Elitzur's Theorem and the Gauge-Invariant Higgs Mechanism

A local gauge transformation that is treated as redundancy does not move the system to a distinct physical vacuum. In the compact, unfixed lattice theories covered by Elitzur’s theorem, a suitable bounded local gauge-variant field functional has zero expectation value even in the thermodynamic construction. A gauge-fixed scalar expectation value can be useful, but it is not by itself a physical order parameter for the local redundancy.

This does not forbid Higgs physics. A Higgs-like regime can contain a massive spin-one state and a scalar state, with their masses read from poles or decay lengths of gauge-invariant correlators. Gauge fixing efficiently exposes this reorganization; gauge-invariant operators state its physical content. The scope here is the distinction itself, a weakly coupled Abelian check, and the bounded phase-diagnostic consequences—not full lattice phase diagrams, electroweak dynamics, or Higgs phenomenology.

Required background. Symmetry Realization and Order Parameters supplies the distinction between theory symmetry, state symmetry, and a physical order parameter. Gauge Fields, Redundancy, and Observable Content supplies the local gauge law, the physical quotient, and the role of gauge fixing.

Helpful background. The Free Maxwell Field and Gauge Redundancy supplies the Gauss constraint and massless spin-one degree count used below.

Local gauge averaging and Elitzur’s theorem

Keep the theorem in its native setting. Consider an unfixed Euclidean lattice path integral with compact gauge group $G$, a local gauge-invariant action, and invariant integration measure. The independent group integration at a lattice site $x$ uses normalized Haar measure,

$$\int_G\mathrm d h_x=1.$$

For a bounded local field functional $\mathcal O$, define its projection under the gauge transformation at that one site:

$$\left(\mathsf P_x\mathcal O\right)[\Phi] =\int_G\mathrm d h_x\, \mathcal O[\Phi^{h_x}].$$

Gauge invariance of the action and measure permits the change of variables $\Phi\mapsto\Phi^{h_x}$. Averaging that equality over $h_x$ gives

$$\begin{aligned} \langle\mathcal O\rangle ={}&\frac{1}{Z} \int\mathcal D\Phi\, e^{-S[\Phi]}\mathcal O[\Phi] \\ ={}&\frac{1}{Z} \int\mathcal D\Phi\, e^{-S[\Phi]} \left(\mathsf P_x\mathcal O\right)[\Phi]. \end{aligned}$$

Therefore,

$$\mathsf P_x\mathcal O=0 \quad\Longrightarrow\quad \langle\mathcal O\rangle=0.$$

For example, a charged scalar field at one site has zero Haar average, whereas $\phi_x^\dagger\phi_x$ survives. More generally, a field functional can contain both invariant and noninvariant pieces; only the piece annihilated by $\mathsf P_x$ is forced to vanish.

The displayed projection is the transparent zero-source mechanism, not the entire thermodynamic theorem. To test spontaneous breaking, let $V$ be a finite lattice volume and couple a parameter $J$ to a bounded local source density that selects the gauge-variant channel. For a nonnegative Euclidean weight, locality and boundedness make the local orbit bound uniform in $V$. The theorem then gives

$$\lim_{J\to0}\lim_{V\to\infty} \langle\mathcal O_x\rangle_{J,V}=0.$$

The thermodynamic limit is taken first, as in the definition of spontaneous breaking; the subsequent zero-source limit still vanishes. Thus a bounded local functional with vanishing local orbit average cannot become a nonzero order parameter for the unfixed local gauge symmetry under these hypotheses. See Elitzur 1975, §§ I–IV, pp. 3978–3982 for the primary result and Fradkin 2013, § 9.6, p. 299 for a specialist formulation.

This argument has no ordinary global-symmetry analogue. A global transformation acts on the entire system and can relate distinct physical states; it cannot be averaged independently at one site while leaving every coupling term fixed. A redundant local transformation instead moves within one gauge orbit.

