nLab circle n-bundle with connection





Differential cohomology

Cohesive \infty-Toposes



We discuss the refinement to higher differential geometry of the concept of a circle group-principal connection (on a circle group-principal bundle). Specifically we indicate how the general abstract definition in terms of cohesion reproduces in the context of smooth cohesion to the representation of circle nn-connections by cocycles in smooth Deligne cohomology (dcct).

In every cohesive (∞,1)-topos H\mathbf{H} there is an intrinsic notion of differential cohomology with coefficients in an abelian group object AHA \in \mathbf{H} that classifies B n1A\mathbf{B}^{n-1}A-principal ∞-bundles with ∞-connection.

Here we discuss the specific realization for H=\mathbf{H} = Smooth∞Grpd the (∞,1)-topos of smooth ∞-groupoids and A=U(1)A = U(1) the circle group.

In this case the intrinsic differential cohomology reproduces ordinary differential cohomology and generalizes it to base spaces that may be smooth manifolds, diffeological spaces, orbifolds and generally smooth ∞-groupoids such as deloopings BG\mathbf{B}G of smooth ∞-groups GG. Differential cocycles on the latter support the ∞-Chern-Weil homomorphism that sends nonabelian ∞-connections to circle nn-bundles whose curvature form realizes a characteristic class in de Rham cohomology.

The ambient context

Let H:=\mathbf{H} := Smooth∞Grpd be the cohesive (∞,1)-topos of smooth ∞-groupoid. As usual, write

(ΠDiscΓcoDisc):SmoothGrpdcoDiscΓDiscΠGrpd (\Pi \dashv Disc \dashv \Gamma \dashv coDisc) : Smooth \infty Grpd \stackrel{\overset{\Pi}{\to}}{\stackrel{\overset{Disc}{\leftarrow}}{\stackrel{\overset{\Gamma}{\to}}{\underset{coDisc}{\leftarrow}}}} \infty Grpd

for the terminal global section (∞,1)-geometric morphism with its extra left adjoint, the intrinsic fundamental ∞-groupoid functor Π\Pi.

From this induced is the path ∞-groupoid adjunction

(Π):SmoothGrpdSmoothGrpd (\mathbf{\Pi} \dashv \mathbf{\flat}) : Smooth \infty Grpd \stackrel{\leftarrow}{\to} Smooth \infty Grpd

and the intrinsic de Rham cohomology adjunction

(Π dR dR):*/SmoothGrpd dRΠ dRSmoothGrpd. (\mathbf{\Pi}_{dR} \dashv \mathbf{\flat}_{dR}) : */Smooth \infty Grpd \stackrel{\overset{\mathbf{\Pi}_{dR}}{\leftarrow}}{\underset{\mathbf{\flat}_{dR}}{\to}} Smooth \infty Grpd \,.

For AA an abelian group object there for each integer nn is the universal curvature characteristic form, given by a cocycle-morphism

curv:B nA dRB n+1A. curv : \mathbf{B}^n A \to \mathbf{\flat}_{dR}\mathbf{B}^{n+1}A \,.

The cocycles for differential cohomology in degree nn with coefficients in AA are the points in the homotopy fiber H diff(,B nA)\mathbf{H}_{diff}(-, \mathbf{B}^n A) of the morphism on cohomology

curv *:H(,B nA)H dR(,B n+1A) curv_* : \mathbf{H}(-, \mathbf{B}^n A) \to \mathbf{H}_{dR}(-, \mathbf{B}^{n+1}A)

induced by this. Every such cocycle H diff(X,B nA)\nabla \in \mathbf{H}_{diff}(X,\mathbf{B}^n A) we may think of as an ∞-connection on the B n1A\mathbf{B}^{n-1}A-principal ∞-bundle classified by the underlying cocycle in H(X,B nA)\mathbf{H}(X, \mathbf{B}^n A).

We consider these constructions in the model H=\mathbf{H} = Smooth∞Grpd. This is the (∞,1)-category of (∞,1)-sheaves

SmoothGrpd:=Sh (,1)(CartSp smooth) Smooth\infty Grpd := Sh_{(\infty,1)}(CartSp_{smooth})

on the site CartSp smooth{}_{smooth} of Cartesian spaces and smooth functions between them. This is a general higher geometry context for differential geometry. For computations we can explicitly present this (∞,1)-category by a local model structure on simplicial presheaves [CartSp op,sSet] proj,loc[CartSp^{op}, sSet]_{proj,loc}

SmoothGrpd([CartSp op,sSet] proj,loc) Smooth \infty Grpd \simeq ([CartSp^{op}, sSet]_{proj, loc})^\circ

as described at presentations of (∞,1)-sheaf (∞,1)-toposes.

In H=\mathbf{H} = Smooth∞Grpd a canonical choice for AA is the circle group

A:=U(1)=/. A := U(1) = \mathbb{R}/\mathbb{Z} \,.

We show how the notion of smooth circle nn-bundles with connection obtained by applying the general setup above to this case reproduces ordinary differential cohomology:

We assume in the following that the reader is familiar with basics of smooth ∞-groupoids.

Flat differential cohomology

The coefficient object for flat differential cohomology in H=\mathbf{H} = Smooth∞Grpd with values in B nU(1)\mathbf{B}^n U(1) is B nU(1)=LConstΓB nU(1)\mathbf{\flat} \mathbf{B}^n U(1) = LConst \Gamma \mathbf{B}^n U(1).

The coefficient object for intrinsic de Rham cohomology is dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1), defined by the (∞,1)-pullback

dRB nU(1) B nU(1) * B nU(1). \array{ \mathbf{\flat}_{dR} \mathbf{B}^n U(1) &\to& \mathbf{\flat} \mathbf{B}^n U(1) \\ \downarrow && \downarrow \\ * &\to& \mathbf{B}^n U(1) } \,.

The following proposition provides models for these objects in in terms of ordinary differential forms.


A fibrant representative in [CartSp op,sSet] proj,cov[CartSp^{op}, sSet]_{proj,cov} of B nU(1)\mathbf{\flat} \mathbf{B}^n U(1) is

B nU(1) chn:=Ξ[C (,U(1))d dRΩ 1()d dRd dRΩ cl n()], \mathbf{\flat}\mathbf{B}^n U(1)_{chn} := \Xi[C^\infty(-,U(1)) \stackrel{d_{dR}}{\to} \Omega^1(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega^n_{cl}(-)] \,,

and a fibrant representative of dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1) is

dRB nU(1) chn:=Ξ[0Ω 1()d dRd dRΩ cl n()]. \mathbf{\flat}_{dR}\mathbf{B}^n U(1)_{chn} := \Xi[0 \to \Omega^1(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega^n_{cl}(-)] \,.

Notice that the complex of sheaves B nU(1)\mathbf{\flat}\mathbf{B}^n U(1) is that which defines flat Deligne cohomology, while that of dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1) is essentially that which defines de Rham cohomology in degree n>1n \gt 1 (see below). Also notice that we denoted by d dRd_{dR} also the differential C (,U(1))d dRlogΩ 1()C^\infty(-,U(1)) \stackrel{d_{dR} log}{\to} \Omega^1(-); this is to stress that we are looking at U(1)U(1) as the quotient /\mathbb{R}/\mathbb{Z}.


Since the global section functor Γ\Gamma amounts to evaluation on the point 0\mathbb{R}^0 and since constant simplicial presheaves on CartSp satisfy descent (on objects in CartSpCartSp!), we have that B nU(1)\mathbf{\flat} \mathbf{B}^n U(1) is represented by the complex of sheaves Ξ[constU(1)00]\Xi[const U(1) \to 0 \to \cdots \to 0]. This is weakly equivalent to Ξ[C (,U(1))d dRΩ 1()d dRd dRΩ cl n()]\Xi[C^\infty(-,U(1)) \stackrel{d_{dR}}{\to} \Omega^1(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega^n_{cl}(-)] by the Poincare lemma applied to each Cartesian space (using the same standard logic that proves the de Rham theorem) in that the degreewise inclusion

constU(1) 0 0 C (,U(1)) d dR Ω 1() Ω cl n() \array{ const U(1) &\to& 0 &\to& \cdots &\to& 0 \\ \downarrow && \downarrow && && \downarrow \\ C^\infty(-,U(1)) &\stackrel{d_{dR}}{\to}& \Omega^1(-) &\to& \cdots &\to& \Omega^n_{cl}(-) }

is objectwise a quasi-isomorphism.

