Wilson-Fisher Fixed Point

Mean-field theory hands you a clean but wrong prediction: critical exponents that are the same in every dimension. Experiments on three-dimensional magnets and fluids disagree, and Onsager's exact two-dimensional Ising solution disagrees violently ( $\beta=\tfrac18$ , not $\tfrac12$ ). For three decades the discrepancy had no systematic cure — there was no small parameter in which to expand the strong fluctuations at a critical point. The Wilson–Fisher fixed point is that cure. It is a genuinely interacting fixed point of the renormalization group that appears below four spatial dimensions and controls the critical behavior of the Ising class and, more broadly, the $O(n)$ models. The mechanism is a balance. At the upper critical dimension $d=4$ the quartic coupling of $\phi^4$ theory is marginal; just below, at $d=4-\varepsilon$ , it becomes relevant and starts to grow near the free-field (Gaussian) fixed point. But growth is not unchecked: the one-loop fluctuation correction feeds back with the opposite sign and bends the flow. Growth and damping cancel at a nonzero coupling $u^*\propto\varepsilon$ — that balance point is the Wilson–Fisher fixed point. Because $u^*$ is small when $\varepsilon$ is small, everything about it can be computed in a controlled expansion, the first systematic theory of non-mean-field criticality. Linearizing the flow about this fixed point gives a discrete spectrum of eigenvalues. Exactly two are relevant — the thermal direction $y_t$ and the ordering-field direction $y_h$ — and every static critical exponent is a function of these two numbers. Their values depend only on the dimension $d$ and the number of order-parameter components $n$ , not on any microscopic lattice detail, which is precisely why disparate materials flow to the same fixed point and share exponents. This work earned Kenneth Wilson the 1982 Nobel Prize and reorganized the whole subject of critical phenomena around the geometry of RG flows.

Key takeaways

The Wilson–Fisher (WF) fixed point is a nontrivial, interacting fixed point of the RG flow for $\phi^4$ scalar field theory that appears below four dimensions, $d<4$ .
It emerges as the Gaussian fixed point loses stability at the upper critical dimension $d_c=4$ : for $\varepsilon=4-d>0$ the quartic coupling is relevant and the flow runs to a new fixed point at $u^*=O(\varepsilon)$ .
The WF point governs the Ising universality class (and the $O(n)$ classes), replacing mean-field exponents with nontrivial ones computed order by order in $\varepsilon$ .
To one loop the correlation-length exponent is $\nu=\tfrac12+\tfrac{n+2}{4(n+8)}\varepsilon+O(\varepsilon^2)$ , and the anomalous dimension $\eta=O(\varepsilon^2)$ ; for $n=1$ , $\nu=\tfrac12+\tfrac{\varepsilon}{12}$ .
It has exactly two relevant directions, temperature ( $t$ ) and field ( $h$ ), giving two independent exponents; every other operator is irrelevant, which is the origin of universality.
Key idea: below four dimensions the quartic coupling of $\phi^4$ theory is relevant at the Gaussian fixed point but self-stabilizes at $u^*=O(\varepsilon)$ ; the resulting interacting fixed point — the Wilson–Fisher point — governs Ising-like criticality with non-mean-field exponents computable in the epsilon-expansion.

Where this lives today

The Wilson–Fisher fixed point is the reference point of an entire modern industry. Its three-dimensional Ising realization — the liquid–vapor critical point, uniaxial ferromagnets, and binary-fluid demixing — has been pinned to five significant figures by the numerical conformal bootstrap, which treats the WF fixed point as a conformal field theory and brackets its exponents ( $\nu\approx0.62999$ , $\eta\approx0.03631$ ) with rigorous inequalities. The $O(2)$ member (superfluid $^4$ He) supplied one of the most precise confrontations of theory and experiment ever performed, the microgravity Lambda-Point measurement of the specific-heat exponent. The same fixed-point thinking now drives functional (non-perturbative) RG studies of quantum critical points in strongly correlated electron systems, the classification of two-dimensional critical phenomena through conformal field theory, and even proposals for asymptotically safe gravity, where a nontrivial ultraviolet fixed point would play the role Wilson–Fisher plays for statistical criticality. Wherever "critical exponents" and "universality class" are used with quantitative intent, the Wilson–Fisher fixed point is the object being described.

