Compact U(1) in 2+1 Dimensions and the Monopole Plasma

In a weakly coupled compact gauge theory in 2+1 dimensions, allowed monopole events can form a dilute Coulomb plasma. Its dual description explains both a magnetic screening mass and an electric string tension. The calculation requires a regulated monopole core, a controlled infrared expansion, and an external electric charge that dynamical matter cannot screen; compactness alone does not establish confinement.

Required background. Dilute instanton ensembles and theta dependence supplies the grand-canonical event sum. Monopoles and dyons supplies the magnetic-charge normalization and semiclassical core.

Helpful background. The Schwinger model and bosonization is a distinct example in which dual variables make screening and mass generation explicit.

The controlled compactification mechanism chain compares this plasma with small-circle non-Abelian mechanisms; the confinement claim–evidence comparison distinguishes the observables each mechanism establishes.

Compactness permits monopole events

Work at zero temperature in three Euclidean dimensions. Absorb the minimal allowed electric probe charge into the connection $A$ , so the unit Wilson loop is $W(C)=\exp(i\oint_C A)$ . Write

\begin{aligned} S_0[A]&=\frac{1}{4e^2}\int\mathrm d^3x\,F_{\mu\nu}F_{\mu\nu} =\frac{1}{2e^2}\int\mathrm d^3x\,B_\mu B_\mu,\\ B_\mu&=\frac12\epsilon_{\mu\nu\rho}F_{\nu\rho}, \qquad \epsilon_{123}=+1. \end{aligned}

with $[e^2]=1$ . For compact $U(1)$ , holonomies are phases. The curvature $F$ is not itself an angle modulo $2\pi$ : its integral on an oriented closed two-surface measures the bundle’s integer magnetic flux. An allowed monopole event satisfies

\int_{S^2}F=2\pi q, \qquad q\in\mathbb Z.

Whether such events occur, and with what weight, is additional ultraviolet information. A broken $SU(2)$ Georgi–Glashow theory with fundamental probes and only adjoint dynamical matter provides a smooth-core realization: the unit probe is unscreened, while the charged gauge particles have electric charge two. Take a positive Higgs mass as well as massive charged gauge particles; the massless Higgs field at the BPS limit cannot be discarded from a photon-only action. Pure compact lattice electrodynamics provides another core regulator, but its strong-coupling expansion is a different calculation.

Let $M$ be a matching scale no larger than the inverse core radius and the heavy masses relevant to the photon-only description, and let $\xi$ be the matched activity per unit volume of each sign of unit monopole. Assume equal positive activities, no extra light fields or unsaturated fermion zero modes, and no Chern–Simons term. The leading dilute semiclassical calculation uses

\frac{\xi}{M^3}\ll1, \qquad \frac{m_\sigma}{M}\ll1, \qquad \frac{m_\sigma}{K}\ll1, \qquad K=\frac{e^2}{4\pi^2}.

The first condition controls core overlap; the second separates the generated mass from heavy physics. The third controls infrared fluctuations about the saddle and does not follow from the first two. Also require the matched omitted core-cluster, higher-harmonic and derivative effects to be small at scale $m_\sigma$ ; the three inequalities alone do not establish this. Self-energy and short-distance contractions are included consistently in the matched activity, rather than counted again as infrared corrections.

One infrared prescription is a periodic box with total magnetic charge zero and vanishing background harmonic flux, followed by its bulk infinite-volume limit. In that box, inverse Laplacians omit the constant mode; their bulk limit below is $G(x)=1/(4\pi|x|)$ . The compact dual zero mode enforces neutrality. On a region with boundary, nonzero total charge instead requires the corresponding boundary flux. These choices cannot be interchanged inside a finite-volume calculation. The dilute plasma construction is developed in Polyakov 1987, § 4, pp. 65–68.

Dualization fixes the normalization

For fixed events, put $\rho(x)=\sum_a q_a\delta^{(3)}(x-x_a)$ , so $\partial_\mu B_\mu=2\pi\rho$ . Implement the local constraint by Fourier integration over a real, dimensionless multiplier $\sigma$ . This local integration over a real lift is distinct from the periodic global zero mode of the dual field, $\sigma\sim\sigma+2\pi$ , which enforces total-charge neutrality. With those global modes treated separately, the first-order integrand is

\exp\!\left\{ -\int\mathrm d^3x\,\frac{B^2}{2e^2} -\frac{i}{2\pi}\int\mathrm d^3x\,\sigma(\partial\cdot B-2\pi\rho) \right\}.

