Rigorous Renormalization Group and Continuum Control

A rigorous renormalization group is a sequence of maps on explicitly normed coordinates, together with an all-scale orbit theorem and a proof that the observables of interest converge. Its central achievement is therefore not a plotted flow or a formal beta function. It is control of the remainder left after each fluctuation field is integrated out, tuning of the finitely many unstable directions, contraction of the irrelevant coordinates, and recovery of correlation functions in a stated infinite-volume or continuum topology.

Helpful background. Wilsonian Coarse Graining and Theory Space supplies the physical interpretation of scale integration. Polchinski’s Exact RG Equation gives a complementary continuous-flow formulation. Regulators, Cutoffs, and Continuum Limits distinguishes cutoff stability, regulator removal, and construction of a continuum theory.

Enter this chapter

The representative setting is a scalar or supersymmetric lattice field on a finite torus of side $L^N$, with covariance split into positive finite-range pieces,

$$C=\sum_{j=1}^{N} C_j.$$

After scales $1,\ldots,j$ have been integrated, the density is written in local-plus-polymer coordinates

$$Z_j(\varphi)=e^{-u_j|\Lambda|}\,(I_j\circ K_j)(\Lambda), \qquad x_j=(V_j,K_j),$$

where $V_j$ contains the selected relevant and marginal local monomials and $K_j$ is the nonlocal remainder. The symbol $\circ$ denotes the circle product over polymers. A rigorous step is a map $x_{j+1}=R_j(x_j)$ between scale-dependent Banach spaces. Its theorem must specify the domains, norms, regulator weights, small-field assumptions, and terminal-scale modifications—not merely the first terms in a perturbative expansion.

The proof architecture has three logically distinct layers. First, a covariance decomposition localizes the fluctuation integration so that separated polymers factorize. Second, localization extracts the new local couplings and leaves a remainder that contracts in a weighted polymer norm. Third, the bare relevant parameters are chosen so that the orbit stays inside the small domain through the mass scale and beyond. Observable insertions then follow their own triangular flow. Only after uniform estimates and convergence are established may one state an infinite-volume limit, a scaling limit, or a universal critical asymptotic.

The dependency map makes those one-way implications explicit. Read the upper chain from the regulated measure toward observable control; the lower branch records that a stable trajectory is obtained by tuning initial data, not by assuming the perturbative orbit is exact.

Finite-range covariance pieces make disjoint fluctuation integrations independent. Localization then separates the finite-dimensional relevant and marginal coordinates from a polymer activity controlled by regulators and a scale-dependent norm. A nonlinear stability theorem tunes the relevant bare parameters and keeps the orbit in the small domain; an observable-sector flow plus terminal-scale estimates is still required to recover correlations. The last conclusion is model-specific: a logarithmic critical asymptotic, a Gaussian scaling limit, or a cutoff-limit bound requires the hypotheses of its own theorem. The diagram is schematic and not to scale. Structured description and source data (JSON)

This architecture is developed with explicit finite-range estimates, norms, maps, flow equations, and global stability in Bauerschmidt, Brydges, and Slade 2019, Chapters 3 and 5–10, pp. 37–50 and 65–168. The method proves considerably more than perturbation theory where its hypotheses are met, but its scope must remain model- and observable-specific. In particular, rigorous logarithmic corrections for four-dimensional weakly coupled lattice $|\varphi|^4$ models do not construct a non-Gaussian four-dimensional continuum QFT.

The chapter sequence

Read the pages in this order.

1. Rigorous RG as a dynamical system defines the scale-indexed Banach spaces, exact integration map, perturbative coordinate map, and contracting remainder.
2. Finite-range decomposition and multiscale integration explains how a lattice Green function is split into positive covariances whose finite range makes polymer factorization possible.
3. Polymer activities and normed RG coordinates constructs the local interaction, circle product, regulators, and weighted norms used to measure the nonperturbative coordinate.
4. Renormalized trajectories and counterterm tuning turns renormalization conditions into a boundary-value problem for the bare mass, vacuum energy, and other relevant parameters.
5. Stable manifolds and relevant–marginal control proves that a tuned nonlinear orbit shadows the approximate marginal flow while the irrelevant coordinate contracts.
6. Universality, critical manifolds, and observable control distinguishes stability of the bulk orbit from convergence of observable insertions and from universality across microscopic actions.
7. Critical four-dimensional φ⁴ and logarithmic corrections derives the role of the marginally irrelevant coupling and states the proved susceptibility and scaling-limit results without promoting them to an interacting continuum construction.
8. Weakly self-avoiding walk and supersymmetric RG relates the walk two-point function to a boson–fermion functional integral and follows the corresponding observable flow.
9. Fermionic multiscale RG and Fermi-surface problems replaces finite-range scalar blocks by sectorized Grassmann kernels and states the geometric and temperature restrictions in a rigorous Fermi-liquid theorem.
10. RG scheme comparison and continuum status separates coordinate changes, approximate conjugacy, regulator independence of selected limits, and the stronger claim that a continuum QFT has been constructed.