The physical picture

At $d=4$ the $\phi^4$ coupling sits on a knife edge: power counting makes it exactly marginal, so whether it grows or shrinks is decided by interactions (one loop) rather than dimensional analysis. Drop below four dimensions and the balance tips — a weak coupling starts to grow. But as it grows, the one-loop fluctuation correction, which reduces the effective coupling, grows faster (it is quadratic in the coupling). The two effects meet at a finite value, and that stationary point is the Wilson–Fisher fixed point. Think of the flow in the $(r,u)$ plane, where $r$ is the tuned mass/temperature parameter and $u$ the quartic coupling. For $d<4$ the Gaussian point at the origin loses control: trajectories near it are pushed along the $u$ direction toward a new attractor at $u^*\sim\varepsilon$ . Tuning $r$ to its critical value places the system on the critical surface, the set of Hamiltonians that flow into the WF point rather than off to high or low temperature. Universality is geometric in this picture. Many different microscopic Hamiltonians are many different starting points, but every one on the critical surface is funneled to the same WF fixed point along the attracting irrelevant directions. Only the two relevant directions — temperature and field — can knock you off it, and they are shared by every member of the class. The epsilon-expansion is the trick that makes this calculable: because $u^*\propto\varepsilon$ is formally small near $d=4$ , perturbation theory is controlled and the exponents come out as Taylor series in $\varepsilon$ .

Figure

The one-loop flow of the quartic coupling: a new fixed point appears at u* ∝ ε below four dimensions

Figure

Correlation-length exponent ν(ε): the one-loop line vs the true 3D Ising value, the outlier at ε = 1

The core equations

The starting point is the Landau–Ginzburg–Wilson action for the coarse-grained order-parameter field. The Wilson–Fisher fixed point is what this action flows to, under the momentum-shell RG, when the spatial dimension is just below four.

S[\phi]=\int\mathrm{d}^d x\,\Big[\tfrac12(\nabla\phi)^2+\tfrac12 r\,\phi^2+u\,\phi^4\Big] \qquad (1)

where: $\phi(x)$ — coarse-grained order-parameter field ( $n$ -component for $O(n)$ ); $r$ — mass/temperature parameter, $\propto(T-T_c)$ at the bare level; $u$ — quartic self-coupling; $d$ — spatial dimension. The Gaussian fixed point (set $u=0$ ) is the reference; the Wilson–Fisher point is the interacting one that branches off it. Power counting at the Gaussian point fixes the engineering dimension of the coupling, which is why four is the pivotal dimension: at $d=4$ the coupling is exactly marginal, and its fate must be settled by loops.

[u]=d-4[\phi]=d-4\cdot\tfrac{d-2}{2}=4-d=\varepsilon \qquad (2)

where: $\varepsilon=4-d$ — deviation below the upper critical dimension; $[\phi]=(d-2)/2$ — canonical field dimension. Positive $[u]$ (for $d<4$ ) means $u$ grows at tree level near the Gaussian point; the loop term is what stops it. The one-loop momentum-shell RG generates the flow of the coupling: a tree-level relevant term $\varepsilon u$ plus a fluctuation term $-Cu^2$ from a single loop with two quartic vertices. Setting the flow to zero gives the fixed point.

\frac{\mathrm{d}u}{\mathrm{d}\ell}=\varepsilon\,u - C\,(n+8)\,u^2+O(u^3) \qquad (3)

where: $\ell=\ln b$ — logarithmic RG scale; $C$ — positive geometric constant from the shell integral ( $C=4K_4$ with $K_4=1/(8\pi^2)$ in the standard normalization); $n$ — number of order-parameter components. The linear term reflects the engineering dimension $[u]=\varepsilon$ ; the quadratic term, with the $O(n)$ combinatorial weight $n+8$ , is the one-loop damping. Solving $\mathrm{d}u/\mathrm{d}\ell=0$ gives two roots: the Gaussian $u=0$ and the interacting Wilson–Fisher coupling, nonzero and proportional to $\varepsilon$ for $d<4$ .

