nLab self-dual higher gauge theory




physics, mathematical physics, philosophy of physics

Surveys, textbooks and lecture notes

theory (physics), model (physics)

experiment, measurement, computable physics

Differential cohomology



The standard action functional for the higher U(1)-gauge field given by a circle n-bundle with connection (P,)(P, \nabla) over a (pseudo) Riemannian manifold (X,g)(X,g) is

XF gF , \nabla \mapsto \int_X F_\nabla \wedge \star_g F_{\nabla} \,,

where F F_\nabla is the curvature (n+1)(n+1)-form. If the dimension

dimX=4k+2 dim X = 4 k + 2

then the Hodge star operator squares to +1+1 (Lorentzian signature) or 1-1 (Euclidean signature) on Ω k+1(X)\Omega^{k+1}(X). Therefore it makes sense in these dimensions to impose the self-duality or chirality constraint

±F =F . \pm F_\nabla = \star F_\nabla \,.

With this duality constraint imposed, one speaks of self-dual higher gauge fields or chiral higher gauge fields or higher gauge fields with self-dual curvature. (These are a higher degree/dimensional generalization of what in Yang-Mills theory are called Yang-Mills instanton field configurations.)

Since imposing the self-duality constraint on the fields that enter the above action functional makes that functional vanish identically, self-dual higher gauge theory is notorious for being subtle in that either it does not have a Lagrangian field theory description, or else a somewhat intricate indirect one (e.g. Witten 96, p. 7, Belov-Moore 06a). But instead one may regard the self-duality condition rather as part of the quantum theory (Witten 96, p.8, Witten 99, section 3 DMW 00b, page 3), namely as a choice of polarization of the phase space of an unconstrained theory in one dimension higher. By such as “holographic principle” the partition function of the self-dual theory on an XX of dimension 4k+24 k +2 is given by the state (wave function) of an abelian higher dimensional Chern-Simons theory in dimension 4k+34 k + 3.

The way this works is understood in much mathematical detail for k=0k = 0, see at AdS3-CFT2 and CS-WZW correspondence. Motivated by this it has been proposed and studied in a fair bit of mathematical detail for k=1k = 1 (Witten 96, Hopkins-Singer 02), see at M5-brane and 6d (2,0)-superconformal QFT and AdS7/CFT6. In both these cases the higher gauge fields are cocycles in ordinary differential cohomology. In (DMW 00, Belov-Moore 06b) it is suggested that similarly taking the self-dual fields to be cocycles in (differential) complex K-theory produces the RR-fields of type II superstring theory in dimension 10.

Definition (outline)

The holographic definition of self-dual abelian higher gauge field theory in dimension 4k+24k+2 given in (Witten 96, Witten 99) is in outline the following.

Imagine the self-dual 2k2k-form field on Σ\Sigma to be coupled to sources given by (2k+1)(2k+1)-form gauge fields, hence by cocycles in ordinary differential cohomology [Σ,B 2k+1𝔾 conn ×][\Sigma, \mathbf{B}^{2k+1}\mathbb{G}^\times_{conn}] of degree (2k+2)(2k+2). One may try to write down a Lagrangian for this coupling (and that is what is discussed in (Belov-Moore 06I)) and the resulting partition function is then locally a function of the source fields, hence a function on the intermediate Jacobian [Σ,B 2k+1𝔾 conn ×][\Sigma, \mathbf{B}^{2k+1}\mathbb{G}^\times_{conn}] and globally (since the action functional will not be strictly gauge invariant) a holomorphic section of a certain holomorphic line bundle on that space.

However, no choice of Lagrangian will enforce a strictly self-dual gauge field (the anti-self dual component will remain in the space of fields, hence the partition function will depend on it). Therefore the idea is to turn this around, reject the idea that self-dual higher gauge theory is directly itself a Lagrangian quantum field theory, and instead define the partition function of the self-dual higher gauge theory (and thereby, indirectly, the self-dual higher gauge theory itself!) as being a certain section of a certain line bundle on the intermediate Jacobian.

The question then is: which line bundle? Some plausibility arguments show that it must be a Theta characteristic bundle, hence a square root of the canonical bundle on the intermediate Jacobian. These have 1-dimensional spaces of holomorphic sections (theta functions) and hence any choice of that already uniquely fixes the would-be partition function, too, up to a choice of global factor. (For that reason later authors often regard the partition function as a line bundle, somewhat abusing the terminology.)

