Nonabelian Surface Holonomy in Higher Gauge Theory

• Nonabelian surface holonomy is the two-dimensional analogue of path holonomy, assigning higher transport via Lie 2-groups and crossed modules.
• The construction paradigms—transport 2-functors, multiplicative integration, and higher Hochschild models—ensure coherence through boundary compatibility and controlled 3-curvature.
• Applications span Wilson-surface observables in gauge theory, monopole flux quantization, and discrete holonomy via surface braid groups, showcasing its versatility.

Nonabelian surface holonomy is the two-dimensional analogue of ordinary path holonomy: instead of assigning parallel transport to loops or paths, it assigns higher transport data to surfaces, typically relative to their boundary. In the modern higher-gauge formulation, the relevant structure is a strict Lie $2$-group or Lie crossed module $(H\xrightarrow{\Phi}G,\triangleright)$, so that $G$-valued $1$-holonomy along paths and $H$-valued $2$-holonomy over surfaces are linked by a boundary map and higher coherence laws (0808.1923, Williams, 4 Dec 2025). The subject also has loop-space, lattice, field-theoretic, homological, probabilistic, and discrete-group incarnations, including Wilson-surface observables, path-curvature formulas for monopoles, higher Hochschild realizations, Young-regular random surfaces, and finite nonabelian holonomy targets arising from surface braid groups (Bak et al., 26 May 2026, Cattaneo, 17 Nov 2025, Abbaspour et al., 2012, Lee et al., 2023, Tan, 2023).

1. Crossed modules, $2$-connections, and the basic notion of $2$-holonomy

A standard starting point is a Lie crossed module $(H\xrightarrow{\Phi}G,\triangleright)$, consisting of Lie groups $G,H$, a Lie group morphism $(H\xrightarrow{\Phi}G,\triangleright)$0, and a smooth action of $(H\xrightarrow{\Phi}G,\triangleright)$1 on $(H\xrightarrow{\Phi}G,\triangleright)$2 by automorphisms, satisfying the Peiffer identities

$(H\xrightarrow{\Phi}G,\triangleright)$3

Differentiation gives a differential crossed module $(H\xrightarrow{\Phi}G,\triangleright)$4, and a $(H\xrightarrow{\Phi}G,\triangleright)$5-connection on a smooth manifold $(H\xrightarrow{\Phi}G,\triangleright)$6 is then a pair

$(H\xrightarrow{\Phi}G,\triangleright)$7

subject to the fake-flatness condition

$(H\xrightarrow{\Phi}G,\triangleright)$8

Its $(H\xrightarrow{\Phi}G,\triangleright)$9-curvature is

$G$0

These are the standard local fields for higher parallel transport (Williams, 4 Dec 2025).

The categorical formulation replaces the path groupoid by a path $G$1-groupoid. In Schreiber–Waldorf’s framework, a gerbe with connection is encoded by a transport $G$2-functor

$G$3

or, in the one-object case, by a smooth $G$4-functor $G$5 for a strict Lie $G$6-group $G$7 (0808.1923). Under the corresponding dictionary, local differential data are pairs $G$8 of forms valued in the crossed-module Lie algebras, constrained by

$G$9

This is the higher analogue of the relation between ordinary connections and holonomy functors (0808.1923).

A complementary cocycle description uses a good cover $1$0 and local data

$1$1

subject to fake-curvature, overlap, and higher cocycle relations. The local $1$2-curvatures

$1$3

satisfy $1$4 and transform by the $1$5-action on overlaps (Glass, 2020). In this language, surface holonomy is a global object assembled from local $1$6-holonomies, edge corrections, and vertex corrections.

2. Construction paradigms: transport $1$7-functors, multiplicative integration, and higher Hochschild models

In the transport-$1$8-functor approach, a surface is represented by a bigon or by a marked closed surface decomposed into bigons, and the $1$9-functor assigns a $H$0-morphism in the target $H$1-category. This abstracts gluing, composition, and smoothness, and was designed to encompass nonabelian differential cocycles, Breen–Messing gerbes, abelian bundle gerbes, and nonabelian bundle gerbes in one framework (0808.1923).

