Compact Gauge Fields, Higgsing, Solitons, and Topological Defects

The previous page explained why compact phase variables cannot be treated as ordinary real scalars. A compact phase has winding sectors; in two dimensions those sectors are vortices, and in three Euclidean dimensions they become vortex worldlines. This page turns the same idea sideways: now the gauge field itself may be compact.

That distinction is not cosmetic. A noncompact Maxwell connection is a real-valued one-form $a_\mu$. For a compact $U(1)$ connection, parallel transport is a circle-valued phase and flux is meaningful modulo $2\pi$. Once a gauge field is compact, charge quantization is built in, monopole events are allowed, and the infrared theory can be radically different from the Gaussian Maxwell theory suggested by expanding the cosine at small field strength.

The second theme of the page is the Higgs mechanism. A condensate can make gauge bosons massive without breaking gauge invariance as a physical redundancy. In Abelian language the phase of a charged scalar is eaten by the gauge field. In non-Abelian language an adjoint scalar can leave a smaller gauge group unbroken, giving massive vector bosons plus a residual massless photon. The same finite-energy logic leads naturally to solitons, vortices, and monopoles.

Helpful background. Lesson 37 develops compact phases, quantized winding, the Villain formulation, and vortex duality. Here the compact variable is a gauge connection, and the same global distinction produces charge and magnetic-flux quantization.

Compact U(1) gauge fields

Compact links and canonical gauge fields

We use Euclidean signature for statistical and instanton-like formulas, and Lorentzian notation only when discussing the one-dimensional kink as a classical field configuration. To keep two common normalizations distinct, $a$ denotes the dimensionless compact connection and $A$ denotes a canonically normalized continuum gauge field:

$$U_{x\mu}=e^{-ia_{x\mu}}, \qquad a_{x\mu}\sim a_{x\mu}+2\pi, \qquad a=eA.$$

The lattice angle is the link integral $a_{x\mu}=\int_x^{x+\hat\mu}a$. A matter field with integer charge label $q$ couples through $e^{-iq a_{x\mu}}$, or equivalently has canonical electric charge $qe$. This sign makes $U_{x\mu}$ transport a field from $x+\hat\mu$ back to $x$. In the Abelian Higgs section $D_\mu=\partial_\mu-ieA_\mu$ for the unit-charge field. In the $SU(2)$ section we use canonical gauge fields and $D_\mu\phi=\partial_\mu\phi+gA_\mu\times\phi$. These choices fix all mass and flux factors below.

The noncompact Maxwell action, written in connection normalization, is

$$S_{\rm Maxwell}={1\over4e^2}\int d^d x\,f_{\mu\nu}f_{\mu\nu}, \qquad f_{\mu\nu}=\partial_\mu a_\nu-\partial_\nu a_\mu,$$

with $a_\mu$ a real field. This is the natural continuum approximation to small fluctuations. But on a lattice, the most basic gauge variable is not $a_{x\mu}$ itself; it is the parallel transporter

$$U_{x\mu}=e^{-ia_{x\mu}}.$$

Gauge transformations act at sites:

$$U_{x\mu}\mapsto e^{i\alpha_x}U_{x\mu}e^{-i\alpha_{x+\hat\mu}},$$

or, in angular variables,

$$a_{x\mu}\mapsto a_{x\mu}+\alpha_{x+\hat\mu}-\alpha_x \quad \text{mod } 2\pi.$$

The plaquette holonomy is

$$e^{-if_{x,\mu\nu}} =U_{x\mu}U_{x+\hat\mu,\nu}U^{-1}_{x+\hat\nu,\mu}U^{-1}_{x\nu},$$

so the flux $f_{x,\mu\nu}$ is also an angular variable. A compact pure-gauge action is therefore

$$S={1\over g_{\rm lat}^2}\sum_p\left(1-\cos f_p\right).$$

For small flux,

$$1-\cos f_p={1\over2}f_p^2+O(f_p^4),$$

so the ordinary Maxwell action is recovered perturbatively. Globally, however, the compact and noncompact theories differ. Fluxes that differ by $2\pi$ are identical in the compact theory, and this allows topological sectors that are invisible in the quadratic expansion.

A compact $U(1)$ gauge field is an angular link variable. The plaquette flux is defined modulo $2\pi$, and a charge-$q$ hopping term is single-valued only when $q$ is an integer in the chosen unit of charge.

