Transgression Forms in Geometry and Gauge Theory

• Transgression forms are secondary geometric objects that explicitly encode the difference between characteristic forms by integrating variations between two connections on a principal bundle.
• They enable gauge invariance by translating bulk field variations into precise boundary terms, thereby ensuring finite action principles in theories like AdS gravity and Chern–Simons models.
• Extensions to mapping spaces, loop spaces, and higher-categorical structures demonstrate their versatility in modeling boundary effects, gerbes, and advanced gauge theories.

Transgression forms are secondary geometric objects that encode the exactness of the difference between characteristic forms attached to two sets of geometric data. In the classical Chern–Weil setting, if $A$ and $\bar A$ are connections on the same principal bundle with curvatures $F$ and $\bar F$, an invariant polynomial $P$ satisfies

$P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$

In that sense, a transgression form is an explicit primitive for a difference of characteristic forms. The term also designates degree-lowering operations obtained by integration along a source manifold, as in loop-space and mapping-space geometry, and more generally natural transformations that pass from bulk data to boundary, lower-degree, or corona data in higher gauge theory, gerbe theory, and coarse homotopy theory (Pais et al., 2023, Vizman, 2011, Bunke, 18 Dec 2025).

1. Chern–Weil origin and the classical formula

For two connections $A$ and $\bar A$ on the same principal bundle, the transgression form used in Chern–Weil theory is

$$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$

with

$F_t=dA_t+A_t^2.$

Its defining property is

$\bar A$0

so the cohomology class of the characteristic form is independent of the chosen connection, while the transgression measures the change at the level of differential forms (Pais et al., 2023).

Ordinary Chern–Simons forms arise as the degenerate case in which one of the two connections is set to zero: $\bar A$1 This identifies Chern–Simons theory as a special instance of transgression theory rather than a separate construction. In the gauge-theoretic literature, the same point is often expressed by the decomposition

$\bar A$2

where $\bar A$3 is an explicit boundary term. Transgression forms also satisfy structural identities such as antisymmetry,

$\bar A$4

and the triangle equation through an intermediate connection, properties that are used repeatedly in gauge-field decompositions and action principles (Salgado et al., 2014).

A common misconception is to treat transgression as merely a convenient rewriting of Chern–Simons theory. The cited works instead use the second connection as essential data: it is what globalizes the construction on nontrivial bundles and what converts quasi-invariance into strict gauge invariance in the simultaneous transformation of both fields (Salgado et al., 2014).

2. Gauge invariance, boundary terms, and transgression actions

In gauge theory on manifolds with boundary, the decomposition

$\bar A$5

has direct dynamical significance. The boundary term $\bar A$6 is not ornamental: it is the term that makes the action

$\bar A$7

genuinely gauge invariant on manifolds with boundary, provided the gauge parameters agree at the boundary,

$\bar A$8

This sharply contrasts with an ordinary Chern–Simons action, which is only quasi-invariant and shifts by an exact form (Pais et al., 2023).

The Hamiltonian analysis of transgression field theory on a manifold with boundary can be carried out while keeping the transgression boundary term throughout Dirac’s algorithm. Using the generalized Poisson bracket of Soloviev and Bering, which preserves both bulk and boundary variations, the boundary contribution in the action is translated into boundary terms in the generators of gauge transformations and spatial diffeomorphisms. In this formulation, the corresponding conserved surface charges are read directly from the symmetry generators, without adding ad hoc regularizing boundary terms to the first-class constraints. When one connection is set to zero, the resulting formulas reduce to the known conserved-charge expressions of higher-dimensional Chern–Simons theories (Pais et al., 2023).

In AdS gravity, transgression-based action principles are finite from the outset. The second connection functions as a regulator or background in a gauge-covariant subtraction scheme, so one obtains a well-defined variational principle, finite Noether charges, and a finite Euclidean action under the asymptotic condition that the AdS gauge curvature be finite. The same framework also organizes gauge and diffeomorphism anomalies with backgrounds: if both connections transform, the transgression is strictly invariant; anomalies appear when the reference connection is held fixed, and the resulting consistent anomalies obey the Wess–Zumino consistency condition (Mora, 2014, Mora, 2014).