A more analytic realization is provided by multiplicative integration. Given a smooth path $H$2, Yekutieli’s $H$3-dimensional multiplicative integral is

$H$4

where $H$5 is an ordered product of exponentials over a dyadic subdivision. For a kite $H$6, consisting of a path and a based surface, the surface holonomy is defined by the limit

$H$7

The zeroth-order term is

$H$8

and higher refinements subdivide the $H$9-simplex into $2$0 small triangles with a fixed surface-ordering scheme. The limit exists, depends smoothly on the data, and yields a transport $2$1-functor

$2$2

whose path and surface assignments satisfy composition, inversion, thin-homotopy invariance, and boundary compatibility (Williams, 4 Dec 2025).

Boundary compatibility is a defining feature: $2$3 Thus $2$4-holonomy in $2$5 projects to the $2$6-holonomy of the boundary loop in $2$7, exactly as required by crossed-module higher gauge theory (Williams, 4 Dec 2025).

A homological model appears in the higher Hochschild approach. Starting from a crossed module $2$8, Abbaspour–Wagemann use the Baez–Schreiber loop-space connection $2$9 and show that, after replacing the crossed module by an equivalent one with $2$0 abelian, the $2$1-holonomy over the torus is represented by a cycle

$2$2

In this formulation, $2$3-holonomy is encoded in higher Hochschild homology in the same way ordinary holonomy is encoded in ordinary Hochschild homology (Abbaspour et al., 2012).

3. Gauge covariance, closed surfaces, and Stokes-type laws

For open surfaces, surface holonomy is naturally relative to a boundary path or marking. For closed surfaces, the nonabelian case is subtler than the abelian one because the raw $2$4-valued quantity depends on choices. In the Schreiber–Waldorf formalism, a marked genus-$2$5 surface produces a naive surface holonomy as a $2$6-morphism in the target $2$7-category, but a canonical group-valued invariant arises only after passing to the reduction

$2$8

The resulting reduced surface holonomy is independent of the chosen reduction, depends only on the equivalence class of the marking, and in the abelian case coincides with the usual gerbe holonomy $2$9 (0808.1923).

Parzygnat sharpened the gauge-invariant target for sphere holonomy by introducing $2$0-conjugacy classes $2$1, the orbits of the $2$2-action on $2$3. For transport $2$4-functors, the gauge-invariant surface holonomy of a sphere is a map

$2$5

independent of the marking, of the chosen trivialization, and of the representative transport $2$6-functor in its equivalence class. The natural map $2$7 is surjective but not injective in general, so $2$8 retains more information than the reduced group (Parzygnat, 2014).

The higher Stokes principle is fundamental. In the multiplicative-integration setting, if $2$9 are two surfaces with common boundary and $(H\xrightarrow{\Phi}G,\triangleright)$0 is a $(H\xrightarrow{\Phi}G,\triangleright)$1-chain with $(H\xrightarrow{\Phi}G,\triangleright)$2, then

$(H\xrightarrow{\Phi}G,\triangleright)$3

This is the global $(H\xrightarrow{\Phi}G,\triangleright)$4-dimensional Stokes theorem, and in the abelian $(H\xrightarrow{\Phi}G,\triangleright)$5 case it becomes the familiar Wess–Zumino phase law

$(H\xrightarrow{\Phi}G,\triangleright)$6

Thus $(H\xrightarrow{\Phi}G,\triangleright)$7-curvature controls the failure of $(H\xrightarrow{\Phi}G,\triangleright)$8-holonomy to depend only on the boundary (Williams, 4 Dec 2025).