A charged lattice field transforms as

$$\psi_x\mapsto e^{iq\alpha_x}\psi_x.$$

The nearest-neighbor hopping term

$$\psi_x^*e^{-iq a_{x\mu}}\psi_{x+\hat\mu}$$

is gauge invariant under small transformations. But compactness asks for more: it must also be single-valued under

$$a_{x\mu}\mapsto a_{x\mu}+2\pi.$$

This requires

$$e^{2\pi i q}=1,$$

so

$$q\in\mathbb Z$$

in units of the minimal charge. Thus charge quantization is not an extra miracle once the gauge group is the compact circle $U(1)$ rather than the additive group $\mathbb R$.

If two matter fields have charges $q_1$ and $q_2$ and both are honest representations of the same compact $U(1)$, then after choosing the minimal charge unit their charges are integers. Hence the ratio

$${q_1\over q_2}$$

is rational. In continuum QED one often hides this by writing arbitrary real couplings. That is fine for local perturbation theory, but it forgets a global assumption: whether the gauge group is $\mathbb R$ or $U(1)$.

For non-Abelian gauge theory, the compactness is even harder to avoid. The familiar groups $SU(N)$, $SO(N)$, and $Sp(N)$ are compact Lie groups. Their Lie algebras describe infinitesimal fields, but Wilson lines live in the group itself:

$$U(C)=\mathcal P\exp\!\left(i\int_C A\right).$$

The local continuum field strength knows about the Lie algebra. The topology of defects, large gauge transformations, and representation quantization knows about the global group.

Compact QED in 2+1 dimensions

In three Euclidean dimensions it is useful to dualize the compact-connection field strength to a vector,

$$b_\mu={1\over2}\epsilon_{\mu\nu\rho}f_{\nu\rho}.$$

For a smooth noncompact gauge field,

$$\partial_\mu b_\mu ={1\over2}\epsilon_{\mu\nu\rho}\partial_\mu f_{\nu\rho}=0,$$

which is just the Bianchi identity. Equivalently, the magnetic flux of a globally defined smooth connection has no sources.

For a compact gauge field the same equation can fail at isolated events:

$$\partial_\mu b_\mu=2\pi\sum_a m_a\delta^{(3)}(x-x_a), \qquad m_a\in\mathbb Z.$$

These events are monopoles in Euclidean spacetime. In $2+1$ dimensions they are instantons: localized tunneling events that change the magnetic flux through space. They are the gauge-field analogue of vortices in a compact scalar.

A compact $U(1)$ action can also contain topological terms. In three dimensions, one often meets the Chern–Simons action

$$S_{\rm CS}={ik\over4\pi}\int a\wedge da,$$

whose gauge invariance under large gauge transformations quantizes $k$ in the standard normalization. The precise allowed levels also depend on whether the theory is a spin theory and on its spectrum of line operators. For the present purpose, the simpler lesson is enough: compactness turns some coefficients and charges into integers because the gauge field is an angle, not a real number.

The monopole plasma of compact QED will become important on the next page. There the Wilson loop diagnoses confinement. Here we only need the kinematic input: compact Maxwell theory has the same small-field Lagrangian as noncompact Maxwell theory, but it has additional monopole sectors. A point monopole needs a short-distance regulator; on the lattice its core action is finite at fixed lattice spacing and enters as the monopole fugacity.

Higgsing in Abelian gauge theory

Let $\Phi$ be a charged complex scalar with Euclidean Lagrangian

$$\mathcal L={1\over4}F_{\mu\nu}F_{\mu\nu}+|D_\mu\Phi|^2+V(|\Phi|^2), \qquad D_\mu\Phi=(\partial_\mu-ieA_\mu)\Phi, \qquad F_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu,$$

where $A_\mu$ is canonically normalized. Suppose the potential has a minimum at

$$|\Phi|={v\over\sqrt2}.$$

Write

$$\Phi={v+h\over\sqrt2}e^{i\chi}.$$

Then

$$|D_\mu\Phi|^2 ={1\over2}(\partial_\mu h)^2 +{1\over2}(v+h)^2(\partial_\mu\chi-eA_\mu)^2.$$

At quadratic order,

$$|D_\mu\Phi|^2 ={1\over2}(\partial_\mu h)^2 +{1\over2}e^2v^2 \left(A_\mu-{1\over e}\partial_\mu\chi\right)^2+\cdots.$$

The combination $A_\mu-\partial_\mu\chi/e$ is gauge invariant. In unitary gauge, $\chi=0$, and the gauge field has a mass term

$${1\over2}e^2v^2 A_\mu A_\mu.$$

The phase field has not disappeared physically; it has supplied the longitudinal polarization of a massive vector boson with

$$m_A=ev.$$

This is the Higgs mechanism in its most economical form.

