differential twisted String and Fivebrane structures

This entry is about the article

We generalize the topological notion of string structure

Then we refine the notion of higher smooth String structure from smooth nonabelian cohomology to


We show


Abelian differential generalized cohomology as developed by Hopkins and Singer has been shown by Freed to formalize the global description of anomaly cancellation problems in higher gauge theories arising in String theory, such as notably the Green-Schwarz mechanism. On the other hand, this mechanism, as well as the Freed- Witten anomaly cancellation, are fundamentally governed by the cohomology classes represented by the relevant nonabelian O(n)O(n)- and U(n)U(n)-principal bundles underlying the tangent bundle and the gauge bundle on target space.

We unify the picture by describing differential nonabelian cohomology and differential twisted cohomology and apply it to these situations. We demonstrate that the Freed-Witten mechanism for the Kalb-Ramond field, the Green-Schwarz mechanism for the Kalb-Ramond field, as well as its magnetic dual version for the H 7H_7-field define cocycles in twisted nonabelian differential cohomology that may be addressed, respectively, as twisted Spin(n)Spin(n)-, twisted String structure - and twisted Fivebrane(n)Fivebrane(n)- structures on target space, where the twist in each case is provided by the obstruction to lifting the gauge bundle through a higher connected cover of U(n)U(n). We work out the (nonabelian) ∞-Lie algebroid valued differential forms data provided by the differential refinements of these twisted cocycles and demonstrate that this reproduces locally the differential form data with the twisted Bianchi identities as known from the string theory literature. The treatment for M-theory leads to models for the CC-field and its dual in differential nonabelian cohomology.



Motivation: background fields

An interesting supply of motivations for and applications of generalized notions of cohomology arises in formal higher energy physics in the context of theories that combine and generalize Maxwell’s theory of the electromagnetic field, Einstein’s theory of the gravitational field and Yang-Mills' theory of general gauge fields.

In order to formalize and study certain phenomena exhibited by such higher background fields – such as

Hopkins and Singer in their seminal work developed the general theory of differential refinements of generalized (Eilenberg-Steenrod) cohomology. Based on this, Freed explained certain subtle effects, previously observed semi-rigorously by physicists, systematically as phenomena exhibited by cocycles in generalized abelian differential cohomology.

This involves notably various twists of one kind of cohomology by another. The most familiar example is twisted K-theory, the cohomology theory that in formal high energy physics describes the RR field.

But the physical applications indicate that this is only the simplest example in a more general theory of twisted generalized cohomology. The next example is the famous Green-Schwarz mechanism in heterotic supergravity, which, as Freed explained, amounts to asserting a kind of twist of a higher differential cohomology class.

While this clarifies some of the structure, it remains noteworthy that the twist in the Green-Schwarz mechanism is fundamentally encoded not in abelian generalized coholomology, but by the class of a GG-principal bundle for a nonabelian group GG, hence by a class in nonabelian cohomology.

The coefficients of ordinary generalized cohomology are spectra. If connective, these are maximally abelian namely symmetric monoidal \infty-groupoids or equivalently – by the homotopy hypothesistopological spaces. In contrast to that, degree nn-nonabelian cohomology allows as coefficients arbitray n-groupoids, i.e. homotopy n-types. Such nonabelian cohomology is traditionally most familiar in the study of 1- and 2-gerbes as well as in the higher Schreier theory of nonabelian group extensions. Here nn may be n=n = \infty.


Using the language of Quillen model categories it was already realized in BrownAHT that generalized (Eilenberg-Steenrod) cohomology, abelian sheaf cohomology as well as nonabelian cohomology all describe hom-sets in homotopy categories of (pre)sheaves with values in infinity-groupoids – called simplicial presheaves

This perspective was later refined by Joyal and Jardine’s study of the model structure on simplicial presheaves. By the recent result of Lurie HTT we know that these constructions model precisely the theory of ∞-stacks in that they are (hypercomplete) models for ∞-stack (∞,1)-toposes.

