nLab
differential string structure

Context

String theory

Differential cohomology

\infty-Chern-Weil theory

Contents

Idea

Where a string structure is a trivialization of a class in integral cohomology, a differential string structure or geometric string structure is the trivialization of this class refined to ordinary differential cohomology:

the first fractional Pontryagin class

12p 1:BSpinB 4 \frac{1}{2} p_1 : B Spin \to B^4 \mathbb{Z}

in the (∞,1)-topos ∞Grpd \simeq Top has a refinement to H=\mathbf{H} = Smooth∞Grpd of the form

12p 1:BSpinB 3U(1) \frac{1}{2}\mathbf{p}_1 : \mathbf{B}Spin \to \mathbf{B}^3 U(1)

– the smooth first fractional Pontryagin class.

The induced morphism on cocycle ∞-groupoids

12p 1:H(X,BSpin)H(X,B 3U(1)) \frac{1}{2}\mathbf{p}_1 : \mathbf{H}(X,\mathbf{B} Spin) \stackrel{}{\to} \mathbf{H}(X,\mathbf{B}^3 U(1))

sends a spin group-principal bundle PP to its corresponding Chern-Simons circle 3-bundle 12p 1(P)\frac{1}{2}\mathbf{p}_1(P).

A choice of trivialization of 12p 1(P)\frac{1}{2}p_1(P) is a string structure. The 2-groupoid of smooth string structures is the homotopy fiber of 12p 1\frac{1}{2}\mathbf{p}_1 over the trivial circle 3-bundle.

By Chern-Weil theory in Smooth∞Grpd this morphism may be further refined to a differential characteristic class 12p^ 1\frac{1}{2}\hat \mathbf{p}_1 that lands in the ordinary differential cohomology H diff(X,B 3U(1))\mathbf{H}_{diff}(X, \mathbf{B}^3 U(1)), classifying circle 3-bundles with connection

12p^ 1:H conn(X,BSpin)H diff(X,B 3U(1)) \frac{1}{2}\hat \mathbf{p}_1 : \mathbf{H}_{conn}(X,\mathbf{B} Spin) \stackrel{}{\to} \mathbf{H}_{diff}(X,\mathbf{B}^3 U(1))

The 2-groupoid of differential string structures is the homotopy fiber of this refinement 12p^ 1\frac{1}{2}\hat \mathbf{p}_1 over the trivial circle 3-bundle with trivial connection or more generally over the trivial circle 3-bundles with possibly non-trivial connection

Such a differential string structure over a smooth manifold XX is characterized by a tuple consisting of

  1. a connection \nabla on a Spin-principal bundle on XX;

  2. a choice of trivial circle 3-bundle with connection (0,H 3)(0, H_3), hence a differential 3-form H 3Ω 3(X)H_3 \in \Omega^3(X);

  3. a choice of equivalence λ\lambda of the Chern-Simons circle 3-bundle with connection 12p^ 1()\frac{1}{2}\hat\mathbf{p}_1(\nabla) of \nabla with this chosen 3-bundle

λ:12p^ 1()(0,H 3). \lambda : \frac{1}{2}\hat \mathbf{p}_1(\nabla) \stackrel{\simeq}{\to} (0,H_3) \,.

More generally, one can consider the homotopy fibers of 12p^ 1\frac{1}{2}\hat \mathbf{p}_1 over arbitrary circle 3-bundles with connection 𝒢^ 4H diff 4(X,B 3U(1))\hat \mathcal{G}_4 \in \mathbf{H}_{diff}^4(X, \mathbf{B}^3 U(1)) and hence replace (0,H 3)(0,H_3) in the above with 𝒢^ 4\hat \mathcal{G}_4. According to the general notion of twisted cohomology, these may be thought of as twisted differential string structures, where the class [𝒢 4]H diff 4(X)[\mathcal{G}_4] \in H^4_{diff}(X) is the twist.

Definition

We will assume that the reader is familiar with basics of the discussion at Smooth∞Grpd. We often write H:=SmoothGrpd\mathbf{H} := Smooth \infty Grpd for short.

Let Spin(n)Spin(n) \in SmoothMfd \hookrightarrow Smooth∞Grpd be the Spin group, for some nn \in \mathbb{N}, regarded as a Lie group and thus canonically as an ∞-group object in Smooth∞Grpd. We shall notationally suppress the nn in the following. Write BSpin\mathbf{B}Spin for the delooping of SpinSpin in Smooth∞Grpd. (See the discussion here). Let moreover B 2U(1)SmoothGrpd\mathbf{B}^2 U(1) \in Smooth \infty Grpd be the circle Lie 3-group and B 3U(1)\mathbf{B}^3 U(1) its delooping.

At Chern-Weil theory in Smooth∞Grpd the following statement is proven (FSS).

Proposition

The image under Lie integration of the canonical Lie algebra 3-cocycle

μ=,[,]:𝔰𝔬b 2 \mu = \langle -,[-,-]\rangle : \mathfrak{so} \to b^2 \mathbb{R}

on the semisimple Lie algebra 𝔰𝔬\mathfrak{so} of the Spin group – the special orthogonal Lie algebra – is a morphism in Smooth∞Grpd of the form

12p 1:BSpinB 3U(1) \frac{1}{2} \mathbf{p}_1 : \mathbf{B}Spin \to \mathbf{B}^3 U(1)

whose image under the the fundamental ∞-groupoid (∞,1)-functor/ geometric realization Π:SmoothGrpd\Pi : Smooth \infty Grpd \to ∞Grpd is the ordinary fractional Pontryagin class

12p 1:BSpinB 4 \frac{1}{2}p_1 : B Spin \to B^4 \mathbb{Z}

in Top. Moreover, the corresponding refined differential characteristic class

12p^ 1:H conn(,BSpin)H diff(,B 3U(1)) \frac{1}{2}\hat \mathbf{p}_1 : \mathbf{H}_{conn}(-,\mathbf{B}Spin) \to \mathbf{H}_{diff}(-, \mathbf{B}^3 U(1))

is in cohomology the corresponding refined Chern-Weil homomorphism

[12p^ 1]:H Smooth 1(X,Spin)H diff 4(X) [\frac{1}{2}\hat \mathbf{p}_1] : H^1_{Smooth}(X,Spin) \to H_{diff}^4(X)

with values in ordinary differential cohomology that corresponds to the Killing form invariant polynomial ,\langle - , - \rangle on 𝔰𝔬\mathfrak{so}.

