Higher-Form Symmetry from Operators and Linking

An internal invertible $p$-form symmetry in a $d$-dimensional QFT is a group $G^{(p)}$ represented by topological operators $U_g(\Sigma^{d-p-1})$ on codimension-$(p+1)$ supports. These operators fuse according to the group law and act on $p$-dimensional charged operators by linking. The degree $p$ therefore fixes three pieces of geometry at once: the charged support has dimension $p$, a small transverse link has dimension $d-p-1$, and the symmetry operator can occupy that link.

This page treats ordinary invertible, group-like, internal higher-form symmetry for $0\leq p\leq d-1$. A finite symmetry need not have a local Noether current, an anomaly can obstruct gauging without erasing the symmetry, and a linked phase does not make the charged operator topological. Actions on higher-dimensional defect categories, non-invertible symmetry, spontaneous breaking, and full higher-gauge foundations are separate questions.

Required background. Support, Codimension, and Operator Data supplies normal links, orientations, genuine-versus-attached operators, and endpoint and junction data. Differential Forms, Integration, Orientation, and Stokes Theorem supplies form degree, integration on oriented cycles, the exterior derivative, and Stokes’ theorem.

Helpful background. Homotopy, Degree, Winding, and Covering Spaces supplies the deformation and winding language used to decide when a linking configuration can be changed without a crossing.

Topological operators define a p-form symmetry

Work first in an oriented Euclidean $d$-manifold $M$. Let $C^p\subset M$ be the closed oriented support of a genuine charged operator. At a smooth interior point of $C$, the normal disk is $D^{d-p}$ and its oriented boundary is

$$S_{\mathrm{link}}^{d-p-1}=\partial D^{d-p}.$$

The support of a symmetry operator has exactly this dimension. For every $g\in G^{(p)}$, the theory supplies a closed topological defect $U_g(\Sigma^{d-p-1})$ satisfying

$$\begin{aligned} U_g\otimes U_h&\simeq U_{gh}, & U_e&\simeq \mathbf 1,\\ U_g(\overline\Sigma)&\simeq U_{g^{-1}}(\Sigma). \end{aligned}$$

Topological means that $U_g$ can be deformed through the complement of all forbidden crossings while carrying its orientation, junctions, endpoints, background data, and any required framing. Invertible means that the two-sided fusion inverse exists; it is stronger than merely having an orientation-reversed drawing. Fusion equivalence also does not choose or normalize the junction operators that realize it.

For a simple charged sector carrying a character $\chi$, a small linking sphere implements the action

$$U_g\!\left(S_{\mathrm{link}}^{d-p-1}\right) \mathcal O_\chi(C^p) =\chi(g)\,\mathcal O_\chi(C^p).$$

For $p=0$, the group can be non-Abelian and the right-hand side is generally a linear representation rather than one character. For $p\geq1$, the local interchange of two codimension-greater-than-one topological generators reverses their fusion order, so an ordinary internal invertible $p$-form symmetry group is Abelian. This locality argument does not turn a non-invertible fusion algebra into a group, and global charge operators on a topologically nontrivial spatial manifold can have additional algebraic data.

These statements and their topological-deformation argument are developed in Bhardwaj et al. 2024, § 2.2.1, arXiv v2, pp. 12–15, Definition 2.1 and eqs. (2.38)–(2.55), Open PDF and in the original operator-first formulation Gaiotto et al. 2015, § 3, arXiv v2, pp. 11–13, especially eqs. (3.1)–(3.4), Open PDF.

A label group is not automatically the faithful physical symmetry. If $\rho_i$ is the action on the $i$th admitted genuine charged sector, define

$$K=\bigcap_i\ker\rho_i\subseteq G^{(p)}.$$

In an Abelian simple sector, $\rho_i$ is just its character $\chi_i$. The faithful group acting on all admitted sectors is $G^{(p)}/K$. Formal probes that are not operators of the chosen global theory cannot by themselves remove this kernel. Conversely, an allowed endpoint or dynamical screening object can identify charge sectors and enlarge $K$.

This common-kernel test is the operator form of the faithfulness criterion in Bhardwaj et al. 2024, § 4.2.2, arXiv v2, pp. 74–75, especially eqs. (4.65)–(4.67), Open PDF.