This leads one to expect that a general theory of smooth cohomology that encompasses abelian as well as nonabelian phenomena concerns a smooth (∞,1)-topos of ∞-stacks on a site such as that of smooth loci: these may be regarded as ∞-Lie groupoids. This perspective on \infty-stacks as the truly general notion of cohomology is implied by Lurie’s very notion of (,1)(\infty,1)-topos as a context that “behaves like topological spaces” . It can be found made explicit for instance in Dug.

We therefore place ourselves in a general context in which all the familiar operations of the homotopy theory of topological spaces – such as notably fibration sequences – make sense and work as expected, but where these spaces may be something more general and/or richer than topological spaces: Specifically, for applications to physics in general and for our applications to differential cohomology in higher gauge theory in particular they may be smooth and wildly infinite-dimensional spaces. The general tool for describing contexts that behave like the category Top of topological spaces but are richer are (∞,1)-toposes of infinity-stacks.

We pick a well-adapted smooth (∞,1)-topos H\mathbf{H} of ∞-Lie groupoids as our context and then study the following questions:

The last of these four questions is our main motivation. Here, however, we don’t go into this last question except that it shall serve to motivate our answer to the first question:

Twisted differential nonabelian cohomology

In the context of general (nonabelian) cohomology we set up notions of


First notice how the (∞,1)-topos H\mathbf{H} comes with its notion of

Differential nonabelian cohomology

Before stating our formulation of differential cohomology the reader familiar with functorial quantum field theory may find it helpful to consider the following motivation. Other readers should skip this motivation.

Motivation from QFT

Recall that, as finally fully formalized in TFTClass, an nn-dimensional topological quantum field theory is an (,n)(\infty,n)-functor

Z:Bord n𝒱 Z : \mathrm{Bord}_n \to \mathcal{V}

on the (∞,n)-category of cobordisms with values in an (∞,n)-category of something like nn-vector spaces.

On the other hand, when we have on a target space object XX a background field under which an nn-dimensional object is charged, we expect a notion of parallel transport and holonomy encoded by an (,n)(\infty,n)-functor

exp():Bord n(X)𝒱 \exp(\int\nabla) : \mathrm{Bord}_n(X) \to \mathcal{V}

from bordisms equipped with maps into XX, that assigns to an nn-dimensional bordism its parallel transport or holonomy, as a morphism in 𝒱\mathcal{V}.

The above suggests that those QFTs ZZ that arise as sigma models in that they are encoded by a background field \nabla are, in some sense, \emph{extensions}

Bord n(X) exp() 𝒱 Z Bord nfieldspace backgroundfield 𝒱 quantumpropagation parameterspace \array{ \mathrm{Bord}_n(X) &\stackrel{\exp(\int\nabla)}{\to}& \mathcal{V} \\ \downarrow & \nearrow_{Z_\nabla} \\ \mathrm{Bord}_n } \;\;\; \array{ field space &\stackrel{background field}{\to} & \mathcal{V} \\ \downarrow & \nearrow_{quantum propagation} \\ parameter space }

along the obvious forgetful (,n)(\infty,n)-functor Bord n(X)Bord(X)\mathrm{Bord}_n(X) \to \mathrm{Bord}(X).

Whatever this extension procedure may be in detail, a necessary prerequisite for studying it is a good grasp of how to encode differential cocycles on XX in terms of functors on Bord n(X)\mathrm{Bord}_n(X).

Formulation in terms of parallel transport

In a series of articles BaezSchr,SWI, SWII, SWIII it was shown that gerbes with connection and more generally higher principal bundles with connection are indeed encoded as morphisms in H\mathbf{H} of the form

:𝒫 n(X)𝒱 \nabla : \mathcal{P}_n(X) \to \mathcal{V}

where the 𝒫 n(X)\mathcal{P}_n(X) is essentially the subobject of Bord n(X)\mathrm{Bord}_n(X) consisting only of bordisms in XX that are homeomorphic to a disk: the path n-groupoid of XX.


