Ultraviolet and Infrared Fixed Points: Criteria and Evidence

A zero of beta functions is necessary for a fixed point, but it is not yet a continuum quantum field theory. One must determine the direction of attraction, show that the zero survives the approximation used to find it, identify a finite physical parameter set, satisfy symmetry and operator-consistency conditions, and exhibit accessible trajectories. This page distinguishes asymptotic freedom, interacting infrared fixed points, walking, and asymptotic safety, then evaluates a supplied gauge–Yukawa candidate without promoting missing evidence into a verdict.

Evidence date. Model-specific literature statements on this page were checked through 3 August 2026. They are restricted to the cited theory, loop order, limit, and assumptions.

Required background. Fixed Points and Linearized RG Flow and Relevant, Marginal, and Irrelevant Directions fix the stability convention. Scheme Transformations and RG Invariants separates coordinate changes from physical data. Helpful background. ‘t Hooft Anomaly Matching supplies a nonperturbative consistency condition for proposed infrared realizations.

Four distinct ultraviolet and infrared behaviors

Use $t=\ln(k/\Lambda)$, so $t$ increases toward the ultraviolet. The following labels describe different mathematical structures.

Asymptotic freedom

For one loop-normalized coupling $a$,

$$\beta(a) =-b_0a^2+O(a^3), \qquad b_0>0,$$

the Gaussian point is ultraviolet-attractive and

$$a(k) \simeq \frac{1}{b_0\ln(k/\Lambda_{\mathrm{dyn}})}.$$

The ultraviolet theory becomes free logarithmically. In a multi-coupling theory, complete asymptotic freedom requires every essential coupling to approach a compatible Gaussian trajectory; one asymptotically free gauge coupling does not guarantee that Yukawa and scalar sectors remain finite. Gross, Wilczek, and Politzer established the non-Abelian gauge mechanism in Gross and Wilczek 1973, pp. 1343–1346 and Politzer 1973, pp. 1346–1349.

Interacting infrared fixed point

At two loops, a schematic gauge beta function is

$$\beta(a) =-b_0a^2-b_1a^3+O(a^4).$$

If $b_0>0$, $b_1<0$, and $a_\star=-b_0/b_1\ll1$, then the nonzero zero is perturbative. Its derivative is

$$B=\beta'(a_\star) =-\frac{b_0^2}{b_1}>0, \qquad \theta=-B<0,$$

so it is attractive toward the infrared in that direction. The Banks–Zaks construction obtains such a controlled family when the one-loop coefficient is parametrically small Banks and Zaks 1982, pp. 189–204. A mass term or another relevant deformation can still drive the theory away; the conformal limit requires those fields to be tuned or forbidden.

Walking without a fixed point

A beta function can remain small over a long interval without vanishing. Near the annihilation of two fixed points, a normal form is

$$\beta(g) =-\delta-(g-g_c)^2, \qquad \delta>0.$$

There is no real zero, but the RG time spent crossing the slow region is

$$|\Delta t| \simeq \int_{-\infty}^{\infty} \frac{dx}{\delta+x^2} =\frac{\pi}{\sqrt\delta}.$$

The associated scale ratio can be exponentially large, $\Lambda_{\mathrm{UV}}/\Lambda_{\mathrm{IR}}\sim e^{\pi/\sqrt\delta}$. Nearly constant couplings and approximate scaling over that interval are walking, not proof of an exact conformal theory. Kaplan and collaborators discuss this fixed-point-loss mechanism in Kaplan et al. 2009, §§ II–III.

Asymptotic safety

An asymptotically safe theory approaches a non-Gaussian fixed point as $k\to\infty$. If

$$B V_I=-\theta_I V_I,$$

then directions with $\theta_I>0$ are UV-attractive. Their span is the local UV critical surface. A finite number of physical positive-$\theta$ directions gives a finite local parameter set; all UV-repulsive directions must be fixed as functions of those parameters. This local condition is necessary, not sufficient: the trajectory must exist globally and the fixed point must define a consistent QFT.

Behavior	UV limit	IR statement	Evidence that distinguishes it
Asymptotic freedom	Gaussian fixed point	strong coupling, mass gap, or another IR endpoint may occur	universal leading coefficient and complete compatible trajectory for all essential couplings
Interacting IR fixed point	often reached from a Gaussian UV theory	non-Gaussian zero attractive after relevant tunings	controlled zero, negative-$\theta$ IR directions, operator consistency, and basin accessibility
Walking	no exact zero required	slow finite-range evolution followed by departure	resolved small beta function, finite crossover scale, and evidence excluding an exact zero
Asymptotic safety	interacting UV fixed point	trajectory must connect to an acceptable low-energy theory	finite UV critical surface, physical spectrum, global trajectory, and independent checks

From a beta-function zero to a QFT claim

A review should proceed in the following order.

