infinity-Chern-Weil theory introduction



Ordinary Chern-Weil theory studies connections on GG-principal bundles over a Lie group GG. In the context of the cohesive (∞,1)-topos Smooth∞Grpd of ∞-Lie groupoids these generalize to ∞-connections on principal ∞-bundles over ∞-Lie groups GG. Accordingly ∞-Chern-Weil theory deals with these higher connections and their relation to ordinary differential cohomology.

Here we describe some introdutcory basics of the general theory in concrete terms.

See ∞-Chern-Weil theory – motivation for some motivation.

Two simplifying special cases of general \infty-Chern-Weil theory are obtained by

  1. restricting attention to low categorical degree ,

    studying principal 1-bundles, principal 2-bundles and maybe 3-bundles; in terms of groupoids, 2-groupoids and maybe 3-groupoids;

  2. restricting attention to infinitesimal aspects

    studying not ∞-Lie groupoids but just their ∞-Lie algebroids. In terms of this it is easy to raise categorical degree to n=n = \infty, but this misses various global cohomological effects (very similar to how rational homotopy theory describes just non-torsion phenomena of genuine homotopy theory).

These are the special cases that this introduction concentrates on.

We start by describing

for low nn in detail, connecting them to standard theory, but presenting everything in such as way as to allow straightforward generalization to the full discussion of principal ∞-bundles.

Then in the same spirit we discuss

for low nn in a fashion that connects to the ordinary notion of parallel transport and points the way to the fully-fledged formulation in terms of the path ∞-groupoid functor.

This leads to differential-form expressions that we shall then finally reformulate in terms of

We end by indicating how under Lie integration this lifts to the full ∞-Chern-Weil theory.

Principal nn-bundles in low dimension

We assume here that the reader has a working knowledge of groupoids and at least a rough idea of 2-groupoids. We first use these notions to motivate some constructions, before discussing the formalization of ∞-groupoid in terms of Kan complexes.

Ordinary smooth principal bundles

Let GG be a Lie group and XX a smooth manifold (all our smooth manifolds are assumed to be finite dimensional and paracompact).

We give a discussion of smooth GG-principal bundles on XX in a manner that paves the way to a straightforward generalization to a description of principal ∞-bundles.

From the group GG we canonically obtain a groupoid that we write BG\mathbf{B}G and call the delooping groupoid of GG. Formally this groupoid is

BG=(G*) \mathbf{B}G = (G \stackrel{\to}{\to} *)

with composition induced from the product in GG. A useful cartoon of this groupoid is

BG={ g 1 = g 2 g 2g 1 } \mathbf{B}G = \left\{ \array{ && \bullet \\ & {}^{\mathllap{g_1}}\nearrow &=& \searrow^{\mathrlap{g_2}} \\ \bullet &&\stackrel{g_2 \cdot g_1 }{\to}&& \bullet } \right\}

where the g iGg_i \in G are elements in the group, and the bottom morphism is labeled by forming the product in the group. (The order of the factors here is a convention whose choice, once and for all, does not matter up to equivalence.)

But we get a bit more, even. Since GG is a Lie group, there is smooth structure on BG\mathbf{B}G that makes it a Lie groupoid, an internal groupoid in the category Diff of smooth manifolds: its collections of objects (trivially) and of morphisms each form a smooth manifold, and all structure maps (source, target, identity, composition) are smooth functions. We shall write

BGLieGrpd \mathbf{B}G \in LieGrpd

for BG\mathbf{B}G regarded as equipped with this smooth structure. Here and in the following the boldface is to indicate that we have an object equipped with a bit more structure – here: smooth structure – than present on the object denoted by the same symbols, but without the boldface. Eventually we will make this precise by having the boldface symbols denote objects in the (∞,1)-topos Smooth∞Grpd which are taken by forgetful functors to objects in ∞Grpd denoted by the corresponding non-boldface symbols.1

Also the smooth manifold XX may be regarded as a Lie groupoid – a groupoid with only identity morphisms. Its cartoon description is simply

X={xidx}. X = \{x \stackrel{id}{\to} x \} \,.

But there are other groupoids associated with XX:

Let {U iX} iI\{U_i \to X\}_{i \in I} be an open cover of XX. To this is canonically associated the Cech groupoid C({U i})C(\{U_i\}). Formally we may write this groupoid as

C({U i})=( i,jU iU jp 2p 1 iU i). C(\{U_i\}) = \left( \coprod_{i,j} U_i \cap U_j \stackrel{\overset{p_1}{\to}}{\underset{p_2}{\to}} \coprod_i U_i \right) \,.

A useful cartoon description of this groupoid is

C({U i})={ (x,j) = (x,i) (x,k)}. C(\{U_i\}) = \left\{ \array{ && (x,j) \\ & \nearrow &=& \searrow \\ (x,i) &&\to&& (x,k) } \right\} \,.

This indicates that the objects of this groupoid are pairs (x,i)(x,i) consisting of a point xXx \in X and a patch U iXU_i \subset X that contains xx, and a morphism is a triple (x,i,j)(x,i,j) consisting of a point and two patches, that both contain the point, in that xU iU jx \in U_i \cap U_j. The triangle in the above cartoon symbolizes the evident way in which these morphisms compose. All this inherits a smooth structure from the fact that the U iU_i are smooth manifolds and the inclusions U iXU_i \to X are smooth functions. hence also C(U)C(U) becomes a Lie groupoid.

There is a canonical functor

C({U i})X:(x,i)x. C(\{U_i\}) \to X \;\; :\;\; (x,i) \mapsto x \,.

This functor is an internal functor in Diff and moreover it is evidently essentially surjective and full and faithful.

However, while essential surjectivity and full-and-faithfulness implies that the underlying bare functor has a homotopy-inverse, that homotopy-inverse never has itself smooth component maps, unless XX itself is a Cartesian space and the chosen cover is trivial.

We do however want to think of C({U i})C(\{U_i\}) as being equivalent to XX even as a Lie groupoid. One says that a smooth functor whose underlying bare functor is an equivalence of groupoids is a weak equivalence of Lie groupoids, which we write as C({U i})XC(\{U_i\}) \stackrel{\simeq}{\to} X. Moreover, we shall think of C(U)C(U) as a good equivalent replacement of XX if it comes from a cover that is in fact a good open cover in that all its non-empty finite intersections U i 0i k:=U i 0U i kU_{i_0 \cdots i_k} := U_{i_0} \cap \cdots \cap U_{i_k} are diffeomorphic to the Cartesian space dimX\mathbb{R}^{dim X}.

We shall discuss later in which precise sense this condition makes C(U)C(U) good in the sense that smooth functors out of C(U)C(U) model the correct notion of morphism out of XX in the context of smooth groupoids (namely it will mean that C(U)C(U) is cofibrant in a suitable model category structure on the category of Lie groupoids). The formalization of this statement is what (∞,1)-topos theory is all about, to which we will come. For the moment we shall be content with accepting this as an ad hoc statement.

Observe that a functor

g:C(U)BG g : C(U) \to \mathbf{B}G

is given in components precisely by a collection of functions

{g ij:U ijG} i,jI \{g_{i j} : U_{i j} \to G \}_{i,j \in I}

such that on each U iU kU jU_i \cap U_k \cap U_j the equality g jkg ij=g ikg_{j k} g_{i j} = g_{i k} of smooth functions holds:

( (x,j) (x,i) (x,k))( g ij(x) g jk(x) g ik(x) ). \left( \array{ && (x,j) \\ & \nearrow && \searrow \\ (x,i) &&\to&& (x,k) } \right) \mapsto \left( \array{ && \bullet \\ & {}^{\mathllap{g_{i j}(x)}}\nearrow && \searrow^{\mathrlap{g_{j k}(x)}} \\ \bullet &&\stackrel{g_{i k}(x)}{\to}&& \bullet } \right) \,.

It is well known that such collections of functions characterize GG-principal bundles on XX. While this is a classical fact, we shall now describe a way to derive it that is true to the Lie-groupoid-context and that will make clear how smooth principal \infty-bundles work.

First observe that in total we have discussed so far spans of smooth functors of the form

C(U) g BG X. \array{ C(U) &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

Such spans of functors, whose left leg is a weak equivalence, are sometimes known, essentially equivalently, as Morita morphisms or generalized morphisms of Lie groupoids, as Hilsum-Skandalis morphisms or groupoid bibundles, or as anafunctors. We are to think of these as concrete models for more intrinsically defined direct morphisms XBGX\to \mathbf{B}G in the (,1)(\infty,1)-topos of \infty-Lie groupoids.

Now consider yet another Lie groupoid canonically associated with GG: we shall write EG\mathbf{E}G for the groupoid whose formal description is

EG=(G×Gp 1G) \mathbf{E}G = \left( G \times G \stackrel{\overset{\cdot}{\to}}{\underset{p_1}{\to}} G \right)

with the evident composition operation. The cartoon description of this groupoid is

EG={ g 2 g 2g 1 1 = g 3g 2 1 g 1 g 3g 1 1 g 3}, \mathbf{E}G = \left\{ \array{ && g_2 \\ & {}^{\mathllap{g_2 g_1^{-1}}}\nearrow &=& \searrow^{\mathrlap{g_3 g_2^{-1}}} \\ g_1 &&\stackrel{ g_3 g_1^{-1}}{\to}&& g_3 } \right\} \,,

This again inherits an evident smooth structure from the smooth structure of GG and hence becomes a Lie groupoid.

There is an evident forgetful functor

EGBG \mathbf{E}G \to \mathbf{B}G

which sends

(g 1g 2)(g 2 1g 1). (g_1 \to g_2) \mapsto (\bullet \stackrel{g_2^{-1} g_1}{\to} \bullet) \,.

Consider then the pullback diagram

P˜ EG C(U) g BG X \array{ \tilde P &\to& \mathbf{E}G \\ \downarrow && \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X }

in the category Grpd(Diff)Grpd(Diff). The object P˜\tilde P is the Lie groupoid whose cartoon description is

P˜={(x,i,g 1) (x,j,g 2=g ij(x)g 1)}, \array{ \tilde P = \left\{ \array{ (x,i,g_1) &&\stackrel{}{\to}&& (x,j,g_2 = g_{i j}(x) g_1 ) } \right\} } \,,

where there is a unique morphism as indicated, whenever the group labels match as indicated. Due to this uniqueness, this Lie groupoid is weakly equivalent to one that comes just from a manifold PP (it is 0-truncated)

P˜P. \tilde P \stackrel{\simeq}{\to} P \,.

This PP is traditionally written as

P=( iU i×G)/, P = \left( \coprod_{i} U_i \times G \right)/{\sim} \,,

where the equivalence relation is precisely that exhibited by the morphisms in P˜\tilde P. This is the traditional way to construct a GG-principal bundle from cocycle functions {g ij}\{g_{i j}\}. We may think of P˜\tilde P as being PP. It is a particular representative of PP in the (,1)(\infty,1)-topos of Lie groupoids.

While it is easy to see in components that the PP obtained this way does indeed have a principal GG-action on it, for later generalizations it is crucial that we can also recover this in a general abstract way. For notice that there is a canonical action

(EG)×GEG (\mathbf{E}G) \times G \to \mathbf{E}G

given by the action of GG on the space of objects, which are themselves identified with GG.

Then consider the pasting diagram of pullbacks

P˜×G EG×G P˜ EG C(U) g BG X. \array{ \tilde P \times G &\to& \mathbf{E}G \times G \\ \downarrow && \downarrow \\ \tilde P &\to& \mathbf{E}G \\ \downarrow && \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

The morphism P˜×GP˜\tilde P \times G \to \tilde P exhibits the principal GG-action of GG on P˜\tilde P.

In summary we find


For {U iX}\{U_i \to X\} a good open cover, there is an equivalence of categories

SmoothFunc(C({U i}),BG)GBund(X) SmoothFunc(C(\{U_i\}), \mathbf{B}G) \simeq G Bund(X)

between the functor category of smooth functors and smooth natural transformations, and the groupoid of smooth GG-principal bundles on XX.

It is no coincidence that this statement looks akin to the maybe more familiar statement which says that equivalence classes of GG-principal bundles are classified by homotopy-classes of morphisms of topological spaces

π 0Top(X,BG)π 0GBund(X), \pi_0 Top(X, \mathbf{B}G) \simeq \pi_0 G Bund(X) \,,

where BG\mathbf{B}G \in Top is the topological classifying space of GG. The category Top of topological spaces, regarded as an (∞,1)-category, is the archetypical (∞,1)-topos the way that Set is the archetypical topos. And it is equivalent to ∞Grpd, the (,1)(\infty,1)-category of bare ∞-groupoids. What we are seeing above is a first indication of how cohomology of bare \infty-groupoids is lifted to a richer (,1)(\infty,1)-topos to cohomology of \infty-groupoids with extra structure.

In fact, all of the statements that we have considered so far become conceptually simpler in the (,1)(\infty,1)-topos. We had already remarked that the anafunctor span XC(U)gBGX \stackrel{\simeq}{\leftarrow} C(U) \stackrel{g}{\to} \mathbf{B}G is really a model for what is simply a direct morphism XBGX \to \mathbf{B}G in the (,1)(\infty,1)-topos. But more is true: that pullback of EG\mathbf{E}G which we considered is just a model for the homotopy pullback of just the point

P˜×G EG×G P˜ EG C(U) g BG X inthemodelcategory P×G G P * X BG . . inthe(,1)topos. \array{ \vdots && \vdots \\ \tilde P \times G &\to& \mathbf{E}G \times G \\ \downarrow && \downarrow \\ \tilde P &\to& \mathbf{E}G \\ \downarrow && \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X \\ {} \\ {} \\ & in\;the\;model\;category & } \;\;\;\;\;\;\; \;\;\;\;\;\;\; \;\;\;\;\;\;\; \array{ \vdots && \vdots \\ P \times G &\to& G \\ \downarrow &\swArrow_{\simeq}& \downarrow \\ P &\to& * \\ \downarrow &\swArrow_{\simeq}& \downarrow \\ X &\stackrel{}{\to}& \mathbf{B}G \\ . \\ . \\ \\ \\ & in\;the\;(\infty,1)-topos } \,.

Cech cocycles

The discussion above of GG-principal bundles was all based on the Lie groupoids BG\mathbf{B}G and EG\mathbf{E}G that are canonically induced by a Lie group GG. We now discuss the case where GG is generalized to a Lie 2-group. The above discussion will go through essentially verbatim, only that we pick up 2-morphisms everywhere. This is the first step towards higher Chern-Weil theory. The resulting generalization of the notion of principal bundle is that of principal 2-bundle. For historical reasons these are known in the literature often as gerbes or as bundle gerbes.

Write U(1)=/U(1) = \mathbb{R}/\mathbb{Z} for the circle group. We have already seen above the groupoid BU(1)\mathbf{B}U(1) obtained from this. But since U(1)U(1) is an abelian group this groupoid has the special property that it still has itself the structure of an group object. This makes it what is called a 2-group. Accordingly, we may form its delooping once more to arrive at a Lie 2-groupoid B 2U(1)\mathbf{B}^2 U(1).

Its cartoon picture is

B 2U(1)={ Id g Id Id } \mathbf{B}^2 U(1) = \left\{ \array{ && \bullet \\ & {}^{\mathllap{Id}}\nearrow & \Downarrow^{\mathrlap{g}}& \searrow^{\mathrlap{Id}} \\ \bullet &&\underset{Id}{\to}&& \bullet } \right\}

for gU(1)g \in U(1). Both horizontal composition as well as vertical composition of the 2-morphisms is given by the product in U(1)U(1).

Let again XX be a smooth manifold with good open cover {U iX}\{U_i \to X\}. The corresponding Cech groupoid we may also think of as a Lie 2-groupoid,

C(U)=( i,j,kU iU jU k i,jU iU j iU i). C(U) = \left( \coprod_{i, j, k} U_i \cap U_j \cap U_k \stackrel{\to}{\stackrel{\to}{\to}} \coprod_{i, j} U_i \cap U_j \stackrel{\to}{\to} \coprod_i U_i \right) \,.

What we see here are the first stages of the full Cech nerve of the cover. Eventually we will be looking at this object in its entirety, since for all degrees this is always a good replacement of the manifold XX, as long as {U iX}\{U_i \to X\} is a good open cover.

So we look now at 2-anafunctors given by spans

C(U) g B 2U(1) X \array{ C(U) &\stackrel{g}{\to}& \mathbf{B}^2 U(1) \\ \downarrow^{\mathrlap{\simeq}} \\ X }

of internal 2-functors. These will model direct morphisms XB 2U(1)X \to \mathbf{B}^2 U(1) in the (,1)(\infty,1)-topos. It is straightforward to read off that the smooth 2-functor g:C(U)B 2U(1)g : C(U) \to \mathbf{B}^2 U(1) is given by the data of a 2-cocycle in the Cech cohomology of XX with coefficients in U(1)U(1). On 2-morphisms it specifies an assignment

g:( (x,j) (x,i) (x,k)){ Id g ijk(x) Id Id } g \;\; : \;\; \left( \array{ && (x,j) \\ & \nearrow &\Downarrow& \searrow \\ (x,i) &&\to&& (x,k) } \right) \;\;\; \mapsto \;\;\; \left\{ \array{ && \bullet \\ & {}^{\mathllap{Id}}\nearrow & \Downarrow^{\mathrlap{g_{i j k}(x)}}& \searrow^{\mathrlap{Id}} \\ \bullet &&\underset{Id}{\to}&& \bullet } \right\}

that is given by a collection of smooth functions

(g ijk:U iU jU kU(1)). (g_{i j k} : U_i \cap U_j \cap U_k \to U(1)) \,.

On 3-morphisms it gives a constraint on these functions, since there are only identity 3-morphisms in B 2U(1)\mathbf{B}^2 U(1):

((x,j) (x,k) (x,i) (x,l)(x,j) (x,k) (x,i) (x,l)) ( g ijk(x) g ikl(x) Id g ijl(x) g jkl(x) ). \begin{aligned} \left( \array{ (x,j) &&\stackrel{}{\to}&& (x,k) \\ \uparrow^{} &&{}^{}\nearrow&& \downarrow^{} \\ (x,i) &&\stackrel{}{\to}&& (x,l) } \;\;\;\; \Rightarrow \;\;\;\; \array{ (x,j) &&\stackrel{}{\to}&& (x,k) \\ \uparrow^{} &&\searrow^{}&& \downarrow^{} \\ (x,i) &&\stackrel{}{\to}&& (x,l) } \right) \\ & \mapsto \left( \array{ \bullet &&\stackrel{}{\to}&& \bullet \\ \uparrow^{} &\Downarrow^{g_{i j k}(x)} &{}^{}\nearrow&\Downarrow^{g_{i k l}(x)}& \downarrow^{} \\ \bullet &&\stackrel{}{\to}&& \bullet } \;\;\;\; \stackrel{Id}{\Rightarrow} \;\;\;\; \array{ \bullet &&\stackrel{}{\to}&& \bullet \\ \uparrow^{} &\Downarrow^{g_{i j l}(x)} &\searrow^{}&\Downarrow^{g_{j k l}(x)}& \downarrow^{} \\ \bullet &&\stackrel{}{\to}&& \bullet } \right) \end{aligned} \,.

This cocycle condition

g ijkg ikl=g ijlg jkl g_{i j k} \cdot g_{i k l} = g_{i j l} \cdot g_{j k l}

is that known from Cech cohomology.

In order to find the circle principal 2-bundle classified by such a cocycle by a pullback operation as before, we need to construct the 2-functor EBU(1)B 2U(1)\mathbf{E} \mathbf{B} U(1) \to \mathbf{B}^2 U(1) that exhibits the universal principal 2-bundle over U(1)U(1). The right choice for EBU(1)\mathbf{E B} U(1) – which we justify systematically in a moment – is indicated by

EBU(1):={ * c 1 g c 2 * c 3=gc 2c 1 } \mathbf{E B}U(1) := \left\{ \array{ && {*} \\ & {}^{\mathllap{c_1}}\nearrow &\Downarrow^{g}& \searrow^{\mathrlap{c_2}} \\ * &&\underset{c_3 = g c_2 c_1}{\to}&& } \right\}

for c 1,c 2,c 3,gU(1)c_1, c_2, c_3, g \in U(1), where all possible composition operations are given by forming the product of these labels in U(1)U(1). The projection EBU(1)B 2U(1)\mathbf{E B}U(1) \to \mathbf{B}^2 U(1) is the obvious one that simply forgets the labels c ic_i of the 1-morphisms and just remembers the labels gg of the 2-morphisms.

Let g:C(U)B 2U(1)g : C(U) \to \mathbf{B}^2 U(1) be a Cech cocycle as above. By the discussion of universal n-bundles we find the corresponding total space object as the pullback

P˜ EBU(1) C(U) g B 2U(1) X. \array{ \tilde P &\to& \mathbf{E}\mathbf{B}U(1) \\ \downarrow && \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}^2 U(1) \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

Unwinding what this means, we see that P˜\tilde P is the 2-groupoid whose objects are that of C(U)C(U), whose morphisms are finite sequences of morphisms in C(U)C(U), each equipped with a label cU(1)c \in U(1), and whose 2-morphisms are generated from those that look like

(x,j) c 1 g ijk(x) c 2 (x,i) c 3 (x,k) \array{ && (x,j) \\ & {}^{\mathllap{c_1}}\nearrow &\Downarrow^{g_{i j k}(x)}& \searrow^{\mathrlap{c_2}} \\ (x,i) &&\stackrel{c_3}{\to}&& (x,k) }

subject to the condition that

c 1c 2=c 3g ijk(x) c_1 \cdot c_2 = c_3 \cdot g_{i j k}(x)

in U(1)U(1). As before for principal 1-bundles PP, where we saw that the analogous pullback 1-groupoid P˜\tilde P was equivalent to the 0-groupoid PP, here we see that this 2-groupoid is equivalent to the 1-groupoid

P=(C(U) 1×U(1)C(U)) P = \left( C(U)_1 \times U(1) \stackrel{\to}{\to} C(U) \right)

with composition law

((x,i)c 1(x,j)c 2(x,k))=((x,i)(c 1c 2g ijk(x))(x,k)). ((x,i) \stackrel{c_1}{\to} (x,j) \stackrel{c_2}{\to} (x,k)) = ((x,i) \stackrel{(c_1 \cdot c_2 \cdot g_{i j k }(x))}{\to} (x,k)) \,.

This is a groupoid central extension

BU(1)PC(U)X. \mathbf{B}U(1) \to P \to C(U) \simeq X \,.