Therefore a fibration in [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj} representing the counit B nU(1)B nU(1)\mathbf{\flat} \mathbf{B}^n U(1) \to \mathbf{B}^n U(1) is the image under Ξ\Xi of

C (,U(1)) Ω 1() Ω cl n() = C (,U(1)) 0 0. \array{ C^\infty(-,U(1)) &\to& \Omega^1(-) &\to & \cdots &\to& \Omega^n_{cl}(-) \\ \downarrow^{\mathrlap{=}} && \downarrow && && \downarrow \\ C^\infty(-, U(1)) &\to& 0 &\to& \cdots &\to& 0 } \,.

We observe that the pullback of this morphism to the point

Ξ[0Ω 1()Ω cl n()] Ξ[C (,U(1))Ω 1()Ω cl n()] Ξ[000] Ξ[C (,U(1))00] \array{ \Xi[0 \to \Omega^1(-) \to \cdots \to \Omega^n_{cl}(-)] &\to& \Xi[C^\infty(-,U(1)) \to \Omega^1(-) \to \cdots \to \Omega^n_{cl}(-)] \\ \downarrow && \downarrow \\ \Xi[0 \to 0 \to \cdots \to 0] &\to& \Xi[C^\infty(-,U(1)) \to 0 \to \cdots \to 0] }

is the pullback over a cospan all whose objects are fibrant and one of whose morphisms is a fibration. Therefore this is a homotopy pullback diagram in [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj} which models the (∞,1)-limit over *B nU(1)B nU(1)* \to \mathbf{B}^n U(1) \leftarrow \mathbf{\flat}\mathbf{B}^n U(1) in PSh (,1)(CartSp)PSh_{(\infty,1)}(CartSp). Since ∞-stackification preserves finite (,1)(\infty,1)-limits this models also the corresponding (,1)(\infty,1)-limit in H\mathbf{H}. Therefore the top left object is indeed a model for dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1).

de Rham cohomology

The intrinsic de Rham cohomology of Smooth∞Grpd with coefficients in \mathbb{R} or U(1)=/U(1) = \mathbb{R}/\mathbb{Z} coincides with the ordinary de Rham cohomology of smooth manifolds and smooth simplicial manifolds in degree greater than 1. This we discuss here. The meaning of the discrepancy in degee 1 and lower is discussed below.

So for this section let nn \in \mathbb{N} with n2n \geq 2.

Above in Flat U(1)-valued differential cohomology we found a fibrant representative of dRB nU(1)SmoothGrpd\mathbf{\flat}_{dR} \mathbf{B}^n U(1) \in Smooth\infty Grpd to be given by

Ξ[Ω 1()d dRΩ 2()d dRΩ cl n()] \Xi[\Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2(-) \stackrel{d_{dR}}{\to} \cdots \to \Omega^n_{cl}(-)]

in [CartSp op,sSet] proj,cov[CartSp^{op}, sSet]_{proj, cov}.


For XSmoothGrpdX \in Smooth\infty Grpd a paracompact smooth manifold we have in for H=SmoothGrpd\mathbf{H} = Smooth \infty Grpd a natural isomorphism

H dR(X,B nU(1)):=π 0H(X, dRB nU(1))H dR n(X), H_{dR}(X,\mathbf{B}^n U(1)) := \pi_0 \mathbf{H}(X,\mathbf{\flat}_{dR} \mathbf{B}^n U(1)) \simeq H_{dR}^n(X) \,,

where on the left we have the intrinsic (∞,1)-topos theoretic notion of de Rham cohomology, and on the right the ordinary notion of de Rham cohomology of a smooth manifold.


Let {U iX}\{U_i \to X\} be a good open cover. At Smooth∞Grpd is discussed that then the Cech nerve C({U i})XC(\{U_i\}) \to X is a cofibrant resolution of XX in [CartSp op,sSet] proj,cov[CartSp^{op}, sSet]_{proj,cov}. Therefore we have

H(X, dRB nU(1))[CartSp op,sSet](C({U i}),Ξ[Ω 1()d dRΩ cl n()]). \mathbf{H}(X,\mathbf{\flat}_{dR} \mathbf{B}^n U(1)) \simeq [CartSp^{op}, sSet](C(\{U_i\}), \Xi[\Omega^1(-) \stackrel{d_{dR}}{\to} \cdots \to \Omega^n_{cl}(-)]) \,.

The right hand is the \infty-groupoid of cocylces in the Cech hypercohomology of the complex of sheaves of differential forms. A cocycle is given by a collection

(C i,B ij,A ijk,,Z i 0,,i n) (C_i, B_{i j}, A_{i j k}, \cdots , Z_{i_0, \cdots, i_n})

of differential forms, with C iΩ cl n(U i)C_i \in \Omega^n_{cl}(U_i), B ijΩ n1(U iU j)B_{i j} \in \Omega^{n-1}(U_i \cap U_j), etc. , such that this collection is annihilated by the total differentoal D=d dR±δD = d_{dR} \pm \delta, where d dRd_{dR} is the de Rham differential and δ\delta the alternating sum of the pullbacks along the face maps of the Cech nerve.

It is a standard result of abelian sheaf cohomology that such cocycles represent classes in de Rham cohomology.

But for the record and since the details of this computation will show up again at some mildly subtle points in further discussion below, we spell this out in some detail.

We can explicitly construct coboundaries connecting such a generic cocycle to one of the form

(F i,0,0,,0) (F_i, 0, 0, \cdots, 0)

by using a partition of unity (ρ iC (X))(\rho_i \in C^\infty(X)) subordinate to the cover {U iX}\{U_i \to X\}, i.e. xU iρ i(x)=0x \in U_i \Rightarrow \rho_i(x) = 0 and iρ i=1\sum_i \rho_i = 1.

For consider

(C i,B ij,A ijk,,Y i 1,,i n,Z i 1,,i n+1) + D(0,0,, i 0ρ i 0Z i 0,i 1,,i n,0) = (C i,B ij,A ijk,,Y i 1,,i n+d dR i 0ρ i 0Z i 0,i 1,,i n,0), \begin{aligned} & (C_i, B_{i j}, A_{i j k}, \cdots , Y_{i_1, \cdots, i_{n}}, Z_{i_1, \cdots, i_{n+1}}) \\ + & D (0, 0, \cdots, \sum_{i_0} \rho_{i_0} Z_{i_0, i_1, \cdots, i_{n}},0) \\ = & (C_i, B_{i j}, A_{i j k}, \cdots , Y_{i_1, \cdots, i_{n}} + d_{dR}\sum_{i_0} \rho_{i_0} Z_{i_0, i_1, \cdots, i_{n}}, 0) \end{aligned} \,,

where we use that from (δZ) i 1,,i n+2=0(\delta Z)_{i_1, \cdots, i_{n+2}} = 0 it follows that

(δρZ) i 1,,i n+1 = i 0ρ i 0 k=1 n+1(1) kZ i 0,i 1,i^ k,,i n+1 = i 0ρ i 0Z i 1,,i n+1 =Z i 1,,i n+1. \begin{aligned} (\delta \sum \rho Z)_{i_1, \cdots, i_{n+1}} &= \sum_{i_0} \rho_{i_0} \sum_{k = 1}^{n+1} (-1)^k Z_{i_0, i_1 \cdots, \hat i_k, \cdots, i_{n+1}} \\ & = \sum_{i_0} \rho_{i_0} Z_{i_1 ,\cdots, i_{n+1}} \\ & = Z_{i_1 ,\cdots, i_{n+1}} \end{aligned} \,.

where I am suppressing some evident signs…

By recurseively adding coboundaries this way, we can annihilate all the higher Cech-components of the original cocycle and arrive at a cocycle of the form (F i,0,,0)(F_i, 0, \cdots, 0).

Such a cocycle being DD-closed says precisely that F i=F| U iF_i = F|_{U_i} for FΩ cl n(X)F \in \Omega^n_{cl}(X) a globally defined closed differential form. Moreover, coboundaries between two cocycles both of this form

(F i,0,,0)=(F i,0,,0)+D(λ i,λ ij,) (F_i, 0, \cdots , 0) = (F'_i, 0, \cdots, 0) + D(\lambda_i, \lambda_{i j}, \cdots)

are necessarily themselves of the form (λ i,λ ij,)=(λ i,0,,0)(\lambda_i, \lambda_{i j}, \cdots) = (\lambda_i, 0 ,\cdots, 0) with λ i=λ| U i\lambda_i = \lambda|_{U_i} for λΩ n1(X)\lambda \in \Omega^{n-1}(X) a globally defined differential nn-form and F=F+d dRλF = F' + d_{dR} \lambda.