The compactness of the phase still matters. If $\chi$ winds at infinity, finite energy requires the gauge field to cancel the phase gradient:

$$D_i\Phi\to0 \qquad (r\to\infty).$$

For a vortex with

$$\Phi\sim {v\over\sqrt2}e^{in\theta},$$

we need

$$e\oint A_i dx^i=2\pi n,$$

or

$$\Phi_B=\int B\,d^2x={2\pi n\over e}.$$

Thus the Abelian Higgs model has quantized magnetic flux tubes. The global vortex of the previous page had logarithmically divergent energy in two dimensions. The local vortex has finite tension because the gauge field screens the angular gradient at long distance.

In the Abelian Higgs model, finite energy requires $D_i\Phi\to0$ far from the vortex core. The gauge field cancels the winding gradient and leaves quantized magnetic flux $\Phi_B=2\pi n/e$.

The vortex core is necessary. If $\Phi$ had nonzero magnitude everywhere, the phase map around the circle could not unwind. At the center the field passes through

$$\Phi=0,$$

where the phase is undefined. This is a recurring pattern for topological defects: the field is forced away from the vacuum manifold in a small core, while outside the core it approaches a nontrivial map into the vacuum manifold.

Adjoint Higgsing and the residual photon

The non-Abelian version is richer because a condensate can leave a subgroup unbroken. Consider an $SU(2)$ gauge theory with an adjoint scalar $\phi^a$, $a=1,2,3$. A schematic Euclidean Lagrangian is

$$\mathcal L={1\over4}F_{\mu\nu}^aF_{\mu\nu}^a +{1\over2}(D_\mu\phi)^a(D_\mu\phi)^a +V(\phi^a\phi^a),$$

with

$$F_{\mu\nu}^a =\partial_\mu A_\nu^a-\partial_\nu A_\mu^a +g\epsilon^{abc}A_\mu^bA_\nu^c, \qquad (D_\mu\phi)^a =\partial_\mu\phi^a+g\epsilon^{abc}A_\mu^b\phi^c.$$

Assume $V$ is minimized at

$$\phi^a\phi^a=v^2.$$

Choose the vacuum direction

$$\phi^a=v\delta^{a3}.$$

Then

$$(D_\mu\phi)^1=gvA_\mu^2, \qquad (D_\mu\phi)^2=-gvA_\mu^1, \qquad (D_\mu\phi)^3=0.$$

Therefore the scalar kinetic term contains

$${g^2v^2\over2}\left[(A_\mu^1)^2+(A_\mu^2)^2\right],$$

but no mass term for $A_\mu^3$. Equivalently,

$$W_\mu^\pm={1\over\sqrt2}(A_\mu^1\mp iA_\mu^2)$$

are massive vector fields, while $A_\mu^3$ is the unbroken $U(1)$ gauge field. Schematically,

$$SU(2)\longrightarrow U(1).$$

At tree level,

$$m_{W^\pm}=gv, \qquad m_{A^3}=0.$$

An adjoint scalar vacuum $\phi^a=v\delta^{a3}$ leaves rotations around the third internal axis unbroken. Two gauge bosons acquire mass, while the $A^3_\mu$ component remains the photon of the residual $U(1)$.

This is the field-theory mechanism behind the phrase “two massive, one massless.” The unbroken field is not picked because it was special in the original Lagrangian. It is picked because the vacuum direction $\phi^a$ defines a stabilizer subgroup. This is a tree-level statement: in $2+1$ dimensions, if the residual $U(1)$ is compact, monopole effects can generate a nonperturbative photon mass, as the next lesson explains.

There is a useful geometric way to say the same thing. Write

$$\phi^a=\rho n^a, \qquad n^a n^a=1.$$

The angular field $n^a$ lives on $S^2$. The covariant derivative contains

$$D_\mu n=\partial_\mu n+gA_\mu\times n.$$

Gauge-field components perpendicular to $n$ can cancel changes of $n$; they become massive after the Higgs field condenses. The component parallel to $n$ generates the rotations that leave $n$ fixed and remains as the Abelian gauge field.