u^*=\frac{\varepsilon}{C\,(n+8)}=O(\varepsilon) \qquad (4)

where: $u^*$ — the Wilson–Fisher fixed-point coupling, $O(\varepsilon)$ ; nonzero and positive only for $\varepsilon>0$ (i.e. $d<4$ ). This is the object the whole article is about: an interacting fixed point at small but finite coupling, which is what makes a controlled expansion possible. Linearizing the temperature direction about the WF point gives the thermal eigenvalue, shifted below its Gaussian value of 2 by the interaction, and hence the correlation-length exponent.

y_t=2-\frac{n+2}{n+8}\,\varepsilon+O(\varepsilon^2)\,,\qquad \nu=\frac{1}{y_t}=\frac12+\frac{n+2}{4(n+8)}\,\varepsilon+O(\varepsilon^2) \qquad (5)

where: $y_t$ — relevant thermal RG eigenvalue at the WF point; $\nu$ — correlation-length exponent, $\xi\sim|t|^{-\nu}$ . For $n=1$ this reads $\nu=\tfrac12+\tfrac{\varepsilon}{12}$ ; the positive correction means the correlation length diverges more slowly than mean-field predicts, because the interaction suppresses large fluctuations. The anomalous dimension is the first quantity that does not appear at one loop; it starts at $O(\varepsilon^2)$ because the one-loop self-energy is momentum-independent.

\eta=\frac{n+2}{2(n+8)^2}\,\varepsilon^2+O(\varepsilon^3) \qquad (6)

where: $\eta$ — anomalous dimension, $G(p)\sim p^{-2+\eta}$ at criticality. For $n=1$ , $\eta=\varepsilon^2/54$ . Its vanishing at one loop is not an accident: it takes a momentum-dependent (two-loop) self-energy to renormalize the gradient term, and only then does the field pick up a nontrivial scaling dimension.

Derivation AThe Wilson-Fisher fixed point from one-loop momentum-shell RG

The goal is to integrate out a thin shell of fast modes, generate the one-loop corrections to the mass and coupling, rescale to restore the cutoff, and locate the nontrivial fixed point at order $\varepsilon$ together with its exchange of stability with the Gaussian point. Start from the Landau–Ginzburg–Wilson action with cutoff $\Lambda$ and split the field $\phi=\phi^<+\phi^>$ into slow modes ( $|q|<\Lambda/b$ ) and fast shell modes ( $\Lambda/b<|q|<\Lambda$ ). The plan is to average over the fast modes to one loop, collect the corrections to $r$ and $u$ , then rescale lengths and fields to canonical form; the fixed point is where the resulting differential flow vanishes.

Step 1. Integrate out the fast modes: the tadpole contraction of two legs of the quartic vertex shifts the mass, $\delta r=4(n+2)\,u\int_{\rm shell}\frac{\mathrm{d}^d q}{(2\pi)^d}\frac{1}{q^2+r}$ . Why: closing two of the four legs of a $\phi^4$ vertex on a fast-mode propagator renormalizes the $\phi^2$ term, and the combinatorial factor $n+2$ counts the index contractions for the $O(n)$ vertex.

Step 2. Collect the one-loop coupling correction: the "fish" diagram with two quartic vertices gives $\delta u=-4(n+8)\,u^2\int_{\rm shell}\frac{\mathrm{d}^d q}{(2\pi)^d}\frac{1}{(q^2+r)^2}$ . Why: two quartic vertices joined by two internal fast-mode lines reduce the effective coupling — the correction is negative — and the $O(n)$ weight of this contraction is $n+8$ .

Step 3. Rescale lengths and fields, $q\to q/b$ and $\phi\to b^{(d-2)/2}\phi$ , to restore the cutoff to $\Lambda$ and the gradient term to canonical form. Why: the rescaling supplies the tree-level engineering-dimension pieces $2r$ and $\varepsilon u$ in the flow, so that the full one-loop flow of the coupling is $\mathrm{d}u/\mathrm{d}\ell=\varepsilon u - 4(n+8)K_4 u^2$ with $K_4=1/(8\pi^2)$ .