Differential cohomology

The intrinsic definition of the ∞-groupoid of cocycles for the intrinsic differential cohomology in H=SmoothGrpd\mathbf{H} = Smooth\infty Grpd with coefficients B nU(1)\mathbf{B}^n U(1) is the object H diff(X,B nU(1))\mathbf{H}_{diff}(X,\mathbf{B}^n U(1)) in the (∞,1)-pullback

H diff(X,B nU(1)) H dR(X,B n+1U(1)) H(X,B nU(1)) curv H dR(X,B n+1U(1)) \array{ \mathbf{H}_{diff}(X,\mathbf{B}^n U(1)) &\to & H_{dR}(X,\mathbf{B}^{n+1} U(1)) \\ \downarrow && \downarrow \\ \mathbf{H}(X,\mathbf{B}^n U(1)) &\stackrel{curv}{\to}& \mathbf{H}_{dR}(X, \mathbf{B}^{n+1} U(1)) }

in ∞Grpd.

We show now that for n1n \geq 1 this reproduces the Deligne cohomology H(X,(n+1) D )H(X,\mathbb{Z}(n+1)_D^\infty) of XX:


For XX a paracompact smooth manifold we have

H diff(X,B nU(1))(H(X,(n+1) D ))× Ω cl n+1(X)H dR n+1int(X). H_{diff}(X,\mathbf{B}^n U(1)) \simeq \left( \;\; H(X,\mathbb{Z}(n+1)_D^\infty) \;\; \right) \times_{\Omega_{cl}^{n+1}(X)} H_{dR}^{n+1}_{int}(X) \,.

Here on the right we have the subset of Deligne cocycles that picks for each integral de Rham cohomology class of XX only one curvature form representative.

We give the proof below after some preliminary expositional discussion.


The restriction to single representatives in each de Rham class is a reflection of the fact that in the above (,1)(\infty,1)-pullback diagram the morphism H dR(X,B n+1U(1))H dR(X,B n+1U(1))H_{dR}(X,\mathbf{B}^{n+1}U(1)) \to \mathbf{H}_{dR}(X,\mathbf{B}^{n+1}U(1)) by definition picks one representative in each connected component. Using the above model of the intrinsic de Rham cohomology in terms of globally defined differential froms, we could easily get rid of this restriction by considering instead of the above (,1)(\infty,1)-pullback the homotopy pullback

H diff(X,B nU(1)) Ω cl n+1(X) H(X,B nU(1)) curv H dR(X,B n+1U(1)) \array{ \mathbf{H}'_{diff}(X,\mathbf{B}^n U(1)) &\to & \Omega_{cl}^{n+1}(X) \\ \downarrow && \downarrow \\ \mathbf{H}(X,\mathbf{B}^n U(1)) &\stackrel{curv}{\to}& \mathbf{H}_{dR}(X, \mathbf{B}^{n+1} U(1)) }

where now the right vertical morphism is the inclusion of the set of objects of our concrete model for the \infty-groupoid H dR(X,B n+1U(1))\mathbf{H}_{dR}(X, \mathbf{B}^{n+1} U(1)). With this definition we get the isomorphism

H diff(X,B nU(1))H(X,(n+1) D ). H'_{diff}(X,\mathbf{B}^n U(1)) \simeq H(X,\mathbb{Z}(n+1)_D^\infty) \,.

From the tradtional point of view of differential cohomology this may be what one expects to see, but from the intrinsic (,1)(\infty,1)-topos theoretic point of view it is quite unnatural – and in fact “evil” – to fix that set of objects of the \infty-groupoid. Of intrinsic meaning is only the set of their equivalences classes.

Circle bundles with connection

Before discussing the full theorem, it is instructive to start by looking at the special case n=1n=1 in some detail, which is about ordinary U(1)U(1)-principal bundles with connection.

This contains in it already all the relevant structure of the general case, but the low categorical degree is more transparently written out and will allow us to pause to highlight some maybe noteworthy aspects of the situation, such as the phenomenon of pseudo-connections below.

In terms of the Dold-Kan correspondence the object BU(1)H\mathbf{B}U(1) \in \mathbf{H} is modeled in [CartSp op,sSet][CartSp^{op}, sSet] by

BU(1)=Ξ(C (,U(1))0). \mathbf{B}U(1) = \Xi(\; C^\infty(-,U(1)) \to 0 \;) \,.

Accordingly we have for the double delooping the model

B 2U(1)=Ξ(C (,U(1))00) \mathbf{B}^2 U(1) = \Xi( \; C^\infty(-,U(1)) \to 0 \to 0 \;)

and for the universal principal 2-bundle

EBU(1)=Ξ(C (,U(1))IdC (,U(1))0). \mathbf{E}\mathbf{B}U(1) = \Xi( \; C^\infty(-,U(1)) \stackrel{Id}{\to} C^\infty(-, U(1)) \to 0 \; ) \,.

In this notation we have also the constant presheaf

B 2U(1)=Ξ(constU(1)00). \mathbf{\flat} \mathbf{B}^2 U(1) = \Xi( \; const U(1) \to 0 \to 0 \; ) \,.

Above we already found the model

dRB 2U(1)=Ξ(0Ω 1()d dRΩ cl 2()). \mathbf{\flat}_{dR} \mathbf{B}^2 U(1) = \Xi(0 \to \Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2_{cl}(-)) \,.

In order to compute the differential cohomology H diff(,BU(1))\mathbf{H}_{diff}(-,\mathbf{B}U(1)) by an ordinary pullback in sSet we also want to resolve the curvature characteristic morphism BU(1) dRB 2U(1)\mathbf{B}U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^2 U(1) by a fibration. We claim that this may be obtained by choosing the resolution BU(1)BU(1) diff,chn\mathbf{B}U(1) \stackrel{\simeq}{\leftarrow} \mathbf{B} U(1)_{diff,chn} given by

BU(1) diff:=Ξ(C (,U(1))Ω 1()d dRIdΩ 1()) \mathbf{B}U(1)_{diff} := \Xi( \; C^\infty(-,U(1)) \oplus \Omega^1(-) \stackrel{d_{dR} \oplus Id}{\to} \Omega^1(-) \; )

with the morphism curv:B diffU(1) dRB 2U(1)curv : \mathbf{B}_{diff}U(1) \to \mathbf{\flat}_{dR}\mathbf{B}^2 U(1) given by

C (,U(1))Ω 1() d dR+Id Ω 1() p 2 d dR Ω 1() d dR Ω cl 2(). \array{ C^\infty(-,U(1)) \oplus \Omega^1(-) &\stackrel{d_{dR} + Id}{\to}& \Omega^1(-) \\ \downarrow^{\mathrlap{p_2}} && \downarrow^{\mathrlap{d_{dR}}} \\ \Omega^1(-) &\stackrel{d_{dR}}{\to}& \Omega^2_{cl}(-) } \,.

By the Poincare lemma applied to each Cartesian space, this is indeed a fibration.

In the next section we give the proof of this (simple) claim. Here in the warmup phase we instead want to discuss the geometric interpretation of this resolution, along the lines of the section curvature characteristics of 1-bundles in the survey-part.


We have the following geometric interpretation of the above models:

dRB 2U(1):U{U * Π 2(U) B 2U(1)}={Π 2(U)B 2U(1)} \mathbf{\flat}_{dR} \mathbf{B}^2 U(1) : U \mapsto \left\{ \array{ U &\to& * \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}^2 U(1) } \right\} = \left\{ \mathbf{\Pi}_2(U) \to \mathbf{B}^2 U(1) \right\}


BU(1) diff:U{U BU(1) Π 2(U) BINN(U(1))}. \mathbf{B}U(1)_{diff} : U \mapsto \left\{ \array{ U &\to& \mathbf{B}U(1) \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}INN(U(1)) } \right\} \,.