Centrally extended groupoids of this kind are known in the literature as bundle gerbes (over the surjective submersion Y=UXY = U \to X ). They may be thought of as given by a line bundle

L (C(U) 1=U× XU) (C(U) 0=U) X \array{ L \\ \downarrow \\ (C(U)_1 = U \times_X U) &\stackrel{\to}{\to}& (C(U)_0 = U) \\ && \downarrow \\ && X }

over the space C(U) 1C(U)_1 of morphisms, and a line bundle morphism

μ g:π 1 *Lπ 2 *Lπ 1 *L \mu_g : \pi_1^* L \otimes \pi_2^* L \to \pi_1^* L

that satisfies an evident associativity law, equivalent to the cocycle codition on gg.

So we see that bundle gerbes are presentations of Lie groupoids that are total spaces of BU(1)\mathbf{B}U(1)-principal 2-bundles.

This is clearly the beginning of a pattern. Next we can form one more delooping and produce the Lie 3-groupoid B 3U(1)\mathbf{B}^3 U(1). A cocycle C(U)B 3U(1)C(U) \to \mathbf{B}^3 U(1) classifies a circle 3-bundle . The total space object P˜\tilde P in the pullback

P˜ EB 2U(1) C(U) g B 3U(1) X \array{ \tilde P &\to& \mathbf{E}\mathbf{B}^2 U(1) \\ \downarrow && \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}^3 U(1) \\ \downarrow^{\mathrlap{\simeq}} \\ X }

is essentially what is known as a bundle 2-gerbe.

String 2-bundles and nonabelian bundle gerbes

Above we saw BU(1)\mathbf{B}U(1)-principal 2-bundles. The groupoid BU(1)\mathbf{B}U(1) is a special case of what is called a Lie 2-group, which is a group object GG in Lie groupoids.

An example of a nonabelian Lie 2-group is the string Lie 2-group StringString, which sits in a fiber sequence of Lie 2-groups of the form

BU(1)StringSpin. \mathbf{B}U(1) \to String \to Spin \,.

A quick way to understand the meaning of this 2-group is from the fact that:

Fact. Given a spin group-principal bundle PXP \to X, its Pontryagin class classifies a circle 3-bundle (a bundle 2-gerbe) called the Chern-Simons circle 3-bundle. The nontriviality of this is precisely the obstruction to lifting the SpinSpin-principal bundle PP to a StringString-principal 2-bundle.

Again, we can construct Lie 2-groupoids equivalent to the total space of a StringString-principal 2-bundle classified by a cocycle g:C(U)BStringg : C(U) \to \mathbf{B}String by forming the pullback.

P˜ EString C(U) g BString X \array{ \tilde P &\to& \mathbf{E}String \\ \downarrow && \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B} String \\ \downarrow^{\mathrlap{\simeq}} \\ X }

These groupoids P˜\tilde P are in the literature known as nonabelian bundle gerbe.

A model for principal \infty-bundles

We have seen above that the theory of ordinary smooth principal bundles is naturally situated within the context of Lie groupoids, and then that the theory of smooth principal 2-bundles is naturally situated within the theory of Lie 2-groupoids. This is clearly the beginning of a pattern in higher category theory where in the next step we see smooth 3-groupoids and so on. Finally the general theory of principal ∞-bundles deals with smooth ∞-groupoids.

A comprehensive discussion of such ∞-Lie groupoids is given there. In this introduction here we will just briefly describe the main tool for modelling these and describe principal \infty-bundles in this model. See also models for ∞-stack (∞,1)-toposes.

We first look at bare ∞-groupoids and then discuss how to equip these with smooth structure.

An ∞-groupoid is first of all supposed to be a structure that has k-morphisms for all kk \in \mathbb{N}, which for k1k \geq 1 go between (k1)(k-1)-morphisms. A useful tool for organizing such collections of morphisms is the notion of a simplicial set. This is a functor on the opposite category of the simplex category Δ\Delta, whose objects are the abstract cellular kk-simplices, denoted [k][k] or Δ[k]\Delta[k] for all kk \in \mathbb{N}, and whose morphisms Δ[k 1]Δ[k 2]\Delta[k_1] \to \Delta[k_2] are all ways of mapping these into each other. So we think of such a simplicial set given by a functor

K:Δ opSet K : \Delta^{op} \to Set

as specifying

  • a set [0]K 0[0] \mapsto K_0 of objects;

  • a set [1]K 1[1] \mapsto K_1 of morphism;

  • a set [2]K 2[2] \mapsto K_2 of 2-morphism;

  • a set [3]K 3[3] \mapsto K_3 of 3-morphism;

and generally

as well as specifying

  • functions ([n][n+1])(K n+1K n)([n] \hookrightarrow [n+1]) \mapsto (K_{n+1} \to K_n) that send n+1n+1-morphisms to their boundary nn-morphisms;

  • functions ([n+1][n])(K nK n+1)([n+1] \to [n]) \mapsto (K_{n} \to K_{n+1}) that send nn-morphisms to identity (n+1)(n+1)-morphisms on them.

The fact that KK is supposed to be a functor enforces that these assignments of sets and functions satisfy conditions that make consistent our interpretation of them as sets of kk-morphisms and source and target maps between these. These are called the simplicial identities.

But apart from this source-target matching, a generic simplicial set does not yet encode a notion of composition of these morphisms.

For instance for Λ 1[2]\Lambda^1[2] the simplicial set consisting of two attached 1-cells

Λ 1[2]={ 1 0 2} \Lambda^1[2] = \left\{ \array{ && 1 \\ & \nearrow && \searrow \\ 0 &&&& 2 } \right\}

and for (f,g):Λ 1[2]K(f,g) : \Lambda^1[2] \to K an image of this situation in KK, hence a pair x 0fx 1gx 2x_0 \stackrel{f}{\to} x_1 \stackrel{g}{\to} x_2 of two composable 1-morphisms in KK, we want to demand that there exists a third 1-morphisms in KK that may be thought of as the composition x 0hx 2x_0 \stackrel{h}{\to} x_2 of ff and gg. But since we are working in higher category theory (and not to be evil), we want to identify this composite only up to a 2-morphism equivalence

x 1 f g x 0 h x 2. \array{ && x_1 \\ & {}^{\mathllap{f}}\nearrow &\Downarrow^{\mathrlap{\simeq}}& \searrow^{\mathrlap{g}} \\ x_0 &&\stackrel{h}{\to}&& x_2 } \,.

From the picture it is clear that this is equivalent to demanding that for Λ 1[2]Δ[2]\Lambda^1[2] \hookrightarrow \Delta[2] the obvious inclusion of the two abstract composable 1-morphisms into the 2-simplex we have a diagram of morphisms of simplicial sets

Λ 1[2] (f,g) K h Δ[2]. \array{ \Lambda^1[2] &\stackrel{(f,g)}{\to}& K \\ \downarrow & \nearrow_{\mathrlap{\exists h}} \\ \Delta[2] } \,.

A simplicial set where for all such (f,g)(f,g) a corresponding such hh exists may be thought of as a collection of higher morphisms that is equipped with a notion of composition of adjacent 1-morphisms.

For the purpose of describing groupoidal composition, we now want that this composition operation has all inverses. For that purpose, notice that for

Λ 2[2]={ 1 0 2} \Lambda^2[2] = \left\{ \array{ && 1 \\ & && \searrow \\ 0 &&\to&& 2 } \right\}

the simplicial set consisting of two 1-morphisms that touch at their end, hence for

(g,h):Λ 2[2]K (g,h) : \Lambda^2[2] \to K

two such 1-morphisms in KK, then if gg had an inverse g 1g^{-1} we could use the above composition operation to compose that with hh and thereby find a morphism ff connecting the sources of hh and gg. This being the case is evidently equivalent to the existence of diagrams of morphisms of simplicial sets of the form

Λ 2[2] (g,h) K f Δ[2]. \array{ \Lambda^2[2] &\stackrel{(g,h)}{\to}& K \\ \downarrow & \nearrow_{\mathrlap{\exists f}} \\ \Delta[2] } \,.

Demanding that all such diagrams exist is therefore demanding that we have on 1-morphisms a composition operation with inverses in KK.

In order for this to qualify as an \infty-groupoid, this composition operation needs to satisfy an associativity law up to coherent 2-morphisms, which means that we can find the relevant tetrahedras in KK. These in turn need to be connected by pentagonators and ever so on. It is a nontrivial but true and powerful fact, that all these coherence conditions are captured by generalizing the above conditions to all dimensions in the evident way:

let Λ i[n]Δ[n]\Lambda^i[n] \hookrightarrow \Delta[n] be the simplicial set – called the iith nn-horn – that consists of all cells of the nn-simplex Δ[n]\Delta[n] except the interior nn-morphism and the iith (n1)(n-1)-morphism.

Then a simplicial set is called a Kan complex, if for all images f:Λ i[n]Kf : \Lambda^i[n] \to K of such horns in KK, the missing two cells can be found in KK- in that we can always find a horn filler σ\sigma in the diagram

Λ i[n] f K σ Δ[n]. \array{ \Lambda^i[n] &\stackrel{f}{\to}& K \\ \downarrow & \nearrow_{\mathrlap{\sigma}} \\ \Delta[n] } \,.

The basic example is the nerve N(C)sSetN(C) \in sSet of an ordinary groupoid CC, which is the simplicial set with N(C) kN(C)_k being the set of sequences of kk composable morphisms in CC. The nerve operation is a full and faithful functor from 1-groupoids into Kan complexes and hence may be thought of as embedding 1-groupoids in the context of general ∞-groupoids.

But we need a bit more than just bare ∞-groupoids. In generalization to Lie groupoids, we need ∞-Lie groupoids. A useful way to encode that an \infty-groupoid has extra structure modeled on geometric test objects that themselves form a category CC is to remember the rule which for each test space UU in CC produces the \infty-groupoid of UU-parameterized families of kk-morphisms in KK. For instance for an ∞-Lie groupoid we could test with each Cartesian space U= nU = \mathbb{R}^n and find the \infty-groupoids K(U)K(U) of smooth nn-parameter families of kk-morphisms in KK.

This data of UU-families arranges itself into a presheaf with values in Kan complexes

K:C opKanCplxsSet K : C^{op} \to KanCplx \hookrightarrow sSet

hence with values in simplicial sets. This is equivalently a simplicial presheaf of sets. The functor category [C op,sSet][C^{op}, sSet] on the opposite category of the category of test objects CC serves as a model for the (∞,1)-category of \infty-groupoids with CC-structure.

While there are no higher morphisms in this functor 1-category that could for instance witness that two \infty-groupoids are not isomorphic, but still equivalent, it turns out that all one needs in order to reconstruct all these higher morphisms (up to equivalence!) is just the information of which morphisms of simplicial presheaves would become invertible if we were keeping track of higher morphism. These would-be invertible morphisms are called weak equivalences and denoted K 1K 2K_1 \stackrel{\simeq}{\to} K_2.

For common choices of CC there is a well-understood way to define the weak equivalences Wmor[C op,sSet]W \subset mor [C^{op}, sSet], and equipped with this information the category of simplicial presheaves becomes a category with weak equivalences . There is a well-developed but somewhat intricate theory of how exactly this 1-cagtegorical data models the full higher category of structured groupoids that we are after, but for our purposes we essentially only need to work inside the category of fibrant objects of a model category structure on simplicial presheaves, which in practice amounts to the fact that we use the following three basic constructions:

  1. ∞-anafunctors – A morphisms XYX \to Y between \infty-groupoids with CC-structure is not just a morphism XYX\to Y in [C op,sSet][C^{op}, sSet], but is a span of such ordinary morphisms

    X^ Y X \array{ \hat X &\to& Y \\ \downarrow^{\mathrlap{\simeq}} \\ X }

    where the left leg is a weak equivalence. This is sometimes called an \infty-anafunctor from XX to YY.

  2. homotopy pullback – For ABpCA \to B \stackrel{p}{\leftarrow} C a diagram, the (∞,1)-pullback of it is the ordinary pullback in [C op,sSet][C^{op}, sSet] of a replacement diagram ABp^C^A \to B \stackrel{\hat p}{\leftarrow} \hat C, where p^\hat p is a good replacement of pp in the sense of the following factorization lemma.

  3. factorization lemma – For p:CBp : C \to B a morphism in [C op,sSet][C^{op}, sSet], a good replacement p^:C^B\hat p : \hat C \to B is given by the composite vertical morphism in the ordinary pullback diagram

    C^ C p B Δ[1] B B, \array{ \hat C &\to& C \\ \downarrow && \downarrow^{\mathrlap{p}} \\ B^{\Delta[1]} &\to& B \\ \downarrow \\ B } \,,

where B Δ[1]B^{\Delta[1]} is the path object of BB: the simplicial presheaf that is over each UCU \in C the simplicial path space B(U) Δ[1]B(U)^{\Delta[1]}.

The principal ∞-bundles that we wish to model are already the main and simplest example of the application of these three items:

Consider an object BG[C op,sSet]\mathbf{B}G \in [C^{op}, sSet] which is an \infty-groupoid with a single object, so that we may think of it as the delooping of an ∞-group GG, let ** be the point and *BG* \to \mathbf{B}G the unique inclusion map. The good replacement of this inclusion morphism is the GG-universal principal ∞-bundle EGBG\mathbf{E}G \to \mathbf{B}G given by the pullback diagram

EG * BG Δ[1] BG BG \array{ \mathbf{E}G &\to& * \\ \downarrow && \downarrow \\ \mathbf{B}G^{\Delta[1]} &\to& \mathbf{B}G \\ \downarrow \\ \mathbf{B}G }

An ∞-anafunctor XX^BGX \stackrel{\simeq}{\leftarrow} \hat X \to \mathbf{B}G we call a cocycle on XX with coefficients in GG, and the (∞,1)-pullback PP of the point along this cocycle, which by the above discussion is the ordinary limit

P EG * BG I BG X^ g BG X \array{ P &\to& \mathbf{E}G &\to& * \\ \downarrow && \downarrow && \downarrow \\ && \mathbf{B}G^I &\to& \mathbf{B}G \\ \downarrow && \downarrow \\ \hat X &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X }

we call the principal ∞-bundle PXP \to X classified by the cocycle.

It is now evident that our discussion of ordinary smooth principal bundles above is the special case of this for BG\mathbf{B}G the nerve of the one-object groupoid associated with the ordinary Lie group GG.

So we find the complete generalization of the situation that we already indicated there, which is summarized in the following diagram:

P˜×G EG×G P˜ EG C(U) g BG X inthemodelcategory P×G G P * X BG . . inthe(,1)topos. \array{ \vdots && \vdots \\ \tilde P \times G &\to& \mathbf{E}G \times G \\ \downarrow && \downarrow \\ \tilde P &\to& \mathbf{E}G \\ \downarrow && \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X \\ {} \\ {} \\ & in\;the\;model\;category & } \;\;\;\;\;\;\; \;\;\;\;\;\;\; \;\;\;\;\;\;\; \array{ \vdots && \vdots \\ P \times G &\to& G \\ \downarrow && \downarrow \\ P &\to& * \\ \downarrow &\swArrow_{\simeq}& \downarrow \\ X &\stackrel{}{\to}& \mathbf{B}G \\ . \\ . \\ \\ \\ & in\;the\;(\infty,1)-topos } \,.

Parallel transport in low dimensions

With a decent handle on principal \infty-bundles as described above we now turn to the description of connections on ∞-bundles. It will turn out that the above cocycle-description of GG-principal \infty-bundles in terms of ∞-anafunctors XX^gBGX \stackrel{\simeq}{\leftarrow} \hat X \stackrel{g}{\to} \mathbf{B}G has, under mild conditions, a natural generalization where BG\mathbf{B}G is replaced by a non-concrete simplicial presheaf BG conn\mathbf{B}G_{conn} which we may think of as the ∞-groupoid of ∞-Lie algebra valued forms. This comes with a canonical map BG connBG\mathbf{B}G_{conn} \to \mathbf{B}G and an \infty-connection \nabla on the \infty-bundle classified by gg is a lift \nabla of gg in the disgram

BG conn X^ g BG X. \array{ && \mathbf{B}G_{conn} \\ & {}^{\mathllap{\nabla}}\nearrow & \downarrow \\ \hat X &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

In the language of ∞-stacks we may think of BG\mathbf{B}G as the \infty-stack (on CartSp) or \infty-prestack (on Diff) GTrivBund()G TrivBund(-) of trivial GG-principal bundles, and of BG conn\mathbf{B}G_{conn} correspondingly as the object GTrivBund ()G TrivBund_{\nabla}(- ) of trivial GG-principal bundles with (non-trivial) connection. In this sense the statement that \infty-connections are cocycles with coefficients in some BG conn\mathbf{B}G_{conn} is a tautology. The real questions are:

  1. What is BG conn\mathbf{B}G_{conn} in concrete formulas?

  2. Why are these formulas what they are? What is the general abstract concept of an \infty-connection? What are its defining abstract properties?

A comprehensive answer to the second question is provided by the general abstract concept of differential cohomology in a cohesive topos. Here in this introduction we will not go into the full abstract theory, but using classical tools we get pretty close. What we describe is a generalization of the concept of parallel transport to higher parallel transport. As we shall see, this is naturally expressed in terms of ∞-anafunctors out of path n-groupoids. This reflects how the full abstract theory arises in the context of an ∞-connected (∞,1)-topos that comes canonically with a notion of fundamental ∞-groupoid in a locally ∞-connected (∞,1)-topos.

Below we begin the discussion of \infty-connections by reviewing the classical theory of connection on a bundle in a way that will make its generalization to higher connections relatively straightforward:

In an analogous way we can then describe certain classes of connections on a 2-bundle – subsuming the notion of connection on a bundle gerbe – in

With that in hand we then revisit the discussion of connections on ordinary bundles. By associating to each bundle with connection its corresponding curvature 2-bundle with connection we obtain a more refined description of connections on bundles, one that is naturally adapted to the construction of curvature characteristic forms in the Chern-Weil homomorphism:

This turns out to be the kind of formulation of connections on an ∞-bundle that drops out of the general abstract theory described at ∞-Chern-Weil homomorphism. In classical terms, its full formulation involves the description of circle n-bundles with connection in terms of Deligne cohomology and the description of the ∞-groupoid of ∞-Lie algebra valued forms in terms of dg-algebra homomorphisms. The first aspect we discuss in

the second in

The combination of these two aspects yields naturally an explicit model for the Chern-Weil homomorphism and its generalization to higher bundles:

Taken together, these constructions allow us to express a good deal of the general \infty-Chern-Weil theory with classical tools. As an example, we describe how the classical Cech-Deligne cocycle construction of the refined Chern-Weil homomorphism (by (BrylinskiMacLaughlin)) drops out from these constructions:

Connections on a principal bundle

There are different equivalent definitions of the classical notion of a connection. One that is useful for our purposes is that a connection \nabla on a GG-principal bundle PXP \to X is a rule tra tra_\nabla for parallel transport along paths: a rule that assigns to each path γ:[0,1]X\gamma : [0,1] \to X a morphism tra (γ):P xP ytra_\nabla(\gamma) : P_x \to P_y between the fibers of the bundle above the endpoints of these paths, in a compatible way:

P x tra (γ) P y tra (γ) P z P x γ y γ z X. \array{ P_x &\stackrel{tra_\nabla(\gamma)}{\to}& P_y &\stackrel{tra_\nabla(\gamma')}{\to}& P_z &&& P \\ && && &&& \downarrow \\ x &\stackrel{\gamma}{\to}& y &\stackrel{\gamma'}{\to}& z &&& X } \,.

In order to formalize this, we introduce a (diffeological) Lie groupoid to be called the path groupoid of XX. (Constructions and results in this section are from ([SWI]).


For XX a smooth manifold let [I,X][I,X] be the set of smooth functions I=[0,1]XI = [0,1] \to X. For UU a Cartesian space, we say that a UU-parameterized smooth family of points in [I,X][I,X] is a smooth map U×IXU \times I \to X. (This makes [I,X][I,X] a diffeological space).

Say a path γ[I,X]\gamma \in [I,X] has sitting instants if it is constant in a neighbourhood of the boundary I\partial I. Let [I,P] si[I,P][I,P]_{si} \subset [I,P] be the subset of paths with sitting instants.

Let [I,X] si[I,X] si th[I,X]_{si} \to [I,X]_{si}^{th} be the projection to the set of equivalence classes where two paths are regarded as equivalent if they are cobounded by a smooth thin homotopy.

Say a UU-parameterized smooth family of points in [I,X] si th[I,X]_{si}^{th} is one that comes from a UU-family of representatives in [I,X] si[I,X]_{si} under this projection. (This makes also [I,X] si th[I,X]_{si}^{th} a diffeological space.)


The passage to the subset and quotient [I,X] si th[I,X]_{si}^{th} of the set of all smooth paths in the above definition is essentially the minimal adjustment to enforce that the concatenation of smooth paths at their endpoints defines the composition operation in a groupoid.


The path groupoid P 1(X)\mathbf{P}_1(X) is the groupoid

P 1(X)=([I,X] si thX) \mathbf{P}_1(X) = ([I,X]_{si}^{th} \stackrel{\to}{\to} X)

with source and target maps given by endpoint evaluation and composition given by concatenation of classes [γ][\gamma] of paths along any orientation preserving diffeomorphism [0,1][0,2][0,1] 1,0[0,1][0,1] \to [0,2] \simeq [0,1] \coprod_{1,0} [0,1] of any of their representatives

[γ 2][γ 1]:[0,1][0,1] 1,0[0,1](γ 2,γ 1)X. [\gamma_2] \circ [\gamma_1] : [0,1] \stackrel{\simeq}{\to} [0,1] \coprod_{1,0} [0,1] \stackrel{(\gamma_2 , \gamma_1)}{\to} X \,.

This becomes an internal groupoid in diffeological spaces with the above UU-families of smooth paths. We regard it as a groupoid-valued presheaf, an object in [CartSp op,Grpd][CartSp^{op}, Grpd]:

P 1(X):U(Diff(U×I,X) si thDiff(U,X)). \mathbf{P}_1(X) : U \mapsto (Diff(U \times I, X)_{si}^{th} \stackrel{\to}{\to} Diff(U,X) ) \,.