Step 4. Solve for the fixed points by setting $\mathrm{d}u/\mathrm{d}\ell=0$ : besides $u=0$ (Gaussian) there is $u^*=\varepsilon/[4(n+8)K_4]\propto\varepsilon$ . Why: the quadratic flow equation has two roots, and the nonzero one exists (is positive) precisely when $\varepsilon>0$ , i.e. below four dimensions — this is the Wilson–Fisher fixed point.

Step 5. Test stability by differentiating the flow: at $u^*$ the slope is $\partial_u(\mathrm{d}u/\mathrm{d}\ell)|_{u^*}=\varepsilon-2\cdot4(n+8)K_4 u^*=-\varepsilon$ , while at the Gaussian point it is $+\varepsilon$ . Why: a negative slope means $u$ is irrelevant at the WF point (trajectories are drawn in), whereas the same eigenvalue is $+\varepsilon>0$ at the Gaussian point (trajectories are pushed out) — the two fixed points exchange stability at $d=4$ , a transcritical bifurcation. *Verified: verify_aq-sm-wilson-fisher-fixed-point.py confirms $[u]=4-d$ , the two roots of the beta function, $u^*\propto\varepsilon$ , and the stability slopes $\mp\varepsilon$ at the WF and Gaussian points.*

Go deeper · Derivation B · 2 minCritical exponents to one loop from the linearized flow

Derivation BCritical exponents to one loop from the linearized flow

The goal is to linearize the $(r,u)$ flow about the Wilson–Fisher point, extract the thermal eigenvalue $y_t$ , and turn it into the exponents $\nu$ , $\gamma$ , and $\alpha$ — correcting along the way a malformed susceptibility relation that the base export carried with a stray " $?$ ". Write $r=r^*+\delta r$ and $u=u^*+\delta u$ and expand the flow to first order. Because the coupling is irrelevant at the WF point, the temperature direction decouples to leading order and its eigenvalue can be read directly from the linearized $r$ -flow.

Step 1. Form the Jacobian of the flow at $(r^*,u^*)$ . To leading order it is triangular: the $u$ direction has eigenvalue $-\varepsilon$ (irrelevant) and decouples, so the thermal eigenvalue is the diagonal $r$ -entry, $y_t=2-4(n+2)K_4\,u^*$ . Why: the off-diagonal coupling of $\delta r$ to $\delta u$ does not affect the eigenvalues at leading order when $u$ is irrelevant, so the relevant eigenvalue is just the $r$ -flow slope.

Step 2. Insert $u^*=\varepsilon/[4(n+8)K_4]$ : the $K_4$ factors cancel and leave $y_t=2-\tfrac{n+2}{n+8}\varepsilon$ . Why: the same geometric constant $K_4$ appears in the mass correction and in the fixed-point coupling, so their ratio is a pure combinatorial number times $\varepsilon$ .

Step 3. Invert to get the correlation-length exponent: $\nu=1/y_t=\tfrac12+\tfrac{n+2}{4(n+8)}\varepsilon+O(\varepsilon^2)$ . Why: expanding $1/(2-a\varepsilon)$ to first order gives $\tfrac12(1+\tfrac{a}{2}\varepsilon)$ , and for $n=1$ this is $\nu=\tfrac12+\tfrac{\varepsilon}{12}\approx0.583$ at $\varepsilon=1$ , versus $\approx0.630$ from resummation.

Step 4. Get the susceptibility from the Fisher relation $\gamma=\nu(2-\eta)$ , which the base export wrote as " $\gamma=2\nu-\eta\nu\,d?$ " — a malformed placeholder. Since $\eta=O(\varepsilon^2)$ , at one loop $\gamma=2\nu=1+\tfrac{n+2}{2(n+8)}\varepsilon$ , i.e. $1+\tfrac{\varepsilon}{6}$ for $n=1$ . Why: the correct scaling relation is $\gamma=\nu(2-\eta)$ ; the anomalous dimension only corrects it at second order, so to one loop the susceptibility exponent is simply twice $\nu$ .

