Fixed Points, Universality, and Continuum Limits

An RG fixed point is a scale-invariant solution of a flow in dimensionless theory space. Its neighborhood organizes far more than zeros of beta functions: the linearized eigenoperators identify which microscopic data must be tuned, nonlinear flow determines crossover and corrections to scaling, and the basin of attraction decides whether a fixed point can define an infrared universality class or an ultraviolet continuum limit.

This chapter develops that chain in one convention. It begins with stability matrices and scaling directions, derives critical exponents and their caveats, constructs critical surfaces, explains universal scaling functions, works the Gaussian-to-Wilson–Fisher benchmark, and ends with an evidence standard for ultraviolet and infrared fixed-point claims. The emphasis is on controlled statements: a beta-function zero is necessary data, not by itself a complete quantum field theory.

Enter through the physical question

If you want to know…	Start with…	The chapter’s exit condition
Whether a beta-function zero is attractive and in which direction	Fixed Points and Linearized RG Flow	A dimensionless fixed point, stability operator, eigenvalues or Jordan blocks, and declared RG-time orientation
Which perturbations are relevant, marginal, irrelevant, dangerous, or redundant	Relevant, Marginal, and Irrelevant Directions	A classification of eigenoperators, including nonlinear analysis of marginal directions
How stability data become $\nu$, $\eta$, $\beta$, $\gamma$, and other critical exponents	Critical Exponents, Scaling Relations, and Hyperscaling Caveats	Named exponents with sign conventions, scaling assumptions, and dimensional or dangerous-irrelevance caveats
How much tuning reaches a critical or continuum trajectory	Critical Surfaces, Crossover, and Corrections to Scaling	A local critical surface, tuning count, crossover scale, and leading correction exponent
Why different microscopic systems can share the same long-distance behavior	Universality Classes and Scaling Functions	Universal data separated from nonuniversal metric factors and a tested scaling-collapse window
How the standard interacting scalar fixed point emerges below four dimensions	Gaussian and Wilson–Fisher Fixed Points	The leading $O(N)$ fixed point, stability data, and an honest epsilon-expansion error boundary
Whether a proposed UV or IR zero supports a continuum-QFT claim	Ultraviolet and Infrared Fixed Points: Criteria and Evidence	A claim–evidence matrix covering scheme stability, operator consistency, accessibility, and independent checks

Readers arriving from perturbative running can begin with the first page after reviewing Multiple Couplings and Coupled RG Flows. Readers using exact or functional flows should also retain the truncation and identity checks from Wilsonian and Functional Renormalization.

Fixed-point data in one convention

Let $g^i(k)$ be dimensionless coordinates and retain the volume convention

$$t\equiv\ln\frac{k}{\Lambda_{\mathrm{UV}}}, \qquad \partial_t g^i=\beta^i(g).$$

Lowering $k$ moves toward the infrared and decreases $t$. A fixed point $g_\star$ satisfies

$$\beta^i(g_\star)=0.$$

For $\delta g^i=g^i-g_\star^i$, linearization gives

$$\partial_t\delta g^i = B^i{}_j\,\delta g^j + O(\delta g^2), \qquad B^i{}_j \equiv \left. \frac{\partial\beta^i}{\partial g^j} \right|_{g_\star}.$$

We define critical exponents $\theta_I$ by

$$B V_I=-\theta_I V_I.$$

For a diagonalizable real flow,

$$\delta g(t) = \sum_I C_I e^{-\theta_I(t-t_0)}V_I.$$

This sign convention makes $\theta_I>0$ a relevant direction: its coefficient grows as $k$ is lowered away from the fixed point. The same perturbation decays as $k\to\infty$, so it is ultraviolet-attractive. The terminology describes the operator’s scaling, while “UV-attractive” or “IR-attractive” describes a direction of motion.

$\theta_I$	Infrared motion as $k$ decreases	Ultraviolet motion as $k$ increases	Linear classification
$\theta_I>0$	Perturbation grows	Perturbation decays	Relevant
$\theta_I<0$	Perturbation decays	Perturbation grows	Irrelevant
$\theta_I=0$	Linear order is silent	Linear order is silent	Marginal; use nonlinear flow

Complex conjugate eigenvalues produce spiraling approach or departure. A non-diagonalizable stability matrix produces powers of $t$ multiplying exponentials. Neither case is captured by listing eigenvalues without eigenvectors or Jordan data.