The order is deliberate. A fixed point has no theorem-level meaning until the space and map are defined; a local beta function does not control the orbit until the remainder estimate and tuned relevant directions are supplied; and bulk stability alone does not identify an observable limit.

Hypotheses and licensed conclusions

Normed RG objects, hypotheses, directional conclusions, and decisive failure tests

Object and domain	Required hypotheses	Licensed conclusion	Excluded converse or upgrade	Adversarial check
Lattice covariance $C=(-\Delta+m^2)^{-1}$ on a torus	A positive finite-range decomposition with uniform derivative bounds and a specified mass-scale regime	Sequential Gaussian integration with exact factorization for sufficiently separated polymers	A momentum-shell partition by itself does not give finite-range independence or positivity of every piece	Put two polymers farther apart than the nominal range and test whether their fluctuation covariance really vanishes
One RG step $R_j:(V_j,K_j)\mapsto(V_{j+1},K_{j+1})$	A scale-dependent domain, localization operator, field and large-set regulators, and differentiable norm estimates	An exact coordinate identity and a contractive bound for the extracted nonperturbative remainder	A truncated beta function neither defines the exact map nor bounds the discarded activity	Delete the large-field regulator and evaluate the norm on a high-amplitude field configuration
All-scale bulk trajectory	Tuned relevant initial coordinates, a controlled marginal recursion, stable-manifold hypotheses, and summable nonlinear errors	An orbit that remains in the small domain and approaches the designated infrared behavior	Forward iteration from generic bare data does not land on the critical manifold, and linear stability is not nonlinear stability	Perturb the bare mass in the unstable direction and check that the deviation grows before the terminal scale
Critical observable with insertions	A controlled bulk orbit, renormalized observable coordinates, coalescence-scale estimates, and convergence in a stated topology	A model-specific two-point asymptotic, susceptibility law, or Gaussian scaling limit	Bulk free-energy control does not automatically prove correlation-function convergence or universality	Insert two sources, follow their mixed coupling, and verify that the claimed amplitude survives the terminal-scale remainder
Weak four-dimensional lattice $|\varphi|^4$ model near criticality	Small positive coupling, dimension four, tuned critical mass, and the lattice and observable hypotheses of the cited theorem	Gaussian critical scaling with proved logarithmic corrections, including the $n$-dependent susceptibility exponent	These results do not yield a non-Gaussian four-dimensional continuum measure or settle every strong-coupling lattice action	Change the dimension or leave the small-coupling domain and test which summability and marginal-flow estimates fail
Supersymmetric walk or sectorized fermionic model	For the walk, an exact boson–fermion representation and supersymmetric localization; for a Fermi surface, regular geometry, sector bounds, and the theorem's temperature range	Walk critical asymptotics or a restricted weak-coupling Fermi-liquid construction with controlled kernels	Scalar polymer estimates do not transfer automatically to Grassmann sectors, nesting, van Hove points, or zero temperature	Move the Fermi level to a singular or nested surface, or break the supersymmetric identity, and locate the first lost estimate
Two regulators or RG coordinate systems	A common microscopic model, explicit maps between domains, uniform error bounds, and separately proved convergence of the same observables	A conditional comparison of selected universal quantities or equivalent limiting correlations	Matching low-order beta coefficients does not prove conjugacy of exact maps, regulator independence, or existence of a continuum QFT	Add an irrelevant term whose accumulated error is not summable and test whether the proposed comparison still closes

Structured table data (JSON) preserves the caption, scoped headers, rows, and reading order.

The table separates three claims that are often collapsed. Step control is a local statement about one exact integration map in one normed domain. Trajectory control is a global statement about a specially tuned orbit. Continuum or critical control is a limit theorem for named observables in a named topology. Each implication needs new hypotheses, and none runs backward without a separate theorem.

The failure map below starts from four valid implication chains and removes one decisive assumption from each. Inspect where the solid conclusion stops: a covariance can remain positive without being finite range, a perturbative local flow can remain correct while the polymer norm diverges, a bulk trajectory can remain controlled while an insertion is not, and two schemes can agree to a fixed perturbative order without defining the same nonperturbative limit.

