Not signed in (Sign In)

Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

  • Sign in using OpenID

Discussion Tag Cloud

Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.

Welcome to nForum
If you want to take part in these discussions either sign in now (if you have an account), apply for one now (if you don't).
    • CommentRowNumber1.
    • CommentAuthorUrs
    • CommentTimeNov 19th 2014
    • (edited Nov 19th 2014)

    I’d like to formalize in cohesive homotopy theory the Penrose-Ward transform for higher bundles as it appears for 3-bundles in section 4 of

    • Christian Saemann, Martin Wolf, Six-Dimensional Superconformal Field Theories from Principal 3-Bundles over Twistor Space (arXiv:1305.4870)

    I think I know in principle how it goes. But at one step I need a fiberwise \flat-modality. This is something that Mike has kept asking about recently and I didn’t give an answer. I still don’t have the answer, but at least now I have a good motivating example.

    So we consider a correspondence

    Z p 1 p 2 X 1 X 2 \array{ && Z \\ & {}^{\mathllap{p_1}}\swarrow && \searrow^{\mathrlap{p_2}} \\ X_1 && && X_2 }

    Let GG be an \infty-group and consider a GG-principal \infty-bundle PX 1P \to X_1 equipped with a trivialization of its pullback along p 1p_1.

    The transform in question is supposed to do the following: from the chosen trivialization on ZZ we are to produce a p 1p_1-fiberwise GG-valued function on ZZ. Associated with this is a p 1p_1-fiberwise Maurer-Cartan form on ZZ which glues to a global form on ZZ and this we are to push down along p 2p_2. (This is an abstract rephrasing of what appears in components in section 4 of the above article.)

    Here I’ll ignore the push-forward for the moment and consider only the first step of producing that fiberwise Maurer-Cartan form.

    On a single fiber x:*Xx \colon \ast \to X it works like this: let gg be the map that modulates PX 1P \to X_1. Then the assumed trivialization of the pullback of this GG-bundle to ZZ gives the diagram

    p 1 1(x) Z * * x X 1 g BG \array{ p_1^{-1}(x) &\longrightarrow& Z &\longrightarrow& \ast \\ \downarrow && \downarrow && \downarrow \\ \ast &\stackrel{x}{\longrightarrow}& X_1 &\stackrel{g}{\longrightarrow}& \mathbf{B}G }

    and, by looping, this induces a map p 1 1(x)Gp_1^{-1}(x) \longrightarrow G. This is the GG-valued function on the fiber. Its fiber MC form is given by the composite

    p 1 1(x)G dRBG. p_1^{-1}(x) \longrightarrow G \longrightarrow \flat_{dR} \mathbf{B}G \,.

    Now we need to generalize this to a construction that does this for all fibers at once.

    Need to think about that.

    • CommentRowNumber2.
    • CommentAuthorDavid_Corfield
    • CommentTimeNov 19th 2014

    But at one step I need a fiberwise \flat-modality.

    Is this the same issue as the one in modal logic I mentioned, where Mike here agreed that

    This is indeed very much the same as the issue that Urs and I had with \flat, which we also weren’t able to internalize or apply in a nonempty context.

    • CommentRowNumber3.
    • CommentAuthorUrs
    • CommentTimeNov 19th 2014

    Yes!

    • CommentRowNumber4.
    • CommentAuthorMike Shulman
    • CommentTimeNov 19th 2014

    What does it mean for a GG-valued function on ZZ to be “p 1p_1-fiberwise”? And can you do the step of producing that function for all fibers at once, before the \flat gets introduced to make the MC form?

    • CommentRowNumber5.
    • CommentAuthorUrs
    • CommentTimeNov 19th 2014
    • (edited Nov 19th 2014)

    Here are some further thoughts on these questions. I’d think the following should give the right construction for the application described in #1. Whether it contains the seed of an answer to what the fiberwise \flat should be I don’t know, but see below.