At the level of derivatives, Glass derived a global formula for the de Rham differential of nonabelian surface holonomy over squares. For a semi-global holonomy $(H\xrightarrow{\Phi}G,\triangleright)$9,

$G,H$0

The bulk term is the transported integral of the local $G,H$1-curvature; $G,H$2 and $G,H$3 are boundary terms built from the curving $G,H$4 and the overlap $G,H$5-forms $G,H$6. For spheres the boundary terms disappear, and the derivative reduces to the basepoint term plus the integrated $G,H$7-curvature (Glass, 2020). This gives a differential form of the principle that surface holonomy varies by $G,H$8-curvature.

4. Alternative realizations and basic obstructions

A persistent obstruction in the subject is that a naive nonabelian replacement of

$G,H$9

by a Lie-algebra-valued two-form with a surface-ordering prescription is not foliation-invariant. In the loop-space approach of Bak and Gustavsson, this obstruction is traced to Teitelboim’s no-go theorem: a straightforward nonabelian generalization using a Lie-algebra-valued two-form $(H\xrightarrow{\Phi}G,\triangleright)$00 cannot be made invariant under arbitrary changes of the foliation of the surface by loops unless the gauge group is abelian (Bak et al., 26 May 2026).

Their resolution is to replace the spacetime $(H\xrightarrow{\Phi}G,\triangleright)$01-form by a loop-algebra-valued $(H\xrightarrow{\Phi}G,\triangleright)$02-form on loop space,

$(H\xrightarrow{\Phi}G,\triangleright)$03

with loop algebra relations

$(H\xrightarrow{\Phi}G,\triangleright)$04

For a foliating family of loops $(H\xrightarrow{\Phi}G,\triangleright)$05, the surface holonomy is defined by

$(H\xrightarrow{\Phi}G,\triangleright)$06

hence

$(H\xrightarrow{\Phi}G,\triangleright)$07

The construction is nonabelian, uses only a one-form connection, and the paper proves invariance under deformations of the foliation and under reparametrizations of the surface (Bak et al., 26 May 2026).

A fully discrete realization is furnished by the lattice Wilson-surface construction on a bipartite hypercubic lattice. The basic datum is a universal plaquette operator

$(H\xrightarrow{\Phi}G,\triangleright)$08

with cyclic symmetry and a unitary compactness condition of the form

$(H\xrightarrow{\Phi}G,\triangleright)$09

Surface holonomy for a worldsheet is obtained by composing these plaquette tensors with index contractions along shared edges. The bipartite lattice permits spike string configurations, which are used to control changes in the number of color indices under string time evolution (Gustavsson, 28 Apr 2026). This is a nonabelian Wilson-surface model in which gauge covariance, backtracking, and local unitarity are enforced algebraically rather than categorically.

These alternative realizations clarify a basic point: nonabelian surface holonomy is not a single formula but a family of compatible constructions. Transport $(H\xrightarrow{\Phi}G,\triangleright)$10-functors and multiplicative integration emphasize crossed modules and $(H\xrightarrow{\Phi}G,\triangleright)$11-connections; loop-space and lattice models emphasize concrete operators acting on extended objects.

5. Field-theoretic, topological, homological, and probabilistic applications

In four-dimensional nonabelian $(H\xrightarrow{\Phi}G,\triangleright)$12 theory with cosmological term, a surface observable is constructed by coupling ambient fields $(H\xrightarrow{\Phi}G,\triangleright)$13 to a $(H\xrightarrow{\Phi}G,\triangleright)$14-dimensional $(H\xrightarrow{\Phi}G,\triangleright)$15-type theory on an embedded surface $(H\xrightarrow{\Phi}G,\triangleright)$16. The surface action is

$(H\xrightarrow{\Phi}G,\triangleright)$17

and the corresponding BV observable is

$(H\xrightarrow{\Phi}G,\triangleright)$18

Its expectation value gives $(H\xrightarrow{\Phi}G,\triangleright)$19-knot invariants, and BV pushforward carries it to a Yang–Mills electric surface operator whose classical leading term is of the form

$(H\xrightarrow{\Phi}G,\triangleright)$20

In this sense, nonabelian surface holonomy appears as a Wilson-surface-type observable and as a concrete realization of nonabelian electric fluxes and ’t Hooft operators (Cattaneo, 17 Nov 2025).