The degree dictionary in d dimensions

For a continuous $U(1)$ $p$-form symmetry, choose a $(p+1)$-form current $j_{p+1}$. In the Euclidean convention used here, its Hodge dual

$$\widetilde j_{d-p-1}=\star j_{p+1}$$

is closed away from charged insertions,

$$\mathrm d\widetilde j_{d-p-1}=0.$$

On a closed oriented $(d-p-1)$-cycle $Y$, the continuous charge and its topological operator can be normalized as

$$Q(Y)=\int_Y\widetilde j_{d-p-1}, \qquad U_\alpha(Y)=\exp\!\bigl(i\alpha Q(Y)\bigr).$$

A background field has form degree $p+1$ and a $p$-form gauge parameter,

$$B_{p+1}\longmapsto B_{p+1}+\mathrm d\Lambda_p.$$

The degree check is that $B_{p+1}\wedge\widetilde j_{d-p-1}$ is a top form. Normalizations, contact terms, global higher-bundle data, and the Ward identity are developed on the next page. A finite $p$-form symmetry generally has no local current at all; its background is discrete cocycle or flat higher-gauge data, not an arbitrary globally defined differential form. This distinction is explicit in Gaiotto et al. 2015, § 3, arXiv v2, pp. 11–13, Open PDF and Schäfer-Nameki 2024, § 2.5, arXiv v2, pp. 23–25, especially eqs. (2.33)–(2.42), Open PDF.

The table is the chapter’s degree cross-check. Its first two columns give the general rule; the last two expose the ordinary-symmetry and four-dimensional one-form limits. The current rows apply only to continuous symmetries.

The degree dictionary for an internal p-form symmetry

Datum	General p-form rule	p = 0 check	p = 1 in d = 4
Charged support	Dimension p	Point operator	Line operator
Symmetry operator	Dimension d − p − 1; codimension p + 1	Hypersurface; codimension 1	Surface; codimension 2
Continuous current	Form degree p + 1	One-form current	Two-form current
Closed dual current	Form degree d − p − 1	Degree d − 1	Two-form in four dimensions
Charge cycle	Dimension d − p − 1	Spatial hypersurface	Closed surface
Background and parameter	Background degree p + 1; parameter degree p	One-form background; scalar parameter	Two-form background; one-form parameter
Top-form coupling check	Background and dual-current degrees add to d	One plus d − 1	Two plus two

The allowed range is also dimensional. For $p=d-1$, the symmetry generators are point supported, the dual current is a zero-form, and cluster or decomposition questions need special care. For $p\geq d$, a codimension-$(p+1)$ support does not exist in $d$ dimensions. Negative-form symmetries, subsystem symmetries, spacetime symmetries, and higher charges on operators of dimension greater than $p$ are outside this page’s definition.

Linking realizes the action

The small-link equation is local. A global linking formula needs more topology. Let $\Sigma^{d-p-1}$ and $C^p$ be disjoint closed oriented supports in a region where $\Sigma$ bounds an oriented $(d-p)$-chain $X$ and where changing $X$ cannot change its intersection with $C$. Define

$$\operatorname{Lk}(\Sigma,C)=X\mathbin{\boldsymbol{\cdot}}C, \qquad \partial X=\Sigma.$$

For a simple Abelian charge sector, keep every other insertion $\mathcal X$ outside the sweep and assume the normalized symmetry defect can be removed after it is unlinked. Topological motion then gives

$$\begin{aligned} &\left\langle U_g(\Sigma)\,\mathcal O_\chi(C)\,\mathcal X \right\rangle\\ &\qquad= \chi(g)^{\operatorname{Lk}(\Sigma,C)} \left\langle\mathcal O_\chi(C)\,\mathcal X\right\rangle . \end{aligned}$$

Reversing either support orientation negates the linking number. In a unitary character sector this inverts the phase. If the supports are not null-homologous in the declared region, if a boundary turns the link into a relative cycle, or if the sweep crosses another insertion, the displayed integer and correlator reduction need not be available. A general non-Abelian $0$-form action or a reducible charged sector is a linear map, not one scalar character.

An anomaly is another distinct qualification. It can obstruct coupling to arbitrary backgrounds or choosing freely reconfigurable topological junctions while the symmetry operators and their action still exist. Thus “anomalous” does not mean “not a symmetry,” and “exact” does not mean “gaugeable.” The network and anomaly ceiling appear already in Gaiotto et al. 2015, § 3, arXiv v2, pp. 12–13, especially eqs. (3.2)–(3.4), Open PDF.

Charge-N matter leaves a finite one-form network

Consider a controlled four-dimensional example. Work on a closed oriented spin Euclidean four-manifold, set $\theta=0$, and hold magnetic backgrounds outside the calculation. Let $\mathfrak a$ be a faithfully normalized compact $U(1)$ connection,

$$\mathfrak a\longmapsto\mathfrak a+\mathrm d\lambda, \qquad \lambda\sim\lambda+2\pi.$$

Assume that the dynamical electric charges generate exactly $N\mathbb Z$ for an integer $N\geq2$—for example, one charge-$N$ matter field and its conjugate, with no smaller electric charge. The spin structure is held fixed and does not produce the linking phase. The Wilson line and its surviving charge are

$$W_n(C)=\exp\!\left(i n\oint_C\mathfrak a\right), \qquad n\in\mathbb Z, \qquad r=[n]_N\in\mathbb Z_N.$$

Why does the continuous electric one-form symmetry reduce to $\mathbb Z_N^{(1)}$? A continuous angle $\varphi$ acts on $W_n$ by $e^{in\varphi}$. The charge-$N$ endpoint must be neutral, so

$$e^{iN\varphi}=1 \quad\Longleftrightarrow\quad \varphi=\frac{2\pi\alpha}{N}, \qquad \alpha\in\mathbb Z_N.$$

Equivalently, charge-$N$ matter identifies the screening classes $n\sim n+N$. This is an equivalence of charge sectors, not an equality of bare or renormalized Wilson-line formulas.