Each dashed branch identifies the first unavailable inference. Long-range covariance pieces spoil exact separation of distant polymer activities; inadequate field or large-set regulators make the nonperturbative norm fail; an untuned mass component grows along the relevant direction; missing observable or terminal-scale estimates leave correlation limits unproved; singular Fermi-surface geometry invalidates sector power counting; and finite-order agreement between schemes does not control accumulated irrelevant errors. The diagram is schematic and not to scale. Structured description and source data (JSON)

What the theorems do—and do not—construct

For weak four-dimensional lattice $|\varphi|^4$ models and weakly self-avoiding walk, the cited results determine critical asymptotics with explicit logarithmic corrections and, for specified fields and test-function topologies, Gaussian scaling limits. Aizenman and Duminil-Copin prove Gaussianity of scaling limits for critical four-dimensional Ising-type and lattice-cutoff $\lambda\phi^4$ models under their stated hypotheses; the result strengthens the triviality picture but is not a construction of an interacting four-dimensional continuum theory Aizenman and Duminil-Copin 2021, Theorem 1.2 and § 1, pp. 163–177.

Other rigorous RG settings have different endpoints. A trajectory for a hierarchical or modified covariance does not automatically transfer to the nearest-neighbor Euclidean model. Ultraviolet stability of a finite-volume scalar measure is not yet Osterwalder–Schrader reconstruction. A weak-coupling Fermi-liquid theorem above an exponentially small temperature does not cover a superconducting instability or the zero-temperature limit. Every page therefore pairs its positive theorem with the smallest adversarial change that breaks the proof.

Functional and exact RG equations remain useful analytical languages, and numerical integrations can reveal candidate fixed points and crossover scales. They become rigorous construction results only after the chosen truncation, normed remainder, regulator dependence, trajectory, and observable limit have been controlled. This chapter supplies the criteria for recognizing when that upgrade has actually occurred.

Review the chapter

For any proposed rigorous RG result, ask the following questions in order.

1. What is the regulated finite-volume or ultraviolet-cutoff object, and is it a measure, Grassmann functional, generating functional, or correlation family?
2. How is the covariance decomposed, and which range, positivity, derivative, and mass-scale estimates are uniform?
3. What are the local coordinates, polymers, regulators, Banach spaces, and norms at scale $j$?
4. Is the RG transformation exact on its stated domain, and what localization produces the perturbative coordinate map?
5. Which term contracts the nonperturbative activity, and are the nonlinear errors summable through all scales?
6. Which bare parameters are tuned, what renormalization condition selects them, and what prevents the orbit from leaving the domain?
7. How are observable insertions renormalized, and what happens at their coalescence and terminal scales?
8. What limit is proved—infinite volume, ultraviolet removal, critical scaling, or a combination—and in which topology?
9. Which microscopic changes are included in the universality statement, and which remain outside its hypotheses?
10. What independent check distinguishes a theorem about the exact map from a calculation of a truncated flow?

Synthesis exercise

A computation finds the same two-loop beta function for two smooth cutoff profiles. In one profile, numerical forward integration approaches the Gaussian fixed point. The bare mass has not been tuned, no polymer or kernel norm is supplied, and no correlation function is shown to converge. Which conclusions are justified?

Solution

The computation supports finite-order perturbative agreement of the two beta functions and may identify a plausible weak-coupling flow. It does not prove that either exact RG transformation exists on a common Banach domain, that their remainders are uniformly small, or that the two maps are conjugate. Without tuning the relevant mass direction, the numerical orbit is not evidence for a critical trajectory. Without all-scale norm estimates and an observable-sector limit, there is no theorem of universality, regulator independence, or continuum construction. The missing work is precisely the covariance and norm control, stable-manifold tuning, observable flow, and convergence statement developed in the chapter.

References

• Aizenman, Michael, and Hugo Duminil-Copin. “Marginal Triviality of the Scaling Limits of Critical 4D Ising and $\phi_4^4$ Models.” Annals of Mathematics 194 (2021): 163–235. DOI; Open PDF. See also the corrigendum.
• Bauerschmidt, Roland, David C. Brydges, and Gordon Slade. “Scaling Limits and Critical Behaviour of the 4-Dimensional $n$-Component $|\varphi|^4$ Spin Model.” Journal of Statistical Physics 157 (2014): 692–742. DOI; Open PDF.
• Bauerschmidt, Roland, David C. Brydges, and Gordon Slade. “Critical Two-Point Function of the 4-Dimensional Weakly Self-Avoiding Walk.” Communications in Mathematical Physics 338 (2015): 169–193. DOI; Open PDF.
• Bauerschmidt, Roland, David C. Brydges, and Gordon Slade. Introduction to a Renormalisation Group Method. Lecture Notes in Mathematics 2242. Singapore: Springer, 2019. DOI; Open PDF.