Definition

For any XX \in Smooth∞Grpd, the 2-groupoid of differential string-structures on XXString diff(X)String_{diff}(X) – is the homotopy fiber of 12p^ 1(X)\frac{1}{2}\hat \mathbf{p}_1(X) over the trivial differential cocycle.

More generally (see twisted cohomology) the 2-groupoid of twisted differential string structures is the (∞,1)-pullback String diff,tw(X)String_{diff,tw}(X) in

String diff,tw(X) H diff 4(X) H conn(X,BSpin) 12p^ 1 H diff(X,B 3U(1)), \array{ String_{diff,tw}(X) &\to& H_{diff}^4(X) \\ \downarrow && \downarrow \\ \mathbf{H}_{conn}(X,\mathbf{B}Spin) &\stackrel{\frac{1}{2}\hat \mathbf{p}_1}{\to}& \mathbf{H}_{diff}(X,\mathbf{B}^3 U(1)) } \,,

where the right vertical morphism is a choice of (any) one point in each connected component (differential cohomology class) of the cocycle ∞-groupoid H diff(X,B 3U(1))\mathbf{H}_{diff}(X,\mathbf{B}^3 U(1)) (the homotopy type of the (∞,1)-pullback is independent of this choice).

More specifically, a geometric string structure is a twisted differential string structure whose differential twist has underlying trivial class.

Note

In terms of local ∞-Lie algebra valued differential forms data this has been considered in (SSSIII), as we shall discuss below.

For the case where the the underlying integral class of the twist is trivial – geometric string structures – something close to this definition, explicitly modeled on bundle 2-gerbes, has been given in (Waldorf). See the discussion below.

Properties

General

Observation

The ∞-groupoid String tw,diff(X)String_{tw,diff}(X) of twisted differential string structures is 2-truncated, hence is a 2-groupoid.

Proof

This follows from the long exact sequence of homotopy groups associated to the defining (∞,1)-pullback, using that

  • H diff(X,B 3U(1))\mathbf{H}_{diff}(X, \mathbf{B}^3 U(1)) is a 3-groupoid;

  • H(X,BSpin)\mathbf{H}(X, \mathbf{B}Spin) is a 1-groupoid;

  • H diff 4(X)H^4_{diff}(X) is a 0-groupoid.

See also (Waldorf, cor. 1.1.5).

Observation

If the underlying integral cohomology class of the twist is trivial, c(tw)=0H 3(X,)c(tw) = 0 \in H^3(X, \mathbb{Z}), then a twtw-twisted differential string structures on a SpinSpin-connection \nabla are characterized by a globally defined 3-form on XX.

This 3-form is the globally defined connection 3-form of an appropriate circle 3-bundle with connection equivalent to the Chern-Simons circle 3-bundle CS()CS(\nabla) whose underlying 3-bundle is by assumption trivial: on a trivial circle nn-bundle every connection may be represented by a globally defined nn-form.

This statement appears as (Waldorf, theorem 1.3.3), where circle 3-bundles are modeled as bundle 2-gerbes. The explicit construction of the globally defined 3-form in this model is spelled out in lemma 3.2.4 there.

Proof

Elements in the defining homotopy pullback from def. 1 over a given connection H(X,BSpin) conn\nabla \in \mathbf{H}(X,\mathbf{B} Spin)_{conn} are chracterized by an element [α]H diff 4(X)[\alpha] \in H^4_{diff}(X) and an equivalence

Ω:CS()α \Omega : CS(\nabla) \stackrel{\simeq}{\to} \alpha

between the corresponding Chern-Simons circle 3-bundle and the given circle 3-bundle α\alpha. In the case at hand, both have underlying trivial class c(CS())=c(α)=0c(CS(\nabla)) = c(\alpha) = 0.

By the characteristic class exact sequence

0Ω 3(X)/Ω int 3(X)H diff 4(X)cH 4(X,) 0 \to \Omega^3(X)/\Omega_{int}^{3}(X) \to H^4_{diff}(X) \stackrel{c}{\to} H^4(X, \mathbb{Z})

any two classes in π 0H diff(X,B 3U(1))H diff 4(X)\pi_0 \mathbf{H}_{diff}(X, \mathbf{B}^3 U(1)) \simeq H^4_{diff}(X) that have trivial underlying class in π 0H(X,B 3U(1))H 4(X,)\pi_0 \mathbf{H}(X, \mathbf{B}^3 U(1)) \simeq H^4(X, \mathbb{Z}) differ by a 3-form modulo a closed 3-form with integral periods.

Therefore both [α][\alpha] as well as [CS()]H diff 4(X)[CS(\nabla)] \in H^4_{diff}(X) are given by a globally defined 3-form modulo an integral form: the global connection 3-form on these trivial circle 3-bundles with connection.

Construction in terms of L L_\infty-Cech cocycles

We use the presentation of the (∞,1)-topos Smooth∞Grpd (as described there) by the local model structure on simplicial presheaves [CartSp smooth op,sSet] proj,loc[CartSp_{smooth}^{op}, sSet]_{proj,loc} to give an explicit construction of twisted differential string structures in terms of Cech-cocycles with coefficients in ∞-Lie algebra valued differential forms.

Proofs not displayed here can be found at differential string structure -- proofs .

Recall the following fact from Chern-Weil theory in Smooth∞Grpd (FSS).

Proposition

The differential fractional Pontryagin class 12p^ 1\frac{1}{2} \hat \mathbf{p}_1 is presented in [CartSp smooth op,sSet] proj[CartSp_{smooth}^{op}, sSet]_{proj} by the top morphism of simplicial presheaves in

cosk 3exp(𝔰𝔬) ChW,smp exp(μ,cs) B 3/ ChW,smp cosk 3exp(𝔰𝔬) diff,smp exp(μ,cs) B 3/ smp BSpin c. \array{ \mathbf{cosk}_3\exp(\mathfrak{so})_{ChW,smp} &\stackrel{\exp(\mu, cs)}{\to}& \mathbf{B}^3 \mathbb{R}/\mathbb{Z}_{ChW,smp} \\ \downarrow && \downarrow \\ \mathbf{cosk}_3\exp(\mathfrak{so})_{diff,smp} &\stackrel{\exp(\mu, cs)}{\to}& \mathbf{B}^3 \mathbb{R}/\mathbb{Z}_{smp} \\ \downarrow^{\mathrlap{\simeq}} \\ \mathbf{B}Spin_{c} } \,.

