Universality, Critical Manifolds, and Observable Control

A rigorous universality statement names an observable and a class of microscopic models, tunes each model to its own critical manifold, and proves that a normalized asymptotic quantity is shared. The exponent or limiting law may be universal while the critical point, metric normalization, field strength, and amplitude remain model dependent. Matching symmetries or low-order beta functions is not enough.

Required background. Stable Manifolds and Relevant–Marginal Control supplies the local critical graph. Renormalized Trajectories and Counterterm Tuning supplies model-dependent tuning.

Helpful background. Thermodynamic Limits, Correlation Decay, and Phase Control supplies infinite-volume correlations. Equivalence, Uniqueness, and Comparison Notions separates a shared exponent from identical theories.

A typed universality theorem

Let $i=1,2$ label two lattice actions with short-range, reflection-positive kinetic operators $Q_i$ whose Fourier symbols have nondegenerate quadratic minima at $p=0$. Let $g_i>0$ be small and tune $\nu_i$ to its model-dependent critical value $\nu_{c,i}$. A susceptibility universality claim has the form

$$\chi_i(t)=A_i\,t^{-1} \bigl(\log t^{-1}\bigr)^{(n+2)/(n+8)} \bigl(1+o(1)\bigr), \qquad t=\nu_i-\nu_{c,i}\downarrow0.$$

The logarithmic exponent is shared; $A_i$ and $\nu_{c,i}$ need not be. If $Q_i(p)=p^TH_ip+O(|p|^4)$, even the continuum distance and field normalization depend on $H_i$ until coordinates are rescaled. The theorem for the standard nearest-neighbor four-dimensional $n$-component $|\phi|^4$ model, including the exponent and Gaussian torus scaling limits, is Bauerschmidt, Brydges, and Slade 2014, Theorems 1.1–1.3, pp. 697–704.

Comparing two different finite-range kinetic terms requires more than citing that result. Each covariance must admit the finite-range and derivative bounds used by the RG; the local extraction must yield the same leading marginal coefficient after metric normalization; and all remainders must obey uniform estimates. Under those hypotheses, differences in $Q_i$ enter irrelevant coordinates and amplitudes, while the ratio controlling the logarithmic exponent is unchanged. Without a source proving those hypotheses for the proposed pair, this is a conditional extension, not a published general theorem.

The quantifiers matter. A genuine theorem chooses a set $\mathfrak M$ of microscopic interactions and constants $g_*,c,C$ that work uniformly for every member in a specified neighborhood. If $\delta K_j$ denotes the difference between two normalized irrelevant coordinates, a typical comparison estimate has the schematic form

$$\|\delta K_{j+1}\|_{j+1} \leq (q+c g_j)\|\delta K_j\|_j +C L^{-\omega j}g_j^2, \qquad q<1,\quad \omega>0.$$

Iterating this inequality makes the microscopic difference summable after the relevant parameters and quadratic metric have been matched. It does not make the two effective actions identical: local coordinate changes and finite early-scale contributions remain in $A_i$. The global-flow estimates behind this separation are developed in Bauerschmidt, Brydges, and Slade 2019, Chapters 7–8, pp. 123–179. A proposed universality class should therefore be tested both for uniformity of the constants and for closure of its covariance and interaction family under the exact step.

Observable coordinates

Introduce a source $\sigma$ coupled to a local field or composite observable and enlarge the RG coordinate:

$$Z_j(\phi,\sigma)= (I_j(V_j,\sigma)\circ K_j(\sigma))(\Lambda_N,\phi).$$

Derivatives at $\sigma=0$ generate correlations. Their relevant source couplings obey linear recurrences driven by the bulk flow, while differentiated polymer norms bound the remainder. For susceptibility, differentiating with respect to the mass coordinate relates $\chi$ to the renormalized mass and the product of mass-renormalization factors. Since $g_j\sim1/(\beta j)$,

$$\prod_{k<j}\bigl(1+\gamma\beta g_k+O(g_k^2)\bigr) \asymp j^\gamma, \qquad \gamma=\frac{n+2}{n+8}.$$

Matching the mass scale $j_m\asymp\log m^{-1}$ turns this power of $j_m$ into the logarithm in $\chi$. The proof also bounds source-dependent $K_j$; the product alone is perturbative evidence, not the theorem.

This is the controlled comparison requested by Landau–Ginzburg–Wilson Quantum Criticality: choose two admissible weak short-range actions, tune each separately, normalize their quadratic forms, and compare the susceptibility exponent rather than their bare couplings or amplitudes.

Independent checks and nonconverses

First, change field normalization $\phi\mapsto c\phi$. The amplitude changes but the logarithmic exponent cannot. Second, differentiate the claimed asymptotic: it must agree with the RG mass-derivative estimate in its domain. Third, verify that the observable source lies in the differentiated Banach map; bulk trajectory control alone does not bound a composite insertion.

Shared susceptibility scaling does not imply equality of all correlations, equality of continuum measures, or a conjugacy of the full RG maps. It proves exactly the named observable asymptotic. A Gaussian scaling limit can coexist with nontrivial logarithmic corrections in unscaled thermodynamic observables.

Adversarial long-range interaction

Replace $Q_i$ by a long-range kernel with $\widehat Q(p)\sim|p|^\alpha$, $\alpha\neq2$. The field dimension and upper critical dimension change; the quartic coordinate may no longer be marginal, and the finite-range bounds above may be unavailable. Symmetry remains $O(n)$, but the short-range exponent has no warrant. A new relevance analysis is required.

Exercise

If $\chi_i(t)\sim A_i t^{-1}(\log t^{-1})^\gamma$, compute a combination that removes the nonuniversal amplitude.

Solution

For fixed $c>0$, the ratio $\chi_i(ct)/\chi_i(t)$ tends to $c^{-1}$; the logarithmic factors have ratio tending to one. To retain $\gamma$, use $t\chi_i(t)/(t_0\chi_i(t_0))$ and compare its logarithmic slope, $d\log(t\chi_i)/d\log\log t^{-1}\to\gamma$. Both constructions eliminate $A_i$ but require the asymptotic regime.

References

• 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. Introduction to a Renormalisation Group Method. Lecture Notes in Mathematics 2242. Singapore: Springer, 2019. DOI; Open PDF.
• Brydges, David C., Gianfausto Guadagni, and Paul K. Mitter. “Finite Range Decomposition of Gaussian Processes.” Journal of Statistical Physics 115 (2004): 415–449. DOI; Open PDF.