And in this presentation the morphism curv:B diffU(1)B 2U(1)curv : \mathbf{B}_{diff}U(1) \to \mathbf{B}^2 U(1) is given over UCartSpU \in CartSp by forming the pasting composite

U BU(1) underlyingcocycle Π 2(U) BINN(U(1)) connection Π 2(U) B 2U(1) curvature \array{ U &\to& \mathbf{B}U(1) &&& underlying\;cocycle \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}INN(U(1)) &&& connection \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}^2 U(1) &&& curvature }

and picking the lowest horizontal morphism.

Here the terms mean the following:

  • INN(U(1))INN(U(1)) is the 2-group Ξ(U(1)U(1))\Xi(U(1) \to U(1)), which is a groupal model for the universal U(1)-principal bundle EU(1)\mathbf{E}U(1);

  • Π 2(U)\mathbf{\Pi}_2(U) is the path 2-groupoid with homotopy class of 2-dimensional paths as 2-morphisms

  • the groupoids of diagrams in braces have as objects commuting diagrams in [CartSp op,sSet][CartSp^{op}, sSet] as indicated, and horizontal 2-morphisms fitting into such diagrams as morphisms.

Using the discussion at 2-groupoid of Lie 2-algebra valued forms (SchrWalII) we have the following:

  1. For XX a smooth manifold, morphisms in [CartSp op,2Grpd][CartSp^{op}, 2Grpd] of the form tra A:Π 2(X)EBU(1)tra_A : \Pi_2(X) \to \mathbf{E}\mathbf{B}U(1) are in bijection with smooth 1-forms AΩ 1(X)A \in \Omega^1(X): the 2-functor sends a path in XX to the the parallel transport of AA along that path, and sends a surface in XX to the exponentiated integral of the curvature 2-form F A=dAF_A = d A over that surface. The Bianchi identity dF A=0d F_A = 0 says precisely that this assignment indeed descends to homotopy classes of surfaces, which are the 2-morphisms in Π 2(X)\Pi_2(X).

  2. Moreover 2-morphisms of the form (λ,α):tra Atra A(\lambda,\alpha) : tra_A \to \tra_{A'} in [CartSp op,2Grpd][CartSp^{op}, 2Grpd] are in bijection with pairs consisting of a λC (X,U(1))\lambda \in C^\infty(X,U(1)) and a 1-form αΩ 1(X)\alpha \in \Omega^1(X) such that A=A+d dRλαA' = A + d_{dR} \lambda - \alpha.

  3. And finally 3-morphisms h:(λ,α)(λ,α)h : (\lambda, \alpha) \to (\lambda', \alpha') are in bijection with hC (X,U(1))h \in C^\infty(X,U(1)) such that λ=λh\lambda' = \lambda \cdot h and α=α+d dRh\alpha' = \alpha + d_{dR} h.

By the same reasoning we find that the coefficient object for flat B 2U(1)\mathbf{B}^2 U(1)-valued differential cohomology is

B 2U(1)=[Π 2(),B 2U(1)]=Ξ(C (,U(1))d dRΩ 1()d dRΩ cl 2()). \mathbf{\flat}\mathbf{B}^2 U(1) = [\Pi_2(-), \mathbf{B}^2U(1)] = \Xi( C^\infty(-,U(1)) \stackrel{d_{dR}}{\to} \Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2_{cl}(-) ) \,.

So by the above definition of differential cohomology in H\mathbf{H} we find that BU(1)\mathbf{B}U(1)-differential cohomology of a paracompact smooth manifold XX is given by choosing any good open cover {U iX}\{U_i \to X\}, taking C({U i})C(\{U_i\}) to be the Cech nerve, which is then a cofibrant replacement of XX in [CartSp op,sSet] proj,cov[CartSp^{op}, sSet]_{proj,cov} and forming the ordinary pullback

H diff(X,BU(1)) H dR 2(X) [CartSp op,sSet](C({U i}),B diffU(1)) curv [CartSp op,sSet](C({U i}), dRB 2U(1)) \array{ \mathbf{H}_{diff}(X,\mathbf{B}U(1)) &\to& H^2_{dR}(X) \\ \downarrow && \downarrow \\ [CartSp^{op},sSet](C(\{U_i\}), \mathbf{B}_{diff}U(1)) &\stackrel{curv}{\to}& [CartSp^{op},sSet](C(\{U_i\}), \flat_{dR}\mathbf{B}^2 U(1)) }

(because the bottom vertical morphism is a fibration, by the fact that our model for B diffU(1) dRB 2U(1)\mathbf{B}_{diff} U(1) \to \flat_{dR}\mathbf{B}^2 U(1) is a fibration, that C({U i})C(\{U_i\}) is cofibrant and using the axioms of the sSet-enriched model category [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj}).


A cocycle in [CartSp op,sSet](C({U i}),B diffU(1))[CartSp^{op},sSet](C(\{U_i\}), \mathbf{B}_{diff}U(1)) is

  1. a collection of functions

    (g ijC (U iU j,U(1))) (g_{i j } \in C^\infty(U_i \cap U_j, U(1)))

    satsifying g ijg jk=g ikg_{i j} g_{j k} = g_{i k} on U iU jU kU_i \cap U_j \cap U_k;

  2. a collection of 1-forms

    (A iΩ 1(U i)) (A_i \in \Omega^1(U_i))
  3. a collection of 1-forms

    (a ijΩ 1(U iU j)) (a_{i j} \in \Omega^1(U_i \cap U_j))

    such that

    A j=A i+d dRg ij+a ij A_j = A_i + d_{dR} g_{i j} + a_{i j}

    on U iU jU_i \cap U_j and

    a ij+a jk=a ik a_{i j} + a_{j k} = a_{i k}

    on U iU jU kU_i \cap U_j \cap U_k.

The curvature-morphism takes such a cocycle to the cocycle

(dA i,a ij,) (d A_i, a_{i j}, )

in the above model [CartSp op,sSet](C({U i}), dRB 2U(1))[CartSp^{op},sSet](C(\{U_i\}), \mathbf{\flat}_{dR}\mathbf{B}^2 U(1)) for intrinsic de Rham cohomology.

Every cocycle with nonvanishing (a ij)(a_{i j}) is in [C({U i}),B diffU(1)][C(\{U_i\}), \mathbf{B}_{diff}U(1)] coboundant to one with vanishing (a ij)(a_{i j})


The first statements are effectively the definition and the construction of the above models. The last statement is as in the above discussion of our model for ordinary de Rham cohomology: given a cocycle with non-vanishing closed a ija_{i j}, pick a partition of unity (ρ iC (X))(\rho_i \in C^\infty(X)) subordinate to the chosen cover and the coboundary given by ( i 0ρ i 0a i 0i)(\sum_{i_0} \rho_{i_0} a_{i_0 i}). This connects (A i,a ij,g ij)(A_i,a_{i j}, g_{i j}) with the cocycle (A i,a ij,g ij)(A'_i, a'_{i j}, g_{i j}) where

A i=A i+ i 0ρ i 0a i 0i A'_i = A_i + \sum_{i_0} \rho_{i_0} a_{i_0 i}


a ij =A jA idg ij =a ij+ i 0(a i 0ia i 0j) =0. \begin{aligned} a'_{i j} & = A'_j - A'_i - d g_{i j} \\ & = a_{i j} + - \sum_{i_0}( a_{i_0 i} - a_{i_0 j} ) \\ & = 0 \end{aligned} \,.

So in total we have found the following story:

  1. In order to compute the curvature characteristic form of a Cech cohomology cocycle g:C({U i})BU(1)g : C(\{U_i\}) \to \mathbf{B}U(1) of a U(1)U(1)-principal bundle, we first lift it

    B diffU(1) (g,) C({U i}) g BU(1) \array{ && \mathbf{B}_{diff}U(1) \\ & {}^{\mathllap{(g,\nabla)}}\nearrow & \downarrow \\ C(\{U_i\}) &\stackrel{g}{\to}& \mathbf{B}U(1) }

    to an equivalent B diffU(1)\mathbf{B}_{diff}U(1)-cocycle, and this amounts to putting (the Cech-representatitve of) a pseudo-connection on the U(1)U(1)-principal bundle.