Observe now that for GG a Lie group and BG\mathbf{B}G its delooping Lie groupoid discussed above, a smooth functor tra:P 1(X)BGtra : \mathbf{P}_1(X) \to \mathbf{B}G sends each (thin-homotopy class of a) path to an element of the group GG

tra:(x[γ]y)(tra(γ)G) tra : (x \stackrel{[\gamma]}{\to} y) \mapsto ( \bullet \stackrel{tra(\gamma) \in G}{\to} \bullet )

such that composite paths map to products of group elements

tra:( y [γ] = [γ] x [γ][γ] z)( tra(γ) = tra(γ) tra(γ)tra(γ) ) tra : \left( \array{ && y \\ & {}^{\mathllap{[\gamma]}}\nearrow &=& \searrow^{\mathrlap{[\gamma']}} \\ x &&\stackrel{[\gamma']\circ [\gamma]}{\to}&& z } \right) \mapsto \left( \array{ && \bullet \\ & {}^{\mathllap{tra(\gamma)}}\nearrow &=& \searrow^{\mathrlap{tra(\gamma')}} \\ \bullet &&\stackrel{tra(\gamma)tra(\gamma')}{\to}&& \bullet } \right)

and such that UU-families of smooth paths induce smooth maps UGU \to G of elements.

There is a classical construction that yields such an assignment: the parallel transport of a Lie-algebra valued 1-form.


Suppose AΩ 1(X,𝔤)A \in \Omega^1(X, \mathfrak{g}) is a degree-1 differential form on XX with values in the Lie algebra 𝔤\mathfrak{g} of GG. Then its parallel transport is the smooth functor

tra A:P 1(X)BG tra_A : \mathbf{P}_1(X) \to \mathbf{B}G

given by

[γ]Pexp( [0,1]γ *A)G, [\gamma] \mapsto P \exp(\int_{[0,1]} \gamma^* A) \; \in G \,,

where the group element on the right is defined to be the value at 1 of the unique solution f:[0,1]Gf : [0,1] \to G of the differential equation

d dRf+γ *Af=0 d_{dR} f + \gamma^*A \wedge f = 0

for the boundary condition f(0)=ef(0) = e.


This construction Atra AA \mapsto tra_A induces an equivalence of categories

[CartSp op,Grpd](P 1(X),BG)BG conn(X), [CartSp^{op},Grpd](\mathbf{P}_1(X), \mathbf{B}G) \simeq \mathbf{B}G_{conn}(X) \,,

where on the left we have the hom-groupoid of groupoid-valued presheaves and where on the right we have the groupoid of Lie-algebra valued 1-forms whose

  • objects are 1-forms AΩ 1(X,𝔤)A \in \Omega^1(X,\mathfrak{g}),

  • morphisms g:A 1A 2g : A_1 \to A_2 are labeled by smooth functions gC (X,G)g \in C^\infty(X,G) such that A 2=g 1Ag+g 1dgA_2 = g^{-1} A g + g^{-1}d g.

This equivalence is natural in XX, so that we obtain another smooth groupoid.


Define BG conn:CartSp opGrpd\mathbf{B}G_{conn} : CartSp^{op} \to Grpd to be the (generalized) Lie groupoid

BG conn:U[CartSp op,Grpd](P 1(),BG) \mathbf{B}G_{conn} : U \mapsto [CartSp^{op}, Grpd](\mathbf{P}_1(-), \mathbf{B}G)

whose UU-parameterized smooth families of groupoids form the groupoid of Lie-algebra valued 1-forms on UU.


This equivalence in particular subsumes the classical facts that parallel transport γPexp( [0,1]γ *A)\gamma \mapsto P \exp(\int_{[0,1]} \gamma^* A)

  • is invariant under orientation preserving reparameterizations of paths;

  • sends reversed paths to inverses of group elements.


There is an evident natural smooth functor XP 1(X)X \to \mathbf{P}_1(X) that includes points in XX as constant paths. This induces a natural morphism BG connBG\mathbf{B}G_{conn} \to \mathbf{B}G that forgets the 1-forms.


Let PXP \to X be a GG-principal bundle that corresponds to a cocycle g:C(U)BGg : C(U) \to \mathbf{B}G under the construction discussed above. Then a connection \nabla on PP is a lift \nabla of the cocycle through BG connBG\mathbf{B}G_{conn} \to \mathbf{B}G.

BG conn C(U) g BG. \array{ && \mathbf{B}G_{conn} \\ & {}^{\mathllap{\nabla}}\nearrow & \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G } \,.

This is equivalent to the traditional definitions.

A morphism :C(U)BG conn\nabla : C(U) \to \mathbf{B}G_{conn} is

  • on each U iU_i a 1-form A iΩ 1(U i,𝔤)A_i \in \Omega^1(U_i, \mathfrak{g});

  • on each U iU jU_i \cap U_j a function g ijC (U iU j,G)g_{i j} \in C^\infty(U_i \cap U_j , G);

such that

  • on each U iU jU_i \cap U_j we have A j=g ij 1(A+d dR)g ijA_j = g_{i j}^{-1}( A + d_{dR} )g_{i j};

  • on each U iU jU kU_i \cap U_j \cap U_k we have g ijg jk=g ikg_{i j} \cdot g_{j k} = g_{i k}.


Let [I,X] si th[I,X] h[I,X]_{si}^{th} \to [I,X]^h the projection onto the full quotient by smooth homotopy classes of paths.

Write Π 1(X)=([I,X] hX)\mathbf{\Pi}_1(X) = ([I,X]^h \stackrel{\to}{\to} X) for the smooth groupoid defined as P 1(X)\mathbf{P}_1(X), but where instead of thin homotopies, all homotopies are divided out.


The above restricts to a natural equivalence

[CartSp op,Grpd](Π 1(X),BG)BG, [CartSp^{op}, Grpd](\mathbf{\Pi}_1(X), \mathbf{B}G) \simeq \mathbf{\flat}\mathbf{B}G \,,

where on the left we have the hom-groupoid of groupoid-valued presheaves, and on the right we have the full sub-groupoid BGBG conn\mathbf{\flat}\mathbf{B}G \subset \mathbf{B}G_{conn} on those 𝔤\mathfrak{g}-valued differential forms whose curvature 2-form F A=d dRA+[AA]F_A = d_{dR} A + [A \wedge A] vanishes.

A connection \nabla is flat precisely if it factors through the inclusion BGBG conn\flat \mathbf{B}G \to \mathbf{B}G_{conn}.

For the purposes of Chern-Weil theory we want a good way to extract the curvature 2-form in a general abstract way from a cocycle :XC(U)BG conn\nabla : X \stackrel{\simeq}{\leftarrow }C(U) \to \mathbf{B}G_{conn}. In order to do that, we first need to discuss connections on 2-bundles.

Connections on principal 2-bundles

There is an evident higher dimensional generalization of the definition of connections on 1-bundles in terms of functors out of the path groupoid discussed above. This we discuss now. We will see that, however, the obvious generalization captures not quite all 2-connections. But we will also see a way to recode 1-connections in terms of flat 2-connections. And that recoding then is the right general abstract perspective on connections, which generalizes to principal ∞-bundles and in fact which in the full theory follows from first principles.

(Constructions and results in this section are from SWII, SWIII)


The path 2-groupoid P 2(X)\mathbf{P}_2(X) is the smooth strict 2-groupoid analogous to P 1(X)\mathbf{P}_1(X), but with nontrivial 2-morphisms given by thin homotopy-classes of disks Δ Diff 2X\Delta^2_{Diff} \to X with sitting instants.

In analogy to the projection P 1(X)Π 1(X)\mathbf{P}_1(X) \to \mathbf{\Pi}_1(X) there is a projection to P 2(X)Π 2(X)\mathbf{P}_2(X) \to \mathbf{\Pi}_2(X) to the 2-groupoid obtained by dividing out full homotopy of disks, relative boundary.


Let GG be a strict Lie 2-group coming from a crossed module ([G 2δG 1],α:G 1Aut(G 2))([G_2 \stackrel{\delta}{\to} G_1], \alpha : G_1 \to Aut(G_2)).Its delooping BG\mathbf{B}G is the strict Lie 2-groupoid coming from the crossed complex [G 2δG 1*][G_2 \stackrel{\delta}{\to} G_1 \stackrel{\to}{\to} *].

BG={ g 1 k g 2 δ(k)g 1g 2 g 1,g 2G 1,kG 2}. \mathbf{B}G = \left\{ \array{ && \bullet \\ & {}^{\mathllap{g_1}}\nearrow & \Downarrow^{\mathrlap{k}}& \searrow^{\mathrlap{g_2}} \\ \bullet &&\underset{\delta(k) g_1 g_2 }{\to}&& \bullet } \;\; | \;\; g_1, g_2 \in G_1, k \in G_2 \right\} \,.

This induces a differential crossed module (𝔤 2δ *𝔤 1)(\mathfrak{g}_2 \stackrel{\delta_*}{\to} \mathfrak{g}_1), the Lie 2-algebra of GG.


For KK an abelian Lie group then BK\mathbf{B}K is the delooping 2-group coming from the crossed module [K1][K \to 1] and BBK\mathbf{B}\mathbf{B}K is the 2-group coming from the complex [K11][K \to 1 \to 1].

A smooth 2-functor Π 2(X)BG\mathbf{\Pi}_2(X) \to \mathbf{B}G now assigns information also to surfaces

( y γ 1 Σ γ 2 x z)( y tra(γ 1) tra(Σ) tra(γ 2) x z) \left( \array{ && y \\ & {}^{\mathllap{\gamma_1}}\nearrow &\Downarrow^{\mathrlap{\Sigma}}& \searrow^{\mathrlap{\gamma_2}} \\ x &&\underset{}{\to}&& z } \right) \mapsto \left( \array{ && y \\ & {}^{\mathllap{tra(\gamma_1)}}\nearrow &\Downarrow^{\mathrlap{tra(\Sigma)}}& \searrow^{\mathrlap{tra(\gamma_2)}} \\ x &&\to&& z } \right)

and thus encodes a higher parallel transport.


There is a natural equivalence of 2-groupoids

[CartSp op,2Grpd](Π 2(X),BG)BG [CartSp^{op}, 2Grpd](\mathbf{\Pi}_2(X), \mathbf{B}G) \simeq \mathbf{\flat} \mathbf{B}G

where on the right we have the 2-groupoid of Lie 2-algebra valued forms whose

  • objects are pairs AΩ 1(X,𝔤 1)A \in \Omega^1(X,\mathfrak{g}_1), BΩ 2(X,𝔤 2)B \in \Omega^2(X,\mathfrak{g}_2) such that the 2-form curvature

    F 2(A,B):=d dRA+[AA]+δ *B F_2(A,B) := d_{dR} A + [A \wedge A] + \delta_* B

    and the 3-form curvature

    F 3(A,B):=d dRB+[AB] F_3(A,B) := d_{dR} B + [A \wedge B]


  • morphisms (λ,a):(A,B)(A,B)(\lambda,a) : (A,B) \to (A',B') are pairs aΩ 1(X,𝔤 2)a \in \Omega^1(X,\mathfrak{g}_2), λC (X,G 1)\lambda \in C^\infty(X,G_1) such that A=λAλ 1+λdλ 1+δ *aA' = \lambda A \lambda^{-1} + \lambda d \lambda^{-1} + \delta_* a and B=λ(B)+d dRa+[Aa]B' = \lambda(B) + d_{dR} a + [A\wedge a]

  • 2-morphisms are… (exercise).

As before, this is natural in XX, so that we that we get a presheaf of 2-groupoids

BG:U[CartSp op,2Grpd](Π 2(U),BG). \mathbf{\flat}\mathbf{B}G : U \mapsto [CartSp^{op}, 2Grpd](\mathbf{\Pi}_2(U), \mathbf{B}G) \,.

If in the above definition we use P 2(X)\mathbf{P}_2(X) instead of Π 2(X)\mathbf{\Pi}_2(X), we obtain the same 2-groupoid, except that the 3-form curvature F 3(A,B)F_3(A,B) is not required to vanish.


Let PXP \to X be a GG-principal 2-bundle classified by a cocycle C(U)BGC(U) \to \mathbf{B}G. Then a structure of a flat connection on a 2-bundle \nabla on it is a lift

BG flat C(U) g BG. \array{ && \mathbf{\flat}\mathbf{B}G \\ & {}^{\mathllap{\nabla_{flat}}}\nearrow & \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G } \,.

For G=BAG = \mathbf{B}A, a connection on a 2-bundle (not necessarily flat) is a lift

[P 2(),BBA] C(U) g BBA. \array{ && [\mathbf{P}_2(-),\mathbf{B}\mathbf{B}A] \\ & {}^{\mathllap{\nabla}}\nearrow & \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}\mathbf{B}A } \,.

We do not state the last definition for general Lie 2-groups GG. The reason is that for general GG 2-anafunctors out of P 2(X)\mathbf{P}_2(X) do not produce the fully general notion of 2-connections that we are after, but yield a special case in between flatness and non-flatness: the case where precisely the 2-form curvature-components vanish, while the 3-form curvature part is unrestricted. This case is important in itself and discussed in detail below.

Only for GG of the form BA\mathbf{B}A does the 2-form curvature necessarily vanish anyway, so that in this case the definition by morphisms out of P 2(X)\mathbf{P}_2(X) happens to already coincide with the proper general one. This serves in the following theorem as an illustration for the toolset that we are exposing, but for the purposes of introducing the full notion of \infty-Chern-Weil theory we will rather focus on flat 2-connections, and then show in Curvature characteristics of 1-bundles how using these one does arrive at a functorial definition of 1-connections that does generalize to the fully general definition of \infty-connections.


Let {U iX}\{U_i \to X\} be a good open cover, a cocycle C(U)[P 2(),B 2A]C(U) \to [\mathbf{P}_2(-), \mathbf{B}^2 A] is a cocycle in Cech cohomology-Deligne cohomology in degree 3.

Moreover, we have a natural equivalence of bicategories

[CartSp op,2Grpd](C(U),[P 2(),B 2U(1)])U(1)Gerb (X), [CartSp^{op}, 2Grpd](C(U), [\mathbf{P}_2(-), \mathbf{B}^2 U(1)]) \simeq U(1) Gerb_\nabla(X) \,,

where on the right we have the bicategory of U(1)U(1)-bundle gerbes with connection.

In particular the equivalence classes of cocycles form the degree-3 ordinary differential cohomology of XX:

H diff 3(X,)π 0([C(U),[P 2(),B 2U(1)]]). H^3_{diff}(X, \mathbb{Z}) \simeq \pi_0( [C(U), [\mathbf{P}_2(-), \mathbf{B}^2 U(1)]]) \,.

A cocycle as above naturally corresponds to a 2-anafunctor

Q B 2U(1) P 2(X). \array{ Q &\to& \mathbf{B}^2 U(1) \\ \downarrow^{\mathrlap{\simeq}} \\ \mathbf{P}_2(X) } \,.

The value of this on 2-morphisms in P 2(X)\mathbf{P}_2(X) is the higher parallel transport of the connection on the 2-bundle.

This appears for instance in the action functional of the sigma model that describes strings charged under a Kalb-Ramond field.

The following example of a flat nonabelian 2-bundle is very degenerate as far as 2-bundles go, but does contain in it the seed of a full understanding of connections on 1-bundles.


For GG a Lie group, its inner automorphism 2-group INN(G)INN(G) is as a groupoid the universal G-bundle EG\mathbf{E}G, but regarded as a 2-group with the group structure coming from the crossed module [GIdG][G \stackrel{Id}{\to} G].

The cartoon presentation of the delooping 2-groupoid BINN(G)\mathbf{B}INN(G) is

BINN(G)={ g 1 k g 2 g 3=g 1g 2k ,g 1,g 2,kG}. \mathbf{B}INN(G) = \left\{ \array{ && \bullet \\ & {}^{\mathllap{g_1}}\nearrow & \Downarrow^{\mathrlap{k}} & \searrow^{\mathrlap{g_2}} \\ \bullet &&\underset{g_3 = g_1 g_2 k}{\to}&& \bullet } \;\; \,, \;\; g_1, g_2, k \in G \right\} \,.

This is the Lie 2-group whose Lie 2-algebra inn(𝔤)inn(\mathfrak{g}) is the one whose Chevalley-Eilenberg algebra is the Weil algebra of 𝔤\mathfrak{g}.


By the above theorem we have that there is a bijection of sets

{Π 2(X)BINN(G)}Ω 1(X,𝔤) \{\mathbf{\Pi}_2(X) \to \mathbf{B} INN(G)\} \simeq \Omega^1(X, \mathfrak{g})

of flat INN(G)INN(G)-valued 2-connections and Lie-algebra valued 1-forms. Under the identifications of this theorem this identification works as follows:

  • the 1-form component of the 2-connection is AA;

  • the vanishing of the 2-form component of the 2-curvature F 2(A,B)=F A+BF_2(A,B) = F_A + B identifies the 2-form component of the 2-connection with the curvature 2-form, B=F AB = - F_A;

  • the vanishing of the 3-form component of the 2-curvature F 3(A,B)=dB+[AB]=d A+[AF A]F_3(A,B) = d B + [A \wedge B] = d_A + [A \wedge F_A] is the Bianchi identity satisfied by any curvature 2-form.

This means that 2-connections with values in INN(G)INN(G) actually model 1-connections and keep track of their curvatures. Using this we see in the next section a general abstract definition of connections on 1-bundles that naturally support the Chern-Weil homomorphism.

Curvature characteristics of 1-bundles

We now describe connections on 1-bundles in terms of their flat curvature 2-bundles . This gives a general abstract notion of connections that generalizes to connections on ∞-bundles and that supports naturally the Chern-Weil homomorphism

Throughout this section GG is a Lie group, BG\mathbf{B}G its delooping 2-groupoid and INN(G)INN(G) its inner automorphism 2-group and BINN(G)\mathbf{B}INN(G) the corresponding delooping Lie 2-groupoid.


Define the smooth groupoid BG diff[CartSp op,Grpd]\mathbf{B}G_{diff} \in [CartSp^{op}, Grpd] as the pullback

BG diff=BG× BINN(G)BINN(G). \mathbf{B}G_{diff} = \mathbf{B}G \times_{\mathbf{B}INN(G)} \mathbf{\flat} \mathbf{B}INN(G) \,.

This is the groupoid-valued presheaf which assigns to UCartSpU \in CartSp the groupoid whose objects are commuting diagrams

U BG Π 2(U) BINN(G), \array{ U &\to& \mathbf{B}G \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}INN(G) } \,,

where the vertical morphisms are the canonical inclusions discussed above, and whose morphisms are compatible pairs of natural transformations

U BG Π 2(U) BINN(G) \array{ U &{{\nearrow \searrow} \atop {\to}}& \mathbf{B}G \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &{{\nearrow \searrow} \atop {\to}}& \mathbf{B} INN(G) }

of the horizontal morphisms.


By the above theorems, we have over any UU \in CartSp that

  • an object in BG diff(U)\mathbf{B}G_{diff}(U) is a 1-form AΩ 1(U,𝔤)A \in \Omega^1(U,\mathfrak{g});

  • a morphism A 1(g,a)A 2A_1 \stackrel{(g,a)}{\to} A_2 is labeled by a function gC (U,G)g \in C^\infty(U,G) and a 1-form aΩ 1(U,𝔤)a \in \Omega^1(U,\mathfrak{g}) such that

    A 2=g 1A 1g+g 1dg+a. A_2 = g^{-1}A_1 g + g^{-1}d g + a \,.

    Notice that this can always be uniquely solved for aa, so that the genuine information in this morphism is just the data given by gg.

  • there are no nontrivial 2-morphisms, even though BINN(G)\mathbf{B}INN(G) is a 2-groupoid: since BG\mathbf{B}G is just a 1-groupoid this is enforced by the commutativity of the above diagram.

From this it is clear that


The projection BG diffBG\mathbf{B}G_{diff} \stackrel{\simeq}{\to} \mathbf{B}G is a weak equivalence.

So BG diff\mathbf{B}G_{diff} is a resolution of BG\mathbf{B}G. We will see that it is the resoluton that supports 2-anafunctors out of BG\mathbf{B}G which represent curvature characteristic classes.


For XC(U)BU(1)X \stackrel{\simeq}{\leftarrow}C(U) \to \mathbf{B}U(1) a cocycle for a U(1)U(1)-principal bundle PXP \to X, we call a lift ps\nabla_{ps} in

BG diff ps C(U) g BG \array{ && \mathbf{B}G_{diff} \\ & {}^{\mathllap{\nabla_{ps}}}\nearrow & \downarrow \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G }

a pseudo-connection on PP.

Pseudo-connections in themselves are not very interesting. But notice that every ordinary connection is in particular a pseudo-connection and we have an inclusion morphism of smooth groupoids

BG connBG diff. \mathbf{B}G_{conn} \hookrightarrow \mathbf{B}G_{diff} \,.

This inclusion plays a central role in the theory. The point is that while BG diff\mathbf{B}G_{diff} is such a boring extenion of BG\mathbf{B}G that it is actually equivalent to BG\mathbf{B}G, there is no inclusion of BG conn\mathbf{B}G_{conn} into BG\mathbf{B}G, but there is into BG diff\mathbf{B}G_{diff}. This is the kind of situation that resolutions are needed for.

It is useful to look at some details for the case that GG is an abelian group such as the circle group U(1)U(1).

In this abelian case the 2-groupoids BU(1)\mathbf{B}U(1), B 2U(1)\mathbf{B}^2 U(1), BINN(U(1))\mathbf{B}INN(U(1)), etc., that so far we noticed are given by crossed complexes are actually given by ordinary chain complexes: we write

Ξ:Ch +sAbKanCplx \Xi : Ch_\bullet^+ \to sAb \to KanCplx

for the Dold-Kan correspondence map that identifies chain complexes with simplicial abelian group and then considers their underlying Kan complexes. Using this map we have the following identifications of our 2-groupoid valued presheaves with complexes of group-valued sheaves

BU(1)=Ξ[C (,U(1))0] \mathbf{B}U(1) = \Xi[C^\infty(-,U(1)) \to 0]
B 2U(1)=Ξ[C (,U(1))00] \mathbf{B}^2 U(1) = \Xi[C^\infty(-,U(1)) \to 0 \to 0]
BINNU(1)=Ξ[C (,U(1))IdC (,U(1))0]. \mathbf{B} INN U(1) = \Xi[C^\infty(-,U(1)) \stackrel{Id}{\to} C^\infty(-,U(1)) \to 0] \,.

For G=AG = A an abelian group, in particular the circle group, there is a canonical morphism BINN(U(1))BBU(1)\mathbf{B} INN(U(1)) \to \mathbf{B}\mathbf{B}U(1).

On the level of chain complexes this is the evident chain map

[C (,U(1)) Id C (,U(1)) 0] [C (,U(1)) 0 0]. \array{ [C^\infty(-,U(1)) &\stackrel{Id}{\to}& C^\infty(-,U(1)) &\to& 0] \\ \downarrow && \downarrow && \downarrow \\ [C^\infty(-,U(1)) &\to& 0 &\to& 0] } \,.