Solitons and finite-energy boundary conditions

A soliton is a classical finite-energy configuration that cannot be deformed to the vacuum without crossing an energy barrier or changing boundary conditions. The simplest example is the kink in one spatial dimension. Take a real scalar field with Lorentzian Lagrangian density

$$\mathcal L={1\over2}(\partial_t\phi)^2-{1\over2}(\partial_x\phi)^2-\lambda(\phi^2-v^2)^2.$$

For a static configuration, the energy is

$$E[\phi]=\int dx\left[{1\over2}(\partial_x\phi)^2+\lambda(\phi^2-v^2)^2\right].$$

Finite energy requires the field to approach a vacuum at spatial infinity:

$$\phi(x\to\pm\infty)=\pm v \quad \text{or} \quad \phi(x\to\pm\infty)=v \quad \text{or} \quad \phi(x\to\pm\infty)=-v.$$

The kink sector is

$$\phi(-\infty)=-v, \qquad \phi(+\infty)=v.$$

It is labeled by the boundary charge

$$Q={\phi(+\infty)-\phi(-\infty)\over2v}.$$

The kink has $Q=1$, the antikink has $Q=-1$, and configurations approaching the same vacuum at both ends have $Q=0$. Continuous finite-energy deformations cannot change $Q$ because they cannot change the asymptotic vacua.

The static equation is

$${d^2\phi\over dx^2}=V'(\phi), \qquad V(\phi)=\lambda(\phi^2-v^2)^2.$$

A representative kink solution is

$$\phi_{\rm K}(x)=v\tanh\left({m(x-X)\over2}\right),$$

where $X$ is its center and $m^2=8\lambda v^2$ is the elementary small-oscillation mass around the vacuum. The exact numerical factor depends on the normalization of $V$, but the hyperbolic tangent shape is universal for the $\phi^4$ double-well kink.

The kink interpolates between the two disconnected vacua of a double-well potential. Translating the kink costs no energy, so the fluctuation operator has a localized zero mode proportional to $d\phi_{\rm K}/dx$.

Now expand around the kink,

$$\phi(x,t)=\phi_{\rm K}(x)+\eta(x,t).$$

The quadratic fluctuation equation has the form

$$\left[-{d^2\over dx^2}+V''(\phi_{\rm K}(x))\right]\psi_n(x)=\omega_n^2\psi_n(x).$$

The zero mode follows without solving the Schrödinger problem. Differentiate the classical equation

$${d^2\phi_{\rm K}\over dx^2}=V'(\phi_{\rm K})$$

with respect to $x$:

$$\left[-{d^2\over dx^2}+V''(\phi_{\rm K})\right]{d\phi_{\rm K}\over dx}=0.$$

Hence

$$\psi_0(x)\propto {d\phi_{\rm K}\over dx}.$$

This mode is not an instability. It is the infinitesimal motion along a family of degenerate solutions labeled by the center $X$. Quantizing it requires replacing the zero-mode amplitude by a collective coordinate $X(t)$.

A true instability would be a negative eigenvalue of the fluctuation operator. For a stable kink, the zero mode is the lowest mode. This can be seen by the node theorem: the derivative of the kink is localized and has no nodes, so it is the ground state of the one-dimensional fluctuation operator.

Continuous moduli and Goldstone zero modes

If the vacuum manifold is continuous, finite-energy solutions often come with additional zero modes. Consider a complex scalar with a global $U(1)$ symmetry,

$$\mathcal L=|\partial_\mu\Phi|^2-V(|\Phi|^2), \qquad V \text{ minimized at } |\Phi|=v.$$

The vacuum manifold is

$$\mathcal M=S^1.$$

A vacuum choice $\Phi=v$ breaks the global symmetry and gives a massless Goldstone field, the phase. In a soliton background, any continuous symmetry that moves one solution to a distinct solution generates a zero mode. Translation gives

$$\delta_X\Phi=\partial_x\Phi_{\rm cl},$$

while a global phase rotation gives

$$\delta_\alpha\Phi=i\Phi_{\rm cl}.$$

Whether this mode is normalizable depends on the dimension and the asymptotic behavior. For an infinite-volume vacuum, the global phase mode is a bulk Goldstone wave rather than a localized collective coordinate. For a localized soliton whose internal orientation can rotate, it can become a genuine internal modulus.

The main physical lesson is that zero modes are not accidental small eigenvalues. They are generated by symmetries of the action that are not symmetries of the chosen classical solution. This is the same logic behind the translational zero mode of the kink, the orientational modes of non-Abelian solitons, and instanton collective coordinates.