For magnetic monopoles, Parzygnat’s path-curvature $(H\xrightarrow{\Phi}G,\triangleright)$21-functor constructs surface holonomy from a path of ordinary holonomies. Given a transport functor $(H\xrightarrow{\Phi}G,\triangleright)$22 and a covering $(H\xrightarrow{\Phi}G,\triangleright)$23-group $(H\xrightarrow{\Phi}G,\triangleright)$24, the path-curvature $(H\xrightarrow{\Phi}G,\triangleright)$25-functor assigns to a bigon $(H\xrightarrow{\Phi}G,\triangleright)$26 the $(H\xrightarrow{\Phi}G,\triangleright)$27-class of the path

$(H\xrightarrow{\Phi}G,\triangleright)$28

This recovers the older loop-of-holonomies description of nonabelian magnetic flux inside the Schreiber–Waldorf formalism. For spheres, the resulting holonomy lies in $(H\xrightarrow{\Phi}G,\triangleright)$29, hence in a central subgroup of the covering group; explicit examples include the Dirac $(H\xrightarrow{\Phi}G,\triangleright)$30 monopole, the $(H\xrightarrow{\Phi}G,\triangleright)$31 monopole with flux $(H\xrightarrow{\Phi}G,\triangleright)$32, and $(H\xrightarrow{\Phi}G,\triangleright)$33 monopoles with flux in the center $(H\xrightarrow{\Phi}G,\triangleright)$34 (Parzygnat, 2014).

The higher-Hochschild viewpoint attaches $(H\xrightarrow{\Phi}G,\triangleright)$35-holonomy to the torus as a homology class. With the Baez–Schreiber loop-space connection $(H\xrightarrow{\Phi}G,\triangleright)$36, the Hochschild cycle

$(H\xrightarrow{\Phi}G,\triangleright)$37

defines a class in higher Hochschild homology associated to $(H\xrightarrow{\Phi}G,\triangleright)$38. This provides a homological algebra model for $(H\xrightarrow{\Phi}G,\triangleright)$39-holonomy and shows that crossed-module surface transport can be encoded in $(H\xrightarrow{\Phi}G,\triangleright)$40 once $(H\xrightarrow{\Phi}G,\triangleright)$41 is replaced by an equivalent abelian representative (Abbaspour et al., 2012).

A further extension treats nonsmooth and random surfaces. The space of surfaces carries horizontal and vertical concatenation, and higher surface representations are formalized as double functors

$(H\xrightarrow{\Phi}G,\triangleright)$42

The construction extends from smooth surfaces to bounded controlled $(H\xrightarrow{\Phi}G,\triangleright)$43-variation surfaces in the Young regime $(H\xrightarrow{\Phi}G,\triangleright)$44, and the resulting matrix surface holonomy satisfies nonabelian Stokes and nonabelian Fubini theorems in that regime. For random surfaces such as fractional Brownian sheets with Hurst parameter $(H\xrightarrow{\Phi}G,\triangleright)$45, expected surface holonomies become characteristic transforms for laws on surface space (Lee et al., 2023). This suggests that nonabelian surface holonomy can function not only as geometric transport but also as a statistical observable.