  2. From that lift the desired curvature characteristic is simply projected out

    B diffU(1) curv dRB 2U(1) (g,) C({U i}) g BU(1), \array{ && \mathbf{B}_{diff}U(1) &\stackrel{curv}{\to}& \mathbf{\flat}_{dR}\mathbf{B}^2 U(1) \\ & {}^{\mathllap{(g,\nabla)}}\nearrow & \downarrow \\ C(\{U_i\}) &\stackrel{g}{\to}& \mathbf{B}U(1) } \,,

    and we find that it lives in the sheaf hypercohomology that models ordinary de Rham cohomology.

  3. Therefore we find that in each cohomology class of curvatures, there is at least one representative which is an ordinary globally defined 2-form. Moreover, the pseudo-connections that map to such a representative are precisely the genuine connections, those for which the (a ij)(a_{i j})-part of the cocycle vaishes.

So we see that ordinary connections on ordinary circle bundles are a means to model the homotopy pullback

H diff(X,BU(1)) H dR 2(X) H(X,BU(1)) H dR(X,BU(1)) \array{ \mathbf{H}_{diff}(X,\mathbf{B}U(1)) &\to& H_{dR}^2(X) \\ \downarrow && \downarrow \\ \mathbf{H}(X,\mathbf{B}U(1)) &\to& \mathbf{H}_{dR}(X,\mathbf{B}U(1)) }

in a 2-step process: first the choice of a pseudo-connection realizes the bottom horizontal morphism as an anafunctor, and then second the restriction imposed by forming the ordinary pullback chooses from all pseudo-connections precisely the genuine connections.

The general version of this story is discussed in detail at differential cohomology in an (∞,1)-topos – Local (pseudo-)connections.

Circle bundles with pseudo-connection

In the above discussion of extracting ordinary connections on ordinary U(1)U(1)-principal bundles from the abstract topos-theoretic definition of differential cohomology, we argued that a certain homotopy pullback may be computed by choosing in the Cech-hypercohomology of the complex of sheaves (Ω 1()d dRΩ cl 2())(\Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2_{cl}(-)) over a manifold XX those cohomology representatives that happen to be represented by globally defined 2-forms on XX. We saw that the homotopy fiber of pseudo-connections over these 2-forms happened to have connected components indexed by genuine connections.

But by the general abstract theory, up to isomorphism the differential cohomology computed this way is guaranteed to be independent of all such choices, which only help us to compute things.

To get a feeling for what is going on, it may therefore be useful to re-tell the analgous story with pseudo-connections that are not genuine connections.

By the very fact that BU(1)B diffU(1)\mathbf{B}U(1) \stackrel{\simeq}{\leftarrow} \mathbf{B}_{diff}U(1) is a weak equivalence, it follows that every pseudo-connection is equivalent to an ordinary connection as cocoycles in [CartSp op,sSet](C({U i}),B diff(G))[CartSp^{op}, sSet](C(\{U_i\}), \mathbf{B}_{diff}(G)).

If we choose a partition of unity (ρ iC (X,))(\rho_i \in C^\infty(X,\mathbb{R})) subordinate to the cover {U iX}\{U_i \to X\}, then we can construct the corresponding coboundary explicitly:

let (A ig ij,a ij)(A_i g_{ij}, a_{i j}) be an arbitrary pseudo-connection cocycle. Consider the Cech-hypercohomology coboundary given by ( i 0ρ i 0a i 0i,0)(\sum_{i_0} \rho_{i_0} a_{i_0 i}, 0). This lands in the shifted cocycle

(A i:=A i+ i 0ρ i 0a i 0i,g ij,a ij), (A'_i := A_i + \sum_{i_0} \rho_{i_0} a_{i_0 i}, g_{i j}, a'_{i j}) \,,

and we can find the new pseudo-components a ija'_{i j} by

a ij=A jA id dRg ij. a'_{i j} = A'_j - A'_i - d_{dR} g_{i j} \,.

Using the computation

i 0ρ i 0(a i 0ia i 0j = i 0ρ i 0(a ii 0+a i 0j = i 0ρ i 0a ij =a ij \begin{aligned} \sum_{i_0} \rho_{i_0} (a_{i_0 i} - a_{i_0 j} &= - \sum_{i_0} \rho_{i_0} (a_{i i_0} + a_{i_0 j} \\ & = \sum_{i_0} \rho_{i_0} a_{i j} \\ & = a_{i j} \end{aligned}

we find that these indeed vanish.

The most drastic example for this is a lift \nabla of a cocycle g=(g ij)g = (g_{i j}) in

B diffU(1) C({U i}) g BU(1) \array{ && \mathbf{B}_{diff} U(1) \\ & {}^{\mathllap{\nabla}}\nearrow & \downarrow \\ C(\{U_i\}) &\stackrel{g}{\to}& \mathbf{B}U(1) }

is one which takes all the ordinary curvature forms to vanish identically

=(A i:=0,g ij,a ij). \nabla = (A_i := 0, g_{i j}, a_{i j}) \,.

This fixes the pseudo-components to be a ij=dg ija_{i j} = - d g_{i j}. By the above discussion, this pseudo-connection with vanishing connection 1-forms is equivalent, as a pseudo-connection, to the ordinary connection cocycle with connection forms (A i:= i 0ρ i 0dg i 0i)(A_i := \sum_{i_0} \rho_{i_0} d g_{i_0 i}). This is a standard formula for equipping U(1)U(1)-principal bundles with Cech cocycle (g ij)(g_{i j}) with a connection.

U(1) 0U(1)_0-groupoid bundles

We saw above that the intrinsic coefficient object dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1) yields ordinary de Rham cohomology in degree n>1n \gt 1. For n=1n = 1 we have that dRBU(1)\mathbf{\flat}_{dR} \mathbf{B}U(1) is given simply by the 0-truncated sheaf of 1-forms, Ω 1():CartSp opSetsSet\Omega^1(-) : CartSp^{op} \to Set \hookrightarrow sSet. Accordingly we have for XX a paracompact smooth manifold

H(X, dRBU(1))=Ω cl 1(X) \mathbf{H}(X, \mathbf{\flat}_{dR}\mathbf{B}U(1)) = \Omega^1_{cl}(X)

instead of H dR 1(X)H^1_{dR}(X).

There is a good reason for this discrepancy: for n1n \geq 1 the object dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1) is the recipient of the intrinsic curvature characteristic morphism

curv B n1U(1):B n1U(1) dRB nU(1). curv_{\mathbf{B}^{n-1} U(1)} : \mathbf{B}^{n-1} U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^n U(1) \,.

For XB n1U(1)X \to \mathbf{B}^{n-1} U(1) a cocycle (an (n2)(n-2)-gerbe without connection), the cohomology class of the composite XB n1U(1) dRB nU(1)X \to \mathbf{B}^{n-1} U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^n U(1) is precisely the obstruction to the existence of a flat extension XB n1U(1)B n1U(1)X \to \mathbf{\flat} \mathbf{B}^{n-1} U(1) \to \mathbf{B}^{n-1} U(1) for the original cocycle.

For n=2n = 2 this is the usual curvature 2-form of a line bundle, for n=3n = 3 it is curvature 3-form of a bundle gerbe, etc. But for n=1n = 1 we have that the original cocycle is just a map of spaces

f:XU(1). f : X \to U(1) \,.

This may be understood as a cocycle for a groupoid principal bundle, for the 0-truncated groupoid with U(1)U(1) as its space of objects. Such a cocycle extends to a flat cocycle precisely if ff is constant as a function. The corresponding curvature 1-form is d dRfd_{dR} f and this is precisely the obstruction to constancy of ff already, in that ff is constant if and only if d dRfd_{dR} f vanishes. Not (necessarily) if it vanishes in de Rham cohomology .

This is the simplest example of a general statement about curvatures of higher bundles: the curvature 1-form is not subject to gauge transformations.

Circle nn-bundles with connection

We now generalize the above discussion on the derivation of the notion of connections on circle bundles from abstract topos-theory to a proof of the full theorem above on the derivation of general Deligne cohomology.

The main step is to model the double (∞,1)-pullback

B nU(1) * dRB n+1U(1) B n+1U(1) * B n+1U(1) \array{ \mathbf{B}^n U(1) &\to& * \\ \downarrow && \downarrow \\ \mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1) &\to& \mathbf{\flat} \mathbf{B}^{n+1} U(1) \\ \downarrow && \downarrow \\ * &\to& \mathbf{B}^{n+1} U(1) }

in H=\mathbf{H} = Smooth∞Grpd that gives the fiber sequence B nU(1)B nU(1) dRB n+1U(1)\mathbf{\flat} \mathbf{B}^n U(1) \to \mathbf{B}^{n} U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^{n+1} U(1) which controls the obstruction theory for flat connections by a homotopy pullback realized suitably as an ordinary pullback of fibrations in [CartSp op,Ch ]Ξ[CartSp op,sSet] proj[CartSp^{op}, Ch_\bullet] \stackrel{\Xi}{\hookrightarrow} [CartSp^{op}, sSet]_{proj}.