Defects from the vacuum manifold

For a global symmetry broken from $G$ to $H$, the vacuum manifold is

$$\mathcal M=G/H,$$

and many defects are classified by its topology. Look in the transverse directions to a defect core. Far from the core, finite energy forces the field onto $\mathcal M$. A defect of codimension $c$ is therefore tested by a map

$$S^{c-1}\longrightarrow \mathcal M,$$

and its topological charge lies in

$$\pi_{c-1}(\mathcal M).$$

The familiar core defects are:

$$\begin{array}{c|c|c} \text{homotopy group} & \text{codimension} & \text{defect} \\ \hline \pi_0(\mathcal M) & 1 & \text{domain wall} \\ \pi_1(\mathcal M) & 2 & \text{vortex or string} \\ \pi_2(\mathcal M) & 3 & \text{monopole} \end{array}$$

For the double well, $\pi_0(\{-v,+v\})$ has two elements and supports walls. A broken global $U(1)$ has $\mathcal M=S^1$ and supports vortices, while $SO(3)\to SO(2)$ gives $\mathcal M=S^2$ and hedgehog sectors.

A three-dimensional texture or skyrmion is classified differently. If the field approaches one fixed vacuum in every direction at infinity, all of spatial infinity is collapsed to a point:

$$\mathbb R^3\cup\{\infty\}\simeq S^3.$$

The field is then a map from compactified whole space, not from a sphere linking a core, and its sectors are measured by $\pi_3(\mathcal M)$. This distinction matters: a skyrmion need not have a singular core.

Core defects are classified by maps from a transverse linking sphere into $\mathcal M$: $\pi_0$, $\pi_1$, and $\pi_2$ give walls, vortices, and monopoles. A three-dimensional skyrmion instead maps compactified whole space $S^3_{\rm space}$ into $\mathcal M$ and is classified by $\pi_3(\mathcal M)$.

For a gauge theory, $G/H$ is a space of gauge-related Higgs orientations rather than a family of physically distinct vacua. The same homotopy calculation can still classify allowed asymptotic gauge and Higgs configurations, but the global form of the gauge group, the spectrum of allowed line operators, and the gauge bundle must be included. One should not interpret the local-gauge case as ordinary spontaneous breaking of a physical redundancy.

Topology is also not the whole stability story. It can prevent a defect from unwinding within the stated boundary conditions, but energetic scaling determines whether it has finite energy, finite tension, or energy that grows with system size.

For a sigma-model field $n(x)\in\mathcal M$ in $D$ spatial dimensions,

$$E\sim {K\over2}\int d^D x\,(\partial_i n)^2.$$

If a configuration has size $R$, then roughly

$$\partial n\sim {1\over R}, \qquad E(R)\sim R^D{1\over R^2}=R^{D-2}.$$

This is Derrick’s scaling estimate for the two-derivative term. In $D=1$ it favors spreading; in $D=2$ it is scale invariant; in $D=3$ it grows with size. Additional terms, gauge fields, potentials, or boundary conditions decide the actual soliton size and energy.

Global monopoles and gauged monopoles

The $O(3)$ vector model in three spatial dimensions has the energy

$$E=\int d^3x\left[{1\over2}(\partial_i\phi^a)^2+{\lambda\over4}(\phi^a\phi^a-v^2)^2\right].$$

The vacuum manifold is

$$\mathcal M=S^2.$$

Since

$$\pi_2(S^2)=\mathbb Z,$$

there are hedgehog sectors. At large radius, the unit-charge configuration is

$$\phi^a(x)\simeq v\,{x^a\over r}.$$

The angular gradient scales like

$$\partial_i\phi^a\sim {v\over r}.$$

Therefore the gradient energy at large radius behaves as

$$E_{\rm grad}\sim \int^R r^2dr\,{v^2\over r^2} \sim v^2 R.$$

A global monopole has linearly divergent energy. It is topologically meaningful but not a finite-energy particle in an infinite system.

Now gauge the $SO(3)$ symmetry, or equivalently consider $SU(2)$ with an adjoint Higgs. The covariant derivative is

$$D_i\phi^a=\partial_i\phi^a+g\epsilon^{abc}A_i^b\phi^c.$$

A gauge field can cancel the angular variation of the hedgehog at infinity:

$$D_i\phi^a\to0 \qquad (r\to\infty).$$

The remaining energy comes from the magnetic field and the core region. The result is the ‘t Hooft–Polyakov monopole: a smooth finite-energy soliton in a theory with

$$SU(2)\longrightarrow U(1).$$

At spatial infinity the normalized Higgs field defines

$$S^2_\infty\longrightarrow SU(2)/U(1)\simeq S^2, \qquad \pi_2(S^2)=\mathbb Z.$$

The gauge field removes the infrared gradient cost without erasing this integer class.