The claim then is that the correct choice of Theta characteristic to use is a quadratic refinement of the intersection pairing (the Beilinson-Deligne cup-product)

[Σ,B 2k+1𝔾 conn ×]×[Σ,B 2k+1𝔾 conn ×][Σ,B 4k+3𝔾 conn ×]B𝔾 conn ×. [\Sigma, \mathbf{B}^{2k+1}\mathbb{G}^\times_{conn}] \times [\Sigma, \mathbf{B}^{2k+1}\mathbb{G}^\times_{conn}] \stackrel{\cup}{\longrightarrow} [\Sigma, \mathbf{B}^{4k+3}\mathbb{G}^\times_{conn}] \stackrel{\int}{\longrightarrow} \mathbf{B}\mathbb{G}^\times_{conn} \,.

Such quadratic refinement turns out to be given for k=0k = 0 by a Spin structure (leading to “Spin Chern-Simons theory”) and for k=1k =1 by a Wu structure (and the latter case is the actual example of interest in (Witten 96)). This is what the central theorem of (Hopkins-Singer 02) establishes rigorously.

One notices now that while the self-dual (4k+2)(4k+2)-dimensional gauge field theory thus itself is not a Lagrangian quantum field theory, it arises holographically form a field theory in dimension 4k+34k+3 that is, namely the intersection pairing above is the Lagrangian for higher dimensional Chern-Simons theory and the holomorphic line bundle which it gives rise to on the intermediate Jacobian, is the prequantum line bundle of Chern-Simons theory, whose holomorphic sections are its quantum states (see at quantization of Chern-Simons theory). From this perspective the square root quadratic refinement is the metaplectic correction in the geometric quantization of Chern-Simobs theory.

This kind of relation

self-dual (4k+2)(4k+2)d gauge theory(4k+3)(4k+3)d Chern-Simons theory
partition function/correlator\leftrightarrowwavefunction/quantum state
conformal blocks\leftrightarrowspace of quantum states

is the hallmark of the holographic principle.


Holographic relation to higher Chern-Simons theory

Idea and examples

Generally, higher dimensional Chern-Simons theory in dimension 4k+34k+3 (for kk \in \mathbb{N}) is holographically related to self-dual higher gauge theory in dimension 4k+24k+2 (at least in the abelian case).

Conformal structure from polarization

We indicate why higher dimensional Chern-Simons theory is – if holographically related to anything – holographically related to self-dual higher gauge theory.

The phase space of higher dimensional Chern-Simons theory in dimension 4k+34k+3 on Σ×\Sigma \times \mathbb{R} can be identified with the space of flat 2k+12k+1-forms on Σ\Sigma. The presymplectic form on this space is given by the pairing

(δB 1,δB 2) ΣδB 1δB 2 (\delta B_1, \delta B_2) \mapsto \int_\Sigma \delta B_1 \wedge \delta B_2

obtained as the integration of differential forms over Σ\Sigma of the wedge product of the two forms.

The geometric quantization of the theory requires that we choose a polarization of the complexification of this space (split the space of forms into “canonical coordinates” and their “canonical momenta”).

One way to achieve this is to choose a conformal structure on Σ\Sigma. The corresponding Hodge star operator

:Ω 2k+1(Σ)Ω 2k+1(Σ) \star : \Omega^{2k+1}(\Sigma) \to \Omega^{2k+1}(\Sigma)

provides the polarization by splitting into self-dual and anti-self-dual forms:

notice that (by this Formula formulas at Hodge star operator) we have on mid-dimensional forms

B=(1) (2k+1)(4k+3)B=B. \star \star B = (-1)^{(2k+1)(4k+3)} B = - B \,.

Therefore it provides a complex structure on Ω 2k+1(Σ)\Omega^{2k+1}(\Sigma) \otimes \mathbb{C}.

We see that the symplectic structure on the space of forms can equivalently be rewritten as

XB 1B 2 = XB 1B 2. \begin{aligned} \int_X B_1 \wedge B_2 & = - \int_X B_1 \wedge \star \star B_2 \end{aligned} \,.

Here on the right now the Hodge inner product of B 1B_1 with B 2\star B_2 appears, which is invariant under applying the Hodge star to both arguments.

We then decompose Ω 2k+1(Σ)\Omega^{2k+1}(\Sigma) into the ±i\pm i-eigenspaces of \star: say BΩ 2k+1(Σ)B \in \Omega^{2k+1}(\Sigma) is imaginary self-dual if

B=iB \star B = i B

and imaginary anti-self-dual if

B=iB. \star B = - i B \,.