Transgression forms also serve as parent actions for lower-dimensional boundary theories. In Poincaré-invariant topological gravity, an even-dimensional topological gravity action is realized as the boundary reduction of a transgression field theory in one higher dimension, with the boundary theory taking the form of a gauged Wess–Zumino–Witten model associated with linear and nonlinear realizations of the Poincaré group (Salgado et al., 2014).

3. Mapping-space transgression and induced differential forms

A second major meaning of transgression is fiber integration on mapping spaces. Let $\bar A$9 be a compact oriented $F$0-manifold, $F$1 a smooth manifold, and $F$2 the Fréchet manifold of smooth maps. The basic construction is the hat pairing

$F$3

which defines a form on $F$4 of degree $F$5 from $F$6 and $F$7. Explicitly, if $F$8, then

$F$9

This construction is natural with respect to diffeomorphisms of both $\bar F$0 and $\bar F$1, and it is compatible with exterior differentiation: $\bar F$2 If $\bar F$3 has boundary, an additional restriction term appears (Vizman, 2011).

The transgression map is the specialization of the hat pairing to the constant function $\bar F$4: $\bar F$5 This is the degree-lowering map usually called transgression in mapping-space geometry. When $\bar F$6, it becomes the standard transgression to the free loop space,

$\bar F$7

The same framework also contains the bar map $\bar F$8 for a volume form $\bar F$9, which preserves degree, and it relates the hat construction on $P$0 to the tilda calculus on the nonlinear Grassmannian $P$1 via

$P$2

The applications listed in the source include symplectic forms on mapping spaces, Hamiltonian actions, momentum maps for branes and fluid dynamics, and current algebras with central extensions (Vizman, 2011).

4. Loop spaces, gerbes, fusion structures, and regression

In the geometry of loop spaces, transgression lifts cohomological operations to fully geometric constructions. For a diffeological abelian bundle gerbe $P$3 with connection over $P$4, transgression produces a principal $P$5-bundle $P$6 whose fiber over a loop $P$7 is the set of isomorphism classes of trivializations of $P$8. The induced connection is determined by surface holonomy, and its curvature is

$P$9

This loop-space bundle carries extra structure absent from an arbitrary principal bundle over $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$0: a fusion product over triples of paths, a compatible connection, and a superficiality or thinness condition controlling dependence on thin homotopies (Waldorf, 2010).

These additional structures make transgression invertible. For gerbes with connection, transgression and regression give an equivalence between the homotopy 1-category of diffeological abelian gerbes with connection over $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$1 and the category of fusion bundles over $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$2 with compatible, symmetrizing, superficial connection (Waldorf, 2010). In the connection-free formulation, the corresponding loop-space objects are thin fusion bundles. A thin structure is built from thin homotopies of loops, namely paths in $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$3 whose adjoint map $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$4 has differential of rank at most $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$5. The resulting equivalence between gerbes on $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$6 and thin fusion bundles on $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$7, up to homotopy of morphisms, is monoidal and natural in $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$8 (Waldorf, 2011).

This loop-space viewpoint has concrete consequences. Lifting problems for central extensions of Lie groups become problems about fusion-preserving thin sections of transgressed bundles; the cited applications include loop-space formulations of spin structures, complex spin structures, and spin connections (Waldorf, 2011). In a related formulation, $P(F)-P(\bar F)=d\mathcal T_{2n+1}[A,\bar A].$9 is spin if and only if $A$0 admits a fusion-preserving orientation (Waldorf, 2010). The same transgression–regression method underlies the construction of string 2-group models: the basic gerbe on a compact, simple, simply connected Lie group transgresses to a bundle over the loop group carrying both a fusion product and the Mickelsson product, and regression turns this into a multiplicative gerbe, hence into a central Lie 2-group extension (Waldorf, 2012).

At the cohomological level, loop-fusion cohomology isolates the image of ordinary transgression. For a compact smooth manifold $A$1 and an abelian group $A$2, the loop-fusion cohomology of the continuous loop space is defined by Čech cochains satisfying multiplicativity constraints with respect to the fusion and figure-of-eight products. The main theorem identifies these groups with the ordinary cohomology of $A$3, and the usual transgression map factors through this identification (Kottke et al., 2013).

5. Higher, categorical, and extended transgression theories

Transgression persists under categorification. For a crossed module of groupoids $A$4, the crossed product groupoid $A$5 carries a natural groupoid homomorphism

$A$6

This map induces a transgression in singular cohomology,

$A$7

and the resulting map coincides with the Tu–Xu transgression map, up to the same sign convention in the alternating-sum formula (Cai, 2020).