A standard ansatz is

$$\phi^a=vH(r){x^a\over r}, \qquad A_i^a={1-K(r)\over gr}\epsilon_{aij}\hat r^j,$$

with boundary behavior

$$H(0)=0, \qquad K(0)=1, \qquad H(\infty)=1, \qquad K(\infty)=0.$$

The residual $U(1)$ magnetic flux is quantized. In conventional normalization the minimal magnetic charge obeys

$$g_m={4\pi\over g},$$

where $g$ is the electric coupling of the $W^\pm$ fields to the unbroken photon in the convention used above. Since the $W^\pm$ have charge $g$, the monopole flux is twice the elementary Dirac unit defined using only those adjoint charges. If fundamental $SU(2)$ probes are allowed, their unbroken charge is $g/2$, and $4\pi/g$ is exactly the minimal Dirac flux. This illustrates why magnetic-charge statements require the global gauge group and the allowed electric representations, not only the Lie algebra.

A global $O(3)$ hedgehog has angular gradients whose energy grows linearly with the system size. In the gauged theory, the gauge field cancels the angular gradient at infinity, leaving a finite-energy monopole with quantized magnetic flux.

This is the non-Abelian cousin of the Abelian Higgs vortex. In both cases the scalar wants to wind in internal space. A global winding costs long-range gradient energy. A gauge field can absorb the winding gradient at infinity, leaving localized field strength and quantized flux.

Skyrmions, textures, and the need for higher derivatives

The same topological logic gives three-dimensional textures. If the field at spatial infinity approaches a fixed value, then compactified space is

$$\mathbb R^3\cup\{\infty\}\simeq S^3.$$

A field valued in $SU(2)\simeq S^3$ defines a map

$$S^3\to S^3,$$

classified by

$$\pi_3(S^3)=\mathbb Z.$$

The winding number can be written as

$$B={1\over24\pi^2}\int d^3x\,\epsilon_{ijk}\operatorname{tr}\! \left(U^{-1}\partial_iU\,U^{-1}\partial_jU\,U^{-1}\partial_kU\right).$$

A two-derivative sigma-model energy alone does not stabilize the size of such a configuration in three dimensions. Under scaling $x\mapsto x/R$, the two-derivative energy grows like $R$, so a configuration can lower its energy by shrinking. The Skyrme model adds a four-derivative term,

$$E_{\rm Skyrme}\sim \int d^3x\,\operatorname{tr} \left( [U^{-1}\partial_iU,U^{-1}\partial_jU]^\dagger [U^{-1}\partial_iU,U^{-1}\partial_jU] \right),$$

which scales like $1/R$. The competition

$$E(R)\sim aR+{b\over R}$$

stabilizes the size. This is another example of a general lesson: topology classifies sectors, but dynamics decides whether a stable finite-size object exists.

Wilson lines as probes of compact gauge dynamics

Gauge-charged fields are not gauge-invariant local observables by themselves. In a gauge theory, the nearest gauge-invariant cousin of a charged two-point function includes a Wilson line:

$$\left\langle \psi^\dagger(x) \exp\left(iq\int_\Gamma a_\mu dx^\mu\right) \psi(y) \right\rangle.$$

The path $\Gamma$ is oriented from $y$ to $x$. Under a gauge transformation, the Wilson line supplies exactly the endpoint phases needed to compensate the transformation of $\psi(y)$ and $\psi^\dagger(x)$.

For a closed curve $C$, the Wilson loop is

$$W_q(C)=\exp\left(iq\oint_C a_\mu dx^\mu\right), \qquad q\in\mathbb Z.$$

In canonical normalization $a=eA$, so the same operator is $\exp(iqe\oint_C A)$. In a compact gauge theory it measures angular holonomy. A charge-$q$ loop cannot distinguish holonomies that differ by $2\pi/q$. In a pure gauge theory, the large-loop behavior distinguishes Coulomb-like perimeter behavior from a confining area law. With dynamical matter able to screen the probe charge, strings can break and the asymptotic Wilson loop need not sharply distinguish Higgs and confining regimes; the matter content and probe representation must be stated.