On the level of 2-groupoids this is the map that forgets the labels on the 1-morphisms

{ g 1 k g 2 kg 2g 1 }{ Id k Id Id }. \left\{ \array{ && \bullet \\ & {}^{\mathllap{g_1}}\nearrow & \Downarrow^{\mathrlap{k}}& \searrow^{\mathrlap{g_2}} \\ \bullet &&\stackrel{k g_2 g_1}{\to}&& } \right\} \;\; \mapsto \;\; \left\{ \array{ && \bullet \\ & {}^{\mathllap{Id}}\nearrow & \Downarrow^{\mathrlap{k}}& \searrow^{\mathrlap{Id}} \\ \bullet &&\stackrel{Id}{\to}&& \bullet } \right\} \,.

In terms of this map INN(U(1))INN(U(1)) serves to interpolate between the single and the double delooping of U(1)U(1). In fact the sequence of 2-functors

BU(1)BINN(U(1))B 2U(1) \mathbf{B}U(1) \to \mathbf{B}INN(U(1)) \to \mathbf{B}^2 U(1)

is a model for the BU(1)\mathbf{B}U(1)-universal principal 2-bundle

BU(1)EBU(1)B 2U(1). \mathbf{B}U(1) \to \mathbf{E} \mathbf{B}U(1) \to \mathbf{B}^2 U(1) \,.

This happens to be an exact sequence of 2-groupoids. Abstractly, what really matters is rather that it is a fiber sequence, meaning that it is exact in the correct sense inside the (∞,1)-category Smooth∞Grpd. For our purposes it is however relevant that this particular model is also exact in the ordinary sense in that we have a commuting diagram

BU(1) * BINN(U(1)) B 2U(1) \array{ \mathbf{B}U(1) &\to& * \\ \downarrow && \downarrow \\ \mathbf{B}INN(U(1)) &\to& \mathbf{B}^2 U(1) }

which is a pullback diagram, exhibitng BU(1)\mathbf{B}U(1) as the kernel of BINN(U(1))B 2U(1)\mathbf{B}INN(U(1)) \to \mathbf{B}^2 U(1).

We shall be interested in the pasting composite of this diagram with the one defining BG diff\mathbf{B}G_{diff} over a domain UU:

U BU(1) * Π 2(U) BINN(U(1)) B 2U(1), \array{ U &\to& \mathbf{B}U(1) &\to& * \\ \downarrow && \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}INN(U(1)) &\to& \mathbf{B}^2 U(1) } \,,

The total outer diagram appearing this way is a component of the following (generalized) Lie 2-groupoid.



dRB 2U(1):=*× B 2U(1)B 2U(1). \mathbf{\flat}_{dR} \mathbf{B}^2U(1) := * \times_{\mathbf{B}^2 U(1)} \mathbf{\flat} \mathbf{B}^2 U(1) \,.

Over any UCartSpU \in CartSp this is the 2-groupoid whose objects are sets of diagrams

U * Π 2(U) B 2U(1). \array{ U &\to& * \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}^2 U(1) } \,.

This are equivalently just morphisms Π 2(U)B 2U(1)\mathbf{\Pi}_2(U) \to \mathbf{B}^2 U(1), which by the above theorems we may identify with closed 2-forms BΩ cl 2(U)B \in \Omega^2_{cl}(U).

The morphisms B 1B 2B_1 \to B_2 in dRB 2U(1)\mathbf{\flat}_{dR} \mathbf{B}^2 U(1) over UU are compatible pseudonatural transformations of the horizontal morphisms

U * Π 2(U) BINN(G), \array{ U &{{\nearrow \searrow} \atop {\to}}& {*} \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &{{\nearrow \searrow} \atop {\to}}& \mathbf{B} INN(G) } \,,

which means that they are pseudonatural transformations of the bottom morphism whose components over the points of UU vanish. These identify with 1-forms λΩ 1(U)\lambda \in \Omega^1(U) such that B 2=B 1+d dRλB_2 = B_1 + d_{dR} \lambda.

Finally the 2-morphisms would be modifications of these, but the commutativity of the above diagram constrains these to be trivial.

In summary this shows that


Under the Dold-Kan correspondence dRB 2U(1)\mathbf{\flat}_{dR} \mathbf{B}^2 U(1) is the sheaf of truncated de Rham complexes

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

Equivalence classes of 2-anafunctors

X dRB 2U(1) X \to \mathbf{\flat}_{dR} \mathbf{B}^2 U(1)

are canonically in bijection with the degree 2 de Rham cohomology of XX.


Notice that – while every globally defined closed 2-form BΩ cl 2(X)B \in \Omega^2_{cl}(X) defines such a 2-anafunctor – not every such 2-anafunctor comes from a globally defined closed 2-form. Some of them assign closed 2-forms B iB_i to patches U 1U_1, that differ by differentials B jB i=d dRλ ijB_j - B_i = d_{dR} \lambda_{i j} of 1-forms λ ij\lambda_{i j} on double overlaps, which themselves satisfy on triple intersections the cocycle condition λ ij+λ jk=λ ik\lambda_{i j} + \lambda_{j k} = \lambda_{i k}. But (using a partition of unity, see there) these non-globally defined forms are always equivalent to globally defined ones.

This simple technical point turns out to play a crucial role in the abstract definition of connections on ∞-bundles: generally, for all nn \in \mathbb{N} the cocycles given by globally defined forms in dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1) constitute curvature characteristic forms of genuine connections. The non-globally defined forms also constitute curvature invariants, but of pseudo-connections. The way the abstract theory finds the genuine connections inside all pseudo-connections is by the fact that we may find for each cocycle in dRB nU(1)\mathbf{\flat}_{dR} \mathbf{B}^n U(1) an equivalent one that does comes from a globally defined form.


There is a canonical 2-anafunctor c^ 1 dR:BU(1) dRB 2U(1)\hat {\mathbf{c}}_1^{dR} : \mathbf{B}U(1) \to \mathbf{\flat}_{dR}\mathbf{B}^2 U(1)

BU(1) diff dRB 2U(1) BU(1), \array{ \mathbf{B}U(1)_{diff} &\to& \mathbf{\flat}_{dR} \mathbf{B}^2 U(1) \\ \downarrow^{\mathrlap{\simeq}} \\ \mathbf{B}U(1) } \,,

where the top morphism is given by forming the pasting-composite with the BU(1)\mathbf{B} U(1)-universal 2-bundle, as described above.


For emphasis, notice that this span is governed by a presheaf of diagrams that over UCartSpU \in CartSp is of the form

U BU(1) transitionfunction Π 2(U) BINN(U) connection Π(U) B 2U(1) curvature. \array{ U &\to& \mathbf{B}U(1) &&& transition\;function \\ \downarrow && \downarrow \\ \mathbf{\Pi}_2(U) &\to& \mathbf{B}INN(U) &&& connection \\ \downarrow && \downarrow \\ \mathbf{\Pi}(U) &\to& \mathbf{B}^2 U(1) &&& curvature } \,.

The top morphisms are the components of the presheaf BU(1)\mathbf{B}U(1). The top squares are those of BU(1) diff\mathbf{B}U(1)_{diff}. Forming the bottom square is forming the bottom morphism, which necessarily satifies the constraint that makes it a components of B 2U(1)\mathbf{\flat}\mathbf{B}^2 U(1).

The interpretation of the stages is as indicated in the diagram:

  1. the top morphism is the transition function of the underlying bundle;

  2. the middle morphism is a choice of (pseudo-)connection on that bundle;

  3. the bottom morphism picks up the curvature of this connection.

We will see that full \infty-Chern-Weil theory is governed by a slight refinement of presheaves of essentially this kind of diagram. We will also see that the three stage process here is really an incarnation of the computation of a connecting homomorphism, reflecting the fact that behind the scenes the notion of curvature is exhibited as the obstruction cocycle to lifts from bare bundles to flat bundles.


For XC(U)gBU(1)X \stackrel{\simeq}{\leftarrow} C(U) \stackrel{g}{\to} \mathbf{B}U(1) the cocycle for a U(1)U(1)-principal bundle as described above, the composition of 2-anafunctors of gg with c^ 1 dR\hat {\mathbf{c}}_1^{dR} yields a cocycle for a 2-form c^ 1 dR(g)\hat {\mathbf{c}}_1^{dR}(g)

BU(1) conn C(V) BU(1) diff dRB 2U(1) C(U) g BU(1) X. \array{ && \mathbf{B}U(1)_{conn} \\ & {}^{\mathrlap{\nabla}}\nearrow & \downarrow \\ C(V) &\stackrel{}{\to}& \mathbf{B} U(1)_{diff} &\to& \mathbf{\flat}_{dR} \mathbf{B}^2 U(1) \\ \downarrow^{\mathrlap{\simeq}} && \downarrow^{\mathrlap{\simeq}} \\ C(U) &\stackrel{g}{\to}& \mathbf{B}U(1) \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

If we take {U iX}\{U_i \to X\} to be a good open cover, then we may assume V=UV = U. We know we can always find a pseudo-connection C(V)BU(1) diffC(V) \to \mathbf{B}U(1)_{diff} that is actually a genuine connection on a bundle in that it factors through the inclusion BU(1) connBU(1) diff\mathbf{B}U(1)_{conn} \to \mathbf{B}U(1)_{diff} as indicated.

The corresponding total map c 1 dR(g)c_1^{dR}(g) represented by c 1 dR()c_1^{dR}(\nabla) is the cocycle for the curvature 2-form of this connection. This represents the first Chern class of the bundle in de Rham cohomology.

For X,AX,A smooth 2-groupoids, write H(X,A)\mathbf{H}(X,A) for the 2-groupoid of 2-anafunctors between them.


Let H dR 2(X)H(X, dRB 2U(1))H_{dR}^2(X) \to \mathbf{H}(X,\mathbf{\flat}_{dR} \mathbf{B}^2 U(1)) be a choice of one closed 2-form representative for each degree-2 de Rham cohomology-class of XX. Then the pullback groupoid H conn(X,BU(1))\mathbf{H}_{conn}(X,\mathbf{B}U(1)) in

H conn(X,BU(1)) H dR 2(X) H(X,BU(1) diff) H(X, dRB 2U(1)) H(X,BU(1))U(1)Bund(X) \array{ \mathbf{H}_{conn}(X,\mathbf{B}U(1)) &\to& H_{dR}^2(X) \\ \downarrow && \downarrow \\ \mathbf{H}(X, \mathbf{B}U(1)_{diff}) &\to& \mathbf{H}(X, \mathbf{\flat}_{dR} \mathbf{B}^2 U(1)) \\ \downarrow^{\mathrlap{\simeq}} \\ \mathbf{H}(X,\mathbf{B}U(1)) \simeq U(1) Bund(X) }

is equivalent to disjoint union of groupoids of U(1)U(1)-bundles with connection whose curvatures are the chosen 2-form representatives.

Circle nn-bundles with connection and Deligne cohomology

For AA an abelian group there is a straightforward generalization of the above constructions to (G=B n1A)(G = \mathbf{B}^{n-1}A)-principal n-bundles with connection for all nn \in \mathbb{N}. We spell out the ingredients of the construction in a way analogous to the above discussion. A first-principles derivation of the objects we consider here is at circle n-bundle with connection. This is content that appeared partly in (SSSIII, FSS). We restrict attention to the circle n-group G=B n1U(1)G = \mathbf{B}^{n-1}U(1).

There is a familiar traditional presentation for ordinary differential cohomology in terms of Cech-Deligne cohomology. We briefly recall how this works and then indicate how this presentation can be derived along the above lines as a presentation of circle n-bundles with connection.


For nn \in \mathbb{N} the Deligne complex is the chain complex of sheaves (on SmoothMfd in general or on CartSp for our purposes here) of abelian groups given as follows

(n+1) D =[C (,/) d dR Ω 1() d dR d dR Ω n1() d dR Ω n() n n1 1 0]. \mathbb{Z}(n+1)^\infty_D = \left[ \array{ C^\infty(-,\mathbb{R}/\mathbb{Z}) &\stackrel{d_{dR}}{\to}& \Omega^1(-) &\stackrel{d_{dR}}{\to}& \cdots &\stackrel{d_{dR}}{\to}& \Omega^{n-1}(-) &\stackrel{d_{dR}}{\to}& \Omega^n(-) \\ n && n-1 && \cdots && 1 && 0 } \right] \,.

This is similar to the nn-fold shifted de Rham complex with two important differences

  1. In degree nn we have the sheaf of U(1)U(1)-valued functions, not of \mathbb{R}-valued functions (= 0-forms). The action of the de Rham differential on this is sometimes written dlog:C (,U(1))Ω 1()d log : C^\infty(-, U(1)) \to \Omega^1(-). But if we think of U(1)/U(1) \simeq \mathbb{R}/\mathbb{Z} then it is just the ordinary de Rham differential applied to any representative in C (,)C^\infty(-, \mathbb{R}) of an element in C (,/)C^\infty(-, \mathbb{R}/\mathbb{Z}).

  2. In degree 0 we do not have closed differential nn-forms (as one would have for the the de Rham complex shifted into non-negative degree), but all nn-forms.

As before we may use of the Dold-Kan correspondence Ξ:Ch +sAbUsSet\Xi : Ch_\bullet^{+} \stackrel{\simeq}{\to} sAb \stackrel{U}{\to} sSet to identify sheaves of chain complexes with simplicial sheaves.

For {U iX}\{U_i \to X\} a good open cover, the Deligne cohomology of XX in degree (n+1)(n+1) is

H diff n+1(X)=π 0[CartSp op,sSet](C({U i}),Ξ(n+1) D ). H_{diff}^{n+1}(X) = \pi_0 [CartSp^{op}, sSet]( C(\{U_i\}), \Xi \mathbb{Z}(n+1)^\infty_D ) \,.

Further using the Dold-Kan correspondence this is equivalently the cohomology of the Cech-Deligne double complex. A Deligne cocycle in degre (n+1)(n+1) then is a tuple

(g i 0,,i n,,A ijk,B ij,C i) (g_{i_0, \cdots, i_n}, \cdots, A_{i j k}, B_{i j}, C_{i})


  • C iΩ n(U i)C_i \in \Omega^n(U_i);

  • B ijΩ n1(U iU j)B_{i j} \in \Omega^{n-1}(U_i \cap U_j);

  • A ijkΩ n2(U iU jU k)A_{i j k } \in \Omega^{n-2}(U_i \cap U_j \cap U_k)

  • and so on

  • g i 0,,i nC (U i 0U i n,U(1))g_{i_0, \cdots, i_n} \in C^\infty(U_{i_0} \cap \cdots \cap U_{i_n} , U(1))

satisfying the cocycle condition

(d dR+(1) degδ)(g i 0,,i n,,A ijk,B ij,C i)=0, (d_{dR} + (-1)^{deg}\delta) (g_{i_0, \cdots, i_n}, \cdots, A_{i j k}, B_{i j}, C_{i}) = 0 \,,

where δ= i(1) ip i *\delta = \sum_{i} (-1)^i p_i^* is the alternating sum of the pullback of forms along the face maps of the Cech nerve.

This is a sequence of conditions of the form

  • C iC j=dB ijC_i - C_j = d B_{i j};

  • B ijB ik+B jk=dA ijkB_{i j} - B_{i k} + B_{j k} = d A_{i j k};

  • and so on

  • (δg) i 0,,i n+1=0(\delta g)_{i_0, \cdots, i_{n+1}} = 0.

For low nn we have seen these conditions in the dicussion of line bundles and of line 2-bundles (bundle gerbes) with connection above. Generally, for any nn \in \mathbb{N}, this is Cech-cocycle data for a circle n-bundle with connection, where

  • C iC_i are the local connection nn-forms;

  • g i 0,,i ng_{i_0, \cdots, i_n} is the transition function of the circle nn-bundle.

We now indicate how the Deligne complex may be derived from differential refinement of cocycles for circle nn-bundles along the lines of the above discussions.


B nU(1) ch:=ΞU(1)[n], \mathbf{B}^n U(1)_{ch} := \Xi U(1)[n] \,,

for the simplicial presheaf given under the Dold-Kan correspondence by the chain complex

U(1)[n]=(C (,U(1))00) U(1)[n] = \left( C^\infty(-,U(1)) \to 0 \to \cdots \to 0 \right)

with the sheaf represented by U(1)U(1) in degree nn.


For {U iX}\{U_i \to X\} an open cover of a smooth manifold XX and C(U)C(U) its Cech nerve, ∞-anafunctors

C(U) g B nU(1) ch X \array{ C(U) &\stackrel{g}{\to}& \mathbf{B}^n U(1)_{ch} \\ \downarrow^{\mathrlap{\simeq}} \\ X }

are in natural bijection with tuples of smooth functions

g i 0i n:U i 0U i n/ g_{i_0 \cdots i_n} : U_{i_0} \cap \cdots \cap U_{i_n} \to \mathbb{R}/\mathbb{Z}


(g) i 0i n+1:= k=0 ng i 0i k1i ki n=0, (\partial g)_{i_0 \cdots i_{n+1}} := \sum_{k = 0}^{n} g_{i_0 \cdots i_{k-1} i_k \cdot i_n} = 0 \,,

that is, to cocycles in degree nn Cech cohomology on UU with values in U(1)U(1).


C(U)Δ 1 (gλg) B nU(1) ch XΔ 1 \array{ C(U)\cdot \Delta^1 &\stackrel{(g \stackrel{\lambda}{\to} g')}{\to}& \mathbf{B}^n U(1)_{ch} \\ \downarrow^{\mathrlap{\simeq}} \\ X \cdot \Delta^1 }

are in natural bijection with tuples of smooth functions

λ i 0i n1:U i 0U i n1/ \lambda_{i_0 \cdots i_{n-1}} : U_{i_0} \cap \cdots \cap U_{i_{n-1}} \to \mathbb{R}/\mathbb{Z}

such that

g i 0i ng i 0i n=(δλ) i 0i n, g'_{i_0 \cdots i_n} - g_{i_0 \cdots i_n} = (\delta \lambda)_{i_0 \cdots i_n} \,,

that is, to Čech coboundaries.

The \infty-bundle PXP \to X classified by such a cocycle we may call a circle n-bundle. For n=1n = 1 this reproduces the ordinary U(1)U(1)-principal bundles that we considered before, for n=2n =2 the bundle gerbes and for n=3n=3 the bundle 2-gerbes.

To equip these circle nn-bundles with connections, we consider the differential refinements B nU(1) diff\mathbf{B}^n U(1)_{diff}, B nU(1) conn\mathbf{B}^n U(1)_{conn} and dRB n+1U(1)\mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1).



dRB n+1U(1) ch:=Ξ(Ω 1()d dRΩ 2()d dRd dRΩ cl n()) \mathbf{\flat}_{dR}\mathbf{B}^{n+1}U(1)_{ch} := \Xi\left( \Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega^n_{cl}(-) \right)

– the image under Ξ\Xi of the truncated de Rham complex – and

B nU(1) diff,ch={() B nU(1) Π() B nINN(U(1))}=Ξ(C (,/) d dR Ω 1() d dR Ω n() Id Id Ω 1() d dR d dR Ω n()) \mathbf{B}^n U(1)_{diff,ch} = \left\{ \array{ (-) &\to& \mathbf{B}^n U(1) \\ \downarrow && \downarrow \\ \mathbf{\Pi}(-) &\to& \mathbf{B}^n INN(U(1)) } \right\} = \Xi \left( \array{ C^\infty(-,\mathbb{R}/\mathbb{Z}) &\stackrel{d_{dR}}{\to}& \Omega^1(-) &\stackrel{d_{dR}}{\to}& \cdots & \to & \Omega^n(-) \\ \oplus & \nearrow_{\mathrlap{Id}} & \cdots & &\cdots& \nearrow_{\mathrlap{Id}} \\ \Omega^1(-) &\stackrel{d_{dR}}{\to}& \cdots &\stackrel{d_{dR}}{\to}& \Omega^n(-) } \right)


B nU(1) conn,ch=Ξ(C (,/)d dRΩ 1()d dRΩ 2()d dRd dRΩ n()) \mathbf{B}^n U(1)_{conn,ch} = \Xi\left( C^\infty(-, \mathbb{R}/\mathbb{Z}) \stackrel{d_{dR}}{\to} \Omega^1(-) \stackrel{d_{dR}}{\to} \Omega^2(-) \stackrel{d_{dR}}{\to} \cdots \stackrel{d_{dR}}{\to} \Omega^n(-) \right)

– the Deligne complex.

There is a canonical morphism

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

We have a pullback diagram

B nU(1) conn,ch Ω cl n+1() B nU(1) diff,ch curv dRB n1U(1) ch B nU(1) ch \array{ \mathbf{B}^n U(1)_{conn,ch} &\to& \Omega^{n+1}_{cl}(-) \\ \downarrow && \downarrow \\ \mathbf{B}^n U(1)_{diff,ch} &\stackrel{curv}{\to}& \mathbf{\flat}_{dR}\mathbf{B}^{n-1}U(1)_{ch} \\ \downarrow^{\mathrlap{\simeq}} \\ \mathbf{B}^n U(1)_{ch} }

in [Cart op,sSet][Cart^{op}, sSet].

This models a homotopy pullback

B nU(1) conn Ω cl n+1() B nU(1) curv dRB n1U(1) \array{ \mathbf{B}^n U(1)_{conn} &\to& \Omega^{n+1}_{cl}(-) \\ \downarrow &\swArrow_{\simeq}& \downarrow \\ \mathbf{B}^n U(1) &\stackrel{curv}{\to}& \mathbf{\flat}_{dR}\mathbf{B}^{n-1}U(1) }

in the (∞,1)-topos H=\mathbf{H} = Smooth∞Grpd and this implies (in particular) for all smooth manifolds XX a homtotopy pullback

H(X,B nU(1) conn) Ω cl n+1(X) H(X,B nU(1)) H(X, dRB n1U(1)). \array{ \mathbf{H}(X,\mathbf{B}^n U(1)_{conn}) &\to& \Omega^{n+1}_{cl}(X) \\ \downarrow &\swArrow_{\simeq}& \downarrow \\ \mathbf{H}(X,\mathbf{B}^n U(1)) &\to& \mathbf{H}(X,\mathbf{\flat}_{dR}\mathbf{B}^{n-1}U(1)) } \,.