1. Complete coordinates. List every gauge, Yukawa, scalar, mass, topological, and other essential coupling allowed by the symmetries at the stated accuracy. A zero in a projected subsystem is not a common zero of the theory.
2. Controlled hierarchy. Identify the small parameter—weak coupling, $\epsilon$, $1/N$, or another expansion—and show that every retained fixed-point coordinate lies inside it. Use a loop ordering consistent with how sectors first feed into one another.
3. Order stability. Recompute coordinates, stability exponents, anomalous dimensions, and vacuum conditions at successive complete orders. Small coordinate drift alone is insufficient if a physical exponent or stability sign is unstable.
4. Physicality. Require real couplings, a bounded scalar potential, acceptable kinetic terms, gauge and BRST consistency, anomaly cancellation, and unitarity bounds for gauge-invariant operators.
5. Physical stability spectrum. Diagonalize the full mixing problem, remove redundant directions, and state the UV or IR orientation. A finite-dimensional truncation cannot establish finiteness of the exact UV critical surface without convergence evidence.
6. Scheme and regulator tests. Under an analytic invertible coupling map, an exact fixed point maps to a fixed point and $B$ changes by similarity. Disappearance under mild admissible reparametrizations signals that truncation errors are comparable to the claimed effect.
7. Global accessibility. Integrate trajectories far enough to show that the fixed point connects to the intended microscopic or infrared theory without encountering an instability, singularity, or unintended phase.
8. Independent observables. Compare operator dimensions, central quantities, amplitude ratios, step-scaling functions, or other invariants with a method whose assumptions and systematics differ.

For perturbative gauge–Yukawa systems, general analyses show why a Yukawa nullcline can change the effective higher-order gauge coefficient and permit weakly coupled UV zeros, while scalar vacuum stability remains a separate condition Bond and Litim 2017, §§ 7–10. General four-loop gauge and three-loop Yukawa beta functions provide higher-order inputs, not an automatic fixed-point verdict Bednyakov and Pikelner 2021.

Worked review of a supplied gauge–Yukawa candidate

Let $\alpha_g$, $\alpha_y$, and $\alpha_u$ be nonnegative loop-normalized gauge, Yukawa, and scalar quartic coordinates. Consider the supplied leading system

$$\begin{aligned} \beta_g &=\alpha_g^2 \left( \varepsilon+6\alpha_g-4\alpha_y \right),\\ \beta_y &=\alpha_y \left( 3\alpha_y-5\alpha_g \right),\\ \beta_u &=4\alpha_u^2 +2\alpha_y\alpha_u -\alpha_y^2, \end{aligned}$$

with a small positive control parameter $\varepsilon$. The signs mimic an infrared-free gauge sector balanced by Yukawa effects, but the coefficients are synthetic and do not specify a microscopic representation.

The interacting Yukawa nullcline gives

$$\alpha_y=\frac53\alpha_g.$$

The gauge equation then fixes

$$\alpha_{g\star}=\frac32\varepsilon, \qquad \alpha_{y\star}=\frac52\varepsilon.$$

The quartic equation has two roots. The positive-potential branch is

$$\boxed{ \alpha_{u\star} =\frac{5(\sqrt5-1)}{8}\varepsilon },$$

while the negative branch fails this simplified stability test.

At the positive fixed point, the stability matrix is

$$B = \begin{pmatrix} \dfrac{27}{2}\varepsilon^2 & -9\varepsilon^2 & 0\\ -\dfrac{25}{2}\varepsilon & \dfrac{15}{2}\varepsilon & 0\\ 0 & \dfrac{5(\sqrt5-5)}{4}\varepsilon & 5\sqrt5\,\varepsilon \end{pmatrix}.$$

Its critical exponents through the first nonzero orders are

$$\begin{aligned} \theta_1 &=\frac32\varepsilon^2+O(\varepsilon^3),\\ \theta_2 &=-\frac{15}{2}\varepsilon+O(\varepsilon^2),\\ \theta_3 &=-5\sqrt5\,\varepsilon. \end{aligned}$$

There is one UV-attractive direction and two UV-repulsive directions, so the supplied truncation has a one-dimensional local UV critical surface. The small $O(\varepsilon^2)$ positive exponent implies especially slow departure along the relevant trajectory.