Let GG be a Lie group, BG\mathbf{B}G the corresponding delooped one-object smooth groupoid in H\mathbf{H} – this is a smooth version of the classifying space of GG – and similarly BAUT(G)\mathbf{B}\mathrm{AUT}(G) the one-object 2-groupoid coming from the automorphism 2-group of GG. Let XX be a manifold, then

H(X,BG)GBund(X)={GprincipalbundlesonX} \mathbf{H}(X, \mathbf{B}G) \simeq G \mathrm{Bund}(X) = \left\{ G -principal bundles on X \right\}
H(𝒫 1(X),BG)GBund (X)={GprincipalbundleswithconnectiononX} \mathbf{H}(\mathcal{P}_1(X), \mathbf{B}G) \simeq G \mathrm{Bund}_\nabla(X) = \left\{ G -principal bundles with connection on X \right\}
H(𝒫 2(X),BAUT(H))HGrb ff(X)={GgerbeswithfakeflatconnectiononX} \mathbf{H}(\mathcal{P}_2(X), \mathbf{B}\mathrm{AUT}(H)) \simeq H \mathrm{Grb}_{\nabla_{\mathrm{ff}}}(X) = \left\{ G-gerbes with fake-flat connection on X \right\}


Urs Schreiber: yes, that needs more detail and polishing – will try to get to that soon

In particular let BBU(1)\mathbf{B}\mathbf{B} U(1) be the smooth 2-groupoid with the Lie group U(1)U(1) in degree 2, then

H(𝒫 2(X),BBU(1))BdlGrb (X)={linebundlegerbeswithgeneralconnection} \mathbf{H}(\mathcal{P}_2(X), \mathbf{B}\mathbf{B}U(1)) \simeq BdlGrb_{\nabla}(X) = \left\{ line bundle gerbes with general connection \right\}

Here the notation means the following:

  • BU(1)\mathbf{B} U(1) is a smooth realization in our smooth (∞,1)-topos H\mathbf{H} of the familiar classifying space for U(1)U(1)-principal bundles. Similarly, B nU(1)\mathbf{B}^n U(1) is a smooth model the iterated classifying space nU(1)\mathcal{B}^n U(1).

  • For XX and AA objects in H smooth\mathbf{H}_{smooth}, the notation H smooth(X,A)\mathbf{H}_{smooth}(X,A) or H(X,A)\mathbf{H}(X,A) for short denotes the topological space or infinity-groupoid of maps from XX to AA, homotopies between maps, homotopies between homotopies, etc.

    In particular, the cohomology itself on XX with coefficients in AA is the the set of homotopy classes of such maps

    H(X,A):=π 0H(X,A). H(X,A) := \pi_0 \mathbf{H}(X,A) \,.
  • Finally, fake flatness is a certain condition on differential structures that is invisible in ordinary connections on a bundle but appears in the generalized context considered here: for AA an n-truncated object the curvature? of a differential cocycle is a list of differential forms in degree 1 through (n+1)(n+1). For n=2n=2 the 2-form curvature is sometimes called the “fake” curvature, whereas the “true” one is the 3-form curvature.

    Also in the case of abelian coefficients B nU(1)\mathbf{B}^n U(1) the fake-flatness condition is empty and plays no role. But in the nonabelian case, as well as in the abelian equivariant case, fake flatness is more restrictive than what one might expect. For full nonabelian differential cohomology the above is slightly too naive and replaced by the following.


For every object XXX \in \mathbf{X} there is an object Π(X)=colim𝒫 n(X)\Pi(X) = colim \mathcal{P}_n(X), the path infinity-groupoid? of XX. The kk-cells of Π(X)\Pi(X) are generated from the existing kk-cells of XX and from kk-dimensional smooth paths (disk-shaped cobordisms) in XX. See path infinity-groupoid? for details.

For any coefficient object AHA \in \mathbf{H} we call H(Π(X),A)\mathbf{H}(\Pi(X), A) the flat differential AA-cohomology of XX or the AA-local systems on XX.

For AA once deloopable there is a morphism

char:H(X,A)H(Π(X),BA) char : \mathbf{H}(X,A) \to \mathbf{H}(\Pi(X),\mathbf{B}A)

whose image P(c)P(c) of an AA-cocycle cc we call the characteristic curvature class of cc.