In strict higher gauge theory based on a Lie crossed module, a 2-connection is a pair

$A$8

with fake curvature and 2-curvature

$A$9

For two 2-connections $\bar A$0 and $\bar A$1, the higher transgression form $\bar A$2 satisfies

$\bar A$3

When the two 2-connections are related by a higher gauge transformation, this transgression produces higher WZW and gauged WZW terms. For the symmetric invariant polynomial associated with the differential crossed module, the pure-gauge higher WZW term vanishes identically and the higher gauged WZW term is exact. Consequently, the higher Chern–Simons action is higher-gauge invariant on closed manifolds, while on manifolds with boundary all gauge dependence is encoded in boundary terms (Song, 12 Apr 2026).

A different extension appears in free differential algebras with one $\bar A$4-form sector. There, the generalized Chern–Weil theorem yields invariant densities $\bar A$5 built from the ordinary curvature and the higher-degree curvature, together with a generalized transgression formula

$\bar A$6

The corresponding generalized Chern–Simons form is obtained by setting the reference fields to zero. The same framework produces two anomaly families, associated respectively with the ordinary gauge parameter and the higher-degree gauge parameter (Salgado, 2021).

6. Specialized analytical, symplectic, modular, and coarse variants

Several recent works use transgression in settings that are not reducible to ordinary principal-bundle Chern–Weil theory. In primitive cohomology on a symplectic manifold $\bar A$7, a primitive superconnection $\bar A$8 acts on pairs of forms, the primitive differential is

$\bar A$9

and for a smooth family $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$0 one has the primitive transgression formula

$$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$1

This implies independence of the associated primitive characteristic class from the chosen primitive superconnection (Zhuang, 31 Dec 2025). In a related mapping-cone setting induced by a closed 2-form, the Mathai–Quillen-type mapping-cone Thom form is closed with respect to the mapping-cone differential, integrates to $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$2 along the fiber, and satisfies an explicit transgression formula under variation of the connection and skew-adjoint endomorphism data (Zhuang, 26 Mar 2026).

Analytical transgression formulas also appear in index theory and supersymmetric field theory. For the $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$3 supersymmetric path integral on the smooth loop space of a closed even-dimensional Riemannian spin manifold with fixed topological spin structure, any smooth family of metrics $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$4 produces a Chern–Simons current $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$5 such that

$$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$6

This transgression formula implies that the resulting loop-space current determines a metric-independent differential-topological invariant essentially given by the $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$7-genus (Boldt et al., 2021). For families of odd-dimensional vertical Dirac operators with a single eigenvalue of multiplicity one crossing zero transversally along a hypersurface $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$8, the generalized transgression formula for the $$\mathcal T_{2n+1}[A,\bar A]=(n+1)\int_0^1 dt\,\big\langle (A-\bar A)F_t^n\big\rangle, \qquad A_t=tA+(1-t)\bar A,$$9-form becomes

$F_t=dA_t+A_t^2.$0

so the usual family-index density acquires an additional current supported on the zero-crossing locus (Wittmann, 2015).

In the theory of modular characteristic forms, even-dimensional cancellation formulas can be transgressed to odd-dimensional secondary forms $F_t=dA_t+A_t^2.$1, whose top-degree components are modular forms over $F_t=dA_t+A_t^2.$2, $F_t=dA_t+A_t^2.$3, and $F_t=dA_t+A_t^2.$4, respectively (Liu et al., 2023). At a different level of abstraction, coarse homotopy theory defines a transgression as a natural transformation

$F_t=dA_t+A_t^2.$5

where $F_t=dA_t+A_t^2.$6 is the Higson corona functor. In that setting, transgression relates coarse homology theories to boundary theories on the corona and is studied together with analytic and topological transgressions and with algebraic and homotopy-theoretic Chern characters (Bunke, 18 Dec 2025).

Across these settings, transgression retains a common structural role: it converts a difference, variation, or asymptotic defect into an exact term, a secondary invariant, or a boundary object. What changes from one domain to another is not that role, but the ambient category in which it is realized—principal bundles, mapping spaces, gerbes, Lie crossed modules, free differential algebras, superconnections, or coarse homology theories.