Four-Dimensional Scalar QFT: Existence and Triviality

Four-dimensional scalar theory sits at its upper critical dimension. Rigorous results prove Gaussian scaling for important ferromagnetic Ising and lattice $\lambda\phi^4$ families and prove precise logarithmic corrections for weakly coupled $n$-component lattice models. They do not prove that every four-dimensional scalar action is trivial, nor do they construct a non-Gaussian, local, positive Wightman $\phi^4_4$ theory. The interaction, component number, stability assumptions, regulator family, scaling, and observable topology are essential parts of each theorem.

Required background. Constructed $P(\phi)_2$ and $\phi^4_3$ models show what full cutoff removal looks like below four dimensions. Critical four-dimensional logarithmic corrections supplies the marginal-flow theorem, and existence, uniqueness, and equivalence claims fix the conclusion types. Helpful background. Regulator removal and renormalized predictions and lines of constant physics supply the physical limiting problem.

Four-dimensional scalar scaling limits

Consider a lattice field $\phi_x\in\mathbb R^n$ with ferromagnetic nearest-neighbor coupling and stable local quartic weight. At criticality define a smeared rescaled field

$$\Phi_a(f)=Z_a^{-1/2}a^4\sum_{x\in a\mathbb Z^4}f(x)\phi_x,$$

where the normalization $Z_a$ is chosen from the two-point scale. A Gaussian scaling limit means that all joint cumulants of $\Phi_a(f_1),\ldots,\Phi_a(f_k)$ above order two vanish in the limit and the limiting moments obey Wick’s rule. It is a statement about this normalized field and this scaling family.

Aizenman and Duminil-Copin’s scalar theorem concerns one real component, nearest-neighbor ferromagnetic coupling, and stable quartic single-site weights. The scaling field is divided by the square root of the variance of a block sum; joint laws of finitely many smeared fields are taken first in the infinite-volume limit $R/\ell\to\infty$, then as the block scale $\ell\to\infty$, allowing varying bare parameters and subsequences. Every such limiting field with the stated decay of the two-point function at large separation is a generalized Gaussian process. The Ising result is included in their analysis. These are the hypotheses and conclusion of Aizenman and Duminil-Copin 2021, arXiv v4, Eqs. (1.10), (1.14), Definition 1.1 and Theorem 1.2, pp. 3–5, PDF.

The proof controls the normalized four-point Ursell function and exponential moments. The published correction changes the exponential-moment bounds (1.26) and (7.2), and requires $\lvert f\rvert$ in the right-hand sides of (6.45) and (6.47); it leaves the theorem’s conclusions unchanged Aizenman and Duminil-Copin 2024, p. 479. The 2021 preprint should therefore be read together with that correction.

This theorem is decisive within its model class. It does not quantify over arbitrary nonpolynomial interactions, non-ferromagnetic measures, gauge-coupled scalar systems, noncommutative geometries, or every multicomponent scaling prescription. “Four-dimensional scalar QFT is trivial” is therefore acceptable only after the model class and the notion of triviality are stated.

Logarithmic interaction traces without a non-Gaussian limit

For weakly coupled $n$-component $|\phi|^4$ on $\mathbb Z^4$, the susceptibility near criticality satisfies

$$\chi(g,\nu_c+t) =A_g t^{-1} \bigl(\log t^{-1}\bigr)^{(n+2)/(n+8)} (1+o(1)).$$

Bauerschmidt, Brydges, and Slade prove this and related specific-heat and finite-volume scaling results for $n\ge1$ and sufficiently small positive bare coupling Bauerschmidt, Brydges, and Slade 2014, arXiv v1, Theorems 1.1–1.3, pp. 6–11, PDF. The logarithm records the marginally irrelevant flow $g_j\sim c/j$. It is compatible with a Gaussian critical field: Gaussianity concerns normalized limiting cumulants, while logarithmic corrections concern how normalization and thermodynamic observables approach the limit.