Step 5. Get the specific-heat exponent from hyperscaling $\alpha=2-d\nu$ with $d=4-\varepsilon$ : expanding gives $\alpha=\tfrac{4-n}{2(n+8)}\varepsilon+O(\varepsilon^2)$ , which for $n=1$ is $\tfrac{\varepsilon}{6}>0$ . Why: for $n<4$ the specific heat diverges (positive $\alpha$ ), for $n=4$ it is marginal at this order, and for $n>4$ it turns into a cusp — a falsifiable $n$ -dependence the expansion predicts. *Verified: verify_aq-sm-wilson-fisher-fixed-point.py confirms $y_t=2-\tfrac{n+2}{n+8}\varepsilon$ , $\nu=\tfrac12+\tfrac{n+2}{4(n+8)}\varepsilon$ (with $\nu(n=1,\varepsilon=1)=7/12$ ), the corrected $\gamma=\nu(2-\eta)$ , and $\alpha=\tfrac{4-n}{2(n+8)}\varepsilon$ .*

Limiting cases

The WF fixed point interpolates smoothly between known regimes as $d$ and $n$ are varied:

Case	Result	Meaning
$d\ge4$ ( $\varepsilon\le0$ )	$u^*\to0$ , $\nu\to\tfrac12$ , $\eta\to0$	The WF point merges with the Gaussian fixed point; mean-field exponents become exact (with logarithmic corrections exactly at $d=4$ ).
Small $\varepsilon=4-d$	$\nu=\tfrac12+\tfrac{n+2}{4(n+8)}\varepsilon$	The controlled one-loop $O(n)$ result; for $n=1$ , $\nu=\tfrac12+\tfrac{\varepsilon}{12}$ .
$d=3$ , $n=1$ (Ising)	$\nu\approx0.630$ , $\eta\approx0.036$ , $\gamma\approx1.237$	Best-known values from resummed high-order series and the conformal bootstrap; the physical realization of the WF point.
$n\to\infty$	$\nu=\dfrac{1}{d-2}$ (for $2<d<4$ )	The spherical-model limit, exactly solvable; provides an all-orders check on the epsilon-series.

Worked examples

Example 1 — The one-loop Ising correlation-length exponent and the honest gap to the true value. This computes the flagship WF number and confronts it with the exact three-dimensional answer.

Step 1. Take the Ising case $n=1$ and evaluate the universal coefficient: $\tfrac{n+2}{4(n+8)}=\tfrac{3}{4\cdot9}=\tfrac{1}{12}$ , so $\nu=\tfrac12+\tfrac{\varepsilon}{12}$ .

Step 2. Set $\varepsilon=1$ to reach three dimensions: $\nu=\tfrac12+\tfrac{1}{12}=\tfrac{7}{12}\approx0.583$ .

Step 3. Compare with the true 3D Ising value $\nu\approx0.630$ from resummed series, Monte Carlo, and the conformal bootstrap: the bare one-loop result undershoots by about 7%.

Step 4. Draw the honest lesson: one loop already captures the qualitative departure from mean-field ( $\nu=\tfrac12\to0.583$ ), but the quantitative gap of $0.630-0.583\approx0.047$ closes only when the divergent series is Borel-resummed. The one-loop Ising exponent is $\nu=7/12\approx0.583$ , the outlier that resummation drives up to $\approx0.630$ .

Example 2 — Exchange of stability at the upper critical dimension. This shows why the WF point exists only below four dimensions.

Step 1. Write the coupling flow $\mathrm{d}u/\mathrm{d}\ell=\varepsilon u - C(n+8)u^2$ and find its roots: $u=0$ and $u^*=\varepsilon/[C(n+8)]$ .

Step 2. Evaluate the stability slope at the Gaussian point: $\partial_u(\mathrm{d}u/\mathrm{d}\ell)|_{u=0}=\varepsilon$ . For $\varepsilon>0$ this is positive, so the Gaussian point is unstable ( $u$ relevant); for $\varepsilon<0$ it is negative (stable).