    First, I think the group object in H /BG\mathbf{H}_{/\mathbf{B}G} in question is G/ adGG/_{ad}G sitting (in H\mathbf{H}) in the homotopy fiber sequence

    G G/ adG BG \array{ G &\longrightarrow& G/_{ad}G \\ && \downarrow \\ && \mathbf{B}G }

    exhibiting the adjoint action of GG on itself. This is defined to be the free loop space object

    G/ adG[Π(S 1),BG] G/_{ad}G \coloneqq [\Pi(S^1), \mathbf{B}G]

    and this makes manifest the \infty-group structure in the slice over BG\mathbf{B}G.

    A trivialization of a GG-principal bundle is in particular a trivialization of its adjoint bundle, and so we may consider generalizing the diagram in #1 to

    Z G/ adG X 1 g BGZ γ P× G adG G/ adG p 1 p 1 X 1 = X 1 g BG \array{ Z &\longrightarrow& G/_{ad} G \\ \downarrow &\swArrow& \downarrow \\ X_1 &\stackrel{g}{\longrightarrow}& \mathbf{B}G } \;\;\; \simeq \;\;\; \array{ Z &\stackrel{\gamma}{\longrightarrow}& P \times^{ad}_G G &\longrightarrow& G/_{ad} G \\ \downarrow^{\mathrlap{p_1}} &\swArrow& \downarrow^{\mathrlap{p_1}} &\swArrow& \downarrow \\ X_1 &\stackrel{=}{\longrightarrow}& X_1 &\stackrel{g}{\longrightarrow}& \mathbf{B}G }

    The adjoint bundle P× G adGP \times^{ad}_G G is the relevant group object in the slice, and γ\gamma is what I had announced as the “p 1p_1-fiberwise function” with values in GG (not claiming that this is necessarily the best terminology).

    Now to get the fiberwise MC form, I’d think the obvious step is the following:

    define ( dRBG)/ adG(\flat_{dR} \mathbf{B}G)/_{ad} G to be the homotopy pushout in

    G G/ adG ( dRBG) ( dRBG)/ adG. \array{ G &\longrightarrow& G/_{ad}G \\ \downarrow && \downarrow \\ (\flat_{dR} \mathbf{B}G) &\longrightarrow& (\flat_{dR} \mathbf{B}G)/_{ad} G } \,.

    One checks, using homotopy pullback stability of homotopy colimits in H\mathbf{H}, that the notation is not just intuitively but also formally justified in that there is a homotopy fiber sequence

    ( dRBG) ( dRBG)/ adG BG \array{ (\flat_{dR} \mathbf{B}G) &\longrightarrow& (\flat_{dR}\mathbf{B}G)/_{ad} G \\ && \downarrow \\ && \mathbf{B}G }

    and a compatible morphism in the slice

    G/ adG ( dRBG)/ adG BG \array{ G/_{ad}G && \longrightarrow && (\flat_{dR} \mathbf{B}G)/_{ad} G \\ & \searrow && \swarrow \\ && \mathbf{B}G }

    which says that the gloabl MC form θ:G dRBG\theta \colon G \to \flat_{dR}\mathbf{B}G is equivariant under the adjoint GG-action, just as it should be.

    That is then, I’d think, the answer to the application in #1: the fiberwise MC form in question is γ *θ\gamma^\ast \theta given by the diagram