Then for imaginary self-dual B 1B_1 and B 2B_2 we find that the symplectic pairing is

(B 1,B 2) =i XB 1B 2 =i X(B 1)(B 2) =+i XB 1B 2. \begin{aligned} (B_1, B_2) &= -i \int_X B_1 \wedge \star B_2 \\ & = -i \int_X (\star B_1) \wedge \star (\star B_2) \\ & = +i \int_X B_1 \wedge \star B_2 \end{aligned} \,.

Therefore indeed the symplectic pairing vanishes on the self-dual and on the anti-selfdual forms. Evidently these provide a decomposition into Lagrangian subspaces.

(See also at Serre duality.)

Therefore a state of higher Chern-Simons theory on Σ\Sigma may locally be thought of as a function of the self-dual forms on Σ\Sigma. Under holography this is (therefore) identified with the correlator of a self-dual higher gauge theory on Σ\Sigma.

Partition function

By the above discussion (…) the partition function of self-dual higher gauge theory is given by (a multiple of) the unique holomorphic section of a square root of the line bundle classified by the secondary intersection pairing. (Witten 96, Hopkins-Singer).

Cohomological description

moduli spaces of line n-bundles with connection on nn-dimensional XX

nnCalabi-Yau n-foldline n-bundlemoduli of line n-bundlesmoduli of flat/degree-0 n-bundlesArtin-Mazur formal group of deformation moduli of line n-bundlescomplex oriented cohomology theorymodular functor/self-dual higher gauge theory of higher dimensional Chern-Simons theory
n=0n = 0unit in structure sheafmultiplicative group/group of unitsformal multiplicative groupcomplex K-theory
n=1n = 1elliptic curveline bundlePicard group/Picard schemeJacobianformal Picard groupelliptic cohomology3d Chern-Simons theory/WZW model
n=2n = 2K3 surfaceline 2-bundleBrauer groupintermediate Jacobianformal Brauer groupK3 cohomology
n=3n = 3Calabi-Yau 3-foldline 3-bundleintermediate JacobianCY3 cohomology7d Chern-Simons theory/M5-brane
nnintermediate Jacobian


Chiral boson in 2d (on the string)

For k=0k = 0 the self-dual theory abelian is that of a scalar field ϕ\phi on a real-2-dimensional surface Σ\Sigma such that dϕ\mathbf{d}\phi is self-dual. For any complex structure on Σ\Sigma, making it a complex torus, this makes ϕ\phi a “chiral” boson, in the sense of a chiral half of the U(1)U(1)-WZW model.

Original discussion is due to Srivastava 1989, GGR 1992. A quick review as a warm-up for the higher dimensional case is in Witten 96, section 2. More discussion is in GBMNV Miao & Müller-Kirsten 2000, see also the references at AdS3-CFT2 and CS-WZW correspondence.

Chiral 2-form in 6d (on the M5-brane)

The worldvolume theory of the M5-brane, the 6d (2,0)-superconformal QFT, contains a self-dual 2-form field. Its AdS7-CFT6 holographic description by 7-dimensional Chern-Simons theory is due to (Witten 96).

Relation between 5d Maxwell theory and self-dual 3-forms in 6d

Consider 5d- and 6d-dimensional Minkowski spacetime equipped with global orthonormal coordinate charts {x κ}\{x^\kappa\}, {x α}\{x^\alpha\}, respectively, adapated to an isometric embedding

4,1 ι 5 5,1 κ= 0,1,2,3,4, α= 0,1,2,3,4, 5 \array{ & \mathbb{R}^{4,1} &\overset{\;\;\;\iota_5\;\;\;}{\hookrightarrow}& \mathbb{R}^{5,1} \\ \kappa = & 0, 1, 2, 3, 4\phantom{,} \\ \alpha = & 0, 1, 2, 3, 4, && 5 }

With this notation, the pullback of differential forms along this embedding is notationally implicit.

Now any differential 3-form H 3H_3 on 5,1\mathbb{R}^{5,1} decomposes as

(1)H 3=F^dx 5+H^ H_3 \;=\; \widehat{F} \wedge d x^{5} + \widehat{H}

for unique differential forms of the form

F^=12F^ κ 1κ 2(x κ,x 5)dx κ 1dx κ 2 \widehat F \;=\; \tfrac{1}{2}\hat F_{\kappa_1 \kappa_2}(x^\kappa, x^5) d x^{\kappa_1} \wedge d x^{\kappa_2}


H^=13!H^ κ 1κ 2κ 3(x κ,x 5)dx κ 1dx κ 2dx κ 3. \widehat{H} \;=\; \tfrac{1}{3!} \widehat{H}_{\kappa_1 \kappa_2 \kappa_3}(x^\kappa, x^5) d x^{\kappa_1} \wedge d x^{\kappa_2} \wedge d x^{\kappa_3} \,.