Take $\varepsilon=0.05$. To test order sensitivity, compare the leading system with the supplied higher-order completion

$$\begin{aligned} \Delta\beta_g&=-\frac12\alpha_g^4,\\ \Delta\beta_y&=-\frac15\alpha_y\alpha_g^2,\\ \Delta\beta_u&=+\frac12\alpha_y^3. \end{aligned}$$

Solving both systems and diagonalizing their full stability matrices gives:

Quantity	Leading system	Supplied next order	Interpretation
$\alpha_{g\star}$	$0.07500$	$0.06945$	$7.4\%$ coordinate shift
$\alpha_{y\star}$	$0.12500$	$0.11608$	$7.1\%$ coordinate shift
$\alpha_{u\star}$	$0.03863$	$0.03435$	positive branch persists; $11.1\%$ shift
relevant $\theta_1$	$+0.003412$	$+0.003415$	sign and small magnitude stable in this test
irrelevant $\theta_2$	$-0.41216$	$-0.38025$	remains UV-repulsive; visible order drift
irrelevant $\theta_3$	$-0.55902$	$-0.50691$	remains UV-repulsive; visible order drift

This passes a limited algebraic test: the zero is weakly coupled, the positive quartic branch persists, the tuning count is unchanged, and the supplied order variation is moderate. It does not pass the full QFT checks. No gauge group or representations were supplied, so gauge anomalies, global anomalies, the actual scalar stability cone, gauge-invariant operator dimensions, and higher-loop coefficients cannot be checked. No independent lattice, bootstrap, functional, or constructive evidence was supplied, and no trajectory to a target infrared theory was demonstrated.

The correct status is therefore controlled fixed-point candidate within the supplied polynomial system; continuum-QFT claim unresolved. Representation and anomaly checks belong with ‘t Hooft Anomaly Matching; model-specific running and conformal-window claims belong with Running and Dynamical Scales in Gauge Theories; operator-spectrum comparisons belong with Controlled Families of Interacting CFTs.

Scheme transformations and false zeros

For an analytic invertible map $g'^a=f^a(g)$,

$$\beta'^a=\frac{\partial f^a}{\partial g^i}\beta^i.$$

An exact common zero maps to a common zero, and its stability spectrum is invariant. Fixed-point coordinates are not. A finite polynomial truncation breaks this exact statement because transforming and then discarding higher-order terms is not the same as transforming the full beta function.

The danger is clearest for

$$\beta(g)=b_1g^2+b_2g^3+b_3g^4+\cdots.$$

The two-term zero $g_\star=-b_1/b_2$ is controlled when $b_1$ is parametrically small, $b_2=O(1)$, and $b_3g_\star^2$ is demonstrably subleading. If $g_\star=O(1)$, a natural $b_3$ term can move or remove it. Stability across a few arbitrary schemes cannot prove existence, but strong instability under mild admissible schemes can disprove the claimed perturbative control.

Observable comparisons should therefore emphasize critical exponents, gauge-invariant operator dimensions, central quantities, and physical amplitude or step-scaling data. Raw $\overline{\mathrm{MS}}$ coordinates are useful reproducibility data, not universal evidence.

Fixed-point evidence matrix

The same evidence route can be decisive for one part of a claim and silent about another. Read each row together with its assumptions and ceiling; independence comes from unlike systematics.

Evidence route	Essential assumptions and control	Direct outputs or observables	Dominant systematics and evidence ceiling
Perturbative expansion	A small $\epsilon$, weak fixed-point coupling, large-$N$ parameter, or other declared expansion; specified renormalization scheme and operator sector	Beta-function zeros, stability eigenvalues, anomalous dimensions, and resummed exponent or amplitude-ratio estimates	Missing orders, asymptotic-series resummation, scheme and operator truncation; controlled local evidence within the expansion domain, not a global existence proof at order-one parameters
Lattice finite-size scaling	A Euclidean discretization in the target basin; controlled critical, continuum, and infinite-volume limits; reflection positivity when used	Correlation lengths, step scaling, Binder-type ratios, spectra, exponents, amplitude ratios, and scaling functions	Cutoff and volume extrapolation, critical tuning, autocorrelation, action dependence, and analytic continuation; strong nonperturbative IR evidence for the simulated universality class
Functional RG	An exact flow equation combined with a declared ansatz, projection, regulator, identity constraints, and convergence tests	Global flow portraits, effective potentials, fixed-point spectra, equations of state, and crossover trajectories	Truncation, projection, regulator dependence, symmetry identities, convexity, and numerics; quantitative candidate evidence unless convergence is independently controlled
Conformal bootstrap	Conformal invariance, crossing, a symmetry sector, unitarity or reflection positivity when imposed, and explicit gap assumptions	Allowed or excluded regions for operator dimensions and OPE coefficients; islands and universal CFT data	Derivative and spin truncations, assumed gaps, navigator or optimization choices, and numerical certification; characterizes or excludes a CFT under stated assumptions but does not supply an RG trajectory
Rigorous or constructive analysis	A precise lattice or continuum model, norm, positivity domain, and theorem hypotheses, often in restricted dimensions or coupling ranges	Existence or nonexistence, controlled continuum correlations, bounds, and in some cases complete RG trajectories	Transfer is limited by theorem hypotheses and model class; strongest conclusion inside the proved domain, with no automatic extension to nearby physical theories