    Z G/ adG θ ( dRBG)/ adG X 1 g BG id BGZ γ *θ P× G ad( dRBG) ( dRBG)/ adG p 1 p 1 X 1 = X 1 g BG. \array{ Z &\longrightarrow& G/_{ad} G &\stackrel{\theta}{\longrightarrow}& (\flat_{dR}\mathbf{B}G)/_{ad} G \\ \downarrow &\swArrow& \downarrow &\swArrow& \downarrow \\ X_1 &\stackrel{g}{\longrightarrow}& \mathbf{B}G &\stackrel{id}{\longrightarrow}& \mathbf{B}G } \;\;\; \simeq \;\;\; \array{ Z &\stackrel{\gamma^\ast \theta}{\longrightarrow}& P \times^{ad}_G (\flat_{dR}\mathbf{B}G) &\longrightarrow& (\flat_{dR} \mathbf{B}G)/_{ad} G \\ \downarrow^{\mathrlap{p_1}} &\swArrow& \downarrow^{\mathrlap{p_1}} &\swArrow& \downarrow \\ X_1 &\stackrel{=}{\longrightarrow}& X_1 &\stackrel{g}{\longrightarrow}& \mathbf{B}G } \,.

    In conclusion, it seems the above tells us what the fiberwise dR\flat_{dR} should be in the special cases that the dependent types are GG-principal bundles. I have no thoughts yet on how to proceed more generally.

    In closing, one remark on how \flat itself appears in this context: notice that BG\flat \mathbf{B}G is also of the form

    BG( dRBG)/G \flat \mathbf{B}G \simeq (\flat_{dR}\mathbf{B}G)/G

    in that it sits (by definition of dR\flat_{dR} !) in a homotopy fiber sequence

    dRBG BG BG. \array{ \flat_{dR} \mathbf{B}G &\longrightarrow& \flat \mathbf{B}G \\ && \downarrow \\ && \mathbf{B}G } \,.

    The action her is the gauge action on flat differentialforms Ag 1dg+g 1AgA \mapsto g^{-1}\mathbf{d}g + g^{-1} A g (i.e. that’s what it is in the standard models, and that’s what we ought to take as the “meaning interpretation” in generality).

    I realize that I am unsure about how BG\flat \mathbf{B}G and ( dRBG)/ adG(\flat_{dR} \mathbf{B}G)/_{ad} G relate. Are they equivalent?!

    • CommentRowNumber6.
    • CommentAuthorMike Shulman
    • CommentTimeNov 20th 2014

    In trying to understand this, I noticed that the link “This statement and its proof is spelled out here” at Maurer-Cartan form needs updating; to what I’m not quite sure.

    Is the morphism on the LHS of your pushout the ordinary MC form? And is the morphism on the RHS of the pushout the same as your “compatible morphism in the slice”?

    • CommentRowNumber7.
    • CommentAuthorMike Shulman
    • CommentTimeNov 20th 2014

    It seems that this construction can be done in much more generality. Given any map f:ABf:A\to B of pointed types (such as BGBG\flat \mathbf{B} G \to \mathbf{B} G), we can consider the pushout of F=fib(f)F = fib(f) and the free loop space LBL B over ΩB\Omega B. The same stability argument should show that we have a fiber sequence FF+ ΩBLBBF \to F +_{\Omega B} L B \to B alongside the obvious one FABF\to A\to B, and we might wonder about the relationship of F+ ΩBLBF+_{\Omega B} L B to AA.

    If BB is connected (like BG\mathbf{B} G) and if we had a map F+ ΩBLBAF +_{\Omega B} L B \to A over BB and under FF, then the five-lemma ought to imply that this map would be an equivalence. There is an obvious map F+ ΩBLBAF +_{\Omega B} L B \to A under FF that is null on LBL B, but it doesn’t commute with the maps to BB.

    Are you sure about that stability-of-colimits argument? It seems right to me too offhand, but the end result is suspicious; it feels like we have a group GG acting on a space FF, and we have some way of constructing some other GG-action that “feels free” relative to just remembering the images of the basepoint of FF under the original action, so why should the new action also be on FF rather than on some bigger space?

    • CommentRowNumber8.
    • CommentAuthorMike Shulman
    • CommentTimeNov 20th 2014

    (Where by “this construction” I meant the construction of dRBG/ adG\flat_{dR} \mathbf{B} G/_{ad} G.)