The distinction depends on which transformations are actually quotiented. A transformation carrying a boundary or asymptotic charge may be a physical global symmetry rather than redundancy. Such transformations must not be included in the local Haar average without first imposing the appropriate boundary phase space and charge conditions.

What gauge fixing changes

Gauge fixing selects or weights representatives of each orbit so that a calculation can be performed. It removes the independent local orbit average used above, and a gauge-fixed expectation value such as

$$\langle\phi\rangle_{\mathrm{gf}}$$

can then be nonzero. Its value, and sometimes even the realization of the symmetry left by the gauge condition, can depend on the gauge choice.

A gauge condition may leave a remnant group—for example, transformations constant in spacetime. That remnant acts as a global symmetry of the gauge-fixed description and can have its own realization. A nonzero $\langle\phi\rangle_{\mathrm{gf}}$ may diagnose that remnant, but it does not show that the original local redundancy relates distinct physical vacua.

Physical conclusions must instead survive the gauge-fixing consistency conditions. Depending on the formulation, they are expressed through gauge-invariant observables or through the corresponding physical-state criterion, such as BRST cohomology. Gauge-dependent propagators and expectation values remain useful intermediate objects; their gauge dependence is the reason they cannot alone define a physical phase.

Two qualifications are essential:

• a genuine global symmetry of the matter theory, such as a custodial or flavor symmetry, can still break and can have a physical order parameter;
• a transformation with a nonzero boundary charge can act nontrivially on physical states even if its local formula resembles a gauge transformation.

Elitzur’s theorem does not erase either physical symmetry. It excludes the unfixed local redundancy from the ordinary global-symmetry pattern.

The Abelian Higgs mechanism in invariant variables

Consider a four-dimensional complex scalar of nonzero integer gauge charge $q$:

$$\begin{aligned} \mathcal L ={}&-\frac14F_{\mu\nu}F^{\mu\nu} +|D_\mu\phi|^2 \\ &-\lambda\left( |\phi|^2-\frac{v^2}{2} \right)^2, \end{aligned}$$

with $\lambda>0$ and

$$D_\mu\phi =\left( \partial_\mu-iqgA_\mu \right)\phi.$$

The site convention $D_\mu=\partial_\mu-igA_\mu$ is applied here with generator $q$. A local transformation acts as

$$\begin{aligned} \phi&\longmapsto e^{iq\alpha(x)}\phi, \\ A_\mu&\longmapsto A_\mu+\frac{1}{g}\partial_\mu\alpha. \end{aligned}$$

On a patch where the scalar modulus is nonzero, write

$$\phi =\frac{\rho}{\sqrt2}e^{i\vartheta}, \qquad \rho=v+\sigma.$$

The phase changes by $\vartheta\mapsto\vartheta+q\alpha$. Hence the local combination

$$B_\mu =A_\mu-\frac{1}{qg}\partial_\mu\vartheta$$

is gauge invariant on that patch. Direct substitution gives

$$\begin{aligned} |D_\mu\phi|^2 ={}&\frac12 \partial_\mu\rho\,\partial^\mu\rho \\ &+\frac12q^2g^2\rho^2 B_\mu B^\mu. \end{aligned}$$

At tree level, expansion about $\rho=v$ gives

$$\begin{aligned} m_B^2&=q^2g^2v^2, \\ m_\sigma^2&=2\lambda v^2. \end{aligned}$$

The first coefficient is the mass term $\tfrac12m_B^2B_\mu B^\mu$; the second is the curvature of the radial potential. This perturbative calculation, including the non-Abelian generalization to a vector mass matrix, is developed in Schwartz 2014, § 28.3, pp. 575–579 and Weinberg 1995, § 21.1, pp. 295–300.