We have commuting diagrams

Ξ(0C (,U(1))Ω 1()d dRIdΩ 1()Ω 2()d dR+Idd dR±IdΩ n()) Ξ(C (,U(1))Id+d dRC (,U(1))Ω 1()d dRIdΩ 1()Ω 2()d dR+IdΩ n1()Ω n()d dR±IdΩ n()) (Id,p 2,p 2,,p 2,d dR) Ξ(0Ω 1()d dRΩ 2()d dRd dRΩ cl n+1()) (C (,U(1))d dRΩ 1()d dRΩ 2()d dRd dRΩ cl n+1()) Ξ(0000) Ξ(C (,U(1))000) \array{ \Xi(\; 0\stackrel{}{\to} {C^\infty(-,U(1)) \atop \oplus \Omega^1(-)} \stackrel{d_{dR} - Id}{\to} {\Omega^1(-) \atop \oplus \Omega^2(-)} \stackrel{d_{dR} + Id}{\to} \cdots \stackrel{d_{dR} \pm Id}{\to} \Omega^n(-) \;) &\to& \Xi(\; C^\infty(-,U(1)) \stackrel{Id + d_{dR}}{\to} {C^\infty(-,U(1)) \atop \oplus \Omega^1(-)} \stackrel{d_{dR} - Id}{\to} {\Omega^1(-) \atop \oplus \Omega^2(-)} \stackrel{d_{dR} + Id}{\to} \cdots { \Omega^{n-1}(-) \atop \oplus \Omega^n(-)} \stackrel{d_{dR} \pm Id}{\to} \Omega^n(-) \;) \\ \downarrow && \downarrow^{\mathrlap{(Id, p_2, p_2, \cdots, p_2,d_{dR})}} \\ \Xi( \; 0 \stackrel{}{\to} \Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega_{cl}^{n+1}(-)) &\to& (C^\infty(-,U(1)) \stackrel{d_{dR}}{\to} \Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega_{cl}^{n+1}(-) \;) \\ \downarrow && \downarrow \\ \Xi( \; 0 \to 0 \to 0 \to \cdots \to 0\;) &\to& \Xi( \; C^\infty(-,U(1)) \to 0 \to 0 \to \cdots \to 0 \;) }

in [CartSp op,sSet] proj[CartSp^{op},sSet]_{proj} where

  • the objects are fibrant models for the corresponding objects in the above (,1)(\infty,1)-pullback diagram;

  • the two right vertical morphisms are fibrations;

  • the two squares are pullback squares.

Therefore this is a homotopy pullback in [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj} that realizes the (,1)(\infty,1)-pullback in question in the (∞,1)-category of (∞,1)-presheaves PSh (,1)(CartSp)PSh_{(\infty,1)}(CartSp). Since ∞-stackification preserves finite (∞,1)-limits, it therefore also presents the above (,1)(\infty,1)-pullback in H=Sh (,1)(CartSp)\mathbf{H} = Sh_{(\infty,1)}(CartSp).


For the lower square we had discussed this already above. For the upper square the same type of reasoning applies. The main point is to find the chain complex in the top right such that it is a resolution of the point and maps by a fibration onto our model for B nU(1)\mathbf{\flat}\mathbf{B}^n U(1). The top right complex is

C (,U(1)) d dR Ω 1() d dR Ω 2() d dR d dR Ω n() id id id id id C (,U(1)) d dR Ω 1() d dR Ω 2() d dR d dR Ω n() \array{ && C^\infty(-,U(1)) &\stackrel{d_{dR}}{\to}& \Omega^1(-) &\stackrel{d_{dR}}{\to}& \Omega^2(-) &\stackrel{d_{dR}}{\to}& \cdots &\stackrel{d_{dR}}{\to}& \Omega^n(-) \\ &{}^{\mathrlap{id}}\nearrow& \oplus &{}^{\mathrlap{id}}\nearrow& \oplus &{}^{\mathrlap{id}}\nearrow& \oplus &{}^{\mathrlap{id}}\nearrow& \cdots & {}^{\mathrlap{id}}\nearrow& \\ C^\infty(-,U(1)) &\stackrel{d_{dR}}{\to}& \Omega^1(-) &\stackrel{d_{dR}}{\to}& \Omega^2(-) &\stackrel{d_{dR}}{\to}& \cdots &\stackrel{d_{dR}}{\to}& \Omega^n(-) }

and the vertical map out of it into C (,U(1))d dRΩ 1()d dRΩ n()d dRΩ cl n+1()C^\infty(-,U(1)) \stackrel{d_{dR}}{\to} \Omega^1(-) \stackrel{}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega^n(-) \stackrel{d_{dR}}{\to} \Omega^{n+1}_{cl}(-) is in positive degree the projection onto the lower row and in degree 0 the de Rham differential. This is manifestly surjective (by the Poincare lemma applied to each object UU \in CartSp) hence this is a fibration.

The pullback object in the top left is in this notation

B nU(1) diff,chn:=Ξ(C (,U(1)) d dR Ω 1() d dR d dR Ω n() id id id Ω 1() d dR d dR Ω n()) \mathbf{B}^n U(1)_{diff,chn} := \Xi \left( \array{ C^\infty(-,U(1)) &\stackrel{d_{dR}}{\to}& \Omega^1(-) &\stackrel{d_{dR}}{\to}& \cdots &\stackrel{d_{dR}}{\to}& \Omega^n(-) \\ \oplus &{}^{\mathrlap{id}}\nearrow& \oplus &{}^{\mathrlap{id}}\nearrow& \cdots & {}^{\mathrlap{id}}\nearrow& \\ \Omega^1(-) &\stackrel{d_{dR}}{\to}& \cdots &\stackrel{d_{dR}}{\to}& \Omega^n(-) } \right)

and in turn the top left vertical morphism curv:B diff nU(1) dRB n+1U(1)curv : \mathbf{B}_{diff}^n U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1) is in positive degree the projection on the lower row and in degree 0 the de Rham differential.

Notice that the evident forgetful morphism B nU(1)B diff nU(1)\mathbf{B}^n U(1) \stackrel{}{\leftarrow} \mathbf{B}^n_{diff} U(1) is indeed a weak equivalence.

With this description we now have the proof of the above theorem

Proof (equivalence of with Deligne cohomology)

Since the above model for curv:B diff nU(1) dRB n+1U(1)curv : \mathbf{B}_{diff}^n U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1) is a fibration and C({U i})C(\{U_i\}) is cofibrant, also

[Cartp op,sSet](C({U i}),B diff nU(1))[Cartp op,sSet](C({U i}), dRB nU(1)) [Cartp^{op}, sSet](C(\{U_i\}), \mathbf{B}^n_{diff}U(1)) \to [Cartp^{op}, sSet](C(\{U_i\}), \mathbf{\flat}_{dR}\mathbf{B}^n U(1))

is a Kan fibration by the fact that [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj} is an sSet QuillensSet_{Quillen}-enriched model category. Therefore the homotopy pullback is computed as an ordinary pullback.

By the above discussion of de Rham cohomology we have that we can assume the morphism H dR n+1(X)[CartSp op,sSet](C({U i}), dRB n+1)H_{dR}^{n+1}(X) \to [CartSp^{op}, sSet](C(\{U_i\}), \mathbf{\flat}_{dR}\mathbf{B}^{n+1}) picks only cocylces represented by globally defined closed differential forms FΩ n+1(X)F \in \Omega^{n+1}(X).

By the nature of the chain complexes apearing in the above proof, we see that the elements inm the fiber over such a globally defined form are precisely the cocycles with values only in the “upper row complex”

C (,U(1))d dRΩ 1()d dRd dRΩ n(). C^\infty(-,U(1)) \stackrel{d_{dR}}{\to} \Omega^1(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega^n(-) \,.

This is precisely the complex of sheaves that defines Deligne cohomology in degree (n+1)(n+1).

