Gauge-Phase Diagnostics from Line Operators

Wilson and ’t Hooft operators diagnose a gauge phase only after three prior questions are answered: which lines are genuine for the global gauge group, which can end on dynamical matter, and which ultraviolet factors have been removed. For a genuine unscreenable line, area, perimeter, and Coulomb behavior encode distinct infrared responses; for a screenable or surface-attached line, the same visual loop can fail as an order parameter.

Required background. Genuine lines, screening, and charge lattices supplies the global line lattice; Coulomb, Higgs, and confining regimes supplies the gauge-invariant phase criteria.

Helpful background. Electric and magnetic one-form symmetries supplies the symmetry interpretation of line laws.

Genuine Wilson–’t Hooft probes

For a closed contour $C$ and a representation $R$, the Wilson operator is

$$W_R(C)=\operatorname{tr}_R\, \mathcal P\exp\!\left(i g\oint_C A_\mu dx^\mu\right).$$

Path ordering is essential in a non-Abelian theory. Parallel transport makes the closed trace locally gauge invariant; the construction and its transformation law are derived in Schwartz 2014, §25.2, pp. 488–493. An open Wilson line is gauge invariant only when its endpoints are attached to fields, defects, or boundaries carrying the conjugate charges.

An ’t Hooft line $H_m(C)$ is defined disorder-theoretically: the path integral is restricted to gauge fields with prescribed magnetic singularity or bundle transition around $C$. Its magnetic label $m$ is a cocharacter modulo the appropriate Weyl action. A dyonic Wilson–’t Hooft line carries a pair $(e,m)$.

“Closed and gauge invariant” is not yet the same as genuine. A genuine line can be defined without choosing an auxiliary surface. The global form of the gauge group and any discrete theta angle select a mutually local lattice of genuine dyonic lines. Schematically, two labels obey a Dirac pairing condition

$$\langle e,m'\rangle-\langle e',m\rangle\in\mathbb Z.$$

For example, pure $SU(N)$ admits the fundamental Wilson line and has an electric $\mathbb Z_N$ one-form symmetry. Passing to $SU(N)/\mathbb Z_N$ removes that fundamental line from the genuine spectrum and permits different magnetic lines. These theories share a Lie algebra and perturbative Feynman rules but not the same line diagnostics. Gaiotto and collaborators spell out this dependence on global form and screening in Gaiotto et al. 2015, §4.2, pp. 17–19.

From a rectangle to the static energy

Work in Euclidean time and take a rectangle $C_{R,T}$ of spatial width $R$ and temporal extent $T$. Inserting static sources in representation $R$ gives a spectral decomposition

$$\langle W_R(R,T)\rangle =\sum_n |c_n(R)|^2e^{-E_n(R)T}.$$

Provided the operator overlaps the ground state in that source sector,

$$E_0(R)=-\lim_{T\to\infty} \frac1T\log\langle W_R(R,T)\rangle.$$

Regulator-dependent endpoint masses and local perimeter terms must be subtracted or compared in differences. The remaining $R$ dependence produces the principal laws:

$$\begin{array}{lll} E_0(R)\sim\sigma R &\Longleftrightarrow& \log\langle W(R,T)\rangle_{\rm ren}\sim-\sigma RT, \\[4pt] E_0(R)\to E_\infty &\Longleftrightarrow& \text{screened or perimeter behavior}, \\[4pt] E_0(R)=E_\infty-\dfrac{c}{R}+\cdots &\Longleftrightarrow& \text{Coulomb behavior in four dimensions}. \end{array}$$

The first equivalence is the area law: a worldsheet of area $RT$ costs energy $\sigma R$ for time $T$. The third contains a nonlocal aspect-ratio dependence and should not be mislabeled as a mere perimeter term.

The order of limits matters. At fixed $R$, take $T$ large enough to project onto the lowest state; then study $R$ and the spatial volume. On a finite lattice, a square-loop fit over a short window cannot by itself distinguish a small string tension from Coulomb behavior or delayed string breaking.

Screening, string breaking, and operator overlap

Suppose dynamical matter can screen the external representation. The static-source Hilbert space then contains both an intact flux-tube state and two separated screened bound states. Their approximate energies behave as

$$E_{\rm string}(R)\simeq2M_{\rm end}+\sigma R, \qquad E_{\rm broken}(R)\simeq2M_{\rm scr}.$$

Near

$$R_b\simeq\frac{2(M_{\rm scr}-M_{\rm end})}{\sigma},$$

mixing produces an avoided crossing, and the true ground-state energy saturates. A thin Wilson loop can have very poor overlap with the broken-string state, so its effective energy may look linear well beyond $R_b$. A reliable variational basis includes Wilson strings and two-meson or screened-endpoint operators.

Equivalently, a Wilson line on which a dynamical field can end is not charged under an exact electric one-form symmetry. Its area law is not protected asymptotically. This screening mechanism and its perimeter-law endpoint are explicit in Fradkin 2013, §9.10, pp. 315–318.