6. Discrete nonabelian holonomy targets from surface braid groups

A discrete group-theoretic viewpoint identifies nonabelian surface holonomy with homomorphisms from surface-related fundamental groups into finite nonabelian targets. For a closed oriented surface $(H\xrightarrow{\Phi}G,\triangleright)$46, the unordered configuration space of $(H\xrightarrow{\Phi}G,\triangleright)$47 points has fundamental group

$(H\xrightarrow{\Phi}G,\triangleright)$48

the surface braid group. Any representation

$(H\xrightarrow{\Phi}G,\triangleright)$49

can be read as a discrete nonabelian holonomy assignment for worldline motions on the surface, with $(H\xrightarrow{\Phi}G,\triangleright)$50 a finite or infinite structure group (Tan, 2023).

The finite-quotient problem asks for the smallest nonabelian groups that occur as quotients of $(H\xrightarrow{\Phi}G,\triangleright)$51. For $(H\xrightarrow{\Phi}G,\triangleright)$52 and $(H\xrightarrow{\Phi}G,\triangleright)$53, if the image of the embedded Artin braid group $(H\xrightarrow{\Phi}G,\triangleright)$54 is noncyclic, then every finite nonabelian quotient $(H\xrightarrow{\Phi}G,\triangleright)$55 satisfies

$(H\xrightarrow{\Phi}G,\triangleright)$56

with equality if and only if $(H\xrightarrow{\Phi}G,\triangleright)$57. If the quotient is braid-reduced, meaning the image of $(H\xrightarrow{\Phi}G,\triangleright)$58 is cyclic, then $(H\xrightarrow{\Phi}G,\triangleright)$59 is $(H\xrightarrow{\Phi}G,\triangleright)$60-step nilpotent and

$(H\xrightarrow{\Phi}G,\triangleright)$61

where $(H\xrightarrow{\Phi}G,\triangleright)$62 is the smallest prime dividing $(H\xrightarrow{\Phi}G,\triangleright)$63, $(H\xrightarrow{\Phi}G,\triangleright)$64 for odd $(H\xrightarrow{\Phi}G,\triangleright)$65, and $(H\xrightarrow{\Phi}G,\triangleright)$66 for $(H\xrightarrow{\Phi}G,\triangleright)$67. Equality occurs precisely for two families of $(H\xrightarrow{\Phi}G,\triangleright)$68-step nilpotent $(H\xrightarrow{\Phi}G,\triangleright)$69-groups, $(H\xrightarrow{\Phi}G,\triangleright)$70 and $(H\xrightarrow{\Phi}G,\triangleright)$71 (Tan, 2023).

These minimal quotients admit a natural holonomy interpretation. The symmetric group $(H\xrightarrow{\Phi}G,\triangleright)$72 gives the smallest permutation-type nonabelian holonomy when classical braiding survives. The braid-reduced minima are central extensions of $(H\xrightarrow{\Phi}G,\triangleright)$73 by a cyclic center, with commutator pairing modeled on a symplectic form; they are the smallest Heisenberg-type finite holonomy targets intrinsic to the surface contribution. This discrete classification shows that nonabelian surface holonomy has an arithmetic and finite-group side alongside its differential-geometric one.

Nonabelian surface holonomy is therefore best understood as a family of higher parallel-transport constructions unified by a common structural pattern: a $(H\xrightarrow{\Phi}G,\triangleright)$74-dimensional holonomy along boundary data, a $(H\xrightarrow{\Phi}G,\triangleright)$75-dimensional transport over surfaces, and a coherence law controlled by $(H\xrightarrow{\Phi}G,\triangleright)$76-curvature or its discrete analogue. Transport $(H\xrightarrow{\Phi}G,\triangleright)$77-functors and crossed modules supply the abstract framework, multiplicative integration supplies an explicit analytic model, loop-space and lattice constructions supply concrete operator realizations, field theory supplies Wilson-surface and flux observables, higher Hochschild homology supplies an algebraic encoding, random-surface theory supplies probabilistic extensions, and surface braid groups supply finite nonabelian targets (0808.1923, Williams, 4 Dec 2025, Bak et al., 26 May 2026, Cattaneo, 17 Nov 2025, Abbaspour et al., 2012, Lee et al., 2023, Tan, 2023).