Here cocycles in H(X,B nU(1) conn)\mathbf{H}(X, \mathbf{B}^n U(1)_{conn}) are modeled by ∞-anafunctors XC(U)gB nU(1) connX \stackrel{\simeq}{\leftarrow} C(U) \stackrel{g}{\to} \mathbf{B}^n U(1)_{conn}, which are in natural bijection with tuples

(C i,B i 0i 1,A i 0i 1,i 2,Z i 0i n1,g i 0i n), \left( C_{i}, B_{i_0 i_1}, A_{i_0 i_1, i_2}, \cdots Z_{i_0 \cdots i_{n-1}}, g_{i_0 \cdots i_{n}} \right) \,,

where C iΩ n(U i)C_i \in \Omega^n(U_i), B i 0i 1Ω n1(U i 0U i 1)B_{i_0 i_1} \in \Omega^{n-1}(U_{i_0} \cap U_{i_1}), etc. such that

C i 0C i 1=dB i 0i 1 C_{i_0} - C_{i_1} = d B_{i_0 i_1}


B i 0i 1B i 0i 2+B i 1i 2=dA i 0i 1i 2, B_{i_0 i_1} - B_{i_0 i_2} + B_{i_1 i_2} = d A_{i_0 i_1 i_2} \,,

etc. This is a cocycle in Cech-Deligne cohomology. We may think of this as encoding a circle n-bundle with connection. The forms (C i)(C_i) are the local connection nn-forms.

Remark. Everything in this construction turns out to follow from general abstract reasoning in every cohesive (∞,1)-topos H\mathbf{H} — except the sheaf Ω cl n()\Omega^n_{cl}(-) of closed nn-forms, which is a non-intrinsic truncation of dRB n+1U(1)\mathbf{\flat}_{dR}\mathbf{B}^{n+1}U(1) whose definition uses concretely the choice of model [CartSp op,sSet][CartSp^{op}, sSet]. But since by the above this object is used to pick homotopy fibers, and since these depend up to equivalence only on the connected component over which they are taken, for fixed XX no information is lost by passing instead to the de Rham cohomology set H dR n+1(X)H_{dR}^{n+1}(X) and choosing a morphism H dR n+1(X)H(X, dRB n+1U(1))H_{dR}^{n+1}(X) \to \mathbf{H}(X, \mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1)) that picks a closed (n+1)(n+1)-form in each cohomology class. Then we can replace the above by the homotopy pullback

H diff(X,B nU(1)) H dR n+1(X) H(X,B nU(1)) H(X, dRB n1U(1)) \array{ \mathbf{H}_{diff}(X,\mathbf{B}^n U(1)) &\to& H^{n+1}_{dR}(X) \\ \downarrow &\swArrow_{\simeq}& \downarrow \\ \mathbf{H}(X,\mathbf{B}^n U(1)) &\to& \mathbf{H}(X,\mathbf{\flat}_{dR}\mathbf{B}^{n-1}U(1)) }

without losing information. And this is defined fully intrinsically.

The definition of \infty-connections on GG-principal \infty-bundles for nonabelian GG may be reduced to this definition, by approximating every GG-cocylce XC(U)BGX \stackrel{\simeq}{\leftarrow} C(U) \to \mathbf{B}G by abelian cocycles by postcomposing with all possible characteristic classes BGB^GB nU(1)\mathbf{B}G \stackrel{\simeq}{\leftarrow} \hat \mathbf{B}G\to \mathbf{B}^n U(1) to extract a circle nn-bundle from it. This is what we turn to now.

The \infty-Chern-Weil homomorphism

We now come to the discussion the Chern-Weil homomorphism and its generalization to the ∞-Chern-Weil homomorphism.

We have seen above GG-principal \infty-bundles for general smooth \infty-groups GG and in particular for abelian groups GG. Naturally, the abelian case is easier and more powerful statements are known about this case. A general strategy for studying nonabelian \infty-bundles therefore is to approximate them by abelian bundles. This is achieved by considering characteristic classes. Roughly, a characteristic class is a map that functorially sends GG-principal \infty-bundles to B nK\mathbf{B}^n K-principal \infty-bundles, for some nn and some abelian group KK. In some cases such an assignment may be obtained by integration of infinitesimal data. If so, then the assignment refines to one of \infty-bundles with connection. For GG an ordinary Lie group this is then what is called the Chern-Weil homomorphism. For general GG we call it the ∞-Chern-Weil homomorphism.

Motivating examples

A simple motivating example for characteristic classes and the Chern-Weil homomorphism is the construction of determinant line bundles.


Let NN \in \mathbb{N}. Consider the unitary group U(N)U(N). By its definition as a matrix Lie group, this comes canonically equipped with the determinant function

det:U(N)U(1) det : U(N) \to U(1)

and by the standard properties of the determinant, this is in fact a group homomorphism. Therefore this has a delooping to a morphism of Lie groupoids

Bdet:BU(N)BU(1). \mathbf{B}det : \mathbf{B}U(N) \to \mathbf{B}U(1) \,.

Under geometric realization this maps to a morphism

Bdet:BU(N)BU(1)K(,2) |\mathbf{B} det| : B U(N) \to B U(1) \simeq K(\mathbb{Z},2)

of topological spaces. This is a characteristic class on the classifying space BU(N)B U(N): the first Chern class (see determinant line bundle for more on this).

By postcomposion with Bdet\mathbf{B}det of the classifying morphisms for principal bundles, it acts on principal bundles: postcomposition of a Cech cocycle

P: C({U i}) (g ij) BU(N) X \array{ P : & C(\{U_i\}) &\stackrel{(g_{i j})}{\to}& \mathbf{B} U(N) \\ & \downarrow^{\mathrlap{\simeq}} \\ & X }

for a U(N)U(N)-principal bundle on a smooth manifold XX with this characteristic class yields the cocycle

detP: C({U i}) (g ij) BU(N) Bdet BU(1) X \array{ det P : & C(\{U_i\}) &\stackrel{(g_{i j})}{\to}& \mathbf{B} U(N) &\stackrel{\mathbf{B}det}{\to}& \mathbf{B}U(1) \\ & \downarrow^{\mathrlap{\simeq}} \\ & X }

for a circle bundle (or its associated line bundle) with transition functions (det(g ij))(det (g_{i j})): the determinant line bundle of PP. The unique class

[detP]H 2(X,) [det P] \in H^2(X, \mathbb{Z})

of this line bundle is a characteristic of the original unitary bundle: its first Chern class c 1(P)c_1(P)

[detP]=c 1(P). [det P] = c_1(P) \,.

This construction directly extends to the case where the bundles carry connections.

We may canonically identify the Lie algebra 𝔲(n)\mathfrak{u}(n) with the matrix Lie algebra of skew-hermitian matrices on which we have the trace operation

tr:𝔲(n)𝔲(1)=i. tr : \mathfrak{u}(n) \to \mathfrak{u}(1) = i \mathbb{R} \,.

This is the differential version of the determinant in that when regarding the Lie algebra as the infinitesimal neighbourhood of the neutral element in U(N)U(N) (see ∞-Lie algebroid for more on this) the determinant becomes the trace under the exponential map

det(1+ϵA)=1+ϵtr(A) det (1 + \epsilon A) = 1 + \epsilon tr(A)

for ϵ 2=0\epsilon^2 = 0.

It follows that for tra :P 1(U i)BU(N)tra_\nabla : \mathbf{P}_1(U_i) \to \mathbf{B}U(N) the parallel transport of a connection on PP locally given by a 1-forms AΩ 1(U i,𝔲(N))A \in \Omega^1(U_i, \mathfrak{u}(N)) by

tra (γ)=𝒫exp [0,1]γ *A tra_\nabla(\gamma) = \mathcal{P} \exp \int_{[0,1]} \gamma^* A

the determinant parallel transport

dettra :P 1(U i)tra BU(N)detBU(1) det tra_\nabla : \mathbf{P}_1(U_i) \stackrel{tra_\nabla}{\to} \mathbf{B} U(N) \stackrel{det}{\to} \mathbf{B}U(1)

is locally given by the formula

dettra (γ)=𝒫exp [0,1]γ *trA det tra_\nabla(\gamma) = \mathcal{P} \exp \int_{[0,1]} \gamma^* tr A

which means that the local connection forms on the determinant line bundle are obtained from those of the unitary bundle by tracing.

(det,tr):{(g ij),(A i)}{(detg ij),(trA i)}. (det,tr) : \{(g_{i j}), (A_i)\} \mapsto \{(det g_{i j}), (tr A_i)\} \,.

This construction extends to a functor

(c^ 1):=(det,tr):U(N)Bund conn(X)U(1)Bund conn(X) (\hat \mathbf{c}_1) := (det, tr) : U(N) Bund_{conn}(X) \to U(1) Bund_{conn}(X)

natural in XX, that sends U(n)U(n)-principal bundles with connection to circle bundles with connection, hence to cocycles in degree-2 ordinary differential cohomology.

This assignment remembers of a unitary bundle one inegral class and its differential refinement:

  • the integral class of the determinant bundle is the first Chern class the U(N)U(N)-bundle

    [c^ 1(P)]=c 1(P); [\hat \mathbf{c}_1(P)] = c_1(P) \,;
  • the curvature 2-form of its connection is a representative in de Rham cohomology of this class

    [F c^ 1(P)]=c 1(P) dR. [F_{\nabla_{\hat \mathbf{c}_1(P)}}] = c_1(P)_{dR} \,.
H diff 2(X) H 2(X,) Ω cl 2(X) c^ 1 c 1(P) trF . \array{ && H^2_{diff}(X) \\ & \swarrow && \searrow \\ H^2(X,\mathbb{Z}) && && \Omega^2_{cl}(X) } \;\;\;\; \array{ && \hat \mathbf{c}_1 \\ & \swarrow && \searrow \\ c_1(P) &&&& tr F_\nabla } \,.

Equivalently this assignment is given by postcomposition of cocycles with a morphism of smooth ∞-groupoids

c^ 1:BU(N) connBU(1) conn. \hat \mathbf{c}_1 : \mathbf{B}U(N)_{conn} \to \mathbf{B}U(1)_{conn} \,.

We say that c^ 1\hat \mathbf{c}_1 is a differential characteristic class, the differential refinement of the first Chern class.

In (BrylinskiMacLaughlin) an algorithm is given for contructing differential characteristic classes on Cech cocycles in this fashion for more general Lie algebra cocycles.

For instance these authors give the following construction for the diffrential refinement of the first Pontryagin class.


Let NN \in \mathbb{N}, write Spin(N)Spin(N) for the Spin group and consider the canonical Lie algebra cohomology 3-cocycle

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

on semisimple Lie algebras, where ,\langle -,- \rangle is the Killing form invariant polynomial.

Let (PX,)(P \to X, \nabla) be a Spin(N)Spin(N)-principal bundle with connection. Let AΩ 1(P,𝔰𝔬(N))A \in \Omega^1(P, \mathfrak{so}(N)) be the Ehresmann connection 1-form on the total space of the bundle.

Then construct a Cech cocycle for Deligne cohomology in degree 4 as follows:

  1. pick an open cover {U iX}\{U_i \to X\} such that there is a choice of local sections σ i:U iP\sigma_i : U_i \to P. Write

    (g ij,A i):=(σ i 1σ j,σ i *A) (g_{i j}, A_i) := (\sigma_i^{-1} \sigma_j, \sigma_i^* A)

    for the induced Cech cocycle.

  2. Choose a lift of this cocycle to an assignment

    • of based paths in Spin(N)Spin(N) to double intersections

      g^ ij:U ij×Δ 1Spin(N), \hat g_{i j} : U_{i j}\times \Delta^1 \to Spin(N) \,,

      with g^ ij(0)=e\hat g_{i j}(0) = e and g^ ij(1)=g ij\hat g_{i j}(1) = g_{i j};

    • of based 2-simplices between these paths to triple intersections

      g^ ijk:U ijk×Δ 2Spin(N); \hat g_{i j k} : U_{i j k}\times \Delta^2 \to Spin(N) \,;

      restricting to these paths in the obvious way;

    • similarly of based 3-simplices between these paths to quadruple intersections

      g^ ijkl:U ijkl×Δ 3Spin(N). \hat g_{i j k l} : U_{i j k l}\times \Delta^3 \to Spin(N) \,.

Such lifts always exists, because the Spin group is connected (because already SO(N)SO(N) is), simply connected (because Spin(N)Spin(N) is the universal cover of SO(N)SO(N)) and also has π 2(Spin(N))=0\pi_2(Spin(N)) = 0 (because this is the case for every compact Lie group).

  1. Define from this a Deligne-cochain by setting

    (g ijkl,A ijk,B ij,C i):=( Δ 3(σ ig^ ijkl) *μ(A)mod, Δ 2(σ ig^ ijk) *cs(A), Δ 1(σ ig^ ij) *cs(A), σ i *μ(A)), (g_{i j k l}, A_{i j k}, B_{i j}, C_{i}) := \left( \array{ \int_{\Delta^3} (\sigma_i \cdot\hat g_{i j k l})^* \mu(A) mod \mathbb{Z}, \\ \int_{\Delta^2} (\sigma_i\cdot \hat g_{i j k})^* cs(A), \\ \int_{\Delta^1} (\sigma_i \cdot \hat g_{i j})^* cs(A), \\ \sigma_i^* \mu(A) } \right) \,,

    where cs(A)=AF A+cA[AA]cs(A) = \langle A \wedge F_A\rangle + c \langle A \wedge [A \wedge A]\rangle is the Chern-Simons form of the connection form AA with respect to the cocyle μ(A)=A[AA]\mu(A) = \langle A \wedge [A \wedge A]\rangle.

They then prove:

  1. This is indeed a Deligne cohomology cocycle;

  2. It represents the differential refinement of the first fractional Pontryagin class of PP.

H diff 4(X) H 4(X,) Ω cl 4(X) 12p^ 1 12p 1 dcs(A) \array{ && H^4_{diff}(X) \\ & \swarrow && \searrow \\ H^4(X,\mathbb{Z}) &&&& \Omega^4_{cl}(X) } \;\;\;\; \array{ && \frac{1}{2} \hat \mathbf{p}_1 \\ & \swarrow && \searrow \\ \frac{1}{2}p_1 &&&& d cs(A) }

In the form in which we have (re)stated this result here the second statement amounts, in view of the first statement, to the observation that the curvature 4-form of the Deligne cocycle is proportional to

F AF AΩ cl 4(X) \langle F_A \wedge F_A \rangle \in \Omega^4_{cl}(X)

which represents the first Pontryagin class in de Rham cohomology. Therefore the key observation is that we have a Deligne cocycle at all. This can be checked directly, if somewhat tediously, by hand. But then the question remains: where does this successful Ansatz come from? And is it natural ? For instance: does this construction extend to a morphism of smooth ∞-groupoids

12p^ 1:BSpin(N) connB 3U(1) conn \frac{1}{2}\hat \mathbf{p}_1 : \mathbf{B} Spin(N)_{conn} \to \mathbf{B}^3 U(1)_{conn}

from Spin-principal bundles with connection to circle 3-bundles with connection?

In the following we give a natural presentation of the ∞-Chern-Weil homomorphism by means of Lie integration of L L_\infty-algebraic data to simplicial presheaves. Among other things, this construction yields an understanding of why this construction is what it is and does what it does. In prop. 22 we reproduce the above example.

The construction proceeds in the following broad steps

  1. The infinitesimal analog of a characteristic class c:BGB nU(1)\mathbf{c} : \mathbf{B}G \to \mathbf{B}^n U(1) is a L-∞ algebra cocycle

    μ:𝔤b n1. \mu : \mathfrak{g} \to b^{n-1} \mathbb{R} \,.
  2. There is a formal procedure of universal Lie integration which sends this to a morphism of smooth ∞-groupoids

    exp(μ):exp(𝔤)exp(b n1)B n \exp(\mu) : \exp(\mathfrak{g}) \to \exp(b^{n-1} \mathbb{R}) \simeq \mathbf{B}^n \mathbb{R}

    presented by a morphism of simplicial presheaves on CartSp.

  3. By finding a Chern-Simons element cscs that witnesses the transgression of μ\mu to an invariant polynomial on 𝔤\mathfrak{g} this construction has a differential refinement to a morphism

    exp(μ,cs):exp(𝔤) connB n conn \exp(\mu,cs) : \exp(\mathfrak{g})_{conn} \to \mathbf{B}^n \mathbb{R}_{conn}

    that sends L L_\infty-algebra valued connections to line n-bundles with connection.

  4. The nn-truncation cosk n+1exp(𝔤)\mathbf{cosk}_{n+1} \exp(\mathfrak{g}) of the object on the left produces the smooth \infty-groups on interest – cosk n+1exp(𝔤)BG\mathbf{cosk}_{n+1} \exp(\mathfrak{g}) \simeq \mathbf{B}G – and the corresponding truncation of exp((μ,cs))\exp((\mu,cs)) carves out the lattice Γ\Gamma of periods in GG of the cocycle μ\mu inside \mathbb{R}. The result is the differential characteristic class

    exp(μ,cs):BG connB n/Γ conn. \exp(\mu,cs) : \mathbf{B}G_{conn} \to \mathbf{B}^n \mathbb{R}/\Gamma_{conn} \,.

    Typically we have Γ\Gamma \simeq \mathbb{Z} such that this then reads

    exp(μ,cs):BG connB nU(1) conn. \exp(\mu,cs) : \mathbf{B}G_{conn} \to \mathbf{B}^n U(1)_{conn} \,.

\infty-Lie theory

We discuss L-∞ algebras and more generally ∞-Lie algebroids – the higher analogs of Lie algebras and Lie algebroids – and their Lie integration to smooth ∞-groupoids presented by simplicial presheaves.

\infty-Lie algebroids

There is a precise sense in which one may think of a Lie algebra 𝔤\mathfrak{g} as the infinitesimal sub-object of the delooping groupoid BG\mathbf{B}G of the corresponding Lie group GG. Without here going into the details of this relation (which needs a little bit of (∞,1)-topos-theory), we want to build certain ∞-Lie groupoids from the knowledge of their infinitesimal subobjects: these subobjects are ∞-Lie algebroids and specifically ∞-Lie algebras – traditionally known as L L_\infty-algebras.

A quick but useful way of formalizing what this means is to observe that ordinary (finite-dimensional) Lie algebras (𝔤,[,])(\mathfrak{g}, [-,-]) are entirely encoded, dually, in their Chevalley-Eilenberg algebras CE(𝔤)=( 𝔤 *,d=[,] *)CE(\mathfrak{g}) = (\wedge^\bullet \mathfrak{g}^*, d = [-,-]^*) : free graded-commutative algebras over the ground field kk (which is \mathbb{R} for our purposes here) on the vector space 𝔤 *[1]\mathfrak{g}^*[1] equipped a differential dd of degree +1 and squaring to 0.

Simply by replacing in this characterization the vector space 𝔤 *\mathfrak{g}^* be an \mathbb{N}-graded vector space, we arrive at the notion of ∞-Lie algebra: the elements of 𝔤[1]\mathfrak{g}[1] in degree kk are the infinitesimal k-morphisms. Moreover, replacing in this characterization the ground field kk by an algebra of smooth functions on a manifold 𝔞 0\mathfrak{a}_0, we obtain the notion of an ∞-Lie algebroid 𝔤\mathfrak{g} over 𝔞 0\mathfrak{a}_0. Morphisms 𝔞𝔟\mathfrak{a} \to \mathfrak{b} of such ∞-Lie algebroids are dually precisely morphisms of dg-algebras CE(𝔞)CE(𝔟)CE(\mathfrak{a}) \leftarrow CE(\mathfrak{b}).

The following definition glosses over some fine print but is entirely sufficient for our present discussion.


The category of ∞-Lie algebroids is the opposite category of the full subcategory of dgAlg

LieAlgbd(dgAlg) op \infty LieAlgbd \subset (dgAlg)^{op}

on graded-commutative cochain dg-algebras in non-negative degree whose underlying graded algebra is a exterior algebra over the degree-0 algebra.

  • A strict \infty-Lie algebra is a dg-Lie algebra (𝔤,,[,])(\mathfrak{g}, \partial, [-,-]) with (𝔤 *, *)(\mathfrak{g}^*, \partial^*) a cochain complex in non-negative degree. With 𝔤 *\mathfrak{g}^* denoting the degreewise dual, the corresponding CE-algebra is CE(𝔤)=( 𝔤 *,d CE=[,] *+ *CE(\mathfrak{g}) = (\wedge^\bullet \mathfrak{g}^*, d_{CE} = [-,-]^* + \partial^*.

  • We had already seen above the infinitesimal approximation of a Lie 2-group: this is a Lie 2-algebra. If the Lie 2-group is a smooth strict 2-group it is encoded equivalently by a crossed module of ordinary Lie groups, and the corresponding Lie 2-algebra is given by a differential crossed module of ordinary Lie algebras.

  • The tangent Lie algebroid TXT X of a smooth manifold XX is the infinitesimal approximation to its fundamental ∞-groupoid. Its CE-algebra is the de Rham complex

    CE(TX)=Ω (X)CE(T X) = \Omega^\bullet(X).

  • For nn \in \mathbb{N}, n1n \geq 1, the Lie nn-algebra b n1b^{n-1}\mathbb{R} is the infinitesimal approximation to B nU()\mathbf{B}^n U(\mathbb{R}) and B n\mathbf{B}^n \mathbb{R}. Its CE-algebra is the dg-algebra on a single generators in degree nn, with vanishing differential.

  • For any \infty-Lie algebra 𝔤\mathfrak{g} there is an \infty-Lie algebra inn(𝔤)inn(\mathfrak{g}) defined by the fact that its CE-algebra is the Weil algebra of 𝔤\mathfrak{g}:

    CE(inn(𝔤))=W(𝔤)=( (𝔤 *𝔤 *[1]),d W 𝔤 *=d CE+σ), CE(inn(\mathfrak{g})) = W(\mathfrak{g}) = (\wedge^\bullet (\mathfrak{g}^* \oplus \mathfrak{g}^*[1]), d_{W}|_{\mathfrak{g}^*} = d_{CE} + \sigma ) \,,

    where σ:𝔤 *𝔤 *[1]\sigma : \mathfrak{g}^* \to \mathfrak{g}^*[1] is the grading shift isomorphism, extended as a derivation.

Lie integration

We discuss Lie integration: a construction that sends an L-∞ algebroid to a smooth ∞-groupoid of which it is the infinitesimal approximation.

The construction we want to describe may be understood as a generalization of the following proposition. This is classical, even if maybe not reflected in the standard textbook literature to the extent it deserves to be (see Lie integration for details and references).