Step 3. Evaluate the slope at the WF point: $\partial_u(\mathrm{d}u/\mathrm{d}\ell)|_{u^*}=\varepsilon-2C(n+8)u^*=\varepsilon-2\varepsilon=-\varepsilon$ . For $\varepsilon>0$ this is negative, so the WF point is stable ( $u$ irrelevant).

Step 4. Conclude: for $\varepsilon>0$ ( $d<4$ ) the two fixed points have swapped roles compared with $\varepsilon<0$ — the Gaussian is unstable and the WF is stable. The fixed points exchange stability exactly at $d=4$ , where $u^*=0$ and they coincide.

Origins · 1 minkenneth wilson and how this was discovered

kenneth wilson

1936–2013, American theoretical physicist · Kenneth Geddes Wilson recast the renormalization group from a technical device of quantum field theory into a general theory of behavior across scales. In two 1971 Physical Review B papers he formulated the momentum-shell RG for critical phenomena, and in 1972, with Michael Fisher, he identified the interacting fixed point below four dimensions and organized the calculation as an expansion in

\varepsilon=4-d

— giving the first systematic, calculable theory of non-mean-field critical exponents. The work explained universality as the geometry of RG flows toward fixed points and earned him the sole 1982 Nobel Prize in Physics "for his theory of critical phenomena in connection with phase transitions." He later pioneered lattice gauge theory and large-scale scientific computing.

History

1937

— Landau's mean-field theory predicts dimension-independent critical exponents

1944

— Onsager's exact 2D Ising solution gives β = 1/8, disproving mean-field universality

1966

— Kadanoff's block-spin scaling picture supplies the physical intuition for fixed points

1971

— Wilson formulates the momentum-shell renormalization group in two Physical Review B papers

1972

— Wilson and Fisher locate the interacting fixed point below four dimensions and introduce the ε = 4 − d expansion (Phys. Rev. Lett. 28, 240)

1974

— Wilson and Kogut's Physics Reports review consolidates the framework

1982

— Wilson receives the Nobel Prize in Physics for the theory of critical phenomena

Going further

References. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 14 (Phase Transitions: Renormalization Group) · M. Kardar, Statistical Physics of Fields, Ch. 5 (Perturbative Renormalization Group) · K. G. Wilson & M. E. Fisher, Phys. Rev. Lett. 28, 240 (1972) · K. G. Wilson & J. Kogut, Phys. Rep. 12, 75 (1974) · J. Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, Ch. 12 · J. Zinn-Justin, Quantum Field Theory and Critical Phenomena, 4th ed.

Provenance. Flow equations, fixed-point coordinates, and one-loop exponents follow the momentum-shell treatment in Kardar (Ch. 5) and Pathria & Beale (Ch. 14); the original results are Wilson & Fisher, Phys. Rev. Lett. 28, 240 (1972) and Wilson & Kogut, Phys. Rep. 12, 75 (1974). A verifier recomputes: (i) $[u]=4-d$ ; (ii) the one-loop coefficients giving $y_t=2-\tfrac{n+2}{n+8}\varepsilon$ and $u^*=\varepsilon/[4(n+8)K_4]$ with $K_4=1/(8\pi^2)$ ; (iii) $\nu=\tfrac12+\tfrac{n+2}{4(n+8)}\varepsilon$ , $\gamma=1+\tfrac{n+2}{2(n+8)}\varepsilon$ , $\eta=O(\varepsilon^2)$ . Quoted 3D Ising values ( $\nu\approx0.630$ , $\eta\approx0.036$ , $\gamma\approx1.237$ ) are standard resummed/bootstrap numbers. The spherical-limit $\nu=1/(d-2)$ is the exact large- $n$ result. Base export corrections: the coupling-flow constant carried a " $?$ " placeholder ( $C=(n+8)/?\,K_d$ ), fixed to $C=4K_4$ ; the susceptibility relation was carried as " $\gamma=2\nu-\eta\nu\,d?$ ", corrected to the standard Fisher relation $\gamma=\nu(2-\eta)$ .

Builds on: Momentum-Shell Renormalization Group · Gaussian Fixed Point and Upper Critical Dimension · Fixed Points of the RG · Linearization About a Fixed Point and Eigenvalue Spectrum · Relevant, Irrelevant, and Marginal Operators. Leads to: Critical Exponents from the RG · The Epsilon-Expansion · Scaling Fields and Scaling Variables · RG Flow in Coupling Space.