Models from \infty-Lie integration

In the previous section we discussed a model

B nU(1) diff,chn curv chn dRB n+1U(1) chn B nU(1) \array{ \mathbf{B}^n U(1)_{diff,chn} &\stackrel{curv_{chn}}{\to}& \mathbf{\flat}_{dR}\mathbf{B}^{n+1} U(1)_{chn} \\ \downarrow^{\mathrlap{\simeq}} \\ \mathbf{B}^n U(1) }

in [CartSp op,sSet][CartSp^{op}, sSet] for the canoncal curvature characteristic class curv:B nU(1) dRB n+1U(1)curv : \mathbf{B}^n U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1) in Smooth∞Grpd with the special property that it did model the abstract (∞,1)-topos-theoretic class under the Dold-Kan correspondence precisely in terms of the familiar Deligne cohomology coefficient complex.

There is another model for the curvature class in [CartSp op,sSet][CartSp^{op}, sSet], one that is useful for constructing the ∞-Chern-Weil homomorphism that maps from nonabelian cohomology in SmoothGrpdSmooth \infty Grpd to U(1)U(1)-valued differential cohomology. This second model is the one naturally adapted to the construction of the object B nU(1)\mathbf{B}^n U(1) by Lie integration from its ∞-Lie algebra b n1b^{n-1} \mathbb{R}. This is described at ∞-Lie groupoid – Lie integration.

For distinguishing the two models, we will indicate the former one by the subscript chn{}_{chn} and the one described now by the subscript simp{}_{simp}.


Here and in the following we adopt for differential forms on simplices the following notational convention:

  • by Ω (Δ n)\Omega^\bullet(\Delta^n) we denote the complex of smooth differential forms on the standard smooth nn-simplex with sitting instants: for every kk \in \mathbb{N} every kk-face of Δ n\Delta^n has a neighbourhood of its boundary such that the form restricted to that neighbourhood is constant in the direction perpendicular to that boundary.

  • for UCartSpU \in CartSp we write Ω (U×Δ k) vert\Omega^\bullet(U \times \Delta^k)_{vert} for the complex of vertical differential forms with respect to the trivial simplex bundle U×Δ kUU \times \Delta^k \to U.


For nn \in \mathbb{N}, define the simplicial presheaf B nU(1) simp[CartSp op,sSet]\mathbf{B}^n U(1)_{simp} \in [CartSp^{op}, sSet] by

B nU(1) simp:=cosk n+1((U,[k])Hom dgAlg(CE(b n1),Ω (U×Δ k) vert))/B n. \mathbf{B}^n U(1)_{simp} := \mathbf{cosk}_{n+1} ( (U, [k]) \mapsto Hom_{dgAlg}( CE(b^{n-1}\mathbb{R}), \Omega^\bullet(U \times \Delta^k)_{vert}) ) /\mathbf{B}^n \mathbb{Z} \,.

Here CE(b n1)CE(b^{n-1}\mathbb{R}) is the Chevalley-Eilenberg algebra of b n1b^{n-1}\mathbb{R}, which is simply the graded-commutative dg-algebra (over \mathbb{R}) on a single generator in degree nn with vanishing differential.

Moreover, cosk n+1()\mathbf{cosk}_{n+1}(-) is the coskeleton-operation and the quotient is by constant nn-forms ωΩ cl n(U×Δ k) vert\omega \in \Omega^n_{cl}(U \times \Delta^k)_{vert} such that Δ nω\int_{\Delta^n}\omega \in \mathbb{Z}. We take the quotient as a quotient of abelian simplicial groups (the group operation is the addition of differential forms).


Under the Dold-Kan correspondence the normalized chain complex of B nU(1) sim\mathbf{B}^n U(1)_{sim} is

N (B nU(1) simp)=(Ω cl n(()×Δ n+1) vert/ k(1) k k *Ω cl n(()×Δ n) vert/00), N_\bullet(\mathbf{B}^n U(1)_{simp}) = ( \cdots \to \Omega^n_{cl}((-)\times \Delta^{n+1})_{vert}/\sim \stackrel{\sum_k (-1)^k \partial_k^* }{\to} \Omega^n_{cl}((-)\times \Delta^{n})_{vert}/\sim \to 0 \to \cdots \to 0 ) \,,

where k:Δ nΔ n+1\partial_k : \Delta^n \to \Delta^{n+1} denotes the embedding of the kkth face of the smooth (n+1)(n+1)-simplex.

Here and in the following we indicate the homologically trivial part of the normalized chain complex of an (n+1)(n+1)-coskeletal simplicial abelian group just by ellipses.


The evident fiber integration of differential forms over simplices

Δ n:Ω (U×Δ n)Ω (U) \int_{\Delta^n} : \Omega^\bullet(U \times \Delta^n) \to \Omega^\bullet(U)

yields a morphism

Δ :B nU(1) simpB nU(1) \int_{\Delta^\bullet} : \mathbf{B}^n U(1)_{simp} \stackrel{\simeq}{\to} \mathbf{B}^n U(1)

in [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj}, which is a weak equivalence.

This is discussed at Lie integration.



B n simp\mathbf{\flat} \mathbf{B}^n \mathbb{R}_{simp} for the simplicial presheaf

B n simp:=cosk n+1((U,[k])Hom dgAlg(CE(b n1),Ω (U×Δ k))) \mathbf{\flat}\mathbf{B}^n \mathbb{R}_{simp} := \mathbf{cosk}_{n+1} ( (U,[k]) \mapsto Hom_{dgAlg}( CE(b^{n-1}\mathbb{R}), \Omega^{\bullet}(U \times\Delta^k) ) )

and write dRB n simp[CartSp op,sSet]\mathbf{\flat}_{dR}\mathbf{B}^n \mathbb{R}_{simp} \in [CartSp^{op}, sSet] for the simplicial presheaf

dRB n simp:=cosk n+1((U,[k])Hom dgAlg(CE(b n1),Ω 1,(U×Δ k))), \mathbf{\flat}_{dR}\mathbf{B}^n \mathbb{R}_{simp} := \mathbf{cosk}_{n+1} ( (U,[k]) \mapsto Hom_{dgAlg}( CE(b^{n-1}\mathbb{R}), \Omega^{\bullet \geq 1, \bullet}(U \times\Delta^k) ) ) \,,

where on the right we have the subcomplex of Ω (U×Δ k)\Omega^\bullet(U \times \Delta^k) on those forms that are non-vanishing on some vector field tangent to UU.

At ∞-Lie groupoid – Lie integrated ∞-groups – Differential coefficients the following is shown:


The evident fiber integration over simplices induces morphisms of simplicial presheaves

Δ :B nU(1) simpB nU(1) chn \int_{\Delta^\bullet} : \mathbf{\flat} \mathbf{B}^n U(1)_{simp} \to \mathbf{\flat} \mathbf{B}^n U(1)_{chn}


Δ : dRB nU(1) simp dRB nU(1) chn \int_{\Delta^\bullet} : \mathbf{\flat}_{dR} \mathbf{B}^n U(1)_{simp} \to \mathbf{\flat}_{dR} \mathbf{B}^n U(1)_{chn}

that are weak equivalences in [CartSp op,sSet] proj[CartSp^{op}, sSet]_{proj}.


Write B nU(1) diff,simp\mathbf{B}^n U(1)_{diff,simp} for the simplicial presheaf given by

B nU(1) diff,simp:=cosk n+1(U,[k]{Ω (U×Δ k) vert CE(b n1) Ω (U×Δ k) W(b n1)})/B n. \mathbf{B}^n U(1)_{diff,simp} := \mathbf{cosk}_{n+1} ( U,[k] \mapsto \left\{ \array{ \Omega^\bullet(U \times \Delta^k)_{vert} &\leftarrow& CE(b^{n-1}\mathbb{R}) \\ \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^k ) &\leftarrow& W(b^{n-1} \mathbb{R}) } \right\} ) / \mathbf{B}^n \mathbb{Z} \,.

Let the morphism

curv simp:B nU(1) diff,simp dRB n+1U(1) simp curv_{simp} : \mathbf{B}^n U(1)_{diff,simp} \to \mathbf{\flat}_{dR}\mathbf{B}^{n+1} U(1)_{simp}

be the one given by postcomposition with the square of dg-algebras

CE(b n1) 0 W(b n1) CE(b n) \array{ CE(b^{n-1}\mathbb{R}) &\leftarrow& 0 \\ \uparrow && \uparrow \\ W(b^{n-1}\mathbb{R}) &\leftarrow& CE(b^n \mathbb{R}) }

described at ∞-Lie algebra cohomology.