Here the middle morphism is the direct Lie integration of the L-∞ algebra cocycle while the top morphisms is its restriction to coefficients for ∞-connections.

In order to compute the homotopy fibers of 12p^ 1\frac{1}{2}\hat \mathbf{p}_1 we now find a resolution of this morphism exp(μ,cs)\exp(\mu,cs) by a fibration in [CartSp smooth op,sSet] proj[CartSp_{smooth}^{op}, sSet]_{proj}. By the fact that this is a simplicial model category then also the hom of any cofibrant object into this morphism, computing the cocycle \infty-groupoids, is a fibration, and therefore, by the general discussion at homotopy pullback, we obtain the homotopy fibers as the ordinary fibers of this fibration.

Presentation of the differential class by a fibration

In order to factor exp(μ,cs)\exp(\mu,cs) into a weak equivalence followed by a fibration, we start by considering such a factorization before differential refinement, on the underlying characteristic class exp(μ)\exp(\mu).

To that end, we replace the Lie algebra 𝔤=𝔰𝔬\mathfrak{g} = \mathfrak{so} by an equivalent but bigger Lie 3-algebra (following SSSIII). We need the following notation:

  • 𝔤=𝔰𝔬\mathfrak{g} = \mathfrak{so}, the special orthogonal Lie algebra (the Lie algebra of the spin group);

  • b 2b^2 \mathbb{R} the line Lie 3-algebra, the single generator in degee 3 of its Chevalley-Eilenberg algebra we denote cCE(b 2)c \in CE(b^2 \mathbb{R}), dc=0d c = 0.

  • ,W(𝔤)\langle -,-\rangle \in W(\mathfrak{g}) is the Killing form invariant polynomial, regarded as an element of the Weil algebra of 𝔰𝔬\mathfrak{so};

  • μ:=,[,]CE(𝔤)\mu := \langle -,[-,-]\rangle \in CE(\mathfrak{g}) the degree 3 Lie algebra cocycle, identified with a morphism

    CE(𝔤)CE(b 2):μ CE(\mathfrak{g}) \leftarrow CE(b^2 \mathbb{R}) : \mu

    of Chevalley-Eilenberg algebras; and normalized such that its continuation to a 3-form on SpinSpin is the image in de Rham cohomology of SpinSpin of a generator of H 3(Spin,)H^3(Spin,\mathbb{Z}) \simeq \mathbb{Z};

  • csW(𝔤)cs \in W(\mathfrak{g}) is a Chern-Simons element interpolating between the two; characterized by the fact that it fits into the commuting diagram

    CE(𝔤) μ CE(b 2) W(𝔤) cs W(b 2) inv(𝔤) , inv(b 2) =CE(b 3) \array{ CE(\mathfrak{g}) &\stackrel{\mu}{\leftarrow}& CE(b^2 \mathbb{R}) \\ \uparrow && \uparrow \\ W(\mathfrak{g}) &\stackrel{cs}{\leftarrow}& W(b^2 \mathbb{R}) \\ \uparrow && \uparrow \\ inv(\mathfrak{g}) &\stackrel{\langle-,-\rangle}{\leftarrow}& inv(b^2 \mathbb{R}) & = CE(b^3 \mathbb{R}) }
  • 𝔤 μ:=𝔰𝔱𝔯𝔦𝔫𝔤\mathfrak{g}_\mu := \mathfrak{string} the string Lie 2-algebra.

Definition

Let (b𝔤 μ)(b\mathbb{R} \to \mathfrak{g}_\mu) denote the L-∞ algebra whose Chevalley-Eilenberg algebra is

CE(b𝔤 μ)=( (𝔤 *bc),d), CE(b\mathbb{R} \to \mathfrak{g}_\mu) = (\wedge^\bullet( \mathfrak{g}^* \oplus \langle b\rangle \oplus \langle c \rangle ), d) \,,

with bb a generator in degree 2, and cc a generator in degree 3, and with differential defined on generators by

d| 𝔤 * =[,] * db =μ+c dc =0. \begin{aligned} d|_{\mathfrak{g}^*} & = [-,-]^* \\ d b & = - \mu + c \\ d c & = 0 \end{aligned} \,.
Proposition

The 3-cocycle CE(𝔤)μCE(b 2) CE(\mathfrak{g}) \stackrel{\mu}{\leftarrow} CE(b^2 \mathbb{R}) factors as

CE(𝔤)(cμ,b0)CE(b𝔤 μ)(cc)CE(b 2):μ, CE(\mathfrak{g}) \stackrel{(c \mapsto \mu, b \mapsto 0)}{\leftarrow} CE(b\mathbb{R} \to \mathfrak{g}_\mu) \stackrel{(c \mapsto c)}{\leftarrow} CE(b^2 \mathbb{R}) : \mu \,,

where the morphism on the left (which is the identity when restricted to 𝔤 *\mathfrak{g}^* and acts on the new generators as indicated) is a quasi-isomorphism.

The point of introducing the resolution (b𝔤 μ)(b \mathbb{R} \to \mathfrak{g}_\mu) in the above way is that it naturally supports the obstruction theory of lifts from 𝔤\mathfrak{g}-connections to string Lie 2-algebra 2-connection

Observation

The defining projection 𝔤 μ𝔤\mathfrak{g}_\mu \to \mathfrak{g} factors through the above quasi-isomorphism (b𝔤 μ)𝔤(b \mathbb{R} \to \mathfrak{g}_\mu) \to \mathfrak{g} by the canonical inclusion

𝔤 μ(b𝔤 μ), \mathfrak{g}_\mu \to (b \mathbb{R} \to \mathfrak{g}_\mu) \,,

which dually on CECE-algebras is given by

t at a t^a \mapsto t^a
bb b \mapsto - b
c0. c \mapsto 0 \,.

In total we are looking at a convenient presentation of the long fiber sequence of the string Lie 2-algebra extension:

(b𝔤 μ) b 2 b 𝔤 μ 𝔤. \array{ && && (b \mathbb{R} \to \mathfrak{g}_\mu) &\to& b^2 \mathbb{R} \\ && & \nearrow & \downarrow^{\mathrlap{\simeq}} \\ b \mathbb{R} &\to& \mathfrak{g}_\mu &\to& \mathfrak{g} } \,.

(The signs appearing here are just unimportant convention made in order for some of the formulas below to come out nice.)