Integration by parts leaves $+iB\cdot\partial\sigma/(2\pi)$ and the charged insertion $e^{i\sum_aq_a\sigma(x_a)}$ . The Gaussian completion is

-\frac{B^2}{2e^2}+\frac{i}{2\pi}B\cdot\partial\sigma =-\frac{1}{2e^2}\left(B-\frac{ie^2}{2\pi}\partial\sigma\right)^2 -\frac{e^2}{8\pi^2}(\partial\sigma)^2.

Thus

S_{\mathrm{dual},0} =\int\mathrm d^3x\,\frac{K}{2}(\partial\sigma)^2, \qquad K=\frac{e^2}{4\pi^2}.

The imaginary conditional mean $ie^2\partial\sigma/(2\pi)$ is a Gaussian integration identity, not an equality between real off-shell Euclidean fields. In particular, two field-strength insertions also receive the Gaussian contact term. Keeping the sign of a fixed charge before summing gives $e^{+i\sigma}$ for $q=+1$ and $e^{-i\sigma}$ for $q=-1$ .

The normalized Gaussian characteristic function gives the Coulomb interaction

S_{\mathrm C} =\frac{1}{2K}\sum_{a\ne b}q_aq_bG(x_a-x_b), \qquad G(x)=\frac{1}{4\pi|x|},

where the sum counts both orders of each pair. Self-contractions are absorbed into $\xi$ using the same core prescription; the gas integrals exclude separations below the matching length $M^{-1}$ , with shorter-distance physics included in matched coefficients. Summing the leading local insertions produces $\xi e^{i\sigma}+\xi e^{-i\sigma}=2\xi\cos\sigma$ . Up to a field-independent normalization,

\begin{aligned} Z&\propto\int\mathcal D\sigma\,e^{-S_{\mathrm{eff}}[\sigma]},\\ S_{\mathrm{eff}} &=\int\mathrm d^3x\left[ \frac{K}{2}(\partial\sigma)^2 +2\xi(1-\cos\sigma)+\Delta\mathcal L \right]. \end{aligned}

Here $\Delta\mathcal L$ has mass dimension three and contains matched higher-derivative and higher-harmonic corrections. Their coefficients depend on the core theory; the leading cosine is not an exact replacement for an arbitrary-density, unregulated point gas. The construction follows Polyakov 1987, § 4, pp. 65–67, with the Gaussian normalization derived explicitly above.

Dropping these corrections and expanding about a minimum gives

m_\sigma^2=\frac{2\xi}{K} =\frac{8\pi^2\xi}{e^2}.

With $y=m_\sigma x$ , the leading action has overall coefficient $K/m_\sigma$ , which explains the separate infrared loop condition. The mass is nonanalytic in weak coupling when $\xi$ is exponentially small. At quadratic order the connected magnetic covariance, with its momentum-conserving delta function removed, is

\langle B_\mu(p)B_\nu(-p)\rangle_c =e^2\left(\delta_{\mu\nu} -\frac{p_\mu p_\nu}{p^2+m_\sigma^2}\right).

The first term is local. The nonlocal part has a massive denominator and decays exponentially at separated points. At $\xi=0$ this becomes the transverse Maxwell covariance, not a longitudinal scalar gradient alone. This contact completion and magnetic screening interpretation agree with Polyakov 1987, § 4, pp. 67–68, Eqs. (4.79)–(4.81). Magnetic screening does not mean that the unscreened external electric probe can be neutralized by the plasma.

A Wilson loop becomes a dual kink

Choose an oriented reference surface $\Sigma$ with boundary $C$ , and define its distributional normal one-form by $\int\mathrm d^3x\,B\cdot\eta_\Sigma=\int_\Sigma F$ . The positive Wilson insertion adds $+iB\cdot\eta_\Sigma$ to the same Gaussian. Its dual kinetic term is therefore

\frac{K}{2}\left(\partial\sigma+2\pi\eta_\Sigma\right)^2.