This sets up the Euclidean worldline representation of charged particles and the Wilson-loop area law. The seed is already visible here. With the Wilson factor above, a charged particle moving along a path $x^\mu(s)$ has the Euclidean interaction

$$S_{E,\rm int} =-iq\int a_\mu(x(s))\,{dx^\mu\over ds}\,ds =-iq\int a_\mu dx^\mu.$$

Therefore the Euclidean weight contains

$$e^{-S_{E,\rm int}}=\exp\left(iq\int a\right),$$

so the first-quantized path integral automatically produces the displayed Wilson line. In pure compact QED, the response of large Wilson loops to monopoles is the cleanest diagnostic of confinement.

Summary

A compact gauge field is locally Maxwell-like but globally different. Its link variables are phases, its plaquette flux is angular, and its matter representations quantize charge. In $2+1$ dimensions compactness allows monopole instantons, the gauge-field analogue of vortices in a compact scalar.

The Higgs mechanism converts phase stiffness into vector-boson mass. In an Abelian Higgs model the scalar phase is eaten, and vortices carry quantized magnetic flux. In an $SU(2)$ theory with an adjoint Higgs, the vacuum leaves a $U(1)$ subgroup unbroken: two gauge bosons become massive, and one photon remains massless at tree level. Compact-monopole effects can change that last conclusion in $2+1$ dimensions.

Solitons and defects are governed by finite-energy boundary conditions. The kink is stabilized by disconnected vacua and has a translation zero mode. Walls, vortices, and monopoles are classified by maps from transverse linking spheres, while a three-dimensional skyrmion maps compactified whole space into its target. Gauge fields can turn long-range global winding energy into localized flux, producing finite-energy objects such as Abelian Higgs vortices and ‘t Hooft–Polyakov monopoles.

Common pitfalls

The quadratic Maxwell action does not determine whether the gauge field is compact. Compact and noncompact $U(1)$ theories have the same small-field expansion but different global sectors.

Gauge symmetry is not literally broken as a physical symmetry. In the Higgs phase, the vacuum choice is a convenient description; the gauge-invariant statement is that the spectrum and long-distance response reorganize, with vector bosons acquiring mass.

Topology classifies sectors, not automatically stable particles. Derrick scaling and gauge fields decide whether the energy is finite and whether the defect has a stable size.

A global vortex or global monopole can be topologically stable but have infrared-divergent energy. Gauging the symmetry can remove the long-range gradient energy by allowing $D_i\phi\to0$ at infinity.

Connection normalization and canonical normalization must not be mixed. The compact connection has action $f^2/(4e^2)$ and integer Wilson-line labels, while $a=eA$ gives the canonical Maxwell action and $D=\partial-ieA$; using one convention for the kinetic term and the other for flux quantization loses factors of $e$.

Exercises

Compactness and integer charge labels

Let $a$ be a compact lattice $U(1)$ gauge field with $a_{x\mu}\sim a_{x\mu}+2\pi$. A charged hopping term is

$$\psi_x^*e^{-iq a_{x\mu}}\psi_{x+\hat\mu}.$$

Show that $q$ must be an integer in units of the minimal charge.

Solution

The link angle $a_{x\mu}$ and $a_{x\mu}+2\pi$ describe the same compact gauge field. Therefore the hopping factor must be single-valued:

$$e^{-iq(a_{x\mu}+2\pi)}=e^{-iq a_{x\mu}}.$$

This requires

$$e^{2\pi iq}=1.$$

Thus

$$q\in\mathbb Z.$$

After choosing the smallest nonzero allowed charge to be $1$, all other charges are integers. If two charges $q_1,q_2$ are present, their ratio is rational.

The adjoint-Higgs mass matrix

Consider an $SU(2)$ gauge theory with an adjoint scalar and covariant derivative

$$(D_\mu\phi)^a=\partial_\mu\phi^a+g\epsilon^{abc}A_\mu^b\phi^c.$$

For the vacuum $\phi^a=v\delta^{a3}$, determine which gauge fields acquire a mass from ${1\over2}(D_\mu\phi)^a(D_\mu\phi)^a$.