For a discrete blocking transformation with scale factor $b>1$, an eigenoperator may instead carry a factor $b^{y_I}$. Matching one blocking step to continuous evolution identifies $y_I=\theta_I$ in the stated convention. Wilson and Kogut develop fixed points, linearized transformations, eigenoperators, and critical surfaces in Wilson and Kogut 1974, §§ 11–12, pp. 152–176.

From local flow to observable scaling

Near an infrared critical point, let $\tau$ be the temperature-like scaling field, $h$ the source conjugate to the order parameter, and $u_a$ irrelevant scaling fields. The singular free-energy density obeys the homogeneity law

$$f_{\mathrm{s}}(\tau,h,\{u_a\}) = b^{-d} f_{\mathrm{s}} \left( b^{y_t}\tau, b^{y_h}h, \{b^{y_a}u_a\} \right), \qquad b>1.$$

The leading eigenvalues give

$$\nu=\frac1{y_t}, \qquad y_h=\frac{d+2-\eta}{2}.$$

Derivatives of the homogeneity law generate scaling relations among the specific-heat, order-parameter, susceptibility, and equation-of-state exponents. These relations require more than a stability matrix: the free energy must have the assumed scaling form, the chosen fields must overlap the leading eigenoperators, and no dangerously irrelevant coupling may enter a denominator or set the ordered-phase amplitude.

Hyperscaling introduces the additional identification that one correlation volume supplies the singular free-energy scale,

$$f_{\mathrm{s}}\sim\xi^{-d}.$$

It can fail above an upper critical dimension, acquire logarithmic modifications at the critical dimension, or require an effective exponent when a dangerously irrelevant variable controls the observable. The dedicated exponent page makes those assumptions explicit rather than treating every algebraic relation as universal.

Critical surfaces and continuum limits

The phrase “critical surface” is used in two closely related orientations.

Problem	Limit being approached	Local surface	What must be tuned
Infrared critical phenomenon	$k\to0$	Stable manifold tangent to irrelevant directions, $\theta_I<0$	All relevant scaling fields that would drive the flow away
Ultraviolet continuum limit	$k\to\infty$	UV critical surface tangent to UV-attractive directions, $\theta_I>0$	Bare data must lie on the surface; its dimension counts free renormalized parameters locally

For an infrared critical point, the codimension of the stable manifold equals the number of relevant tunings. A small residual relevant component sets a crossover scale rather than reaching the fixed point exactly. Irrelevant components decay but leave corrections of order $L^{-\omega}$ or $k^\omega$, with $\omega=-\theta_{\mathrm{irr}}>0$ for the leading irrelevant direction.

For an ultraviolet fixed point, a trajectory on the UV critical surface approaches $g_\star$ as the cutoff is removed. A finite number of UV-attractive directions is the local predictivity criterion often invoked in asymptotic-safety discussions. It is not enough: the full trajectory must remain well defined, satisfy symmetry and unitarity requirements, and connect to the intended infrared theory.

Linearization controls only a neighborhood. Folded manifolds, nearby fixed points, walking regions, first-order boundaries, and strong-coupling singularities require the nonlinear flow. The chapter’s shared phase portrait keeps tangent-space statements visually distinct from global trajectory claims.

Universal data and coordinate data

A universality class is specified by more than spatial dimension and a familiar symmetry label. Relevant degrees of freedom, locality or interaction range, internal and spacetime symmetry, boundary conditions, conservation laws for dynamic questions, and the basin of attraction all matter.

At a controlled fixed point, the following distinctions are essential:

Data	Typical status
Critical exponents and operator scaling dimensions	Universal once the theory, sector, and normalization convention are fixed
Properly normalized scaling functions and amplitude ratios	Universal after nonuniversal metric factors are removed
Fixed-point coupling coordinates	Scheme, field-coordinate, and regulator dependent
Individual amplitudes and crossover scales	Usually nonuniversal
Number of physical relevant directions	Universal after redundant directions are quotiented
Truncated stability eigenvalues	Approximations requiring order, regulator, and projection checks

An analytic change of coupling coordinates maps the stability matrix by similarity at an exact fixed point, preserving its spectrum. A singular map, a finite truncation, or an inconsistent change of scheme can spoil that argument. The fixed-point evidence page therefore asks whether the zero and its physical data persist under controlled transformations, not whether raw coordinates remain numerically identical.