Using the parameter $v$ to expand the classical potential is not a claim that the unfixed quantum expectation value of $\phi$ equals $v/\sqrt2$. Moreover, $B_\mu$ is only a polar-coordinate variable: it can fail at zeros of $\phi$, around defects, or when no single phase chart exists. A polynomial gauge-invariant vector operator avoids that local-coordinate limitation:

$$\begin{aligned} \mathcal J_\mu &\equiv i\left[ \phi^\dagger D_\mu\phi -(D_\mu\phi)^\dagger\phi \right] \\ &=qg\rho^2B_\mu. \end{aligned}$$

Expanding this exact relation about $\rho=v$ gives

$$\mathcal J_\mu =qg\left(v^2+2v\sigma+\sigma^2\right)B_\mu.$$

In a weakly coupled Higgs-like regime, the term linear in $B_\mu$ shows that the connected gauge-invariant correlator $\langle\mathcal J_\mu(x)\mathcal J_\nu(0)\rangle_{\mathrm c}$ has nonzero overlap with the massive vector state at leading order. Likewise,

$$\phi^\dagger\phi =\frac{v^2}{2}+v\sigma+O(\sigma^2)$$

interpolates the radial channel. Pole positions of these physical correlators, or their exponential decay lengths in Euclidean signature, state the gauge-invariant mass-generation result. Beyond weak coupling, an operator can also overlap with multiparticle states, and a perturbative pole can broaden or disappear; the spectrum must be determined in the declared regime. The gauge-invariant spectral interpretation is discussed in Fradkin 2013, §§ 9.6 and 9.10, p. 299 and pp. 314–318.

The degree count supplies a separate check. A massless vector has two physical polarizations and a complex scalar has two real components. In the weak Higgs-like description, the same four degrees of freedom appear as a massive vector with three polarizations plus one radial scalar. “The Goldstone is eaten” is shorthand for this gauge-fixed reorganization. There is no physical Goldstone theorem for a broken local redundancy.

Global sources versus residual finite gauge structure

The preceding explicit-breaking example used

$$\Delta\mathcal L_{\mathrm{global}} =\epsilon(\phi+\phi^\dagger)$$

and

$$\Delta\mathcal L_{\mathrm{global}} =h\phi^N+h^*(\phi^\dagger)^N$$

as genuine explicit breakings of a global $U(1)$. The second deformation retains a finite subgroup of the original phase rotations.

Once the same $U(1)$ is gauged and $\phi$ is charged, neither expression is gauge invariant with a fixed numerical coefficient. Those expressions do not descend to functions on gauge orbits and therefore cannot be imported as physical deformations of the same gauge theory. Either can appear after gauge fixing, or inside a larger gauge-invariant coupling to additional charged background or dynamical data, but a temporary gauge-fixed source selects a representative rather than a physical vacuum orientation.

A formally transforming charged spurion organizes a covariant family of backgrounds. Holding it at a fixed noninvariant value introduces background structure; it does not turn gauge redundancy into a physical symmetry that has been explicitly broken. By contrast, gauge-invariant deformations built from $\phi^\dagger\phi$ can change masses and move the theory between regimes without selecting a gauge orientation.

A different construction can leave discrete gauge structure. Suppose the minimal electric charge is normalized to one and a scalar $\Phi_N$ has gauge charge $N$. A nonzero scalar configuration is unchanged when

$$e^{iN\alpha}=1, \qquad \alpha=\frac{2\pi k}{N}.$$

This $\mathbb Z_N$ is residual gauge structure, not a global $\mathbb Z_N$ with $N$ physical vacuum orientations. Here $N$ labels the gauge charge, not the power of the global anisotropy above. Its physical meaning depends on the global form, allowed matter, and genuine line spectrum; it can be detected through flux, line, defect, and topological data. See Global Form, Matter Representations, and the Faithful Gauge Group and Genuine Line Spectra, Discrete Theta Data, and Theory Specification.

Physical phase diagnostics and bounded continuity

A Higgs-like regime should be characterized by physical information, chosen for the theory at hand:

• poles, thresholds, and decay lengths of gauge-invariant correlators;
• screening of admissible charges and the behavior of genuine line operators;
• realization of genuine global or higher-form symmetries, when present;
• defects, topological sectors, and thermodynamic nonanalyticities.