Representative primary analyses illustrate the distinct ceilings: the epsilon expansion constructs a perturbative fixed point Wilson and Fisher 1972, pp. 240–243; finite-size lattice scaling controls volume and correction terms Hasenbusch 2010, §§ II–V; effective-average-action studies expose truncation and regulator choices Berges, Tetradis, and Wetterich 2002, §§ 2–3, pp. 245–287; bootstrap bounds assume crossing and unitarity El-Showk et al. 2012, §§ II–IV; and rigorous construction can establish a complete trajectory for a precisely defined modified model Abdesselam 2007, pp. 727–772.

No single row should be silently upgraded into all the others. A bootstrap island does not construct an RG trajectory; a stable perturbative zero does not supply nonperturbative existence; a functional fixed point does not establish truncation convergence; and a rigorous theorem does not automatically transfer beyond its hypotheses.

Dated perturbative benchmark and evidence levels

For one specific four-dimensional $SU(N_c)$ gauge–Yukawa model in a controlled Veneziano expansion, a 2023 calculation used four-loop gauge plus three-loop Yukawa and quartic beta functions to study the UV zero, conformal window, unitarity, and connecting trajectories Litim et al. 2023, §§ II–V. An independent 2024 analysis of the same model and loop hierarchy found a weakly interacting UV fixed point with one relevant direction and included finite-$N_c$ effects and operator mixing Bednyakov and Mukhaeva 2024, §§ III–V.

As reviewed through 3 August 2026, those works justify the label high-order perturbatively supported for that declared model and expansion window. They do not establish that arbitrary gauge–Yukawa theories are asymptotically safe, determine a universal conformal-window boundary outside the analyzed limits, or settle gravity-assisted fixed points. Those broader claims require their own dated evidence in the relevant subject volumes.

Useful status language is:

Status	Minimum meaning
Controlled perturbative fixed point	explicit small parameter, stable complete orders, physical spectrum and potential within the perturbative domain
Nonperturbatively supported fixed point	compatible invariant data from controlled nonperturbative methods, with stated continuum and truncation limits
Fixed-point candidate	a zero and local spectrum exist in a declared approximation, but one or more decisive checks remain open
Walking regime	slow finite-range running is supported, without claiming an exact zero
Excluded under stated assumptions	a theorem, bound, or controlled calculation rules out the claim inside a precisely stated domain
Unresolved	existing evidence does not control the relevant systematics or disagrees without a resolved cause

These labels attach conditions and dates to the scientific statement; they are not editorial completion states.

Common pitfalls

Calling any beta-function zero asymptotic safety. A UV fixed point also needs a finite physical UV critical surface, operator consistency, and a global trajectory.

Confusing walking with conformality. A small beta function over many decades can arise from nearby complex or annihilated fixed points. Only an exact common zero supports scale invariance at arbitrarily long distances.

Testing only coupling coordinates across loops. Coordinates are scheme dependent. Stability exponents, anomalous dimensions, potential stability, and physical observables must also remain controlled.

Transferring a model-specific result. Loop stability in one gauge group, representation, large-$N$ limit, or regulator truncation does not establish the same fixed point in another theory.

Exercises

1. Classify a two-loop gauge zero

For $\beta(a)=-b_0a^2-b_1a^3$ with $b_0>0$ and $b_1<0$, find the nonzero zero and classify its gauge direction.

Solution

The zero is $a_\star=-b_0/b_1>0$. Its derivative is

$$\beta'(a_\star) =-\frac{b_0^2}{b_1}>0,$$

so $\theta=-\beta'(a_\star)<0$. It is irrelevant and therefore attractive toward the infrared. Perturbative control additionally requires $a_\star\ll1$ in the chosen loop normalization.

2. Derive the walking hierarchy

For $\beta(g)=-\delta-(g-g_c)^2$, integrate across the slow region and find the scale hierarchy as $\delta\to0^+$.