Screening is representation dependent. In $SU(N)$ theory, adjoint fields preserve the $N$-ality class of a Wilson probe, whereas fundamental dynamical matter can screen every electric $N$-ality. The correct question is therefore not “Does the Wilson loop have an area law?” but “Which genuine unscreenable line, if any, has which renormalized asymptotic law?”

Phase classification with electric and magnetic lines

When both electric and magnetic genuine lines exist, their joint behavior is more informative than either alone.

Infrared response	Electric line	Magnetic or dyonic line	Qualified interpretation
Coulomb	Coulomb law	Coulomb law for an admissible dual probe	Gapless photon; a continuous one-form symmetry is spontaneously broken
Electric confinement	Area law for unscreenable electric charge	Perimeter behavior for the condensed or screened magnetic channel	Electric flux tube and nonzero string tension
Higgs-like electric screening	Perimeter behavior	Often an area law for an unscreenable magnetic probe	Electric charges screened; magnetic flux may form vortices
Oblique confinement	Area law for some electric/magnetic probes	Perimeter behavior for a particular dyonic combination	Condensed object carries both electric and magnetic charge
Discrete topological regime	Laws depend on the genuine line lattice	Nontrivial linking phases can survive	Large-loop magnitude alone is insufficient; include braiding and ground-state data

This table is a map, not a universal theorem. Matter content can remove one of the lines, and a discrete gauge theory can have perimeter-law lines whose mutual linking phase carries the essential information.

For a genuine line charged under an exact one-form symmetry, an area law means the symmetry is unbroken. Perimeter or Coulomb behavior means it is spontaneously broken after removable local geometric terms are treated properly. In four dimensions, breaking a continuous one-form symmetry gives a gapless photon. The precise statement, including dimensional restrictions, appears in Gaiotto et al. 2015, §5, pp. 28–33.

Line-operator classification procedure

Before making a phase claim, record the following in order.

1. Theory data: spacetime dimension, global group $G$, matter representations, theta and discrete theta parameters, and boundary conditions.
2. Candidate charges: electric weights, magnetic cocharacters, and allowed dyonic pairs.
3. Genuineness: which candidates require an attached surface and which mutually local set defines the theory.
4. Screening quotient: which lines can end on dynamical fields or defects.
5. Renormalization: subtraction of perimeter, endpoint, and cusp divergences, using one prescription across the comparison.
6. Infrared extraction: the $T\to\infty$ static energy, followed by large $R$ and infinite-volume limits.
7. Cross-check: agreement with the gauge-invariant spectrum, exact one-form symmetry realization, defect content, and string-breaking channels.

The conclusion should name the actual line: “the nonzero-$N$-ality Wilson line has tension $\sigma$” is testable; “Wilson loops confine” is not sufficiently specified.

Independent checks and failure diagnoses

• Center-charge check: verify that attaching each dynamical representation leaves the claimed screening class unchanged.
• Surface-independence check: deform any auxiliary surface. A phase acquired under deformation signals that the operator is not genuine without extra data.
• Pairing check: compute the electric–magnetic pairing of the proposed genuine lines; a nonintegral pairing means they cannot all be simultaneously genuine.
• Spectral check: compare $-T^{-1}\log\langle W(R,T)\rangle$ at increasing $T$ with a multichannel variational extraction.
• Shape check: use several aspect ratios. Area, perimeter, and Coulomb terms scale differently.
• Finite-size check: repeat at multiple volumes and keep the loop well separated from its periodic images.

Common pitfalls

Ignoring line renormalization. A bare perimeter divergence can dominate the logarithm of a loop. Subtract local divergences before assigning an infrared law.

Calling every closed loop genuine. Global form can force a Wilson or ’t Hooft operator to be attached to a surface. Only the declared genuine lattice supports the symmetry diagnosis.

Missing string breaking. Poor overlap with screened states can fake an area law. Include operators for every energetically allowed screening channel.

Inferring a phase from one line. Dyonic, magnetic, or topological data may distinguish phases with the same electric-loop magnitude.

Next, running and dynamical scales explains why perturbative gauge dynamics generates a scale, but not by itself which row of the phase table is realized. Numerical loop extraction is treated in Wilson and Polyakov loops, static energies, and screening diagnostics.

References

• Fradkin, Eduardo. Field Theories of Condensed Matter Physics. 2nd ed. Cambridge University Press, 2013, §9.10, pp. 315–318. DOI.
• Gaiotto, Davide, Anton Kapustin, Nathan Seiberg, and Brian Willett. “Generalized Global Symmetries.” Journal of High Energy Physics 2015, no. 2 (2015): 172, §§4.2 and 5, pp. 17–19 and 28–33. DOI. Open PDF.
• Schwartz, Matthew D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014, §25.2, pp. 488–493. DOI.