Proposition

The image under Lie integration of the above factorization is

exp(μ):cosk 3exp(𝔤)cosk 3exp(b𝔤 μ)B 3/ c \exp(\mu) : \mathbf{cosk}_3\exp(\mathfrak{g}) \to \mathbf{cosk}_3\exp(b \mathbb{R} \to \mathfrak{g}_\mu) \to \mathbf{B}^3 \mathbb{R}/\mathbb{Z}_c

where the first morphism is a weak equivalence followed by a fibration in the model structure on simplicial presheaves [CartSp smooth op,sSet] proj[CartSp_{smooth}^{op}, sSet]_{proj}.

Proof

To see that the left morphism is objectwise a weak homotopy equivalence, notice that a [k][k]-cell of exp(b𝔤 μ)\exp(b \mathbb{R} \to \mathfrak{g}_\mu) consists of a triple (A,B,C)(A,B,C), where AA is a vertical flat 𝔤\mathfrak{g}-valued 1-form on U×Δ kU\times\Delta^k, BB is a vertical 2-form and CC a 3-form on U×Δ kU\times\Delta^k, such that dB=Cμ(A,A,A)d B=C-\mu(A,A,A) and dC=0d C=0, since AA is flat. Therefore CC is uniquely determined by AA and BB, and there are no conditions on BB. This means that a [k][k]-cell of exp(b𝔤 μ)\exp(b \mathbb{R} \to \mathfrak{g}_\mu) is identified with a pair consisting of a based smooth function f:Δ kSpinf : \Delta^k \to Spin and a vertical 2-form BΩ si,vert 2(U×Δ k)B \in \Omega^2_{si,vert}(U \times \Delta^k), (both suitably with sitting instants perpendicular to the boundary of the simplex). Since there is no further condition on the 2-form, it can always be extended from the boundary of the kk-simplex to the interior (for instance simply by radially rescaling it smoothly to 0). Accordingly the simplicial homotopy groups of exp(b𝔤 μ)(U)\exp(b \mathbb{R} \to \mathfrak{g}_\mu)(U) are the same as those of exp(𝔤)(U)\exp(\mathfrak{g})(U). The morphism between them is the identity in ff and picks B=0B = 0 and is hence clearly an isomorphism on homotopy groups.

We turn now to discussing that the second morphism is a fibration. The nontrivial degrees of the lifting problem

Λ[k] i cosk 3exp(b𝔤 μ)(U) Δ[k] B 3/ c(U) \array{ \Lambda[k]_i &\to& \mathbf{cosk}_3\exp(b\mathbb{R} \to \mathfrak{g}_\mu)(U) \\ \downarrow && \downarrow \\ \Delta[k] &\to& \mathbf{B}^3 \mathbb{R}/\mathbb{Z}_c(U) }

are k=3k = 3 and k=4k = 4.

Notice that a 33-cell of B 3/ c(U)\mathbf{B}^3 \mathbb{R}/ \mathbb{Z}_c(U) is a smooth function U/U \to \mathbb{R}/\mathbb{Z} and that the morphism exp(b𝔤 μ)B 3/ c\exp(b\mathbb{R} \to \mathfrak{g}_\mu) \to \mathbf{B}^3 \mathbb{R}/\mathbb{Z}_c sends the pair (f,B)(f,B) to the fiber integration Δ 3(f *θ[θθ]+dB)mod\int_{\Delta^3}(f^* \langle \theta \wedge [\theta \wedge \theta]\rangle + d B) mod \mathbb{Z}.

Our lifting problem in degree 3, has given a function c:U×Δ 3/c : U \times \Delta^3 \to \mathbb{R}/\mathbb{Z} and a smooth function (with sitting instants at the subfaces) f:U×Λ i 3Spinf : U \times \Lambda^3_i \to Spin together with a 2-form BB on the horn U×Λ i 3U \times \Lambda^3_i.

By pullback along the standard continuous retract Δ 3Λ i 3\Delta^3 \to \Lambda^3_i which is non-smooth only where ff has sitting instants, we can always extend ff to a smooth function f:U×Δ 3Spinf' : U \times \Delta^3 \to Spin with the property that Δ 3(f) *θ[θθ]=0\int_{\Delta^3} (f')^* \langle \theta \wedge [\theta \wedge \theta]\rangle = 0. (Following the general discussion at Lie integration.)

In order to find a horn filler for the 2-form component, consider any smooth 2-form with sitting instants and non-vanishing integeral on Δ 2\Delta^2, regarded as the missing face of the horn. By multiplying it with a suitable smooth function on UU we can obtain an extension B˜Ω si,vert 3(U×Δ 3)\tilde B \in \Omega^3_{si,vert}(U \times \partial \Delta^3) of BB to all of U×Δ 3U \times \partial \Delta^3 with the property that its integral over Δ 3\partial \Delta^3 is the given cc. By the Stokes theorem it remains to extend B˜\tilde B to the interior of Δ 3\Delta^3 in any way, as long as it is smooth and has sitting instants.

To that end, we can find in a similar fashion a smooth UU-parameterized family of closed 3-forms HH with sitting instants on Δ 3\Delta^3, whose integral over Δ 3\Delta^3 equals cc. Since by sitting instants this 3-form vanishes in a neighbourhood of the boundary, the standard formula for the Poincare lemma applied to it produces a 2-form BΩ si,vert 2(U×Δ 3)B' \in \Omega^2_{si, vert}(U \times \Delta^3) with dB=Cd B' = C that itself is radially constant at the boundary. By construction the difference B˜B| Δ 3\tilde B - B'|_{\partial \Delta^3} has vanishing surface integral. By the discussion at Lie integration it follows that the difference extends smoothly and with sitting instants to a closed 2-form B^Ω si,vert 2(U×Δ 3)\hat B \in \Omega^2_{si,vert}(U \times \Delta^3). Therefore the sum

B+B^Ω si,vert 2(U×Δ 3) B' + \hat B \in \Omega^2_{si,vert}(U \times \Delta^3)

equals BB when restricted to Λ i k\Lambda^k_i and has the property that its integral over Δ 3\Delta^3 equals cc. Together with our extension ff', this constitutes a pair that solves the lifting problem.

