org/10.1007/BF01015730
12. Riccardo Guida and Jean Zinn-Justin, 'Critical exponents of the N-vector model', Journal of Physics A: Mathematical and General 31, 8103 (1998). DOI: 10.1088/0305-4470/31/40/006, https://doi.org/10.1088/0305-4470/31/40/006; arXiv: cond-mat/9803240, https://arxiv.org/abs/cond-mat/9803240
13. Mikhail V. Kompaniets and Erik Panzer, 'Minimally subtracted six-loop renormalization of O(n)-symmetric phi^4 theory and critical exponents', Physical Review D 96, 036016 (2017). DOI: 10.1103/PhysRevD.96.036016, https://doi.org/10.1103/PhysRevD.96.036016; arXiv: 1705.06483, https://arxiv.org/abs/1705.06483
14. Filip Kos, David Poland, David Simmons-Duffin and Alessandro Vichi, 'Precision islands in the Ising and O(N) models', Journal of High Energy Physics 2016, 36 (2016). DOI: 10.1007/JHEP08(2016)036, https://doi.org/10.1007/JHEP08(2016)036
15. Claudio Bonanno, Andrea Cappelli, Mikhail Kompaniets, Satoshi Okuda and Kay Jörg Wiese, 'Benchmarking the Ising Universality Class in $3 \le d < 4$ dimensions', SciPost Physics 14, 135 (2023). DOI: 10.21468/SciPostPhys.14.5.135, https://doi.org/10.21468/SciPostPhys.14.5.135; arXiv: 2210.03051, https://arxiv.org/abs/2210.03051
16. Giacomo Sberveglieri, Marco Serone and Gabriele Spada, 'Renormalization Scheme Dependence, RG Flow and Borel Summability in phi^4 Theories in d<4', arXiv: 1905.02122 (2019), https://arxiv.org/abs/1905.02122
17. Léonard Ferdinand, Razvan Gurau, Carlos I. Perez-Sanchez and Fabien Vignes-Tourneret, 'Borel summability of the 1/N expansion in quartic O(N)-vector models', Annales Henri Poincaré 25, 2037 (2023). DOI: 10.1007/s00023-023-01350-w, https://doi.org/10.1007/s00023-023-01350-w; arXiv: 2209.09045, https://arxiv.org/abs/2209.09045
18. Harold Erbin, Vincent Lahoche and Mohamed Tamaazousti, 'Constructive expansion for quartic vector fields theories. I. Low dimensions', arXiv: 1904.05933 (2019), https://arxiv.org/abs/1904.05933
19. Nils Berglund and Tom Klose, 'Perturbation theory for the $\Phi^4_3$ measure, revisited with Hopf algebras', arXiv: 2207.08555 (2022), https://arxiv.org/abs/2207.08555
20. Oliver Schnetz, 'Seven loops $\phi^4$', arXiv: 2212.03663 (2022), https://arxiv.org/abs/2212.03663
21. Abouzeid M. Shalaby, 'Critical Exponents of the O(N)-symmetric $\phi^4$ Model from the $\varepsilon^7$ Hypergeometric-Meijer Resummation', arXiv: 2005.12714 (2020), https://arxiv.org/abs/2005.12714
22. Gerald V. Dunne and Max Meynig, 'Instantons or Renormalons? A Comment on phi4_4 Theory in the MS Scheme', arXiv: 2111.15554 (2021), https://arxiv.org/abs/2111.15554
23. Edouard Brezin, 'Should we worry about renormalons in the epsilon-expansion?', arXiv: 2301.01174 (2023), https://arxiv.org/abs/2301.01174