In the case that H 3H_3 has vanishing Lie derivative along the x 5x^5-direction,

(2) 5H 3=0 \mathcal{L}_5 H_3 \;=\; 0

then also these components forms do not depend on x 5x^5 are actualls pullbacks of differential forms on 4,1\mathbb{R}^{4,1}.

In terms of this decomposition, the 6d Hodge dual of H 3H_3 is equivalently given by the 5d Hodge duals of these components as (best seen by the relation to Hodge pairing according to this Prop.)

(3) 6H 3=( 5H^)dx 5 5F^ \star_6 H_3 \;=\; \big( \star_5 \widehat{H}\big) \wedge d x^{5} - \star_5 \widehat{F}

Since the Hodge star operator squares to unity in the special case that it is applied to differential 3-forms on 6d Minkowski spacetime (by this Prop.)

6 6H 3=+H 3 \star_6 \star_6 H_3 \;=\; + H_3

we may ask for H 3H_3 to he Hodge self-dual. By (3) this means equivalently that its 5d components are 5d Hodge duals of each other:

(H 3= 6H 3)H 3=F^dx 5+H^(H^= 5F^). \big( H_3 \;=\; \star_{6} H_3 \big) \;\;\; \overset{ H_3 = \widehat{F} \wedge d x^5 + \widehat{H} }{ \Leftrightarrow } \;\;\; \big( \widehat{H} = \star_5 \widehat{F} \big) \,.

It follows that if there is no x 5x^5-dependence (2) then the condition that H 3H_3 be a closed and self-dual 3-form is equivalent to its 5d components F^\widehat{F} (H^\widehat{H}) being the (dual) field strength/Faraday tensor satisfying the Maxwell equations of D=5 Maxwell theory (without source current):

dH 3=0 6H 3=H 3}AD=6 self-dual 3-form theory 5H 3=0H 3=F^dx 5+{ dF^=0 d 5F^=0AD=5 Maxwell theory \underset{ \color{blue} { {\phantom{A}} \atop {\text{D=6 self-dual 3-form theory}} } }{ \left. \array{ & d H_3 = 0 \\ & \star_6 H_3 = H_3 } \right\} } \;\;\; \overset{ {\mathcal{L}_{5} H_3 = 0} \atop {H_3 = \widehat{F}\wedge d x^5 + \cdots} }{ \Leftrightarrow } \;\;\; \underset{ \color{blue} { {\phantom{A}} \atop \text{D=5 Maxwell theory} } }{ \left\{ \array{ & d \widehat{F} = 0 \\ & d \star_5 \widehat{F} = 0 } \right. }

This may be summarized as saying that the massless part of the Kaluza-Klein reduction of self-dual 3-form theory from 6d to 5d is D=5 Maxwell theory.

Essentially this relation underlies the formulation of the M5-brane via the Perry-Schwarz Lagrangian.

RR-Fields in 10d (on the “9-brane”)

The RR-field in type II string theory (by the idea of K-theory classification of D-brane charge) are meant to be self-dual as cocycles in (differential) topological K-theory (cf. at pre-metric electromagnetism the section on RR-fields).

The proper discussion of these fields hence involves generalizing the above story from ordinary cohomology and principally polarized intermediate Jacobians to an analog in Whitehead-generalized cohomology such as topological K-theory and an appropriate generalized form of intermediate Jacobians (cf. Hodge-filtered differential cohomology).

(Of course one may, as a warmup, “approximate” K-theory classes on a 10-manifold by integral cohomology lifts of the degree-5 component of their Chern character and hence discuss that as abelian self-dual higher gauge theory in dimension 10. This is discussed in (Witten 99, section 4)).

The relevant analog of the intermediate Jacobian for K-theory on a 10-manifold is naturally taken to be

K(X) /K(X) K(X)\otimes_{\mathbb{Z}} \mathbb{R}/ K(X)

(where the action is via the Chern character/realification KKK \to K \otimes \mathbb{R}).

This is proposed and discussed in (Witten 99, section 4.3, Moore-Witten 99, section 3, DMW 00, section 7.1, Belov-Moore 06b, section 5, MPS 11):

where for ordinary cohomology the (quadratic refinement of the) intersection product (= Deligne-Beilinson cup product followed by fiber integration in ordinary differential cohomology) provided the symplectic structure/polarization of the intermediate Jacobian, here it is tensoring of Dirac operators followed by fiber integration in differential K-theory, hence the index map in K-theory.