The extension problem in degree 4 amounts to a similar construction: by coskeletalness the condition is that for a given c:U/c : U \to \mathbb{R}/\mathbb{Z} and a given vertical 2-form on U×Δ 3U \times \partial \Delta^3 such that its integral equals cc, as well as a function f:U×Δ 3Spinf : U \times \partial \Delta^3 \to Spin, we can extend the 2-form and the function along U×Δ 3U×Δ 3U \times \partial \Delta^3 \to U \times \Delta^3. The latter follows from the fact that π 2Spin=0\pi_2 Spin = 0 which guarantees a continuous filler (with sitting instants), and using the Steenrod-Wockel approximation theorem to make this smooth. We are left with the problem of extending the 2-form, which is the same problem we discussed above after the choice of B˜\tilde B.

We now proceed to extend this factorization to the exponentiated differential coefficients (see connection on an ∞-bundle).

Proposition

(presentation of the differential class by a fibration)

Under Lie integration the above factorization of the Lie algebra cocycle
maps to the factorization

exp(μ,cs):cosk 3exp(𝔤) ChWcosk 3exp((b𝔤 μ)) ChWB 3U(1) ChW,ch \exp(\mu, cs) : \mathbf{cosk}_3 \exp(\mathfrak{g})_{ChW} \stackrel{\simeq}{\to} \mathbf{cosk}_3 \exp((b \mathbb{R} \to \mathfrak{g}_\mu))_{ChW} \to \mathbf{B}^3 U(1)_{ChW,ch}

of exp(μ,cs)\exp(\mu,cs) in [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj}, where the first morphism is a weak equivalence and the second a fibration.

Proof

The following proof makes use of details discussed at differential string structure -- proofs .

We discuss that the first morphism is an equivalence. Clearly it is injective on homotopy groups: if a sphere of AA-data cannot be filled, then also adding the (B,C)(B,C)-data does not yield a filler. So we need to check that it is also surjective on homotopy groups: if the AA-data can be filled, then also the corresponding (B,C)(B,C)-data has a filler. Since the curvature HH is horizontal it is already extended. We may extend BB in any smooth way to U×Δ kU \times \Delta^k (for instance by rescaling it smoothly to zero at the center of the kk-simplex) and then take the equation dB=CS(A)+C+Hd B = - CS(A) + C + H to define the extension of CC.

We now check that the second morphism is a fibration. It is itself the composite

cosk 3exp(b𝔤 μ) ChWexp(b 2) ChW/ Δ B 3/ ChW,ch. \mathbf{cosk}_{3} \exp(b \mathbb{R} \to \mathfrak{g}_\mu)_{ChW} \to \exp(b^2 \mathbb{R})_{ChW}/\mathbb{Z} \stackrel{\int_{\Delta^\bullet}}{\to} \mathbf{B}^3 \mathbb{R}/\mathbb{Z}_{ChW,ch} \,.

Here the second morphism is a degreewise surjection of simplicial abelian groups, hence a degreewise surjection under the normalized chain complex functor, hence is itself already a projective fibration. Therefore it is sufficient to show that the first morphism here is a fibration.

In degree k=0k = 0 to k=3k = 3 the lifting problems

Λ[k] i exp(b𝔤 μ) smp,ChW(U) Δ[k] exp(b 2) smp,ChW/(U) \array{ \Lambda[k]_i &\to & \exp(b\mathbb{R} \to \mathfrak{g}_{\mu})_{smp,ChW}(U) \\ \downarrow && \downarrow \\ \Delta[k] &\to& \exp(b^2 \mathbb{R})_{smp,ChW}/\mathbb{Z}(U) }

may all be equivalently reformulated as lifting against a cylinder D kD k×[0,1]D^k \hookrightarrow D^k \times [0,1] by using the sitting instants of all forms.

We have then a 3-form CΩ si 3(U×D k1×[0,1])C \in \Omega^3_{si}(U \times D^{k-1}\times [0,1]) with horizontal curvature 𝒢Ω 4(U)\mathcal{G} \in \Omega^4(U) and differential form data (A,B)(A,B) on U×D k1U \times D^{k-1} given. We may always extend AA along the cylinder direction [0,1][0,1] (its vertical part is equivalently a based smooth function to SpinSpin which we may extend constantly). HH has to be horizontal so it is to be constantly extended along the cylinder.

We can then use the kind of formula that proves the Poincare lemma to extend BB. Let Ψ:(D k×[0,1])×[0,1](D k×[0,1])\Psi : (D^k \times [0,1]) \times [0,1] \to (D^k \times [0,1]) be a smooth contraction. Then while d(HCS(A)+C)d(H - CS(A) + C) may be non-vanishing, by horizonatlity of their curvature characteristic forms we still have that ι tΨ t *d(HCS(A)+C)\iota_{\partial_t} \Psi_t^* d(H - CS(A) + C) vanishes (since the contraction vanishes).

Therefore the 2-form

B˜:= [0,1]ι tΨ t *(HCS(A)+C) \tilde B := \int_{[0,1]} \iota_{\partial_t} \Psi_t^*(H - CS(A) + C)

satisfies dB˜=(HCS(A)+C)d \tilde B = (H - CS(A) + C). It may however not coincide with our given BB at t=0t = 0. But the difference BB˜| t=0B - \tilde B|_{t = 0} is a closed form on the left boundary of the cylinder. We may find some closed 2-form on the other boundary such that the integral around the boundary vanishes. Then the argument from the proof of the Lie integration of the line Lie n-algebra applies and we find an extension λ\lambda to a closed 2-form on the interior. The sum

B^:=B˜+λ \hat B := \tilde B + \lambda

then still satisfies dB^=HCS(A)Cd \hat B = H - CS(A) - C and it coincides with BB on the left boundary.

Notice that here B˜\tilde B indeed has sitting instants: since HH, CS(A)CS(A) and CC have sitting instants they are constant on their value at the boundary in a neighbourhood perpendicular to the boundary, which means for these 3-forms in the degrees 3\leq 3 that they vanish in a neighbourhood of the boundary, hence that the above integral is towards the boundary over a vanishing integrand.

In degree 4 the nature of the lifting problem

Λ[4] i cosk 3exp(b𝔤 μ)(U) Δ[4] B 3/ ChW,ch \array{ \Lambda[4]_i &\to& \mathbf{cosk}_3\exp(b\mathbb{R} \to \mathfrak{g}_\mu)(U) \\ \downarrow && \downarrow \\ \Delta[4] &\to& \mathbf{B}^3 \mathbb{R}/\mathbb{Z}_{ChW,ch} }

starts out differently, due to the presence of cosk 3\mathbf{cosk}_3, but it then ends up amounting to the same kind of argument:

We have four functions U/U \to \mathbb{R}/\mathbb{Z} which we may realize as the fiber integration of a 3-form CC on U×(Δ[4]δ iΔ[3])U \times (\partial \Delta[4] \setminus \delta_i \Delta[3]), and we have a lift to (A,B,C,H)(A,B,C, H)-data on U×(Δ[4]δ i(Δ[3]))U \times (\partial \Delta[4]\setminus \delta_i(\Delta[3])) (the boundary of the 4-simplex minus one of its 3-simplex faces).