Given a class PH(Π(X),BA)P \in \mathbf{H}(\Pi(X),\mathbf{B}A) we call the corresponding PP-twisted flat differential AA-cohomology H P(Π(X),A)\mathbf{H}^{P}(\Pi(X),A) the differential A-valued cohomology with curvature PP.

The last clause uses the following definition of twisted cohomology.

Twisted nonabelian cohomology

Recall that the main point of having an (infinity,1)-topos H\mathbf{H} that replaces the ordinary (infinity,1)-category Top is that familiar notions of homotopy theory all apply as expected in H\mathbf{H}. In particular, there is a notion of homotopy pullback and fibration sequence in H synth\mathbf{H}_{synth}.

Let FGAF \to G \to A be such a fibration sequence in H\mathbf{H}. We have left-exactness of the Hom for any object XHX \in \mathbf{H} we obtain the fibration sequence

H(X,F)H(X,G)obstr H(X,A) \array{ \mathbf{H}(X,F) \to \mathbf{H}(X,G) \stackrel{obstr}{\to} & \mathbf{H}(X,A) }

of \infty-groupoids or topological spaces, that characterizes FF-cocycles as those GG-cocycles whose obstructing AA-cocycle is trivializable.


For cH(X,A)c \in \mathbf{H}(X,A) any possibly nontrivial AA-cocycle on XX, define the cc-twisted FF-cohomology H c(X,A)\mathbf{H}^c(X,A) to be the homotopy pullback

H(X,F) * *c H(X,G) obstr H(X,A). \array{ \mathbf{H}(X,F) &\to& {*} \\ \downarrow && \downarrow^{{*}\mapsto c} \\ \mathbf{H}(X,G) &\stackrel{obstr}{\to}& \mathbf{H}(X,A) } \,.

Examples and Applications

We discuss the following exampleS and applicationS of twisted differential cohomology .


Urs yes, I will work on this

  • fibration sequence: BU(n)BPU(n)B 2U(1)\mathbf{B}U(n) \to \mathbf{B} PU(n) \to \mathbf{B}^2 U(1)

    • twisting cocycle: lifting gerbe;

    • twisted cocycle: twisted bundles / gerbe modules

    • twisted Bianchi identity: dF =H 3d F_\nabla = H_3

    • occurence: Freed-Witten anomaly cancellation on D-brane

  • fibration sequence: BString(n)BSpin(n)12p 1B 3U(1)\mathbf{B}String(n) \to \mathbf{B} Spin(n) \stackrel{\frac{1}{2}p_1}{\to} \mathbf{B}^3 U(1)

    • twisting cocycle: Chern-Simons 2-gerbe;

    • twisted cocycle: twisted nonabelian String-gerbe with conection

    • twisted Bianchi identity: dH 3F F d H_3 \propto \langle F_\nabla \wedge F_\nabla \rangle

    • occurence: Green-Schwarz anomaly cancellation

    • Proof.

    • for aspects of the twisted case see also

      • AsJu

      • for aspects of the untwisted case see also (Wa)

  • fibration sequence: BFivebrane(n)BString(n)16p 2B 7U(1)\mathbf{B}Fivebrane(n) \to \mathbf{B} String(n) \stackrel{\frac{1}{6}p_2}{\to} \mathbf{B}^7 U(1)

  • fibration sequence: B 2U(1)B(U(1) 2)B 2\mathbf{B}^2 U(1) \to \mathbf{B} (U(1) \to \mathbb{Z}_2) \stackrel{}{\to} \mathbf{B} \mathbb{Z}_2

    • twisting cocycle: 2\mathbb{Z}_2-orbifold;

    • twisted cocycle: orientifold gerbe / Jandl gerbe with connection

    • occurence: unoriented string

    • unwrap the above abstract nonsense and use the above results to find SchrSchwWal and the bosonic part of DiFrMo

this is the end of section 1) Plan

2) Differential nonabelian cohomology

3) Twisted topological structures

4) L L_\infty-Connections

5) Twisted differential structures

Revised on October 22, 2009 07:27:25 by Urs Schreiber (