The set of square diagrams of dg-algebras above is over (U,[k])(U,[k]) the set of nn-forms ω\omega on U×Δ kU \times \Delta^k whose curvature (n+1)(n+1)-form dωd \omega has no component with all legs along Δ k\Delta^k.


The morphism given by fiber integration of differential forms over the simplex factor fits into a diagram

B nU(1) diff,simp curv simp dRB n+1U(1) simp Δ Δ B nU(1) diff,chn curv chn dRB n+1U(1) chn, \array{ \mathbf{B}^n U(1)_{diff,simp} &\stackrel{curv_{simp}}{\to}& \mathbf{\flat}_{dR}\mathbf{B}^{n+1}U(1)_{simp} \\ \downarrow^{\mathrlap{\int}_{\Delta^\bullet}} && \downarrow^{\mathrlap{\int}_{\Delta^\bullet}} \\ \mathbf{B}^n U(1)_{diff,chn} &\stackrel{curv_{chn}}{\to}& \mathbf{\flat}_{dR}\mathbf{B}^{n+1}U(1)_{chn} } \,,

where the vertical morphisms are weak equivalences.


Fiber integration induces a weak equivalence

Δ :B n diff,simpB n diff,chn \int_{\Delta^\bullet} : \mathbf{B}^n \mathbb{R}_{diff,simp} \stackrel{\simeq}{\to} \mathbf{B}^n \mathbb{R}_{diff, chn}

Observe that B n diff,simp\mathbf{B}^n \mathbb{R}_{diff,simp} is the pullback of dRB n+1 simpB n+1 simp\mathbf{\flat}_{dR} \mathbf{B}^{n+1}\mathbb{R}_{simp} \to \mathbf{\flat}\mathbf{B}^{n+1} \mathbb{R}_{simp} along the evident forgetful morphism from

(U,[k]){Ω (U×Δ k)W(b n1)}. (U,[k]) \mapsto \{\Omega^\bullet(U \times \Delta^k) \leftarrow W(b^{n-1} \mathbb{R})\} \,.

This forgetful morphism is evidently a fibration (because it is a degreewise surjection under Dold-Kan), hence this pullback models the homotopy fiber of dRB n+1B n+1\mathbf{\flat}_{dR} \mathbf{B}^{n+1} \mathbb{R} \to \mathbf{\flat} \mathbf{B}^{n+1} \mathbb{R}. Since by the above fiber integration gives a weak equivalence of pulback diagrams the claim follows.


Write B nU(1) conn,simpB nU(1) diff,simp\mathbf{B}^n U(1)_{conn,simp} \hookrightarrow \mathbf{B}^n U(1)_{diff,simp} for the sub-presheaf which over (U,[k])(U,[k]) is the set of those forms ω\omega on U×Δ kU \times \Delta^k such that the curvature dωd \omega has no leg along Δ k\Delta^k.


Under fiber integration over simplices, B nU(1) conn,simp\mathbf{B}^n U(1)_{conn,simp} is quasi-isomorphic to the Deligne cohomology-complex.

B nU(1) conn,simp Δ U(1)(n) D connection X^ B nU(1) diff,simp Δ B nU(1) diff,chn pseudoconnection dRB n+1U(1) simp Δ dRB n+1U(1) chn curvature. \array{ && \mathbf{B}^n U(1)_{conn,simp} &\stackrel{\int_{\Delta^\bullet}}{\to^\simeq}& U(1)(n)_D^\infty &&& connection \\ & \nearrow & \downarrow && \downarrow \\ \hat X &\to& \mathbf{B}^n U(1)_{diff,simp} &\stackrel{\int_{\Delta^\bullet}}{\to^\simeq}& \mathbf{B}^n U(1)_{diff,chn} &&& pseudo-connection \\ & \searrow & \downarrow && \downarrow \\ && \mathbf{\flat}_{dR}\mathbf{B}^{n+1}U(1)_{simp} &\stackrel{\int_{\Delta^\bullet}}{\to^\simeq}& \mathbf{\flat}_{dR}\mathbf{B}^{n+1}U(1)_{chn} &&& curvature } \,.

In summary this gives us the following alternative perspective on connections on B n1U(1)\mathbf{B}^{n-1}U(1)-principal ∞-bundles: such a connection is a cocycle with values in the B n\mathbf{B}^n \mathbb{Z}-quotient of the (n+1)(n+1)-coskeleton of the simplicial presheaf which over (U,[k])(U,[k]) is the set of diagrams of dg-algebras

C (U)Ω (Δ k) CE(b n1) underlyingcocycle Ω (U)Ω (Δ k) W(b n1) connection Ω (U)C (Δ k) CE(b n) curvature \array{ C^\infty(U)\otimes \Omega^\bullet(\Delta^k) &\leftarrow& CE(b^{n-1}\mathbb{R}) &&& underlying\;cocycle \\ \uparrow && \uparrow \\ \Omega^\bullet(U) \otimes \Omega^\bullet(\Delta^k) &\leftarrow& W(b^{n-1}\mathbb{R}) &&& connection \\ \uparrow && \uparrow \\ \Omega^\bullet(U)\otimes C^\infty(\Delta^k) &\leftarrow& CE(b^n \mathbb{R}) &&& curvature }

where the restriction to the top morphism is the underlying cocycle and the restriction to the bottom morphism the curvature form.

The generalization to such diagram cocycles from b n1b^{n-1}\mathbb{R} to general ∞-Lie algebras 𝔤\mathfrak{g} we discuss below in ∞-Lie algebra valued connections.

In homotopy type theory

We discuss the formulation of the above in the homotopy type theory-internal language of the (∞,1)-topos H=\mathbf{H} = Smooth∞Grpd.

Given the two functions

Ω cl (n+1) dRB n+1U(1) \Omega^{(n+1)}_{cl} \to \mathbf{\flat}_{dR}\mathbf{B}^{n+1} U(1)

(inclusion of the set of closed (n+1)(n+1)-forms into the (n+1)(n+1)-groupoid of de Rham cocycles)


curv:B nU(1) dRB n+1U(1) curv : \mathbf{B}^n U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^{n+1} U(1)

(the universal curvature class / Maurer-Cartan form of the circle (n1)(n-1)-group)

the smooth moduli ∞-stack of circle nn-bundles with connection from above is expressed in homotopy type theory as

B nU(1) conn{P:B nU(1)|FΩ cl n+1.curv(P)=F}. \mathbf{B}^n U(1)_{conn} \simeq \left\{ P : \mathbf{B}^n U(1) | \exists F \in \Omega^{n+1}_{cl} . curv(P) = F \right\} \,.

Spelled out this expresses B nU(1) conn\mathbf{B}^n U(1)_{conn} as

See the discussion at homotopy pullback for why this is indeed interpreted by the homotopy pullback B nU(1)× dRB n+1U(1)Ω cl n+1\mathbf{B}^n U(1) \times_{\mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1)} \Omega^{n+1}_{cl}.


For n=1n = 1 a circle nn-bundle with connection in the sense discussed here is indeed an ordinary hermitian line bundle or equivalently U(1)U(1)-principal bundle with connection.

For n=2n = 2 a circle 2-bundle with connection is equivalent to a bundle gerbe with connection (at least over a smooth manifold. Over an orbifold the definition given here does produce the correct equivariant cohomology, which is different from that of bundle gerbes that are equivariant in the ordinary sense.)

Classes of examples of higher circle bundles with connection are provided by ∞-Chern-Weil theory which provides homomorphisms of the form

H conn(X,BG)H diff(X,B nU(1)). \mathbf{H}_{conn}(X,\mathbf{B}G) \to \mathbf{H}_{diff}(X, \mathbf{B}^n U(1)) \,.

See for instance

for the class of circle 3-bundles that arise as differential refinements of degree 4 characteristic classes such as the Pontryagin class.



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


The above discussion is from

Discussion of Deligne cohomology as classifying highe bundle gerbes (bundle 2-gerbes, etc.) with connection:

Last revised on September 3, 2020 at 15:01:41. See the history of this page for a list of all contributions to it.