We observe that we can

  • always extend CC smoothly to the remaining 3-face such that its fiber integration there reproduces the signed difference of the four given functions corresponding to the other faces (choose any smooth 3-form with sitting instants and with non-vanishing integral and rescale smoothly);

  • fill the AA-data horizonatlly due to the fact that π 2(Spin)=0\pi_2 (Spin) = 0.

  • the HH-form is already horizontal, hence already filled.

Moreover, by the fact that the 2-form BB already is defined on all of Δ[4]δ i(Δ[3])\partial \Delta[4] \setminus \delta_i(\Delta[3]) its fiber integral over the boundary Δ[3]\partial \Delta[3] coincides with the fiber integral of HCS(A)+CH - CS(A) + C over Δ[4]δ i(Δ[3])\partial \Delta[4] \setminus \delta_i (\Delta[3])). But by the fact that we have lifted CC and the fact that μ(A vert)=CS(A)| Δ 3\mu(A_{vert}) = CS(A)|_{\Delta^3} is an integral cocycle, it follows that this equals the fiber integral of CCS(A)C - CS(A) over the remaining face.

Use then as above the vertical Poincare lemma-formula to find B˜\tilde B on U×Δ 3U \times \Delta^3 with sitting instants that satisfies the equation dB=HCS(A)+Cd B = H - CS(A) + C there. Then extend the closed difference BB˜| 0B - \tilde B|_{0} to a closed smooth 2-form on Δ 3\Delta^3. As before, the difference

B^:=B˜+λ \hat B := \tilde B + \lambda

is an extension of BB that constitutes a lift.

Explicit Cech cocycles

Corollary

For any XX \in SmoothMfd \hookrightarrow Smooth∞Grpd, for any choice of differentiaby good open cover with corresponding cofibrant presentation X^=C({C i})[CartSp smooth op,sSet] proj\hat X = C(\{C_i\})\in [CartSp_{smooth}^{op}, sSet]_{proj} we have that the 2-groupoids of twisted different String structuress are presented by the ordinary fibers of the morphism of Kan complexes

[CartSp op,sSet](X^,exp(μ,cs)):[CartSp op,sSet](X^,cosk 3exp(b𝔤 μ) ChW)[CartSp op,sSet](X^,B 3U(1) ChW). [CartSp^{op}, sSet](\hat X,\exp(\mu,cs)) : [CartSp^{op}, sSet](\hat X, \mathbf{cosk}_3 \exp(b \mathbb{R} \to \mathfrak{g}_\mu)_{ChW}) \to [CartSp^{op}, sSet](\hat X, \mathbf{B}^3 U(1)_{ChW}) \,.

over any basepoints in the connected components of the Kan complex on the right, which correspond to the elements
[C^ 3]H diff 4(X)[\hat \mathbf{C}_3] \in H_{diff}^4(X) in the ordinary differential cohomology of XX.

Proof

Since [CartSp smooth op,sSet] proj[CartSp_{smooth}^{op}, sSet]_{proj} is a simplicial model category the morphism [CartSp op,sSet](X^,exp(μ,cs))[CartSp^{op}, sSet](\hat X,\exp(\mu,cs)) is a fibration because exp(μ,cs)\exp(\mu,cs) is and X^\hat X is cofibrant.

It follows from the discussion at homotopy pullback that the ordinary pullback of simplicial presheaves

String diff,tw(X) H diff 4(X) [CartSp op,sSet](X^,cosk 3exp(b𝔤 μ) ChW) [CartSp op,sSet](X^,B 3U(1) ChW) \array{ String_{diff,tw}(X) &\to& H_{diff}^4(X) \\ \downarrow && \downarrow \\ [CartSp^{op}, sSet](\hat X, \mathbf{cosk}_3 \exp(b \mathbb{R} \to \mathfrak{g}_\mu)_{ChW}) &\to& [CartSp^{op}, sSet](\hat X, \mathbf{B}^3 U(1)_{ChW}) }

is a presentation for the defining (∞,1)-pullback for String diff,tw(X)String_{diff,tw}(X), as defined above.

We unwind the explicit expression for a twisted differential string structure under this equivalence.

Any twisting cocycle is in the above presentation given by a Cech Deligne-cocycle (as discussed at circle n-bundle with connection)

H^ 3=((H 3) i,) \hat \mathbf{H}_3 = ((H_3)_i, \cdots)

with local connection 3-form (H 3) iΩ 3(U i)(H_3)_i \in \Omega^3(U_i) and globally defined curvature 4-form 𝒢 4Ω 4(X)\mathcal{G}_4 \in \Omega^4(X).

Note

A differential string structure on XX twisted by this cocycles is on patches U iU_i a morphism

Ω (U i)W˜(b𝔤 μ) \Omega^\bullet(U_i) \leftarrow \tilde W(b\mathbb{R}\to \mathfrak{g}_\mu)

in dgAlg, subject to some horizontality constraints. The components of this are over each U iU_i a collection of differential forms of the following structure

(F ω= dω+12[ωω] H 3= B:=dB+CS(ω)C 3 𝒢 4= dC 3 dF ω= [ωF ω] dH 3= F ωF ω𝒢 4 d𝒢 4= 0) it a ω a r a F ω a b B c C 3 h H 3 g 𝒢 4|(r a= dt a+12C a bct bt c h= db+csc g= dc dr a= C a bct br a dh= ,g dg= 0). \left( \array{ F_\omega =& d \omega + \frac{1}{2}[\omega \wedge \omega] \\ H_3 =& \nabla B := d B + CS(\omega) - C_3 \\ \mathcal{G}_4 =& d C_3 \\ d F_\omega =& - [\omega \wedge F_\omega] \\ d H_3 =& \langle F_\omega \wedge F_\omega\rangle - \mathcal{G}_4 \\ d \mathcal{G}_4 =& 0 } \right)_i \;\;\;\; \stackrel{ \array{ t^a & \mapsto \omega^a \\ r^a & \mapsto F^a_\omega \\ b & \mapsto B \\ c & \mapsto C_3 \\ h & \mapsto H_3 \\ g & \mapsto \mathcal{G}_4 } }{\leftarrow}| \;\;\;\; \left( \array{ r^a =& d t^a + \frac{1}{2}C^a{}_{b c} t^b \wedge t^c \\ h = & d b + cs - c \\ g =& d c \\ d r^a =& - C^a{}_{b c} t^b \wedge r^a \\ d h =& \langle -,-\rangle - g \\ d g =& 0 } \right) \,.