For 𝔤\mathfrak{g} a (finite-dimensional) Lie algebra, let exp(𝔤)[CartSp op,sSet]\exp(\mathfrak{g}) \in [CartSp^{op}, sSet] be the simplicial presheaf given by the assignment

exp(𝔤):UHom dgAlg(CE(𝔤),Ω (U×Δ ) vert), \exp(\mathfrak{g}) : U \mapsto Hom_{dgAlg}(CE(\mathfrak{g}), \Omega^\bullet(U \times \Delta^\bullet)_{vert}) \,,

in degree kk of dg-algebra homomorphisms from the Chevalley-Eilenberg algebra of 𝔤\mathfrak{g} to the dg-algebra of vertical differential forms with respect to the trivial bundle U×Δ kUU \times \Delta^k \to U.


Shortly we will be considering variations of such assignments that are best thought about when writing out the hom-sets on the right here as sets of arrows; as in

exp(𝔤):(U,[k]){Ω vert (U×Δ k) A vert CE(𝔤)}). \exp(\mathfrak{g}) : (U,[k]) \mapsto \left\{ \array{ \Omega^\bullet_{vert}(U \times \Delta^k) &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) } \right\} ) \,.

For 𝔤\mathfrak{g} an ordinary Lie algebra it is an ancient (see Chern-Weil theory – history) and simple but important observation that dg-algebra morphisms Ω (Δ k)CE(𝔤)\Omega^\bullet(\Delta^k) \leftarrow CE(\mathfrak{g}) are in natural bijection with Lie-algebra valued 1-forms that are flat in that their curvature 2-forms vanish: the 1-form itself determines precisely a morphism of the underlying graded algebras, and the respect for the differentials is exactly the flatness condition. It is this elementary but similarly important observation that historically led Eli Cartan to Cartan calculus and the algebraic formulation of Chern-Weil theory.

One finds that it makes good sense to generally, for 𝔤\mathfrak{g} any ∞-Lie algebra or even ∞-Lie algebroid, think of Hom dgAlg(CE(𝔤),Ω (Δ k))Hom_{dgAlg}(CE(\mathfrak{g}), \Omega^\bullet(\Delta^k)) as the set of ∞-Lie algebroid valued differential forms whose curvature forms (generally a whole tower of them) vanishes.


Let GG be the simply-connected Lie group integrating 𝔤\mathfrak{g} according to Lie's three theorems and BG[CartSp op,Grpd]\mathbf{B}G \in [CartSp^{op}, Grpd] its delooping Lie groupoid regarded as a groupoid-valued presheaf on CartSp. Write τ 1()\tau_1(-) for the truncation operation that quotients out 2-morphisms in a simplicial presheaf to obtain a presheaf of groupoids.

We have an isomorphism

BG=τ 1exp(𝔤). \mathbf{B}G = \tau_1 \exp(\mathfrak{g}) \,.

To see this, observe that the presheaf exp(𝔤)\exp(\mathfrak{g}) has as 1-morphisms UU-parameterized families of 𝔤\mathfrak{g}-valued 1-forms A vertA_{vert} on the interval, and as 2-morphisms UU-parameterized families of flat 1-forms on the disk, interpolating between these. By identifying these 1-forms with the pullback of the Maurer-Cartan form on GG, we may equivalently think of the 1-morphisms as based smooth paths in GG and 2-morphisms smooth homotopies relative endpoints between them. Since GG is simply-connected this means that after dividing out 2-morphisms only the endpoints of these paths remain, which identify with the points in GG.

The following proposition establishes the Lie integraiton of the shifted 1-dimensional abelian L-∞ algebras b n1b^{n-1} \mathbb{R}.


For nn \in \mathbb{N}, n1n \geq 1. Write

B n ch:=Ξ[n] \mathbf{B}^n \mathbb{R}_{ch} := \Xi \mathbb{R}[n]

for the simplicial presheaf on CartSp that is the image of the sheaf of chain complexes represented by \mathbb{R} in degree nn and 0 in other degrees, under the Dold-Kan correspondence Ξ:Ch +sAbsSet\Xi : Ch_\bullet^+ \to sAb \to sSet.

Then there is a canonical morphism

Δ :exp(b n1)B n ch \int_{\Delta^\bullet} : \exp(b^{n-1}\mathbb{R}) \stackrel{\simeq}{\to} \mathbf{B}^n \mathbb{R}_{ch}

given by fiber integration of differential forms along U×Δ nUU \times \Delta^n \to U and this is an equivalence (a global equivalence in the model structure on simplicial presheaves).

The proof of this statement is discussed at Lie integration.

This statement will make an appearance repeatedly in the following discussion. Whenever we translate a construction given in terms exp()\exp(-) into a more convenient chain complex representation.

Characteristic classes from Lie integration

We now describe characteristic classes and then furhter below curvature characteristic forms on GG-bundles in terms of Lie integration to simplicial presheaves. For that purpose it is useful for a moment to ignore the truncation issue – to come back to it later – and consider these simplicial presheaves untruncated.

To see characteristic classes in this picture, write CE(b n1)CE(b^{n-1} \mathbb{R}) for the commutative real dg-algebra on a single generator in degree nn with vanishing differential. As our notation suggests, this we may think as the Chevalley-Eilenberg algebra of a higher Lie algebra – the ∞-Lie algebra b n1b^{n-1} \mathbb{R} – which is an Eilenberg-MacLane object in the homotopy theory of ∞-Lie algebras, representing ∞-Lie algebra cohomology in degree nn with coefficients in \mathbb{R}.

Restating this in elementary terms, this just says that dg-algebra homomorphisms

CE(𝔤)CE(b n1):μ CE(\mathfrak{g}) \leftarrow CE(b^{n-1}\mathbb{R}) : \mu

are in natural bijection with elements μCE(𝔤)\mu \in CE(\mathfrak{g}) of degree nn, that are closed, d CE(𝔤)μ=0d_{CE(\mathfrak{g})} \mu = 0. This is the classical description of a cocycle in the Lie algebra cohomology of 𝔤\mathfrak{g}.


Every such \infty-Lie algebra cocycle μ\mu induces a morphism of simplicial presheaves

exp(μ):exp(𝔤)exp(b n) \exp(\mu) : \exp(\mathfrak{g}) \to \exp(b^n \mathbb{R})

given by postcomposition

Ω vert (U×Δ l)A vertCE(𝔤)μCE(b n). \Omega^\bullet_{vert}(U \times \Delta^l) \stackrel{A_{vert}}{\leftarrow} CE(\mathfrak{g}) \stackrel{\mu}{\leftarrow} CE(b^n \mathbb{R}) \,.

(first Pontryagin class)

Assume 𝔤\mathfrak{g} to be a semisimple Lie algebra, let ,\langle -,-\rangle be the Killing form and μ=,[,]\mu = \langle -,[-,-]\rangle the corresponding 3-cocycle in Lie algebra cohomology. We may assume without restriction that this cocycle is normalized such that its left-invariant continuation to a 3-form on GG has integral periods. Observe that since π 2(G)\pi_2(G) is trivial we have that the 3-coskeleton of exp(𝔤)\exp(\mathfrak{g}) is equivalent to BG\mathbf{B}G. By the inegrality of μ\mu, the operation of exp(μ)\exp(\mu) on exp(𝔤)\exp(\mathfrak{g}) followed by integration over simplices, as in prop. 14, descends to an ∞-anafunctor from BG\mathbf{B}G to B 3U(1)\mathbf{B}^3 U(1), as indicated on the right of this diagram in [CartSp op,sSet][CartSp^{op}, sSet]

exp(𝔤) exp(μ) exp(b n1) Δ C(V) g^ cosk 3exp(𝔤) Δ cosk 3exp(μ) B 3/ C(U) g BG X. \array{ && \exp(\mathfrak{g}) &\stackrel{\exp(\mu)}{\to}& \exp(b^{n-1}\mathbb{R}) \\ && \downarrow && \downarrow^{\mathrlap{\int_{\Delta^\bullet}}} \\ C(V) & \stackrel{\hat g}{\to}& \mathbf{cosk}_3 \exp(\mathfrak{g}) &\stackrel{\int_{\Delta^\bullet}\mathbf{cosk}_3 \exp(\mu)}{\to}& \mathbf{B}^3 \mathbb{R}/\mathbb{Z} \\ \downarrow^{\mathrlap{\simeq}}&& \downarrow^{\mathrlap{\simeq}} \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

Precomposing this – as indicated on the left of the diagram – with another \infty-anafunctor XC(U)gBGX \stackrel{\simeq}{\leftarrow}C(U)\stackrel{g}{\to} \mathbf{B}G for a GG-principal bundle , hence a collection of transition functions {g ij:U iU jG}\{g_{i j} : U_i \cap U_j \to G\} amounts to choosing (possibly on a refinement VV of the cover UU of XX)

  • on each V iV jV_i \cap V_j a lift g^ ij\hat g_{i j} of g ijg_{i j} to a familly of smooth based paths in GGg^ ij:(V iV j)×Δ 1G\hat g_{i j} : (V_i \cap V_j) \times \Delta^1 \to G – with endpoints g ijg_{i j};

  • on each V iV jV kV_i \cap V_j \cap V_k a smooth family g^ ijk:(V iV jV k)×Δ 2G\hat g_{i j k} : (V_i \cap V_j \cap V_k) \times \Delta^2 \to G of disks interpolating between these paths;

  • on each V iV jV kV lV_i \cap V_j \cap V_k \cap V_l a a smooth family g^ ijkl:(V iV jV kV l)×Δ 3G\hat g_{i j k l} : (V_i \cap V_j \cap V_k \cap V_l) \times \Delta^3 \to G of 3-balls interpolating between these disks.

On this data the morphism Δ exp(μ)\int_{\Delta^\bullet} \exp(\mu) acts by sending each 3-cell to the number

g^ ijkl Δ 3g^ ijkl *μmod, \hat g_{i j k l} \mapsto \int_{\Delta^3} \hat g_{i j k l}^* \mu \;\; mod \mathbb{Z} \,,

where μ\mu is regarded in this formula as a closed 3-form on GG.

We say this is Lie integration of Lie algebra cocycles.


The Cech cohomology cocycle obtained this way is the first Pontryagin class of the GG-bundle classified by GG.

We shall show this below, as part of our L L_\infty-algebraic reconstruction of the above motivating example. In order to do so, we now add differential refinement to this Lie integration of characteristic classes.

L L_\infty-algebra valued connections

Above we described ordinary connections on bundles as well as connections on 2-bundles in terms of parallel transport over paths and surfaces, and showed how such is equivalently given by cocycles with coefficients in Lie-algebra valued differential forms and Lie 2-algebra valued differential forms, respectively.

Notably we saw (here) for the case of ordinary U(1)U(1)-principal bundles, that the connection and curvature data on these is encoded in presheaves of diagrams that over a given test space UU \in CartSp look like

U BU(1) transitionfunction Π(U) BINN(U) connection Π(U) B 2U(1) curvature \array{ U &\to& \mathbf{B}U(1) &&& transition\;function \\ \downarrow && \downarrow \\ \mathbf{\Pi}(U) &\to& \mathbf{B}INN(U) &&& connection \\ \downarrow && \downarrow \\ \mathbf{\Pi}(U) &\to& \mathbf{B}^2 U(1) &&& curvature }

together with a constraint on the bottom morphism.

It is in the form of such a kind of diagram that the general notion of connections on ∞-bundles may be modeled. In the full theory of differential cohomology in a cohesive topos this follows from first principles, but for our present introductory purpose we shall be content with taking this simple situation of U(1)U(1)-bundles together with the notion of Lie integration as sufficient motivation for the constructions considered now.

So we pass now to what is to some extent the reverse construction of the one considered before: we define a notion of ∞-Lie algebra valued differential forms and show how by a variant of Lie integration these integrate to coefficient objects for connections on ∞-bundles.

In the main entry ∞-Chern-Weil theory we discuss how this dg-algebraic construction follows from a general abstract definitions of differential cohomology in a cohesive topos.

The material of this section is due to (SSSI) and (FSS).

Curvature characteristics and Chern-Simons forms

For GG a Lie group, we have described above connections on GG-principal bundles in terms of cocycles with coefficients in the Lie-groupoid of Lie-algebra valued forms BG conn\mathbf{B}G_{conn}

BG conn connection BG diff pseudoconnection ps C(U) g BG transitionfunction X. \array{ && \mathbf{B}G_{conn} &&& connection \\ & {}^{\mathllap{\nabla}}\nearrow & \downarrow \\ && \mathbf{B}G_{diff} &&& pseudo-connection \\ & {}^{\mathllap{\nabla_{ps}}}\nearrow & \downarrow^{\mathrlap{\simeq}} \\ C(U) &\stackrel{g}{\to}& \mathbf{B}G &&& transition\;function \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

In this context we had derived Lie algebra valued forms from the parallel transport description BG conn=[P 1(),BG]\mathbf{B}G_{conn} = [\mathbf{P}_1(-), \mathbf{B}G]. We now turn this around and use Lie integration to construct parallel transport from Lie-algebra valued forms. The construction is such that it generalizes verbatim to ∞-Lie algebra valued forms.

For that purpose notice that another classical dg-algebra associated with 𝔤\mathfrak{g} is its Weil algebra W(𝔤)W(\mathfrak{g}).


The Weil algebra W(𝔤)\mathrm{W}(\mathfrak{g}) is the free dg-algebra on the graded vector space 𝔤 *\mathfrak{g}^*, meaning that there is a natural isomorphism

Hom dgAlg(W(𝔤),A)Hom Vect (𝔤 *,A), \mathrm{Hom}_{\mathrm{dgAlg}}(W(\mathfrak{g}), A) \simeq \mathrm{Hom}_{\mathrm{Vect}_{\mathbb{Z}}}(\mathfrak{g}^*, A) \,,

which is singled out among the isomorphism class of dg-algebras with this property by the fact that the projection of graded vector spaces 𝔤 *𝔤 *[1]𝔤 *\mathfrak{g}^* \oplus \mathfrak{g}^*[1] \to \mathfrak{g}^* extends to a dg-algebra homomorphism

CE(𝔤)W(𝔤):i *. CE(\mathfrak{g}) \leftarrow W(\mathfrak{g}) : i^* \,.

(Notice that general the dg-algebras that we are dealing with are semi-free dgas in that only their underlying graded algebra is free, but not the differential).

The most obvious realization of the free dg-algebra on 𝔤 *\mathfrak{g}^* is (𝔤 *𝔤 *[1])\wedge^\bullet (\mathfrak{g}^* \oplus \mathfrak{g}^*[1]) equipped with the differential that is precisely the degree shift isomorphism σ:𝔤 *𝔤 *[1]\sigma : \mathfrak{g}^* \to \mathfrak{g}^*[1] extended as a derivation. This is not the Weil algebra on the nose, but is of course isomorphic to it. The differential of the Weil algebra on (𝔤 *𝔤 *[1])\wedge^\bullet (\mathfrak{g}^* \oplus \mathfrak{g}^*[1]) is given on the unshifted generators by the sum of the CE-differential with the shift isomorphism

d W(𝔤) 𝔤 *=d CE(𝔤)+σ. d_{W(\mathfrak{g})}|_{\mathfrak{g}^*} = d_{CE(\mathfrak{g})} + \sigma \,.

This uniquely fixes the differential on the shifted generators – a phenomenon known (at least after mapping this to differential forms, as we discuss below) as the Bianchi identity.

Using this, we can express also the presheaf BG diff\mathbf{B}G_{diff} from def 7 in diagrammatic fashion.


For GG a simply connected Lie group, the presheaf BG diff[CartSp op,Grpd]\mathbf{B}G_{diff} \in [CartSp^{op}, Grpd] is isomorphic to

BG diff=τ 1(exp(𝔤) diff:(U,[k]){Ω vert (U×Δ k) A vert CE(𝔤) Ω (U×Δ k) A W(𝔤)}) \mathbf{B}G_{diff} = \tau_1 \left( \exp(\mathfrak{g})_{diff} : (U,[k]) \mapsto \left\{ \array{ \Omega^\bullet_{vert}(U \times \Delta^k) &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) \\ \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^k) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) } \right\} \right) \,

where on the right we have the 1-truncation of the simplicial presheaf of diagrams as indicated, where the vertical morphisms are the canonical ones.

Here over a given UU the bottom morphism in such a diagram is an arbitrary 𝔤\mathfrak{g}-valued 1-form AA on U×Δ kU \times \Delta^k. This we can decompose as A=A U+A vertA = A_U + A_{vert}, where A UA_U vanishes on tangents to Δ k\Delta^k and A vertA_{vert} on tangents to UU. The commutativity of the diagram asserts that A vertA_{vert} has to be such that the curvature 2-form F A vertF_{A_{vert}} vanishes when both its arguments are tangent to Δ k\Delta^k.

On the other hand, there is in the above no further constraint on A UA_U. Accordingly, as we pass to the 1-truncation of exp(𝔤) diff\exp(\mathfrak{g})_{diff} we find that morphisms are of the form (A U) 1g(A U) 2(A_U)_1 \stackrel{g}{\to} (A_U)_2 with (A U) i(A_U)^i arbitrary. This is the definition of BG diff\mathbf{B}G_{diff}.


We see below that it is not a coincidence that this is reminiscent to the first condition on an Ehresmann connection on a GG-principal bundle, which asserts that restricted to the fibers a connection 1-form on the total space of the bundle has to be flat. Indeed, the simplicial presheaf BG diff\mathbf{B}G_{diff} may be thought of as the (,1)(\infty,1)-sheaf of pseudo-connections on trivial \infty-bundles. Imposing on this also the second Ehresmann condition will force the pseudo-connection to be a genuine connection.

We now want to lift the above construction exp(μ)\exp(\mu) of characteristic classes by Lie integration of Lie algebra cocycles μ\mu from plain bundles classified by BG\mathbf{B}G to bundles with (pseudo-)connection classified by BG diff\mathbf{B}G_{diff}. By what we just said we therefore need to extend exp(μ)\exp(\mu) from a map on just exp(𝔤)\exp(\mathfrak{g}) to a map on exp(𝔤) diff\exp(\mathfrak{g})_{diff}.

This is evidently achieved by completing a square in dgAlg of the form

CE(𝔤) μ CE(b n1) W(𝔤) cs W(b n1) \array{ CE(\mathfrak{g}) &\stackrel{\mu}{\leftarrow}& CE(b^{n-1} \mathbb{R}) \\ \uparrow && \uparrow \\ W(\mathfrak{g}) &\stackrel{cs}{\leftarrow}& W(b^{n-1} \mathbb{R}) }

and defining exp(μ) diff:exp(𝔤) diffexp(b n1) diff\exp(\mu)_{diff} : \exp(\mathfrak{g})_{diff} \to \exp(b^{n-1}\mathbb{R})_{diff} to be the operation of forming pasting composites with this.

Here W(b n1)W(b^{n-1}\mathbb{R}) is the Weil algebra of the Lie n-algebra b n1b^{n-1} \mathbb{R}. This is the dg-algebra on two generators cc and kk, respectively, in degree nn and (n+1)(n+1) with the differential given by d W(b n1):ckd_{W(b^{n-1} \mathbb{R})} : c \mapsto k.

The commutativity of this diagram says that the bottom morphism takes the degree-nn generator cc to an element csW(𝔤)cs \in W(\mathfrak{g}) whose restriction to the unshifted generators is the given cocycle μ\mu.

As we shall see below, any such choice cscs will extend the characteristic cocycle obtained from exp(μ)\exp(\mu) to a characteristic differential cocycle, exhibiting the \infty-Chern-Weil homomorphism. But only for special nice choices of cscs will this take genuine \infty-connections to genuine \infty-connections – instead of to pseudo-connections. As we discuss in the full ∞-Chern-Weil theory, this makes no difference in cohomology. But in practice it is useful to fine-tune the construction such as to produce nice models of the \infty-Chern-Weil homomorphism given by genuine \infty-connections.

This is achieved by imposing the following additional constraint on the choice of extension cscs of μ\mu:


For μCE(𝔤)\mu \in CE(\mathfrak{g}) a cocycle and csW(𝔤)cs \in W(\mathfrak{g}) a lift of μ\mu through W(𝔤)CE(𝔤)W(\mathfrak{g}) \leftarrow CE(\mathfrak{g}), we say that W(𝔤)\langle -\rangle \in W(\mathfrak{g}) is an invariant polynomial in transgression with μ\mu if

  • both \langle -\rangle as well as d W(𝔤)d_{W(\mathfrak{g})}\langle - \rangle sit entirely in the shifted generators, in that 𝔤 *[1]W(𝔤)\in \wedge^\bullet \mathfrak{g}^*[1] \hookrightarrow W(\mathfrak{g}).

For 𝔤\mathfrak{g} a Lie algebra, this definition of invariant polynomials is equivalent to the traditional one.

To see this explicitly, let {t a}\{t^a\} be a basis of 𝔤 *\mathfrak{g}^* and {r a}\{r^a\} the corresponding basis of 𝔤 *[1]\mathfrak{g}^*[1]. Write {C a bc}\{C^a{}_{b c}\} for the structure constants of the Lie bracket in this basis.

Then for P=P (a 1,,a k)r a 1r a k r𝔤 *[1]P = P_{(a_1 , \cdots , a_k)} r^{a_1} \wedge \cdots \wedge r^{a_k} \in \wedge^{r} \mathfrak{g}^*[1] an element in the shifted generators, the condition that it is d W(𝔤)d_{W(\mathfrak{g})}-closed is equivalent to

C c(a 1 bP b,,a k)t cr a 1r a k, C^{b}_{c (a_1} P_{b, \cdots, a_k)} t^c \wedge r^{a_1} \wedge \cdots \wedge r^{a_k} \,,

where the parentheses around indices denotes symmetrization, as usual, so that this is equivalent to

iC c(a i bP a 1a i1ba i+1,a k)=0 \sum_{i} C^{b}_{c (a_i} P_{a_1 \cdots a_{i-1} b a_{i+1} \cdots, a_k)} = 0

for all choices of indices. This is the component-version of the familiar invariance statement

iP(t 1,,t i1,[t c,t i],t i+1,,t k)=0 \sum_i P(t_1, \cdots, t_{i-1}, [t_c, t_i], t_{i+1}, \cdots , t_k) = 0

for all t 𝔤t_\bullet \in \mathfrak{g}.


Write inv(𝔤)W(𝔤)inv(\mathfrak{g}) \subset W(\mathfrak{g}) (or W(𝔤) basicW(\mathfrak{g})_{basic}) for the sub-dgalgebra on invariant polynomials.


We have W(b n1)CE(b n)W(b^{n-1}\mathbb{R}) \simeq CE(b^n \mathbb{R}).