Solution

In the chosen vacuum, $\partial_\mu\phi^a=0$, so

$$(D_\mu\phi)^a=g\epsilon^{ab3}A_\mu^b v.$$

The components are

$$(D_\mu\phi)^1=g\epsilon^{123}A_\mu^2v=gvA_\mu^2,$$ $$(D_\mu\phi)^2=g\epsilon^{213}A_\mu^1v=-gvA_\mu^1,$$

and

$$(D_\mu\phi)^3=0.$$

Therefore

$${1\over2}(D_\mu\phi)^a(D_\mu\phi)^a ={g^2v^2\over2}\left[(A_\mu^1)^2+(A_\mu^2)^2\right].$$

The fields $A_\mu^1$ and $A_\mu^2$, or equivalently $W_\mu^\pm$, have tree-level mass $gv$. The field $A_\mu^3$ remains massless at tree level and is the gauge field of the unbroken $U(1)$.

The kink translation zero mode

For a static kink satisfying

$${d^2\phi_{\rm K}\over dx^2}=V'(\phi_{\rm K}),$$

show that $d\phi_{\rm K}/dx$ is a zero mode of the quadratic fluctuation operator

$$\mathcal O=-{d^2\over dx^2}+V''(\phi_{\rm K}).$$

Solution

Differentiate the classical equation with respect to $x$:

$${d^3\phi_{\rm K}\over dx^3}=V''(\phi_{\rm K}){d\phi_{\rm K}\over dx}.$$

Rearranging gives

$$- {d^2\over dx^2}\left({d\phi_{\rm K}\over dx}\right) +V''(\phi_{\rm K}){d\phi_{\rm K}\over dx}=0.$$

Thus

$$\mathcal O {d\phi_{\rm K}\over dx}=0.$$

The zero mode exists because translating the kink center changes the solution but not its energy.

Infrared energy of a global monopole

Estimate the large-distance energy of a global monopole in three spatial dimensions with

$$\phi^a\simeq v{x^a\over r}$$

and energy density ${1\over2}(\partial_i\phi^a)^2$.

Solution

At large $r$, the field changes only angularly. A unit vector on the sphere has angular derivatives of order $1/r$, so

$$\partial_i\phi^a\sim {v\over r}.$$

The gradient energy out to radius $R$ is therefore

$$E(R)\sim \int^R r^2dr\,{v^2\over r^2}.$$

Hence

$$E(R)\sim v^2 R.$$

The energy grows linearly with system size. A global monopole is topologically meaningful but not a finite-energy particle in infinite volume.

Magnetic flux of an Abelian Higgs vortex

In the Abelian Higgs model, suppose that far from a vortex core

$$\Phi\to {v\over\sqrt2}e^{in\theta}.$$

Use $D_i\Phi\to0$ to derive flux quantization.

Solution

With

$$D_i=\partial_i-ieA_i,$$

finite energy requires

$$D_i\Phi\to0.$$

Far from the core,

$$\partial_i\Phi\simeq i(\partial_i n\theta)\Phi,$$

so the condition becomes

$$\partial_i(n\theta)-eA_i\to0.$$

Integrating around a large circle gives

$$e\oint A_i dx^i=\oint \partial_i(n\theta)dx^i=2\pi n.$$

By Stokes’ theorem,

$$\Phi_B=\int B\,d^2x=\oint A_i dx^i={2\pi n\over e}.$$

Thus the magnetic flux is quantized.

References

• H. B. Nielsen and P. Olesen, “Vortex-Line Models for Dual Strings,” Nuclear Physics B 61 (1973), 45–61.
• A. M. Polyakov, “Particle Spectrum in Quantum Field Theory,” JETP Letters 20 (1974), 194–195.
• A. M. Polyakov, Gauge Fields and Strings, Contemporary Concepts in Physics, vol. 3 (Harwood Academic Publishers, 1987), Chapters 4–6.
• G. ’t Hooft, “Magnetic Monopoles in Unified Gauge Theories,” Nuclear Physics B 79 (1974), 276–284.

Further reading

• S. Coleman, Aspects of Symmetry: Selected Erice Lectures (Cambridge University Press, 1985), especially the lectures on solitons and instantons.
• N. Manton and P. Sutcliffe, Topological Solitons (Cambridge University Press, 2004).
• R. Rajaraman, Solitons and Instantons: An Introduction to Solitons and Instantons in Quantum Field Theory (North-Holland, 1982).
• M. Srednicki, Quantum Field Theory (Cambridge University Press, 2007), Sections 84–85 and 92–93.
• A. Zee, Quantum Field Theory in a Nutshell, 2nd ed. (Princeton University Press, 2010), chapters on symmetry breaking, vortices, monopoles, instantons, and duality.