See at intermediate Jacobian – For complex K-theory.

The following table lists classes of examples of square roots of line bundles

line bundlesquare rootchoice corresponds to
canonical bundleTheta characteristicover Riemann surface and Hermitian manifold (e.g.Kähler manifold): spin structure
density bundlehalf-density bundle
canonical bundle of Lagrangian submanifoldmetalinear structuremetaplectic correction
determinant line bundlePfaffian line bundle
quadratic secondary intersection pairingpartition function of self-dual higher gauge theoryintegral Wu structure



Original articles on the general issue include

Surveys include

  • Greg Moore, A Minicourse on Generalized Abelian Gauge Theory, Self-Dual Theories, and Differential Cohomology, Lectures at Simons Center for Geometry and Physics (2011) (pdf)

  • Richard Szabo, Quantization of Higher Abelian Gauge Theory in Generalized Differential Cohomology (arXiv:1209.2530)

An approach based on supergeometry:

For the M5-sigma model

The sigma-model description of the (single) M5-brane of Green-Schwarz action functional-type was found in covariant form in


and following the non-covariant form of the self-duality mechanism (Perry-Schwarz action) of

Discussion in the superembedding approach;

Amplification that the resulting B-field on the M5-brane is not naively self-dual:

Discussion of the equivalence of these superficially different action functionals is in

On the self-dual higher gauge field on the M5-brane lifted to superspace in the D'Auria-Fré formulation of supergravity:

For ordinary U(1)U(1)-higher gauge fields / ordinary differential cohomology

Self-duality for higher abelian gauge fields (in ordinary differential cohomology):

Original reference on self-dual/chiral fields include

The chiral boson in 2d:

A quick exposition of the basic idea is in

A precise formulation of the phenomenon in terms of ordinary differential cohomology and differential K-theory:

The idea of describing self-dual higher gauge theory by holography with abelian higher dimensional Chern-Simons theory in one dimension higher originates in

Conceptual aspects of this are also discussed in section 6.2 of

Motivated by this the ordinary differential cohomology of self-dual fields had been discussed in

Discussion of S-duality in 6d self-dual higher gauge theory via non-commutative-deformation:

Discussion of the conformal blocks and geometric quantization of self-dual higher gauge theories is in

  • Kiyonori Gomi, An analogue of the space of conformal blocks in (4k+2)(4k+2)-dimensions (pdf)

For the case of nonabelian self-dual 1-form gauge fields see the references at Yang-Mills instanton.

Another proposal for a Lagrangian for self-dual higher gauge fields, at the cost of an unusual kinds of decoupled auxiliary fields:

Enhancement to superspace (in D'Auroa-Fre formulation of supergravity?):

Further discussion:

  • Chris Hull, Covariant Action for Self-Dual pp-Form Gauge Fields in General Spacetimes [arXiv:2307.04748]

  • Sujiphat Janaun, Anajak Phonchantuek, Pichet Vanichchapongjaroen, Nonlinear chiral forms in the Sen formulation [arXiv:2404.05380]

For RR-fields / differential K-theory

Discussion of (Lagrangian densities for) D=10 type II supergravity with “duality-symmetric”/“democratic”/“pregeometric” for of the RR-fields:

Enhancement of the self-duality constraint on pregeometric RR-fields from (twisted) de Rham cohomology to (twisted) topological K-theory (under the hypothesized K-theory classification of D-brane charge) in terms of a quadratic form on differential K-theory:

An indication of a more refined discussion in twisted differential KR-theory:

See at orientifold for more on this.

Expressing the self-duality of pregeometric RR-fields in terms of 11d Chern-Simons theory:

Some review:

  • Richard Szabo, section 3.6 and 4.6 of: Quantization of Higher Abelian Gauge Theory in Generalized Differential Cohomology, ESI 2385 (2012) [[arXiv:1209.2530, pdf]]

Discussion in the context of flux quantization (here: D-brane charge quantization in K-theory):

For the 11d supergravity C-field

Formulation of the equations of motion of D=11 supergravity in superspace on fields including a flux density G 7G_7 a priori independent of the flux density G 4G_4 of the supergravity C-field:

Discussion of Lagrangian densities for D=11 supergravity with an a priori independent dual C-field field and introduction of the “duality-symmetric” terminology:

Discussion in the context of shifted C-field flux quantization:

Last revised on May 21, 2024 at 10:47:58. See the history of this page for a list of all contributions to it.