Using this, we can now encode the two conditions on the extension cscs of the cocycle μ\mu as the commutativity of this double square diagram

CE(𝔤) μ CE(b n1) cocycle W(𝔤) cs W(b n1) ChernSimonselement inv(𝔤) inv(b n1) invariantpolynomial. \array{ CE(\mathfrak{g}) &\stackrel{\mu}{\leftarrow}& CE(b^{n-1} \mathbb{R}) &&& cocycle \\ \uparrow && \uparrow \\ W(\mathfrak{g}) &\stackrel{cs}{\leftarrow}& W(b^{n-1} \mathbb{R}) &&& Chern-Simons\;element \\ \uparrow && \uparrow \\ inv(\mathfrak{g}) &\stackrel{\langle -\rangle}{\leftarrow}& inv(b^{n-1} \mathbb{R}) &&& invariant\;polynomial } \,.

In such a diagram, we call cscs the Chern-Simons element that exhibits the transgression between μ\mu and \langle - \rangle.

We shall see below that under the \infty-Chern-Weil homomorphism, Chern-Simons elements give rise to the familiar Chern-Simons forms – as well as their generalizations – as local connection data of secondary characteristic classes realized as circle n-bundles with connection.


What this diagram encodes is the construction of the connecting homomorphism for the long exact sequence in cohomology that is induced from the short exact sequence

ker(i *)W(𝔤)CE(𝔤) ker(i^*) \to W(\mathfrak{g}) \to CE(\mathfrak{g})

subject to the extra constraint of basic elements.

d W μ cs CE(𝔤) W(𝔤) inv(𝔤). \array{ && \langle - \rangle &\leftarrow& \langle - \rangle \\ && \uparrow^{\mathrlap{d_{W}}} \\ \mu &\leftarrow& cs \\ \\ \\ CE(\mathfrak{g}) &\leftarrow& W(\mathfrak{g}) &\leftarrow& inv(\mathfrak{g}) } \,.

To appreciate the construction so far, recall the

Classical fact

For GG a compact Lie group, the rationalization BGk\mathbf{B}G \otimes k of the classifying space BG\mathbf{B}G is the rational space whose Sullivan model is given by the algebra inv(𝔤)inv(\mathfrak{g}) of invariant polynomials on the Lie algebra 𝔤\mathfrak{g}.

This means that we may think of the consztructons so far in terms of the following picture:

deloopedgroup BG 𝔤 CE(𝔤) ChevalleyEilenbergalgebra deloopedgroupaluniversalbundle BEG inn(𝔤) W(𝔤)=CE(inn(𝔤)) Weilalgebra rationalizedclassifyingspace iB n i ib n i1 inv(𝔤) algebraofinvariantpolynomials Lieintegration \array{ delooped\; \infty-group &&& \mathbf{B}G && \mathfrak{g} && CE(\mathfrak{g}) &&& Chevalley-Eilenberg\;algebra \\ &&& \downarrow && \downarrow && \uparrow \\ delooped\;groupal\;universal\;\infty-bundle &&& \mathbf{B E}G && inn(\mathfrak{g}) && W(\mathfrak{g}) = CE(inn(\mathfrak{g})) &&& Weil\;algebra \\ &&& \downarrow && \downarrow && \uparrow \\ rationalized\;classifying\;space &&& \prod_i \mathbf{B}^{n_i} \mathbb{R} && \prod_i b^{n_i-1} \mathbb{R} && inv(\mathfrak{g}) &&& algebra\;of\;invariant\;polynomials \\ \\ &&& &\stackrel{Lie integration}{\leftarrow}& }
  • For 𝔤\mathfrak{g} a semisimple Lie algebra, ,\langle -,-\rangle the Killing form invariant polynomial, there is a Chern-Simons element csW(𝔤)cs \in W(\mathfrak{g}) witnessing the transgression to the cocycle μ=16,[,]\mu = - \frac{1}{6} \langle -,[-,-] \rangle. Under a 𝔤\mathfrak{g}-valued form Ω (X)W(𝔤):A\Omega^\bullet(X) \leftarrow W(\mathfrak{g}) : A this maps to the ordinary degree 3 Chern-Simons form

    cs(A)=AdA+13A[AA]. cs(A) = \langle A \wedge d A\rangle + \frac{1}{3} \langle A \wedge [A \wedge A]\rangle \,.

\infty-Connections from Lie integration

We have seen above for 𝔤\mathfrak{g} an \infty-Lie algebroid the object exp(𝔤) diff\exp(\mathfrak{g})_{diff} that classifies pseudo-connections on exp(𝔤)\exp(\mathfrak{g})-principal \infty-bundles and serves to support the \infty-Chern-Weil homomorphism. We now discuss the genuine ∞-connections among these pseudo-connections. From the point of view of the general abstract theory these are particularly nice representatives of more intrinsically defined structures.

For XX a smooth manifold and 𝔤\mathfrak{g} an ∞-Lie algebra or more generally an ∞-Lie algebroid, a \infty-Lie algebroid valued differential form on XX is a morphism of dg-algebras

Ω (X)W(𝔤):A \Omega^\bullet(X) \leftarrow W(\mathfrak{g}) : A

from the Weil algebra of 𝔤\mathfrak{g} to the de Rham complex of XX. Dually this is a morphism of ∞-Lie algebroids

A:TXinn(𝔤) A : T X \to inn(\mathfrak{g})

from the tangent Lie algebroid to the inner automorphism ∞-Lie algebra.

Its curvature is the composite of morphisms of graded vector spaces

Ω (X)AW(𝔤)F ()𝔤 *[1]:F A. \Omega^\bullet(X) \stackrel{A}{\leftarrow} W(\mathfrak{g}) \stackrel{F_{(-)}}{\leftarrow} \mathfrak{g}^*[1] : F_{A} \,.

Precisely if the curvatures vanish does the morphism factor through the Chevalley-Eilenberg algebra

(F A=0)( CE(𝔤) A flat Ω (X) A W(𝔤)) (F_A = 0) \;\;\Leftrightarrow \;\; \left( \array{ && CE(\mathfrak{g}) \\ & {}^{\mathllap{\exists A_{flat}}}\swarrow & \uparrow \\ \Omega^\bullet(X) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) } \right)

in which case we call AA flat.

The curvature characteristic forms of AA are the composite

Ω (X)AW(𝔤)F ()inv(𝔤):F A, \Omega^\bullet(X) \stackrel{A}{\leftarrow} W(\mathfrak{g}) \stackrel{\langle F_{(-)} \rangle}{\leftarrow} inv(\mathfrak{g}) : \langle F_A\rangle \,,

where inv(𝔤)W(𝔤)inv(\mathfrak{g}) \to W(\mathfrak{g}) is the inclusion of the invariant polynomials.


For UU a smooth manifold, the \infty-groupoid of 𝔤\mathfrak{g}-valued forms (see ∞-groupoid of ∞-Lie-algebra valued forms) is the Kan complex

exp(𝔤) conn(U):[k]{Ω (U×Δ k)AW(𝔤)vΓ(TΔ k):ι vF A=0} \exp(\mathfrak{g})_{conn}(U) : [k] \mapsto \left\{ \Omega^\bullet(U \times \Delta^k) \stackrel{A}{\leftarrow} W(\mathfrak{g}) \;\; | \;\; \forall v \in \Gamma(T \Delta^k) : \iota_v F_A = 0 \right\}

whose k-morphisms are 𝔤\mathfrak{g}-valued forms AA on U×Δ kU \times \Delta^k with sitting instants, and with the property that their curvature vanishes on vertical vectors.

The canonical morphism

exp(𝔤) connexp(𝔤) \exp(\mathfrak{g})_{conn} \to \exp(\mathfrak{g})

to the untruncated Lie integration of 𝔤\mathfrak{g} is given by restriction of AA to vertical differential forms (see below).


Here we are thinking of U×Δ kUU \times \Delta^k \to U as a trivial bundle.

The first Ehresmann condition can be identified with the conditions on lifts \nabla in ∞-anafunctors

exp(𝔤) conn C(U) g exp(𝔤) X \array{ && \exp(\mathfrak{g})_{conn} \\ & {}^{\mathllap{\nabla}}\nearrow & \downarrow \\ C(U) &\stackrel{g}{\to}& \exp(\mathfrak{g}) \\ \downarrow^{\mathrlap{\simeq}} \\ X }

that define connections on ∞-bundles.

Curvature characteristics

For Aexp(𝔤) conn(U,[k])A \in \exp(\mathfrak{g})_{conn}(U,[k]) a 𝔤\mathfrak{g}-valued form on U×Δ kU \times \Delta^k and for W(𝔤)\langle - \rangle \in W(\mathfrak{g}) any invariant polynomial, the corresponding curvature characteristic form F AΩ (U×Δ k)\langle F_A \rangle \in \Omega^\bullet(U \times \Delta^k) descends down to UU.


It is sufficient to show that for all vΓ(TΔ k)v \in \Gamma(T \Delta^k) we have

  1. ι vF A=0\iota_v \langle F_A \rangle = 0;

  2. vF A=0\mathcal{L}_v \langle F_A \rangle = 0.

The first condition is evidently satisfied if already ι vF A=0\iota_v F_A = 0. The second condition follows with Cartan calculus and using that d dRF A=0d_{dR} \langle F_A\rangle = 0:

vF A=dι vF A+ι vdF A=0. \mathcal{L}_v \langle F_A \rangle = d \iota_v \langle F_A \rangle + \iota_v d \langle F_A \rangle = 0 \,.

For a general \infty-Lie algebra 𝔤\mathfrak{g} the curvature forms F AF_A themselves are not necessarily closed (rather they satisfy the Bianchi identity), hence requiring them to have no component along the simplex does not imply that they descend. This is different for abelian \infty-Lie algebras: for them the curvature forms themselves are already closed, and hence are themselves already curvature characteristics that do descent.

It is useful to organize the 𝔤\mathfrak{g}-valued form AA, together with its restriction A vertA_{vert} to vertical differential forms and with its curvature characteristic forms in the commuting diagram

Ω (U×Δ k) vert A vert CE(𝔤) gaugetransformation Ω (U×Δ k) A W(𝔤) 𝔤valuedform Ω (U) F A inv(𝔤) curvaturecharacteristicforms \array{ \Omega^\bullet(U \times \Delta^k)_{vert} &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) &&& gauge\;transformation \\ \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^k) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) &&& \mathfrak{g}-valued\;form \\ \uparrow && \uparrow \\ \Omega^\bullet(U) &\stackrel{\langle F_A\rangle}{\leftarrow}& inv(\mathfrak{g}) &&& curvature\;characteristic\;forms }

in dgAlg.

The commutativity of this diagram is implied by ι vF A=0\iota_v F_A = 0.


Write exp(𝔤) CW(U)\exp(\mathfrak{g})_{CW}(U) for the \infty-groupoid of 𝔤\mathfrak{g}-valued forms fitting into such diagrams.

exp(𝔤) CW(U):[k]{Ω (U×Δ k) vert A vert CE(𝔤) Ω (U×Δ k) A W(𝔤) Ω (U) F A inv(𝔤)}. \exp(\mathfrak{g})_{CW}(U) : [k] \mapsto \left\{ \array{ \Omega^\bullet(U \times \Delta^k)_{vert} &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) \\ \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^k) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) \\ \uparrow && \uparrow \\ \Omega^\bullet(U) &\stackrel{\langle F_A\rangle}{\leftarrow}& inv(\mathfrak{g}) } \right\} \,.

If we just consider the top horizontal morphism in this diagram we obtain the object

exp(𝔤)(U):[k]{Ω (U×Δ k) vert A vert CE(𝔤)} \exp(\mathfrak{g})(U) : [k] \mapsto \left\{ \array{ \Omega^\bullet(U \times \Delta^k)_{vert} &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) } \right\}

discussed in detail at Lie integration. If we consider the top square of the diagram we obtain the object

exp(𝔤) diff(U):[k]{Ω (U×Δ k) vert A vert CE(𝔤) Ω (U×Δ k) A W(𝔤)}. \exp(\mathfrak{g})_{diff}(U) : [k] \mapsto \left\{ \array{ \Omega^\bullet(U \times \Delta^k)_{vert} &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) \\ \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^k) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) } \right\} \,.

This forms a resolution of exp(𝔤)\exp(\mathfrak{g}) and may be thought of as the \infty-groupoid of pseudo-connections.

We have an evident sequence of morphisms

exp(𝔤) conn genuineconnections exp(𝔤) CW pseudoconnectionwithglobalcurvaturecharacteristics exp(𝔤) diff pseudoconnections exp(𝔤) barebundles, \array{ \exp(\mathfrak{g})_{conn} &&& genuine\;connections \\ \downarrow \\ \exp(\mathfrak{g})_{CW} &&& pseudo-connection\;with\;global\;curvature\;characteristics \\ \downarrow \\ \exp(\mathfrak{g})_{diff} &&& pseudo-connections \\ \downarrow^{\mathrlap{\simeq}} \\ \exp(\mathfrak{g}) &&& bare bundles } \,,

where we label the objects by the structures they classify, as discussed at ∞-Chern-Weil theory.

Here the botton morphism is a weak equivalence and the others are monomorphisms.

Notice that in full ∞-Chern-Weil theory the fundamental object of interest is really exp(𝔤) diff\exp(\mathfrak{g})_{diff} – the object of pseudo-connections. The other objects only serve the purpose of picking particularly nice representatives:

the object exp(𝔤) CW\exp(\mathfrak{g})_{CW} may contain pseudo-connections, those for which at least the associated circle n-bundles with connection given by the \infty-Chern Weil homomorphism are genuine connections.

This distinction is important: over objects XX \in Smooth∞Grpd that are not smooth manifolds but for instance orbifolds, the genuine connections for higher Lie algebras do not exhaust all nonabelian differential cocycles. This just means that not every differential class in this case does have a nice representative in the usual sense.

1-Morphisms: integration of infinitesimal gauge transformations

The 1-morphisms in exp(𝔤)(U)\exp(\mathfrak{g})(U) may be thought of as gauge transformations between 𝔤\mathfrak{g}-valued forms. We unwind what these look like concretely.


Given a 1-morphism in exp(𝔤)(X)\exp(\mathfrak{g})(X), represented by 𝔤\mathfrak{g}-valued forms

Ω (U×Δ 1)W(𝔤):A \Omega^\bullet(U \times \Delta^1) \leftarrow W(\mathfrak{g}) : A

consider the unique decomposition

A=A U+(A vert:=λdt), A = A_U + ( A_{vert} := \lambda \wedge d t) \; \; \,,

with A UA_U the horizonal differential form component and t:Δ 1=[0,1]t : \Delta^1 = [0,1] \to \mathbb{R} the canonical coordinate.

We call λ\lambda the gauge parameter . This is a function on Δ 1\Delta^1 with values in 0-forms on UU for 𝔤\mathfrak{g} an ordinary Lie algebra, plus 1-forms on UU for 𝔤\mathfrak{g} a Lie 2-algebra, plus 2-forms for a Lie 3-algebra, and so forth.

We describe now how this enccodes a gauge transformation

A 0(s=0)λA U(s=1). A_0(s=0) \stackrel{\lambda}{\to} A_U(s = 1) \,.

By the nature of the Weil algebra we have

ddsA U=d Uλ+[λA]+[λAA]++ι sF A, \frac{d}{d s} A_U = d_U \lambda + [\lambda \wedge A] + [\lambda \wedge A \wedge A] + \cdots + \iota_s F_A \,,

where the sum is over all higher brackets of the ∞-Lie algebra 𝔤\mathfrak{g}.

In the Cartan calculus for 𝔤\mathfrak{g} an ordinary Lie algebra one writes the corresponding second Ehremsnn condition ι sF A=0\iota_{\partial_s} F_A = 0 equivalently

sA=ad λA. \mathcal{L}_{\partial_s} A = ad_\lambda A \,.

Define the covariant derivative of the gauge parameter to be

λ:=dλ+[Aλ]+[AAλ]+. \nabla \lambda := d \lambda + [A \wedge \lambda] + [A \wedge A \wedge \lambda] + \cdots \,.

In this notation we have

  • the general identity

    (1)ddsA U=λ+(F A) s \frac{d}{d s} A_U = \nabla \lambda + (F_A)_s
  • the horizontality or rheonomy constraint or second Ehresmann condition ι sF A=0\iota_{\partial_s} F_A = 0, the differential equation

    (2)ddsA U=λ. \frac{d}{d s} A_U = \nabla \lambda \,.

This is known as the equation for infinitesimal gauge transformations of an \infty-Lie algebra valued form.


By Lie integration we have that A vertA_{vert} – and hence λ\lambda – defines an element exp(λ)\exp(\lambda) in the ∞-Lie group that integrates 𝔤\mathfrak{g}.

The unique solution A U(s=1)A_U(s = 1) of the above differential equation at s=1s = 1 for the initial values A U(s=0)A_U(s = 0) we may think of as the result of acting on A U(0)A_U(0) with the gauge transformation exp(λ)\exp(\lambda).


(connections on ordinary bundles)

For 𝔤\mathfrak{g} an ordinary Lie algebra with simply connected Lie group GG and for BG conn\mathbf{B}G_{conn} the groupoid of Lie algebra-valued forms we have an isomorphism

τ 1exp(𝔤) conn=BG conn \tau_1 \exp(\mathfrak{g})_{conn} = \mathbf{B}G_{conn}

To see this, first note that the sheaves of objects on both sides are manifestly isomorphic, both are the sheaf of Ω 1(,𝔤)\Omega^1(-,\mathfrak{g}). For morphisms, observe that for a form Ω (U×Δ 1)W(𝔤):A\Omega^\bullet(U \times \Delta^1) \leftarrow W(\mathfrak{g}) : A which we may decompose into a horizontal and a verical pice as A=A U+λdtA = A_U + \lambda \wedge d t the condition ι tF A=0\iota_{\partial_t} F_A = 0 is equivalent to the differential equation

tA=d Uλ+[λ,A]. \frac{\partial}{\partial t} A = d_U \lambda + [\lambda, A] \,.

For any initial value A(0)A(0) this has the unique solution

A(t)=g(t) 1(A+d U)g(t), A(t) = g(t)^{-1} (A + d_{U}) g(t) \,,

where g:[0,1]Gg : [0,1] \to G is the parallel transport of λ\lambda:

t(g (t) 1(A+d U)g(t)) = g(t) 1(A+d U)λg(t)g(t) 1λ(A+d U)g(t) \begin{aligned} & \frac{\partial}{\partial t} \left( g_(t)^{-1} (A + d_{U}) g(t) \right) \\ = & g(t)^{-1} (A + d_{U}) \lambda g(t) - g(t)^{-1} \lambda (A + d_{U}) g(t) \end{aligned}

(where for ease of notaton we write actions as if GG were a matrix Lie group).

In particular this implies that the endpoints of the path of 𝔤\mathfrak{g}-valued 1-forms are related by the usual cocycle condition in BG conn\mathbf{B}G_{conn}

A(1)=g(1) 1(A+d U)g(1). A(1) = g(1)^{-1} (A + d_U) g(1) \,.

In the same fashion one sees that given 2-cell in exp(𝔤)(U)\exp(\mathfrak{g})(U) and any 1-form on UU at one vertex, there is a unique lift to a 2-cell in exp(𝔤) conn\exp(\mathfrak{g})_{conn}, obtained by parallel transporting the form around. The claim then follows from the previous statement of Lie integration that τ 1exp(𝔤)=BG\tau_1 \exp(\mathfrak{g}) = \mathbf{B}G.

Differential characteristic classes from Lie integration

We have now the ingredients in hand to produce a construction of differential characteristic classes – the refined ∞-Chern-Weil homomorphism – in terms of Lie integration of differential refinements of L L_\infty-algebra cocycles.

We first consider the local construction that produces the de Rham cohomology data of the differential characteristic classes. Since this turns out to be a generalization of the construction of the action functional of Chern-Simons theory, we speak of

Applying a coskeleton-truncation to this construction carves out the period lattice of the L L_\infty-algebra cocycle inside the line \mathbb{R}, which yields to the fully-fledged differential characteristic classes, typically called secondary characteristic classes

In full ∞-Chern-Weil theory the \infty-Chern-Weil homomorphism is conceptually very simple: for every nn there is canonically a morphism of ∞-Lie groupoids B nU(1) dRB n+1U(1)\mathbf{B}^n U(1) \to \mathbf{\flat}_{dR}\mathbf{B}^{n+1}U(1) where the object on the right classifies ordinary de Rham cohomology in degree n+1n+1. For GG any ∞-group and any characteristic class c:BGB n+1U(1)\mathbf{c} : \mathbf{B}G \to \mathbf{B}^{n+1}U(1), the \infty-Chern-Weil homomorphism is the operation that takes a GG-principal ∞-bundle XBGX \to \mathbf{B}G to the composite XBGB nU(1) dRB n+1U(1)X \to \mathbf{B}G \to \mathbf{B}^n U(1) \to \mathbf{\flat}_{dR} \mathbf{B}^{n+1}U(1).

All the construction that we consider here in this introduction serve to present this abstract operation. The \infty-connections that we considered yield resolutions of B nU(1)\mathbf{B}^n U(1) and BG\mathbf{B}G in terms of which the abstract morphisms are modeled as ∞-anafunctors.

\infty-Chern-Simons functionals

We have considered above ∞-connections in terms of dg-algebra homomorphisms and Chern-Simons elements witnessing the transgression of cocycles to invariant polynomials in terms of dg-algebra homomorphisms. There is an evident way to compose these two constructions.


Let 𝔤\mathfrak{g} be an L-∞ algebra and μ:𝔤b n1\mu : \mathfrak{g} \to b^{n-1}\mathbb{R} a cocycle in its L-∞ algebra cohomology, which transgresses to an invariant polynomial \langle -\rangle, witnessed by a Chern-Simons element cscs.