The chapter’s two threaded calculations

The first thread is a two-coupling polynomial flow. Each page reuses it for a different purpose: locating a nontrivial zero, diagonalizing the stability matrix, distinguishing marginal nonlinear flow, drawing critical and crossover trajectories, and testing how coordinate changes affect invariant data. The numbers are deliberately simple enough to check by hand.

The second thread is the $O(N)$ scalar theory in $d=4-\epsilon$. With

$$U(\rho) = m^2\rho+\frac{\lambda}{6}\rho^2, \qquad g\equiv\frac{\lambda}{8\pi^2},$$

the leading quartic beta function is

$$\beta_g = -\epsilon g + \frac{N+8}{6}g^2 + O(g^3).$$

It has the Gaussian fixed point $g_\star=0$ and the interacting solution

$$g_\star = \frac{6\epsilon}{N+8} + O(\epsilon^2).$$

The dedicated benchmark page derives the mass and field exponents consistently and states which claims are controlled only when $\epsilon\ll1$. Wilson and Fisher introduced the expansion below four dimensions in Wilson and Fisher 1972, pp. 240–243.

Evidence grows in layers

A credible fixed-point claim accumulates mutually distinct checks.

Layer	Question	What a pass establishes
Algebraic zero	Do the declared beta functions vanish?	A candidate fixed point of that finite system
Local stability	Is the stability spectrum stable under order, scheme, regulator, and projection changes?	Controlled local scaling data within the explored family
Operator consistency	Are symmetry identities, redundant directions, boundedness, and unitarity constraints satisfied?	Compatibility with necessary QFT structure
Trajectory accessibility	Does a basin or critical surface connect the candidate to admissible microscopic or infrared data?	More than an isolated formal zero
Independent evidence	Do perturbation theory, lattice, functional methods, bootstrap constraints, or rigorous results agree where their domains overlap?	Cross-method support with distinct systematics

No layer substitutes for a later one. A small critical exponent does not prove walking over a specified range; a stable beta zero does not prove reflection positivity; a finite truncation cannot establish a continuum limit by solver convergence alone.

Limits of fixed-point reasoning

Stop the linear analysis when the perturbation is no longer small compared with the fixed-point scale or when nonlinear terms compete with the smallest retained eigenvalue.

Stop the exponent claim when the sign convention, field normalization, redundant sector, or hyperscaling assumption is unresolved.

Stop the universality claim when metric factors have not been removed, the scaling window has not been demonstrated, or the proposed systems differ in relevant data.

Stop the continuum-QFT claim when the zero is lost under admissible scheme or truncation changes, the critical surface is inaccessible, a symmetry or unitarity constraint fails, or no observable beyond beta functions has been checked.

Boundaries and handoffs

This chapter develops fixed-point flow, scaling directions, critical exponents, critical surfaces, static universality, the leading Wilson–Fisher benchmark, and evidence classification. For conformal multiplets continue to Conformal Symmetry and Representations; for continuum extrapolation use Lattice and Hamiltonian Field Theory; for material universality classes use Many-Body and Quantum Matter; and for dynamic universality use Thermal and Nonequilibrium QFT.

The next chapter, Effective Field Theory: Construction and Power Counting, uses relevant and irrelevant operators for a different task: organizing a controlled expansion away from a fixed point or heavy threshold. RG relevance and EFT power counting inform one another but are not synonyms.

References

• Fisher, Michael E. “The Renormalization Group in the Theory of Critical Behavior.” Reviews of Modern Physics 46 (1974): 597–616. DOI
• Wegner, Franz J. “Corrections to Scaling Laws.” Physical Review B 5 (1972): 4529–4536. DOI
• Wilson, Kenneth G., and Michael E. Fisher. “Critical Exponents in 3.99 Dimensions.” Physical Review Letters 28 (1972): 240–243. DOI
• Wilson, Kenneth G., and John Kogut. “The Renormalization Group and the Epsilon Expansion.” Physics Reports 12 (1974): 75–200. DOI