The sheet source and field are defined with a common regulator; the kinetic expression is not a literal square of a delta distribution. Near a planar portion, choose the normal coordinate $z$ so $\eta_\Sigma=\mathrm dz\,\delta(z)$ . A branch representative cancels the singular sheet with $[\sigma]_{z=0}=-2\pi$ . This is distinct from a continuous lift: setting $\varphi=\sigma+2\pi H(z)$ , with $H$ the Heaviside step function, removes that cut and gives neighboring asymptotic values $0$ and $2\pi$ , which represent the same compact value rather than distinct vacua. The potential is unchanged because $\cos\varphi=\cos\sigma$ . Reversing the loop reverses the wall orientation but not its tension. The reference sheet is not a physical infinitesimal string; the saddle spreads over a width $m_\sigma^{-1}$ .

For a large nearly planar loop, the continuous wall obeys

K\varphi''=2\xi\sin\varphi, \qquad \varphi(-\infty)=0, \quad \varphi(+\infty)=2\pi.

With $m_\sigma^2=2\xi/K$ , the solution is

\varphi_{\mathrm{kink}}(z) =4\arctan e^{m_\sigma(z-z_0)}.

Multiplying the equation by $\varphi'$ gives the first integral

\frac{K}{2}(\varphi')^2=2\xi(1-\cos\varphi).

The action per unit area is therefore

\begin{aligned} \sigma_{\mathrm{string}} &=\int_{-\infty}^{\infty}\mathrm dz \left[\frac{K}{2}(\varphi')^2+2\xi(1-\cos\varphi)\right]\\ &=\int_0^{2\pi}\mathrm d\varphi\, \sqrt{4K\xi(1-\cos\varphi)}\\ &=8K m_\sigma =8\sqrt{2K\xi}>0. \end{aligned}

After the common short-distance perimeter renormalization, the leading semiclassical large-loop behavior is

\langle W(C)\rangle \sim\exp[-\sigma_{\mathrm{string}}A_{\min}(C)].

The linear dimensions and curvature radii of the loop must greatly exceed $m_\sigma^{-1}$ , which itself exceeds the core scale. Take the bulk infinite-volume limit before the asymptotic loop limit, and retain the small $m_\sigma/K$ condition for this classical tension. Perimeter terms, wall fluctuations and matched corrections are not included in $8K m_\sigma$ . If dynamical matter can screen the probe, a flux tube may break and this need not be the asymptotic area law. In the stated adjoint-only $SU(2)$ realization, the unit fundamental probe is protected against that screening; even electric charges are not.

The noncompact counterexample

Noncompact Maxwell theory uses the same local quadratic action but does not sum over monopole sectors. There is then a continuous shift symmetry $\sigma\mapsto\sigma+\alpha$ , so $\xi=0$ and no cosine is generated:

m_\sigma=0, \qquad \sigma_{\mathrm{string}}=0.

The photon remains massless and static charges have the logarithmic Coulomb interaction characteristic of two spatial dimensions. This counterexample shows why the local Maxwell Lagrangian alone does not encode the confinement claim. The global field space and allowed disorder events are physical data.

The fixed-regulator infrared result does not prove a nonzero mass gap in the strict continuum limit of pure compact lattice theory at fixed $e^2$ ; that limit requires tracking the core action and fugacity as the cutoff is removed. Nor does this calculation prove confinement in undeformed four-dimensional Yang–Mills. Its dimension, Abelian infrared field, calculable monopole activity, unscreened probe and scale hierarchy are essential. The small-circle calculation rebuilds the relevant hypotheses in a different setting.

Exercises

1. Dimensional check. Verify the mass dimensions of $K$ , $\xi$ , $m_\sigma$ , and $\sigma_{\mathrm{string}}$ .

Solution

Because $\sigma$ is dimensionless in three dimensions, $[(\partial\sigma)^2]=2$ and the Lagrangian has dimension three, so $[K]=1$ . The cosine is dimensionless, hence $[\xi]=3$ . Thus $[m_\sigma^2]=[\xi/K]=2$ and $[\sigma_{\mathrm{string}}]=[\sqrt{K\xi}]=2$ , the correct dimension for energy per unit length.

2. Remove compactness. In the same zero-temperature, infinite-volume Maxwell problem with no dynamical charged matter, forbid monopole operators. Starting from the effective action, identify which two conclusions fail. Use the continuous Wilson-induced lift when considering the interpolation energy.