Then let

exp(μ,cs):exp(𝔤) connexp(b n1) conn \exp(\mu,cs) : \exp(\mathfrak{g})_{conn} \to \exp(b^{n-1}\mathbb{R})_{conn}

be the morphism of simplicial presheaves obtained by forming pasting composites of the defining diagrams in dgAlg of these structures:

over UCartSpU \in CartSp and [k]Δ[k] \in \Delta the morphism exp(μ,cs)\exp(\mu,cs) sends an element Aexp(𝔤) conn(U) kA \in \exp(\mathfrak{g})_{conn}(U)_k to the element cs(A)exp(b n1) conncs(A) \in \exp(b^{n-1}\mathbb{R})_{conn} given explicitly as follows

(Ω vert (U×Δ k) A vert CE(𝔤) transitionfunction/Cechcocycle Ω (U×Δ k) A W(𝔤) connection Ω (U) F A inv(𝔤) curvaturecharacteristics)(CE(𝔤) μ CE(b n1) cocycle W(𝔤) cs W(b n1) ChernSimonselement inv(𝔤) inv(b n1) invariantpolynomial) \left( \array{ \Omega^\bullet_{vert}(U \times \Delta^k) &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) &&& transition\;function\;/\;Cech\;cocycle \\ \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^{k}) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) &&& connection \\ \uparrow && \uparrow \\ \Omega^\bullet(U) &\stackrel{\langle F_A\rangle}{\leftarrow}& inv(\mathfrak{g}) &&& curvature\;characteristics } \right) \circ \left( \array{ CE(\mathfrak{g}) &\stackrel{\mu}{\leftarrow}& CE(b^{n-1} \mathbb{R}) &&& cocycle \\ \uparrow && \uparrow \\ W(\mathfrak{g}) &\stackrel{cs}{\leftarrow}& W(b^{n-1} \mathbb{R}) &&& Chern-Simons\;element \\ \uparrow && \uparrow \\ inv(\mathfrak{g}) &\stackrel{\langle -\rangle}{\leftarrow}& inv(b^{n-1} \mathbb{R}) &&& invariant\;polynomial } \right)
=(Ω (U×Δ k) vert A vert CE(𝔤) μ CE(b n1) :μ(A vert) characteristicclass Ω (U×Δ k) A W(𝔤) cs W(b n1) :cs(A) ChernSimonsform Ω (U) F A inv(𝔤) inv(b n1) :F A curvaturecharacteristicform). = \; \; \; \left( \array{ \Omega^\bullet(U \times \Delta^k)_{vert} &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) &\stackrel{\mu}{\leftarrow}& CE(b^{n-1} \mathbb{R}) & : \mu(A_{vert}) &&& characteristic\;class \\ \uparrow && \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^k) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) &\stackrel{cs}{\leftarrow}& W(b^{n-1} \mathbb{R}) & : cs(A) &&& Chern-Simons\;form \\ \uparrow && \uparrow && \uparrow \\ \Omega^\bullet(U) &\stackrel{\langle F_A\rangle}{\leftarrow}& inv(\mathfrak{g}) &\stackrel{\langle -\rangle}{\leftarrow}& inv(b^{n-1} \mathbb{R}) & : \langle F_A\rangle &&& curvature\;characteristic\;form } \right) \,.

By restriction to the top two layers of these diagrams this analogously yields a morphism

exp(μ,cs):exp(𝔤) diffexp(b n1) diff. \exp(\mu, cs): \exp(\mathfrak{g})_{diff} \to \exp(b^{n-1}\mathbb{R})_{diff} \,.

Analogously, projection onto the third horizontal layer gives amorphism

exp(μ,cs):exp(b n1) diff dRexp(b n) smp Δ dRB n+1 ch \exp(\mu,cs) : \exp(b^{n-1}\mathbb{R})_{diff} \to \mathbf{\flat}_{dR}\exp(b^{n} \mathbb{R})_{smp} \underoverset{\int_{\Delta^\bullet}}{\simeq}{\to} \mathbf{\flat}_{dR} \mathbf{B}^{n+1} \mathbb{R}_{ch}

to the de Rham coefficient object.


The morphism exp(μ,cs)\exp(\mu,cs) carries 𝔤\mathfrak{g}-valued connections \nabla locally given by 𝔤\mathfrak{g}-valued forms AA to b n1b^{n-1}\mathbb{R}-valued connections whose higher parallel transport over an nn-dimensional smooth manifold Σ\Sigma is locally given by the integral Σcs(A)\int_\Sigma cs(A) of the Chern-Simons form cs(A)cs(A) over Σ\Sigma. This assignment A Σcs(A)A \mapsto \int_\Sigma cs(A) is the action functional for an ∞-Chern-Simons theory defined by the invariant polynomial W(𝔤)\langle -\rangle \in W(\mathfrak{g}). Therefore we may regard exp(μ,cs)\exp(\mu,cs) as being the Lagrangian for this ∞-Chern-Simons theory.

In total, this construction constitutes an \infty-anafunctor

exp(𝔤) diff exp(μ) diff dRB n+1 ch exp(𝔤). \array{ \exp(\mathfrak{g})_{diff} &\stackrel{\exp(\mu)_{diff}}{\to}& \mathbf{\flat}_{dR} \mathbf{B}^{n+1}\mathbb{R}_{ch} \\ \downarrow^{\mathrlap{\simeq}} \\ \exp(\mathfrak{g}) } \,.

Postcomposition with this is the simple \infty-Chern-Weil homomorphism: it sends a cocycle

C(U) exp(𝔤) X \array{ C(U) &\to& \exp(\mathfrak{g}) \\ \downarrow^{\mathrlap{\simeq}} \\ X }

for an exp(𝔤)\exp(\mathfrak{g})-principal ∞-bundle to the curvature form represented by

C(V) (g,) exp(𝔤) diff exp(μ) diff exp(b n1) diff dRB n+1 ch C(U) g exp(𝔤) X. \array{ C(V) &\stackrel{(g,\nabla)}{\to}& \exp(\mathfrak{g})_{diff} &\stackrel{\exp(\mu)_{diff}}{\to}& \exp(b^{n-1}\mathbb{R})_{diff} &\stackrel{}{\to}& \mathbf{\flat}_{dR} \mathbf{B}^{n+1}\mathbb{R}_{ch} \\ \downarrow^{\mathrlap{\simeq}} && \downarrow^{\mathrlap{\simeq}} \\ C(U) &\stackrel{g}{\to}& \exp(\mathfrak{g}) \\ \downarrow^{\mathrlap{\simeq}} \\ X } \,.

For 𝔤\mathfrak{g} an ordinary Lie algebra the image under τ 1()\tau_1(-) of this diagram constitutes the ordinary Chern-Weil homomorphism in that:

for gg the cocycle for a GG-principal bundle, any ordinary connection on a bundle constitutes a lift (g,)(g,\nabla) to the tip of the anafunctor and the morphism represented by that is the Cech-hypercohomology cocycle on XX with values in the truncated de Rham complex given by the globally defined curvature characteristic form F F \langle F_\nabla \wedge \cdots \wedge F_\nabla\rangle.

This construction however discards the information in the choice of connection and in the Chern-Simons form of this connection. Below we lift this construction to one that produces the full secondary characteristic classes in ordinary differential cohomology of the refined \infty-Chern-Weil homomorphism.

Secondary characteristic classes

So far we discussed the untruncated coefficient object exp(𝔤) conn\exp(\mathfrak{g})_{conn} for 𝔤\mathfrak{g}-valued ∞-connections. The real object of interest is the kk-truncated version τ kexp(𝔤) conn\tau_k \exp(\mathfrak{g})_{conn} where kk \in \mathbb{N} is such that τ kexp)𝔤BG\tau_k \exp)\mathfrak{g} \simeq \mathbf{B}G is the delooping of the \infty-Lie group in question.

Under such a truncation, the integrated \infty-Lie algebra cocycle exp(μ):exp(𝔤)exp(b n1)exp(\mu) : exp(\mathfrak{g}) \to exp(b^{n-1}\mathbb{R}) will no longer be a simplicial map. Instead, the periods of μ\mu will cut out a lattice Γ\Gamma in \mathbb{R}, and exp(μ)\exp(\mu) does descent to the quotient of \mathbb{R} by that lattice

exp(μ):τ kexp(𝔤)B n/Γ. \exp(\mu) : \tau_k \exp(\mathfrak{g}) \to \mathbf{B}^n \mathbb{R}/\Gamma \,.

We now say this again in more detail.

Suppose 𝔤\mathfrak{g} is such that the (n+1)(n+1)-coskeleton cosk n+1exp(𝔤)BG\mathbf{cosk}_{n+1} \exp(\mathfrak{g}) \stackrel{\simeq}{\to} \simeq \mathbf{B}G for the desired GG. Then the periods of μ\mu over (n+1)(n+1)-balls cut out a lattice Γ\Gamma \subset \mathbb{R} and thus we get an ∞-anafunctor

cosk n+1exp(𝔤) diff B n/Γ diff dRB n+1/Γ BG \array{ \mathbf{cosk}_{n+1} \exp(\mathfrak{g})_{diff} &\to& \mathbf{B}^{n}\mathbb{R}/\Gamma_{diff} &\to& \mathbf{\flat}_{dR} \mathbf{B}^{n+1} \mathbb{R}/\Gamma \\ \downarrow^{\mathrlap{\simeq}} \\ \mathbf{B}G }

This is curvature characteristic class. We may always restrict to genuine \infty-connections and refine

cosk n+1exp(𝔤) conn B n/Γ conn cosk n+1exp(𝔤) diff B n/Γ diff dRB n+1/Γ BG \array{ \mathbf{cosk}_{n+1} \exp(\mathfrak{g})_{conn} &\to& \mathbf{B}^{n}\mathbb{R}/\Gamma_{conn} \\ \downarrow && \downarrow \\ \mathbf{cosk}_{n+1} \exp(\mathfrak{g})_{diff} &\to& \mathbf{B}^{n}\mathbb{R}/\Gamma_{diff} &\to& \mathbf{\flat}_{dR} \mathbf{B}^{n+1} \mathbb{R}/\Gamma \\ \downarrow \\ \mathbf{B}G }

which models the refined \infty-Chern-Weil homomorphism with values in ordinary differential cohomology

H conn(X,BG)H conn(X,B n+1/Γ) \mathbf{H}_{conn}(X,\mathbf{B}G) \to \mathbf{H}_{conn}(X, \mathbf{B}^{n+1} \mathbb{R}/\Gamma)

We can now reproduce our motivating example of the Brylinski-McLaughlin construction of the the differential refinement of the first fractional Pontryagin class as a special case of the presentation of the \infty-Chern-Weil homomorphism by Lie integrated simplicial presheaves.


Let 𝔤=𝔰𝔬(n)\mathfrak{g} = \mathfrak{so}(n) be the special orthogonal Lie algebra, μ=,[,]\mu = \langle -,[-,-]\rangle the canonical Lie algebra cohomology 3-cocycle and csW(𝔤)cs \in W(\mathfrak{g}) the standard Chern-Simons element witnessing the transgression to the Killing form invariant polynomial.

Then for XX any smooth manifold, the Lie integration of (μ,cs)(\mu,cs) presents a morphism morphism

exp(μ)H conn(X,BSpin(n))H conn(X,B 3U(1)) \exp(\mu) \mathbf{H}_{conn}(X, \mathbf{B}Spin(n)) \to \mathbf{H}_{conn}(X, \mathbf{B}^3 U(1))

that sends SpinSpin-principal bundles with connection to their Chern-Simons circle 3-bundle with connection and as such represents a differential refinement of the first fractional Pontryagin class

exp(μ,cs)=12p^ 1. \exp(\mu,cs) = \frac{1}{2}\hat \mathbf{p}_1 \,.

Moreover, the defining presentation on simplicial presheaves of exp(μ,cs)\exp(\mu,cs) given by the \infty-anafunctor

exp(𝔤) diff exp(μ) diff exp(b n1) diff Δ C(V) (g^,^) cosk 3exp(𝔤) diff B 3U(1) diff C(U) (g,) BG diff X \array{ && \exp(\mathfrak{g})_{diff} &\stackrel{\exp(\mu)_{diff}}{\to}& \exp(b^{n-1}\mathbb{R})_{diff} \\ && \downarrow && \downarrow^{\mathrlap{\int_{\Delta^\bullet}}} \\ C(V) &\stackrel{(\hat g,\hat \nabla)}{\to}& \mathbf{cosk}_3\exp(\mathfrak{g})_{diff} &\to& \mathbf{B}^3 U(1)_{diff} \\ \downarrow^{\mathrlap{\simeq}} && \downarrow \\ C(U) &\stackrel{(g,\nabla)}{\to}& \mathbf{B}G_{diff} \\ \downarrow^{\mathrlap{\simeq}} \\ X }

exhibits exactly the Brylinski-MacLaughlin algorithm for constructing Cech-cocycle representatives for this class.

This is due to (FSS)

By feeding in more general transgressive ∞-Lie algebra cocycles through this machine, we obtain cocycles for more general differential characteristic classes. For instance the next one is the second fractional Pontryagin class of smooth String principal 2-bundles with connection (FSS). Moreover, these constructions naturally yield the full cocycle \infty-groupoids, not just their cohomology sets. This allows to form the homotopy fibers of the \infty-Chern-Weil homomorphism and thus define differential string structures etc., and twisted differential string structures etc. (SSSIII).


This section gives a concise summary of the constructions introduced above.

For connections on GG-principal 1-bundles

We have the following diffeological 1- or 2-groupoids.

Let GG be a Lie group. We have the following Lie groupoids associated with that

  • BG\mathbf{B}G – the coefficient for GG-principal bundles;

  • INN(G)=G//GINN(G) = G//G – the inner automorphism 2-group of GG, a groupal model for the universal principal bundle;

  • BINN(G)\mathbf{B}INN(G) – the coefficient for INN(G)INN(G)-principal 2-bundle;

  • BG conn:=Hom Grpd(Diffeo)(P 1(),BG)\mathbf{B}G_{conn} := Hom_{Grpd(Diffeo)}(\mathbf{P}_1(-), \mathbf{B}G) – the coefficient for GG-principal bundles with connection;

  • BG:=Hom Grpd(Diffeo)(Π 2(),BINN(G))\mathbf{\flat} \mathbf{B}G := Hom_{Grpd(Diffeo)}(\Pi_2(-), \mathbf{B}INN(G)) the coefficient for flat GG-principal bundles with flat connection;

  • BINN(G):=[Π 2(),BINN(G)]\mathbf{\flat} \mathbf{B}INN(G) := [\Pi_2(-), \mathbf{B}INN(G)] the coefficient for flat INN(G)INN(G)-principal 2-bundles;

  • BG diff:=BINN(G)× BINN(G)BG\mathbf{B}G_{diff} := \mathbf{\flat}\mathbf{B}INN(G) \times_{\mathbf{B}INN(G)} \mathbf{B}G – the coefficient for GG-principal bundles with pseudo-connection;

We have the following morphisms between these:

  • XP 1(X)X \to \mathbf{P}_1(X) – inclusion of constant paths into all paths;

  • P 1(X)Π 1(X)\mathbf{P}_1(X) \to \mathbf{\Pi}_1(X) – sends thin homotopy-classes of paths to their full homotopy classes;

  • BGBG conn\mathbf{\flat}\mathbf{B}G \to \mathbf{B}G_{conn} – the morphism which forgets that a connection is flat;

  • BG connBG\mathbf{B}G_{conn} \to \mathbf{B}G – forgets the connection on a GG-bundle, induced locally by UP 1(U)U \to \mathbf{P}_1(U);

  • BG connBINN(G)\mathbf{B}G_{conn} \to \mathbf{\flat} \mathbf{B}INN(G) – the morphism that fills in the integrated curvature between paths enclosing a surface;

  • BG connBG diff\mathbf{B}G_{conn} \to \mathbf{B}G_{diff} the morphism that regards an ordinary connection as a special case of a pseudo-connection, induced as a morphism into a pullback by the two morphisms BG connBG\mathbf{B}G_{conn} \to \mathbf{B}G and BG connBINN(G)\mathbf{B}G_{conn} \to \mathbf{\flat} \mathbf{B}INN(G);

For connections on GG-principal \infty-bundles

For 𝔤\mathfrak{g} an ∞-Lie algebra or more generally an ∞-Lie algebroid and exp(𝔤)[CartSp op,sSet]\exp(\mathfrak{g}) \in [CartSp^{op},sSet] its untruncated Lie integration, the simplicial presheaf exp(𝔤) conn\exp(\mathfrak{g})_{conn} of ∞-Lie algebra valued differential forms is such that lifts \nabla

exp(𝔤) conn C(U) g exp(𝔤) X \array{ && \exp(\mathfrak{g})_{conn} \\ & {}^{\nabla}\nearrow & \downarrow \\ C(U) &\stackrel{g}{\to}& \exp(\mathfrak{g}) \\ \downarrow^{\mathrlap{\simeq}} \\ X }

of exp(𝔤)\exp(\mathfrak{g})-cocycles gg constitute a connection on an ∞-bundle on the principal ∞-bundle defined by gg:

exp(𝔤) connexp(𝔤) conn:(U,[k]){Ω vert (U×Δ n) A vert CE(𝔤) transitionfunction/Cechcocycle firstEhresmanncondition Ω (U×Δ n) A W(𝔤) connection secondEhresmanncondition Ω (U) F A inv(𝔤) curvaturecharacteristics}. \exp(\mathfrak{g})_{conn} \subset \exp(\mathfrak{g})_{conn'} : (U,[k]) \mapsto \left\{ \array{ \Omega^\bullet_{vert}(U \times \Delta^n) &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) &&& transition\;function\;/\;Cech\;cocycle \\ \uparrow && \uparrow &&&& first\;Ehresmann\;condition \\ \Omega^\bullet(U \times \Delta^n) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) &&& connection \\ \uparrow && \uparrow &&&& second\; Ehresmann\;condition \\ \Omega^\bullet(U) &\stackrel{\langle F_A\rangle}{\leftarrow}& inv(\mathfrak{g}) &&& curvature\;characteristics } \right\} \,.

For fixed UU \in CartSp and kΔk \in \Delta the sets on the right are sets of ∞-Lie algebra valued differential forms on U×Δ kU \times \Delta^k subject two conditions:

  1. restricted to the fibers the forms become flat and coincide with the forms that define the transition functions;

  2. their curvature characteristic forms F A\langle F_A \rangle descend to the base.

The subsheaf exp(𝔤) connexp(𝔤) conn\exp(\mathfrak{g})_{conn} \hookrightarrow \exp(\mathfrak{g})_{conn'} is that for every curvature form F AF_A has no component along the simplicial directions.

Here Ω (U×Δ k) vert\Omega^\bullet(U \times \Delta^k)_{vert} are the vertical differential forms on the trivial simplex bundle U×Δ kUU \times \Delta^k \to U and on the right we have the canonical sequence Chevalley-Eilenberg algebra \leftarrow Weil algebra \leftarrow invariant polynomials and all morphisms are dg-algebra morphisms.

CE(𝔤) G ChevalleyEilenbergalgebra W(𝔤) EG Weilalgebra inv(𝔤) BG algebraofinvariantpolynomials. \array{ CE(\mathfrak{g}) &&& G &&& Chevalley-Eilenberg\;algebra \\ \uparrow &&& \downarrow \\ W(\mathfrak{g}) &&& \mathbf{E}G &&& Weil\;algebra \\ \uparrow &&& \downarrow \\ inv(\mathfrak{g}) &&& \mathbf{B}G &&& algebra\;of\;invariant\;polynomials } \,.

A triple consisting of

is exhibited by a commuting diagram

CE(𝔤) μ CE(b k) cocycle W(𝔤) cs μ W(b k) ChernSimonselement inv(𝔤) μ inv(b k) invariantpolynomial \array{ CE(\mathfrak{g}) &\stackrel{\mu}{\leftarrow}& CE(b^{k} \mathbb{R}) &&& cocycle \\ \uparrow && \uparrow \\ W(\mathfrak{g}) &\stackrel{cs_\mu}{\leftarrow}& W(b^k \mathbb{R}) &&& Chern-Simons\;element \\ \uparrow && \uparrow \\ inv(\mathfrak{g}) &\stackrel{\langle -\rangle_\mu}{\leftarrow}& inv(b^k \mathbb{R}) &&& invariant\;polynomial }

in dgAlg.

The \infty-Chern-Weil homomorphism at this untruncated level is postcomposition with the lift of

exp(μ):exp(𝔤)exp(b n1) \exp(\mu) : \exp(\mathfrak{g}) \to \exp(b^{n-1}\mathbb{R})

to the map

exp(μ) conn:exp(𝔤) connexp(b n1) conn \exp(\mu)_{conn} : \exp(\mathfrak{g})_{conn} \to \exp(b^{n-1}\mathbb{R})_{conn}

given by forming the pasting composites

Ω (U×Δ n) vert A vert CE(𝔤) μ CE(b k) :μ(A vert) characteristicclass Ω (U×Δ n) A W(𝔤) cs μ W(b k) :cs μ(A) ChernSimonsform Ω (U) F A inv(𝔤) μ inv(b k) :F A μ curvaturecharacteristicform. \array{ \Omega^\bullet(U \times \Delta^n)_{vert} &\stackrel{A_{vert}}{\leftarrow}& CE(\mathfrak{g}) &\stackrel{\mu}{\leftarrow}& CE(b^k \mathbb{R}) & : \mu(A_{vert}) &&& characteristic\;class \\ \uparrow && \uparrow && \uparrow \\ \Omega^\bullet(U \times \Delta^n) &\stackrel{A}{\leftarrow}& W(\mathfrak{g}) &\stackrel{cs_\mu}{\leftarrow}& W(b^k \mathbb{R}) & : cs_\mu(A) &&& Chern-Simons\;form \\ \uparrow && \uparrow && \uparrow \\ \Omega^\bullet(U) &\stackrel{\langle F_A\rangle}{\leftarrow}& inv(\mathfrak{g}) &\stackrel{\langle -\rangle_\mu}{\leftarrow}& inv(b^k \mathbb{R}) & : \langle F_A\rangle_\mu &&& curvature\;characteristic\;form } \,.

This produces a b n1b^{n-1}\mathbb{R}-valued connections with local connection forms the Chern-Simons forms CS μ(A)CS_\mu(A) and with curvature the curvature characteristic form μ\langle - \rangle_\mu.

Under truncation exp(𝔤)τ nexp(𝔤)BG\exp(\mathfrak{g}) \to \tau_n \exp(\mathfrak{g}) \simeq \mathbf{B}G this decends under suitable conditions to the genuine refine \infty-Chern-Weil homomorphism

exp(μ) conn:BG conn=τ nexp(𝔤) conn(B n/Γ) conn \exp(\mu)_{conn} : \mathbf{B}G_{conn} = \tau_n \exp(\mathfrak{g})_{conn} \to (\mathbf{B}^n \mathbb{R}/\Gamma)_{conn}

that sends principal \infty-bundles with connection to circle n-bundles with connection.


The text of this entry is reproduced from the introduction of

A commented list of further related references is at


  1. There are actually two such forgetful functors, Γ\Gamma and Π\Pi. The first sends BG\mathbf{B}G to BG discB G_{disc}, which in topology is known as K(G,1)K(G,1). The other sends BG\mathbf{B}G to the classifying space BGB G. (see ∞-Lie groupoid – geometric realization). This distinction is effectively the origin of differential cohomology.

Revised on April 29, 2013 13:26:08 by Urs Schreiber (