Volume Reduction, Large-N Equivalence, and Their Hypotheses

Large-$N$ volume reduction says that a center-neutral sector can become independent of spacetime volume. Orbifold equivalence similarly identifies a common projection-neutral sector of two theories. Neither statement applies to all observables: factorization, unbroken center and projection symmetries, translation realization, and the order of $N$, volume, lattice-spacing, continuum, and infrared limits are indispensable hypotheses.

Required background. Large-N limits, normalizations, and orders of limits supplies fixed ’t Hooft data and noncommuting-limit warnings. Large-N factorization and master-field claims supplies the connected-correlator suppression used to close loop equations.

Helpful background. Line operators, screening, and symmetry diagnostics supplies center charge and winding-loop tests.

The one-site reduction and its center symmetry

Consider $SU(N)$ lattice gauge theory with link matrices $U_\mu(x)$ and Wilson action

$$S_{\mathrm W} = -2bN \sum_x\sum_{\mu<\nu} \operatorname{Re}\operatorname{tr} \left[ U_\mu(x)U_\nu(x+\hat\mu) U_\mu^\dagger(x+\hat\nu)U_\nu^\dagger(x) \right],$$

where $b=1/\lambda_{\mathrm t}(a)$ is held fixed as $N\to\infty$ at lattice spacing $a$. The one-site Eguchi–Kawai model replaces all links in direction $\mu$ by one matrix $U_\mu$:

$$S_{\mathrm{EK}} = -2bN \sum_{\mu<\nu} \operatorname{Re}\operatorname{tr} \left( U_\mu U_\nu U_\mu^\dagger U_\nu^\dagger \right).$$

It has a $(\mathbb Z_N)^d$ center symmetry

$$U_\mu\longmapsto z_\mu U_\mu, \qquad z_\mu^N=1.$$

A reduced Wilson word

$$\widehat W(C) = \frac1N\operatorname{tr} \left( U_{\mu_1}^{s_1}\cdots U_{\mu_\ell}^{s_\ell} \right), \qquad s_j=\pm1,$$

has center charge determined by its net winding in each direction. Contractible closed loops are neutral. An open path or a winding Polyakov loop is generally charged and lies outside the equivalence sector.

Why loop equations can agree

Varying one link in a Wilson-loop expectation gives a Schwinger–Dyson or loop equation. In the extended lattice theory, translation relates loops based at different sites. In the reduced model, all sites have been identified, so its loop equation contains additional terms represented by open Wilson words.

If $(\mathbb Z_N)^d$ is unbroken,

$$\left\langle \frac1N\operatorname{tr}U_\mu^k \right\rangle =0$$

for nonzero center charge. If normalized neutral loops also factorize,

$$\langle W(C_1)W(C_2)\rangle = \langle W(C_1)\rangle\langle W(C_2)\rangle +O(N^{-2}),$$

then the unwanted charged terms vanish and the neutral loop hierarchies close identically at $N=\infty$. Matching equations is useful only with matching boundary data and saddle selection; otherwise the same formal hierarchy may admit different solutions.

Eguchi and Kawai derive this reduction from the large-$N$ loop equations Eguchi and Kawai 1982, pp. 1063–1065. The same logic extends to volume and orbifold equivalences when the appropriate discrete symmetries are realized Kovtun, Ünsal, and Yaffe 2007, §§2–4.

First application: testing Eguchi–Kawai reduction

Before replacing a large lattice by one site, perform four tests.

1. Match held data. Use the same gauge group sequence, $b=1/\lambda_{\mathrm t}(a)$, matter representation, masses in lattice units, regulator, and state.
2. Restrict the observable. Map a contractible Wilson loop to the corresponding reduced word. Do not infer a charged Polyakov loop, momentum-carrying non-neutral operator, or finite-$N$ spectrum.
3. Test center symmetry. The eigenphases of every $U_\mu$ must remain center-symmetric. Nonzero charged traces diagnose breaking.
4. Declare limits. Establish the large-$N$ neutral-sector relation at regulated fixed parameters, then take continuum or infrared limits only along sequences on which symmetry and uniform error control persist.

The original one-site model fails in dimensions greater than two in the weak-coupling regime because the link eigenvalues can clump and break center symmetry. Then

$$\left\langle \frac1N\operatorname{tr}U_\mu \right\rangle\ne0,$$

the extra open-loop terms survive, and reduced and extended loop equations differ. This is a failure of a hypothesis, not a finite-$N$ correction to an otherwise valid equivalence.