Thus there are at least three distinct statements:

1. a finite regulated quartic measure exists;
2. selected critical observables have rigorously controlled logarithmic asymptotics;
3. normalized scaling fields converge, and every scaling limit in the theorem’s class is Gaussian.

None of these constructs a non-Gaussian $\phi^4_4$ Wightman theory. Conversely, Gaussianity in the proved ferromagnetic class is much stronger than a positive perturbative beta function: it is a nonperturbative statement about scaling-limit correlations.

First application: test a proposed continuum trajectory

Return to regulator removal. Suppose a proposed trajectory tunes the bare mass and coupling as $a\to0$ and exhibits a slowly running four-point coupling on accessible lattices. A continuum construction must control convergence along a declared trajectory or subsequence, identify the renormalized field normalization, and establish the complete named reconstruction axioms for the limiting hierarchy, including reflection positivity. Tightness is one way to obtain convergent subsequences; it does not by itself identify the limiting theory. A surviving higher connected correlation of the defining scalar field would exclude a Gaussian limit for that field. Existence along one controlled trajectory does not require uniqueness across all trajectories or subsequences; that stronger conclusion needs a separate proof.

The rigorous logarithmic theorem can verify the predicted critical exponent in its weak-coupling lattice class. The marginal-triviality theorem then says that the normalized field scaling limits in its ferromagnetic class are Gaussian. Neither theorem supplies a different nontrivial trajectory outside its hypotheses, and finite-lattice running cannot override the Gaussianity result inside them.

What remains open at the cutoff date

At the evidence cutoff 2026-08-10, there is no accepted construction of a non-Gaussian, local, reflection-positive four-dimensional continuum $\lambda\phi^4$ theory with the usual stable polynomial interaction. There are proposals and models with changed interactions or geometries, but their conclusions must not be imported into the standard lattice ferromagnetic theorem class. Likewise, the available Gaussianity results should not be enlarged into a no-go theorem for every conceivable four-dimensional scalar QFT.

The open obligation for a claimed nontrivial model is constructive: produce the limiting object with its reconstruction axioms, prove non-Gaussianity of the defining scalar field, and state whether uniqueness of the continuum limit is additionally established. A non-Gaussian nonlinear composite of a free field alone does not establish interacting dynamics; the constructive-QFT comparison gives an explicit counterexample. The open obligation for a universal triviality claim is classificatory: prove that its hypotheses cover every intended stable local scalar construction.

Failure tests

Perturbative overreach. A positive one-loop beta function suggests Landau behavior but does not control all scales or define the continuum measure. Full triviality does not follow from that calculation alone.

Finite-lattice overreach. Smooth continuum extrapolations of several observables are evidence, not tightness and convergence of the complete hierarchy. The strongest surviving claim is the measured scaling behavior with stated uncertainties.

Theorem overreach. Applying the Aizenman–Duminil-Copin result to a non-ferromagnetic or nonpolynomial model without showing its random-current hypotheses changes the theorem. The Gaussian conclusion then has not been proved.

Exercises

Can the susceptibility contain a nontrivial logarithmic correction while the scaling field is Gaussian?

Solution

Yes. The logarithm modifies the scale-dependent field and mass normalization as the marginal coupling flows to zero. After the corresponding normalization, higher connected cumulants can still vanish and the limiting field can obey Wick’s rule. Logarithmic approach to a Gaussian fixed point is not a non-Gaussian continuum interaction.

References

• Aizenman, Michael. “Proof of the Triviality of $\phi^4_d$ Field Theory and Some Mean-Field Features of Ising Models for $d>4$.” Physical Review Letters 47 (1981): 1–4. DOI.
• 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; corrigendum 199 (2024): 479. DOI; Open PDF, arXiv v4; Corrigendum; Open PDF, complete one-page correction in publisher preview.
• 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, arXiv v1 with its own pagination.