Here we are indicating on the right the generators and their relation in W˜(b𝔤 μ)\tilde W(b\mathbb{R} \to \mathfrak{g}_\mu) and on the left their images and the images of the relations in Ω (U i)\Omega^\bullet(U_i). This are first the definitions of the curvatures themselves and then the Bianchi identities satisfied by these.

Differential string structures and fermionic string quantum anomalies

The Pfaffian line bundle controlling the fermionic path integral of the heterotic superstring propagating on target XX trivializes precisely if the target has a (geometric) string structure.

One shows that the Pfaffian line bundle on the worldsheet is isomorphic as a bundle with connection with the transgression of the differential string structure on the target space to the mapping space [Σ,X][\Sigma,X]. So the target space having a (differential) string structure is a sufficient condition for the cancellation of the quantum anomaly.

(First argued in Killingback, later made precise in (Bunke)).

The Green-Schwarz mechanism in heterotic supergravity

We discuss the application of twisted differential string structures in supergravity and string theory.

Local differential form data as in note 2 above is known in theoretical physics in the context of the Green-Schwarz mechanism for 10-dimensional supergravity.

In this context

  • ω\omega is called the spin connection;

  • the components ((H 3) i,)((H_3)_i, \cdots) of the above cocycle are known as the 𝒢^ 4\hat \mathcal{G}_4-twisted Kalb-Ramond field.

In this application the twisting cocycle 𝒢^ 4H diff 4(X)\hat \mathcal{G}_4 \in H^4_{diff}(X) is itself the Chern-Simons circle 3-bundle of a unitary group-principal bundle with local connection form AΩ 1(U,𝔲)A \in \Omega^1(U, \mathfrak{u}). Therefore in this case C 3=CS(A)C_3 = CS(A) and the above local form data becomes

H 3=dB+CS(ω)CS(A) H_3 = d B + CS(\omega) - CS(A)
dH 3=F ωF ωF AF A. d H_3 = \langle F_\omega \wedge F_\omega \rangle - \langle F_A \wedge F_A \rangle \,.

Since H 3H_3 is the would-be curvature of a circle 2-bundle with connection, this is the first higher Maxwell equation that exhibits

j mag:=F ωF ωF AF A j_{mag} := \langle F_\omega \wedge F_\omega \rangle - \langle F_A \wedge F_A \rangle

as the magnetic charge distribution that twists this 2-bundle. This may be interpreted as the magnetic charge density of a classical background density of magnetic fivebranes. For more details on this see Green-Schwarz mechanism.

More precisely, the twisted differential string structure of the Green-Schwarz mechanism in heterotic supergravity for fixed gauge bundles are therefore given by the (∞,1)-pullback

GSBackground fixedgaugebundle(X) π 0H conn(X,BU) c^ 2 H conn(X,BSpin) 12p^ 1 H dR(X,B 3U(1)). \array{ GSBackground_{fixed gauge bundle}(X) &\to& \pi_0 \mathbf{H}_{conn}(X, \mathbf{B}U) \\ \downarrow &\swArrow_{\simeq}& \downarrow^{\mathrlap{\hat \mathbf{c}_2}} \\ \mathbf{H}_{conn}(X,\mathbf{B}Spin) &\stackrel{\frac{1}{2}\hat \mathbf{p}_1}{\to}& \mathbf{H}_{dR}(X, \mathbf{B}^3 U(1)) } \,.

Clearly, if we take into account also gauge transformations of the gauge bundle, we should replace this by the full

GSBackground(X) H conn(X,BU) c^ 2 H conn(X,BSpin) 12p^ 1 H dR(X,B 3U(1)). \array{ GSBackground(X) &\to& \mathbf{H}_{conn}(X, \mathbf{B}U) \\ \downarrow &\swArrow_{\simeq}& \downarrow^{\mathrlap{\hat \mathbf{c}_2}} \\ \mathbf{H}_{conn}(X,\mathbf{B}Spin) &\stackrel{\frac{1}{2}\hat \mathbf{p}_1}{\to}& \mathbf{H}_{dR}(X, \mathbf{B}^3 U(1)) } \,.

The look of this diagram makes manifest how in this situation we are looking at the structures that homotopically cancel the differential classes 12p^\frac{1}{2}\hat \mathbf{p} and c^ 2\hat \mathbf{c}_2 against each other.

More discussion of this is in (SSSIII).

Since H dR(X,B 3U(1))\mathbf{H}_{dR}(X, \mathbf{B}^3 U(1)) is abelian, we may consider the corresponding Mayer-Vietoris sequence by realizing GSBackground(X)GSBackground(X) equivalently as the homotopy fiber of the difference of differential cocycles 12p^ 1c^ 2\frac{1}{2}\hat \mathbf{p}_1 - \hat \mathbf{c}_2.

GSBackground(X) * H conn(X,BSpin×BU) 12p^ 1c^ 2 H dR(X,B 4U(1)). \array{ GSBackground(X) &\to& * \\ \downarrow &\swArrow_{\simeq}& \downarrow \\ \mathbf{H}_{conn}(X,\mathbf{B}Spin \times \mathbf{B}U) &\stackrel{\frac{1}{2}\hat \mathbf{p}_1-\hat \mathbf{c}_2}{\to}& \mathbf{H}_{dR}(X, \mathbf{B}^4 U(1)) } \,.

Indeed, the above explicit presentation by simplicial presheaves generalizes immediately to describe this case, realizing UU-twisted differential string structures equivalently as differential “untwisted UU-twisted-string-structures”.

We may usefully formalize this further by defining the String c 2String^{\mathbf{c}_2}-2-group to be the homotopy fiber

BString c 2 * G(Spin×U) 12p 1c 2 B 3U(1). \array{ \mathbf{B} String^{\mathbf{c}_2} &\to& * \\ \downarrow && \downarrow \\ \mathbf{G}(Spin \times U) &\stackrel{\frac{1}{2}\mathbf{p}_1 - \mathbf{c}_2}{\to}& \mathbf{B}^3 U(1) } \,.