Quenched, twisted, adjoint-fermion, and double-trace-deformed constructions can stabilize center symmetry in selected regimes. Each changes the theory or its global data and requires its own projection, symmetry, and continuum checks; the word “stabilized” is not a proof of equivalence.

Shared calculation. The large-N counting and topology map supplies the adjoint surface count used in loop factorization. The large-N scaling comparison lists the neutral-sector and order-of-limits tests.

Orbifold equivalence in neutral sectors

Let a discrete group $\Gamma$ act on a parent theory and let projection by $\Gamma$ define a daughter. Large-$N$ equivalence compares:

• parent operators invariant under $\Gamma$;
• daughter operators invariant under the corresponding theory-space permutation or related projection symmetry;
• states in which both required symmetries are unbroken.

Correlation functions of mapped neutral single-trace operators can then agree after the coupling and volume map is applied. Charged operators, twisted sectors not included in the map, finite-$N$ observables, and phases with broken projection symmetry are not covered.

The symmetry requirement is two-sided. An unbroken parent projection symmetry is insufficient if the daughter permutation symmetry breaks, and conversely. The nonperturbative equivalence conditions and their necessity are formulated in Kovtun, Ünsal, and Yaffe 2007, §§2.1–2.3 and §4.

Orders of limits and continuum claims

A safe regulated statement has the form

$$\lim_{N\to\infty} \left[ \langle O\rangle_{\mathrm{extended},N,a,L} -\langle O'\rangle_{\mathrm{reduced},N,a,L'} \right] =0$$

for a mapped neutral pair and parameters inside an unbroken-symmetry phase. A continuum claim additionally needs a sequence $a\to0$ along which a fixed physical scale is held constant. A volume-independent infinite-volume observable may require

$$\lim_{a\to0} \lim_{N\to\infty} \langle O\rangle_{N,a,L}$$

with center unbroken at every regulated point. Interchanging these operations demands a uniform bound; it is not licensed by the formal loop equations.

At finite $N$, corrections are typically $O(N^{-2})$ for normalized neutral single traces in an adjoint theory, but infrared enhancement, tunneling among center sectors, or an $N$-dependent approach to a symmetry boundary can invalidate that estimate. Translation breaking or nonuniform momentum resolution must also be checked when extracting local continuum quantities from a reduced model.

Common pitfalls

Checking the action but not the state. Center symmetry can be present in the Lagrangian and broken by the dominant saddle. Measure charged loops or eigenphase distributions.

Comparing observables outside the neutral sector. A winding loop carries center charge; equivalence of contractible loops does not predict it.

Taking the continuum limit through a broken phase. Symmetry must remain realized along the entire scaling trajectory, and large-$N$ errors must be uniform on that trajectory.

Exercises

1. Under $U_\mu\mapsto z_\mu U_\mu$, find the center charge of

$$\frac1N\operatorname{tr} \left(U_1U_2U_1^\dagger U_2^\dagger\right)$$

and of $N^{-1}\operatorname{tr}U_1^k$.

Solution

The plaquette word contains one $U_\mu$ and one $U_\mu^\dagger$ for each direction, so all phases cancel and it is neutral. The second word transforms by $z_1^k$ and is charged unless $k=0$ modulo $N$.

2. Explain why factorization alone does not prove reduction.

Solution

Factorization closes products of neutral loop expectations, but the reduced loop equations also contain open or winding words. Those vanish only if the relevant center symmetry is unbroken. Matching boundary data and saddle selection are also required.

3. A charged Polyakov loop becomes nonzero as $a\to0$ along a proposed reduced sequence. What conclusion follows?

Solution

The center-symmetry hypothesis fails along that sequence. Neutral-sector volume reduction cannot be continued to the claimed continuum limit without changing the construction or trajectory and re-establishing all equivalence conditions.

References

• Eguchi, T., and Kawai, H. (1982). “Reduction of Dynamical Degrees of Freedom in the Large N Gauge Theory.” Physical Review Letters 48, 1063–1066. doi:10.1103/PhysRevLett.48.1063.
• Kovtun, P., Ünsal, M., and Yaffe, L. G. (2007). “Volume Independence in Large $N_c$ QCD-Like Gauge Theories.” Journal of High Energy Physics 2007(06), 019. doi:10.1088/1126-6708/2007/06/019. Open PDF.