Solution

With $x=g-g_c$,

$$|\Delta t| =\int_{-\infty}^{\infty}\frac{dx}{\delta+x^2} =\frac{\pi}{\sqrt\delta}.$$

Since RG time is a logarithm of scale,

$$\frac{\Lambda_{\mathrm{UV}}}{\Lambda_{\mathrm{IR}}} \sim \exp\left(\frac{\pi}{\sqrt\delta}\right).$$

The hierarchy diverges as the two real fixed points meet at $\delta=0$, even though no real fixed point remains for $\delta>0$.

3. Reproduce the supplied fixed point

Solve the three leading gauge–Yukawa beta functions on their interacting, positive-quartic branch.

Solution

The interacting Yukawa nullcline is $\alpha_y=5\alpha_g/3$. Substitution in $\beta_g=0$ gives $\varepsilon-(2/3)\alpha_g=0$, hence $\alpha_g=3\varepsilon/2$ and $\alpha_y=5\varepsilon/2$. Writing $\alpha_u=r\alpha_y$, the quartic equation becomes

$$4r^2+2r-1=0,$$

so $r=(-1\pm\sqrt5)/4$. The positive branch gives

$$\alpha_u =\frac{\sqrt5-1}{4}\alpha_y =\frac{5(\sqrt5-1)}8\varepsilon.$$

4. Test a claimed zero

A paper reports an order-one zero of a three-loop beta function, but the zero disappears after an admissible near-identity scheme transformation and no observable exponent is given. Which status is justified?

Solution

At most unresolved truncated zero. The order-one coordinate has no small control parameter, its disappearance shows that omitted terms are competitive, and no scheme-invariant stability or observable data support it. One cannot infer either a fixed point or walking without further analysis of the full flow.

References

• Abdesselam, Abdelmalek. “A Complete Renormalization Group Trajectory Between Two Fixed Points.” Communications in Mathematical Physics 276 (2007): 727–772. DOI. Open PDF.
• Banks, Tom, and Alexander Zaks. “On the Phase Structure of Vector-Like Gauge Theories with Massless Fermions.” Nuclear Physics B 196 (1982): 189–204. DOI.
• Bednyakov, Alexander V., and Alfiia I. Mukhaeva. “Asymptotic Safety in the Litim–Sannino Model at Four Loops.” Physical Review D 109 (2024): 065030. DOI. Open PDF.
• Bednyakov, Alexander V., and Andrey F. Pikelner. “Four-Loop Gauge and Three-Loop Yukawa Beta Functions in a General Renormalizable Theory.” Physical Review Letters 127 (2021): 041801. DOI.
• Berges, Jürgen, Nikolaos Tetradis, and Christof Wetterich. “Non-Perturbative Renormalization Flow in Quantum Field Theory and Statistical Physics.” Physics Reports 363 (2002): 223–386. DOI. Open PDF.
• Bond, Andrew D., and Daniel F. Litim. “Theorems for Asymptotic Safety of Gauge Theories.” European Physical Journal C 77 (2017): 429. DOI. Open PDF.
• El-Showk, Sheer, Miguel F. Paulos, David Poland, Slava Rychkov, David Simmons-Duffin, and Alessandro Vichi. “Solving the 3D Ising Model with the Conformal Bootstrap.” Physical Review D 86 (2012): 025022. DOI. Open PDF.
• Gross, David J., and Frank Wilczek. “Ultraviolet Behavior of Non-Abelian Gauge Theories.” Physical Review Letters 30 (1973): 1343–1346. DOI.
• Hasenbusch, Martin. “Finite Size Scaling Study of Lattice Models in the Three-Dimensional Ising Universality Class.” Physical Review B 82 (2010): 174433. DOI. Open PDF.
• Kaplan, David B., Jong-Wan Lee, Dam T. Son, and Mikhail A. Stephanov. “Conformality Lost.” Physical Review D 80 (2009): 125005. DOI. Open PDF.
• Litim, Daniel F., Nahzaan Riyaz, Emmanuel Stamou, and Tom Steudtner. “Asymptotic Safety Guaranteed at Four-Loop Order.” Physical Review D 108 (2023): 076006. DOI. Open PDF.
• Politzer, H. David. “Reliable Perturbative Results for Strong Interactions?” Physical Review Letters 30 (1973): 1346–1349. DOI.
• Wilson, Kenneth G., and Michael E. Fisher. “Critical Exponents in 3.99 Dimensions.” Physical Review Letters 28 (1972): 240–243. DOI.