We have then that GSBackground(X)GSBackground(X) is the 3-groupoid of untwisted differential BString c 2\mathbf{B}String^{\mathbf{c}_2}-structures.

GSBackground(X) * 0 H conn(B(Spin×U)) 12p^ 1c 2 H dR(X,B 4U(1)). \array{ GSBackground(X) &\to& * \\ \downarrow && \downarrow^{0} \\ \mathbf{H}_{conn}(\mathbf{B} (Spin \times U)) &\stackrel{\frac{1}{2}\hat \mathbf{p}_1 - \mathbf{c}_2}{\to}& \mathbf{H}_{dR}(X, \mathbf{B}^4 U(1)) } \,.

More on this in (FiSaSc).

This is supposed to be (see section 12 of (DFM)) the restriction to the boundary of the supergravity C-field, which is the (,1)(\infty,1)-pullback

CField(Y) H dR 4(Y) dC 3 H(Y,B(Spin×U)) 12p^ 1c 2 H(Y,B 4U(1)). \array{ C Field(Y) &\to& H^4_{dR}(Y) \\ \downarrow && \downarrow^{\mathrlap{d C_3}} \\ \mathbf{H}(Y,\mathbf{B} (Spin \times U)) &\stackrel{\frac{1}{2}\hat \mathbf{p}_1 - \mathbf{c}_2}{\to}& \mathbf{H}(Y, \mathbf{B}^4 U(1)) } \,.

where YY is 11-dimensional with paritalY=X\parital Y = X. Notice that here in the bottom left we have bundles without connection, or equivalently (when computing the homotopy pullback by an ordinary pullback along a fibration) with pseudo-connections.

By the discussion at supergravity C-field under a shift of connection 1 2\nabla_1 \mapsto \nabla_2 the CC-field transforms as

C 2=C 1+CS( 1, 2), C_2 = C_1 + CS(\nabla_1, \nabla_2) \,,

where on the right we have the relative Chern-Simons form. This vanishes precisely on the genuine gauge transformations. Hence as we restrict from 11-dimensions to 10, two things happen:

  1. the supergravity CC-field vanishes,

  2. the gauge bundles develop genuine connections.

Relation to string 2-connections

By the discussion at connection on an ∞-bundle we have that for 𝔤\mathfrak{g} an L-∞ algebra and

BG:=cosk n+1exp(𝔤) \mathbf{B}G := \mathbf{cosk}_{n+1} \exp(\mathfrak{g})

the delooping of the smooth Lie n-group obtained from it by Lie integration, the coefficient for ∞-connections on GG-principal ∞-bundles is

BG conn:=cosk n+1exp(𝔤) conn, \mathbf{B}G_{conn} := \mathbf{cosk}_{n+1} \exp(\mathfrak{g})_{conn} \,,

where on the very right we have the simplicial presheaf

exp(𝔤):(U,[k]){A:CE(𝔤)Ω (U×Δ n)|AisverticallyflatandF Aishorizontal}. \exp(\mathfrak{g}) : (U,[k]) \mapsto \left\{ A : CE(\mathfrak{g}) \to \Omega^\bullet(U \times \Delta^n) | A\;is\;vertically\;flat\; and\;F_A\;is \; horizontal \right\} \,.

(See ∞-Chern-Weil homomorphism for details).

Proposition

The 2-groupoid of entirely untwisted differential string structures on XX (the twist being 0H diff 4(X)0 \in H^4_{diff}(X)) is equivalent to that of string 2-group principal 2-bundles with 2-connection:

String diff,tw=0(X)String2Bund (X). String_{diff, tw = 0}(X) \simeq String 2Bund_{\nabla}(X) \,.
Proof

By the above discussion of Cech cocycles we compute String diff,tw=0(X)String_{diff, tw = 0}(X) as the ordinary fiber of the morphism of simplicial presheaves

[CartSp op,sSet](C({U i}),cosk 3exp(b𝔤 μ))[CartSp op,sSet](C({U i}),B 3U(1) diff) [CartSp^{op}, sSet]( C(\{U_i\}), \mathbf{cosk}_3 \exp(b \mathbb{R} \to \mathfrak{g}_\mu)) \to [CartSp^{op}, sSet]( C(\{U_i\}), \mathbf{B}^3 U(1)_{diff})

over the identically vanishing cocycle.

In terms of the component formulas spelled out in the above discussion of the GS-mechanism, this amounts to restricting to those cocyles for which n each degree the equations

C=0 C = 0
G=0 G = 0

holds.

Comparing this to the explicit formulas for exp(b𝔤 μ)\exp(b \mathbb{R} \to \mathfrak{g}_\mu) and exp(b𝔤 μ) conn\exp(b \mathbb{R} \to \mathfrak{g}_\mu)_{conn} in the above we see that these cocycles are exactly those that factor through the canonical inclusion

𝔤 μ(b𝔤 μ) \mathfrak{g}_\mu \to (b \mathbb{R} \to \mathfrak{g}_\mu)

from observation 6 of the string Lie 2-algebra into the mapping cone Lie 3-algebra of the extension b𝔤 μ𝔤b \mathbb{R} \to \mathfrak{g}_\mu \to \mathfrak{g} that defines it.

Whitehead tower of the orthogonal group

References

A discussion of differential string structures in terms of bundle 2-gerbes is given in

The description of the gauge transformations of the supergravity C-field is in section 3 of

E. Diaconescu, Greg Moore, Dan Freed, The MM-theory 3-form and E 8E_8-gauge theory (arXiv:hep-th/0312069)

The local data for the ∞-Lie algebra valued differential forms for the description of twisted differential string structures as above was given in

The full Cech-Deligne cocycles induced by this (but not yet the homotopy fibers over them) were discussed in

The 2-group String c 2String^{\mathbf{c}_2} and its differential structurs, etc. are discussed in

A comprehensive discussion including all the formal background and the applications is attempted at

in section 4.2.

The relation to quantum anomaly cancellation in heterotic string theory has been first discussed in

  • Killingback, World-sheet anomalies and loop geometry Nuclear Physics B Volume 288, 1987, Pages 578-588

and given a rigorous treatment in

  • Ulrich Bunke, String structures and trivialisations of a Pfaffian line bundle (arXiv)

Revised on March 21, 2014 09:15:04 by Urs Schreiber (89.204.138.115)