A nonzero $\langle\phi^\dagger\phi\rangle$ is gauge invariant, but it is not an order parameter for broken gauge redundancy. It can remain nonzero and vary smoothly across regions that differ only by useful Higgs-like or confinement-like descriptions.

Fradkin and Shenker proved a precise nonuniversal continuity statement. For the fixed-length compact lattice gauge–Higgs models and fundamental matter representations in their analysis, there is an analytic region connecting confinement-like and Higgs-like parameter regimes for local gauge-invariant observables. Consequently those regimes are not universally separated by a thermodynamic phase boundary in that class of models. See Fradkin and Shenker 1979, §§ I.D, II.B, IV, and Appendix, pp. 3684–3687 and 3694–3696, with the representation-dependent interpretation in Fradkin 2013, §§ 9.10 and 9.12, pp. 314–318 and 321–322.

The theorem establishes an analytic connecting region, not an analytic entire phase diagram. It does not extend unchanged to adjoint or higher-charge matter, residual discrete gauge structure, Coulomb regions, every nonlocal or topological diagnostic, or an arbitrary continuum limit. Gauge-invariant massive correlations alone also do not prove that two points can be joined without a phase transition.

Scope and failure tests

Compact lattice theorem. The Haar-average proof and the cited thermodynamic theorem concern compact local gauge groups with the stated lattice action, invariant measure, nonnegative Euclidean weight, locality, and boundedness assumptions. A noncompact group, a complex or sign-indefinite weight, or an arbitrary continuum regularization requires its own argument.

Local field functional. Elitzur’s conclusion applies to the local gauge-variant component with zero local orbit average. It does not force gauge-invariant dressings, Wilson lines, boundary observables, or topological operators to vanish.

Redundancy group. Transformations carrying physical boundary or asymptotic charges are not automatically averaged away. The redundancy subgroup must be identified before applying the theorem.

Weak-coupling expansion. The polar variables and pole matching above assume a stable weakly coupled Higgs-like regime. They are not a nonperturbative spectral construction, and they need not provide a simple local interpolator for every perturbative field in every representation.

Physical global symmetries. A gauge theory can possess genuine global, discrete, or higher-form symmetries. Their spontaneous breaking is a physical question and must be analyzed with their own invariant order parameters.

Common pitfalls

“Gauge symmetry breaks and the Goldstone is eaten.” This familiar phrase compresses a useful gauge-fixed calculation. The physical statement is that the spectrum reorganizes into a massive vector and remaining scalar modes, without spontaneous breaking of a local redundancy.

“A nonzero gauge-fixed scalar expectation value is a Higgs order parameter.” It can diagnose a remnant symmetry of that gauge condition, but its value is gauge dependent. Physical phase claims require gauge-invariant data.

“Elitzur’s theorem forbids a Higgs regime.” It forbids the specified local gauge-variant order parameter in the unfixed measure. It does not forbid massive gauge-invariant states, screening, or Higgs-like dynamics.

“The one-line Haar average is the whole theorem.” It proves the transparent finite-volume zero-source identity. The thermodynamic statement also uses a nonnegative weight, locality, boundedness, and the volume-then-source order of limits.

“Higgs-like and confining regimes are always the same phase.” The Fradkin–Shenker result is restricted to particular lattice actions, representations, and analytic domains. Other matter content or invariant diagnostics can yield genuine distinctions.

“A charge-$N$ Higgs field produces $N$ global vacua.” The residual $\mathbb Z_N$ is gauge structure under the stated global-form assumptions. Gauge-related scalar orientations are not distinct physical vacua.

“The global scalar source can be carried unchanged into the gauged theory.” A fixed charged source is gauge variant. It must be removed after gauge fixing or embedded in a larger gauge-invariant construction.

Check your understanding

These questions are for self-study and are not graded.