Let $U_\alpha(\Sigma)$ be the normalized topological generator on a closed oriented surface. Assume the residual electric $\mathbb Z_N^{(1)}$ symmetry is exact on this restricted background and that its invertible group-like surface network has coherent topological junction data. Take $C$ and $\Sigma$ to be disjoint in a linking ball, keep $\mathcal X$ outside the sweep, and normalize a positive unit link of $U_1$ with $W_1$ to give $e^{2\pi i/N}$. Then

$$\begin{aligned} &\left\langle U_\alpha(\Sigma)\,W_n(C)\,\mathcal X \right\rangle\\ &\qquad= \exp\!\left[ \frac{2\pi i}{N}\,\alpha r\, \operatorname{Lk}(\Sigma,C) \right] \left\langle W_n(C)\,\mathcal X\right\rangle . \end{aligned}$$

The screening quotient and its topological surface realization are developed in Bhardwaj et al. 2024, § 3.2.1, arXiv v2, pp. 29–31, especially eqs. (3.10), (3.16)–(3.19), Open PDF.

The support degrees now read directly from the table: $W_n$ is a charged line, $U_\alpha$ is a codimension-two surface, and a junction of surface sheets is a line. With two incoming surfaces and one outgoing surface meeting along an oriented line $K$, write

$$U_\alpha\otimes U_\beta \simeq U_{\alpha+\beta\;\mathrm{mod}\;N},$$ $$\mathcal I_{\alpha,\beta}^{\ \gamma}(K): U_\alpha\otimes U_\beta\longrightarrow U_\gamma, \qquad \alpha+\beta-\gamma=0\pmod N.$$

The incidence condition is necessary, not a construction or normalization of $\mathcal I$. Its linking check is the character identity

$$\exp\!\left[ \frac{2\pi i}{N}\,r(\alpha+\beta-\gamma) \right]=1.$$

For an endpoint check, orient a path $P$ from $y$ to $x$. A charge-$N$ open Wilson line transforms as

$$W_N(P:y\to x) \longmapsto e^{iN[\lambda(x)-\lambda(y)]}W_N(P:y\to x).$$

If $\psi_N\mapsto e^{iN\lambda}\psi_N$, the composite

$$\overline\psi_N(x)\,W_N(P:y\to x)\,\psi_N(y)$$

is gauge invariant. Hence the charge class of $W_N$ is screened, whereas $W_1$ cannot end on the declared charge-$N$ matter and detects the faithful $\mathbb Z_N^{(1)}$ action. Adding charge-one matter would remove this electric one-form symmetry entirely. None of these facts decides the magnetic symmetry, the phase of the theory, or whether the finite symmetry should be gauged.

Matter, dimension, and boundaries can change the answer

The operator definition is exact only after its domain is fixed.

• Screening changes the faithful group. If a charged $p$-operator can end on admitted lower-dimensional dynamical data, its charge may become trivial or identified with another charge. The surviving symmetry is dual to the unscreened genuine charge sectors, not to every formal probe label.

• Finite and continuous symmetries have different local data. A finite symmetry can have topological operators, characters, backgrounds, and anomalies without a local conserved current.

• Boundaries require relative data. A small sphere can become a hemisphere or relative cycle. A symmetry operator may meet the boundary only on declared boundary defects, and charge can flow into boundary degrees of freedom.

• Anomalies preserve the symmetry but obstruct operations. Failure of a freely reconfigurable junction network can obstruct gauging. It does not by itself make every symmetry operator nontopological.

• Framing is conditional. Orientation is needed for the signed linking action above. A normal framing is extra data only in a theory or regulator whose operators are framing-sensitive.

• Several symmetries need not form a higher group. Coexisting $0$-form and $1$-form groups form an ordinary product unless their background transformations or extension data are coupled.

The endpoint interpretation and its relation to screening are also explained in Schäfer-Nameki 2024, § 2.4, arXiv v2, pp. 20–22, Open PDF. This page stops at the ordinary invertible $p$-form action; higher charges on larger defects and their higher-representation structure belong to the mathematical handoff below.

Common pitfalls

Calling any codimension-$(p+1)$ defect a symmetry operator. The defect must be topological on its declared deformation domain, invertible, equipped with a group law, and act consistently on charged sectors. Codimension alone does none of this.