What does it mean for a fixed point to be "interacting" rather than "free"?

At a free (Gaussian) fixed point the coupling constant is zero, so the field is a collection of non-interacting Gaussian fluctuations and all correlation functions factorize. At the Wilson–Fisher fixed point the coupling is nonzero, $u^*\propto\varepsilon$ , so fluctuations genuinely scatter off one another; correlation functions do not factorize and the field acquires an anomalous dimension. "Interacting" is the statement that the fixed-point action is not quadratic — it is the reason the exponents differ from mean-field.

d=4-\varepsilon

a real physical dimension?

No — dimension is analytically continued as a formal device. The loop integrals are defined for continuous $d$ (dimensional regularization), so $\varepsilon=4-d$ is a bookkeeping parameter, not a physical small coupling you can dial in the lab. The justification is empirical and asymptotic: the resulting series for the exponents, once resummed, reproduces the exponents measured in genuine integer dimensions $d=3$ and $d=2$ . The non-integer dimension is a scaffold; the answers it produces at $\varepsilon=1$ and $\varepsilon=2$ are what get compared to experiment.

Why does the combinatorial factor

n+8

appear, and

n+2

elsewhere?

Both count contractions of the $O(n)$ -symmetric quartic vertex $(\phi^2)^2$ . The mass renormalization comes from a tadpole that closes two legs of one vertex; summing over the index structure gives $n+2$ . The coupling renormalization comes from a one-loop diagram with two vertices joined by two internal lines; the three ways to pair the external legs, each with its index sum, add up to $n+8$ (eight index-independent channels plus $n$ from a closed index loop). These are the fingerprints of the $O(n)$ symmetry in the exponents, and they are exactly what makes $\nu$ and $\eta$ depend on $n$ .

How do you get from one relevant eigenvalue

y_t

to all the exponents?

There are two relevant eigenvalues, $y_t$ and $y_h$ , and every static exponent is a function of these and the dimension: $\nu=1/y_t$ , $2-\alpha=d/y_t$ , $\beta=(d-y_h)/y_t$ , $\gamma=(2y_h-d)/y_t$ , $\delta=y_h/(d-y_h)$ , and $\eta=d+2-2y_h$ . So once the linearized RG at the WF point supplies $y_t$ and $y_h$ , the full exponent set follows algebraically, and the scaling relations (Rushbrooke, Widom, Fisher, Josephson) hold automatically because they are identities in $y_t,y_h,d$ . That is why "two independent exponents" is the correct count.

Why is the correction to

\nu

positive — why does the interaction slow the divergence?

Physically, a positive quartic coupling penalizes large excursions of the order parameter, suppressing the biggest fluctuations. Suppressed fluctuations mean the correlation length grows more gradually as the critical point is approached, so $\nu$ increases above its mean-field value $\tfrac12$ . Formally, the interaction lowers the thermal eigenvalue $y_t$ below $2$ by $\tfrac{n+2}{n+8}\varepsilon$ , and since $\nu=1/y_t$ , a smaller $y_t$ means a larger $\nu$ . The sign of the shift is a direct consequence of the coupling being positive (a stable, bounded-below potential).

Does the Wilson–Fisher fixed point control the two-dimensional Ising model?

It is the same universality class in the sense of being the Ising ( $n=1$ ) fixed point continued to $d=2$ ( $\varepsilon=2$ ), but the epsilon-expansion at $\varepsilon=2$ is far outside its controlled regime and does not reproduce the exact 2D exponents accurately by bare truncation. The exact 2D Ising exponents ( $\beta=\tfrac18$ , $\gamma=\tfrac74$ , $\nu=1$ , $\eta=\tfrac14$ ) come from Onsager's exact solution and from two-dimensional conformal field theory, not from the $\varepsilon$ -series. So the WF fixed point is the right conceptual object, but the practical tools for $d=2$ are exact solutions and CFT rather than the small- $\varepsilon$ expansion.