1. Prove the local Haar-average identity for a field functional with $\mathsf P_x\mathcal O=0$. Which assumptions are visible in the proof, and which additional hypotheses and order of limits enter the full thermodynamic theorem?
2. Derive the transformation of $B_\mu$, its tree-level mass, and the polynomial relation $\mathcal J_\mu=qg\rho^2B_\mu$. Check dimensions in four spacetime dimensions.
3. Classify three scalar constructions: the linear source in the global theory, the same fixed source after gauging, and a charge-$N$ scalar leaving residual finite gauge structure.

Check

1. Gauge invariance of $S$ and $\mathcal D\Phi$ makes

   $$\begin{aligned} \langle\mathcal O\rangle &= \frac{1}{Z} \int\mathcal D\Phi\, e^{-S}\mathcal O \\ &= \frac{1}{Z} \int\mathcal D\Phi\, e^{-S}\mathsf P_x\mathcal O \\ &=0. \end{aligned}$$

   The displayed zero-source proof uses compact normalized Haar measure, an invariant action and measure, and the vanishing orbit projection. The full source-defined theorem additionally uses a nonnegative Euclidean weight, locality, and boundedness to obtain a volume-uniform bound. It takes $V\to\infty$ at fixed $J$ before sending $J\to0$.

2. Since $\vartheta\mapsto\vartheta+q\alpha$ and $A_\mu\mapsto A_\mu+g^{-1}\partial_\mu\alpha$,

   $$\begin{aligned} B_\mu' &= A_\mu+\frac1g\partial_\mu\alpha \\ &\quad-\frac{1}{qg} \partial_\mu(\vartheta+q\alpha) \\ &=B_\mu. \end{aligned}$$

   The angular kinetic term is $\tfrac12q^2g^2\rho^2B_\mu B^\mu$, so expansion at $\rho=v$ gives $m_B^2=q^2g^2v^2$. Direct substitution into the current gives

   $$\mathcal J_\mu =qg\rho^2B_\mu.$$

   In four dimensions, $[\rho]=[B_\mu]=1$ and $[g]=0$, so both sides have dimension three. At fixed $v$, $m_B\to0$ as $g\to0$, but the variable $B_\mu$ itself is singular in that limit; at exactly zero coupling the scalar phase becomes a physical global-symmetry coordinate.

3. The global linear source is a genuine explicit breaking of a physical $U(1)$. The same fixed charged source does not descend to gauge orbits after gauging, so it is not an admissible gauge-invariant deformation of the same theory. A charge-$N$ scalar can instead leave residual $\mathbb Z_N$ gauge structure, whose physical content lies in invariant line, flux, defect, or topological observables rather than $N$ gauge-related scalar orientations.

What follows

• Coulomb, Higgs, and Confining Regimes develops screening, force laws, spectra, and line-operator diagnostics.
• The Higgs Doublet and Electroweak Symmetry Breaking develops the electroweak scalar sector and its physical observables.
• Gauge Ensembles and Renormalized Observables develops numerical gauge-invariant spectrum and phase diagnostics.
• The Faddeev–Popov Construction begins the consistency treatment of interacting gauge-fixed path integrals.

References

• Elitzur, Shmuel. “Impossibility of Spontaneously Breaking Local Symmetries.” Physical Review D 12, no. 12 (1975): 3978–3982. DOI.
• Fradkin, Eduardo. Field Theories of Condensed Matter Physics. 2nd ed. Cambridge: Cambridge University Press, 2013. DOI.
• Fradkin, Eduardo, and Stephen H. Shenker. “Phase Diagrams of Lattice Gauge Theories with Higgs Fields.” Physical Review D 19, no. 12 (1979): 3682–3697. DOI.
• Schwartz, Matthew D. Quantum Field Theory and the Standard Model. Cambridge: Cambridge University Press, 2014. DOI.
• Weinberg, Steven. The Quantum Theory of Fields, Volume II: Modern Applications. Cambridge: Cambridge University Press, 1995. DOI.