Assuming a finite symmetry has a Noether current. The topological operator and linking action are primary. Local currents belong to continuous symmetries and can disappear when a continuous group is reduced to a finite subgroup.

Using one linked phase as a complete classification. The phase records one character pairing. It does not determine all charged sectors, anomaly data, junction coherence, spin or framing dependence, gauging choices, or the infrared phase.

Equating an anomaly with explicit breaking. An anomalous symmetry still acts. The obstruction is to a compatible background or gauging operation under the stated hypotheses.

Forgetting genuine and boundary data. An attached probe, an operator that can end, and a boundary-absorbed charge need not define the same faithful $p$-form symmetry as a closed genuine operator in the bulk.

Check your understanding

Use each answer outline to check the stated assumptions as well as the result.

1. In $d=4$, fill every degree for $p=1$ and explain why the charged object and the symmetry operator have different dimensions.

   Answer outline

   The charged support has dimension one, so it is a line. Its normal disk is three-dimensional and its local link is $S^2$. The symmetry operator is therefore a two-dimensional surface of codimension two. A continuous current and its Hodge dual are both two-forms; the charge cycle and background field are also two-dimensional or degree two, respectively.

2. Show that reversing either support in a unitary character-valued linking action inverts the phase.

   Answer outline

   Reversing $\Sigma$ reverses the spanning-chain orientation, while reversing $C$ reverses the intersection sign. Either operation sends $\operatorname{Lk}(\Sigma,C)$ to its negative. A unit-modulus character to the negative integer power is the inverse, equivalently its complex conjugate.

3. Suppose a finite group $G^{(p)}$ acts on two genuine charged sectors through representations $\rho_1$ and $\rho_2$. What is the faithful group seen by those sectors?

   Answer outline

   Compute $K=\ker\rho_1\cap\ker\rho_2$. Elements of $K$ act trivially on both sectors, so the faithful group is $G^{(p)}/K$. For Abelian simple sectors, replace $\rho_i$ by the corresponding character $\chi_i$. Adding another genuine charged sector can shrink the kernel; adding endpoint data can instead identify or screen sectors.

4. For $N=6$, take $n=8$, $\alpha=4$, and a positive unit link. Compute the phase and check the junction $U_4\otimes U_5\to U_3$.

   Answer outline

   The surviving charge is $r=[8]_6=2$, so the phase is $e^{2\pi i(4)(2)/6}=e^{2\pi i/3}$. The signed junction incidence is $4+5-3=6=0\pmod 6$. This checks charge compatibility but does not construct or normalize the junction operator.

5. Which rows of the degree table remain meaningful for a finite $p$-form symmetry?

   Answer outline

   The charged support, symmetry-operator dimension and codimension, charge cycle, linking action, and background degree remain. A discrete higher-gauge background replaces an unrestricted differential form. The local current and closed-dual-current rows need not exist.

6. Diagnose the claim: “A theory has both a $0$-form group and a $1$-form group, so its symmetry is a higher group.”

   Answer outline

   Coexistence only gives a candidate product. Higher-group mixing requires coupled background transformations or equivalent extension/Postnikov data. The correct repair route is the higher-group background page linked below.

Continue to currents, phases, gauging, and higher groups

Higher-Form Currents, Charges, Backgrounds, and Ward Identities is the immediate next step: it derives the current/background dictionary, contact terms, large transformations, and finite-symmetry qualifications.

For concrete gauge-theory realizations, continue to Electric and Magnetic One-Form Symmetries. Breaking Higher-Form Symmetry and Diagnosing Phases owns area/perimeter qualifications and phase diagnostics, while Gauging a Higher-Form Symmetry owns the sum over higher-gauge backgrounds and the resulting operator projection or attachment.

Higher-Group Symmetry and Coupled Backgrounds gives the test for genuine mixing rather than a direct product. The theorem-level and higher-categorical formulation belongs to Invertible p-Form Symmetries and Topological Defects.

References

• Bhardwaj, Lakshya, Lea E. Bottini, Ludovic Fraser-Taliente, Liam Gladden, Dewi S. W. Gould, Arthur Platschorre, and Hannah Tillim. “Lectures on Generalized Symmetries.” Physics Reports 1051 (2024): 1–87. DOI. Open PDF, arXiv v2.

• Gaiotto, Davide, Anton Kapustin, Nathan Seiberg, and Brian Willett. “Generalized Global Symmetries.” Journal of High Energy Physics 2015, no. 2 (2015): 172. DOI. Open PDF, arXiv v2.

• Schäfer-Nameki, Sakura. “ICTP Lectures on (Non-)Invertible Generalized Symmetries.” Physics Reports 1063 (2024): 1–55. DOI. Open PDF, arXiv v2.