    • CommentRowNumber9.
    • CommentAuthorMike Shulman
    • CommentTimeNov 20th 2014

    Wait, now I don’t believe the stability argument any more. At least, the one I had in mind will I think only show that the fiber of F+ ΩBLBBF+_{\Omega B} L B\to B is the pushout of ΩB\Omega B and F×ΩBF\times \Omega B along ΩB×ΩB\Omega B \times \Omega B, so the fiber of dRBG/ adG\flat_{dR}\mathbf{B} G/_{ad} G would be the pushout of GG and dRBG×G\flat_{dR}\mathbf{B} G \times G along G×GG\times G. But maybe you had a different argument in mind, could you spell it out?

    • CommentRowNumber10.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014

    I have fixed some of the links, thanks for the alert. The general cohesive defininition is at cohesive infinity-topos – structures – Maurer-Cartan forms and curvature characteristic forms, the smooth discussion is at smooth infinity-groupoid – structures – Maurer-Cartan forms and curvature characteristic forms.

    (These links broke when I split off the “…structures” entries. Apparently more such links are broken, will need to fix them eventually.)

    • CommentRowNumber11.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014
    • (edited Nov 20th 2014)

    [I see I am overlapping with your comments. I’ll retain what I typed, but the first bit you just reproduced in #9.]

    Let’s see. So we want to compute the homotopy fiber of the bottom vertical map in

    G G/ adG dRBG Q p BG \array{ G &\longrightarrow& G/_{ad}G \\ \downarrow && \downarrow \\ \flat_{dR} \mathbf{B}G &\longrightarrow& Q \\ && \downarrow^{\mathrlap{p}} \\ && \mathbf{B}G }

    where QQ is the colimit under the top left span diagram. The separate pullbacks are clearly these

    G×G G G× dRBG fib(p) * \array{ G \times G &\longrightarrow& G \\ \downarrow && \downarrow \\ G \times \flat_{dR} \mathbf{B}G &\longrightarrow& fib(p) \\ && \downarrow \\ && \ast }

    The point to be careful about is what the top map is. Chasing through the diagrams (which I won’t draw here), it is projection on the second factor. So we may write the span as the pointwise product of two spans

    G×G (*,id) *×G (id,θ) G× dRBG fib(p) * \array{ G \times G &\stackrel{(\ast,id)}{\longrightarrow}& \ast \times G \\ \downarrow^{\mathrlap{(id,\theta)}} && \downarrow \\ G \times \flat_{dR} \mathbf{B}G &\longrightarrow& fib(p) \\ && \downarrow \\ && \ast }

    and then all maps in the span are product maps.

    [ okay, it seems we agree to this point. Maybe there is disagreement about the next step, where I am claiming that: ]

    From this I conclude that fib(p) dRBGfib(p) \simeq \flat_{dR} \mathbf{B}G.

    Let me check my reasoning…

    • CommentRowNumber12.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014

    From #6 (probably clear by now):

    Is the morphism on the LHS of your pushout the ordinary MC form?

    Yes.

    • CommentRowNumber13.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014
    • (edited Nov 20th 2014)

    re #7, thanks for this comment here:

    If BB is connected (like BG\mathbf{B} G) and if we had a map F+ ΩBLBAF +_{\Omega B} L B \to A over BB and under FF, then the five-lemma ought to imply that this map would be an equivalence. There is an obvious map F+ ΩBLBAF +_{\Omega B} L B \to A under FF that is null on LBL B, but it doesn’t commute with the maps to BB.

    I had been wondering about this. Thanks for mentioning the use of the five-lemma here. My problem was that I was unable (unable, that is, while sitting on a RER B train from Paris Charles-de-Gaulle to downtown) to produce the sensible map from G/ adGG/_{ad}G to BG\flat \mathbf{B}G.

    • CommentRowNumber14.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014

    re #9, #11

    right, I suppose my last step there lacks justification. Hm.

    But I’ll have to call it quits now. Very late here.

    • CommentRowNumber15.
    • CommentAuthorMike Shulman
    • CommentTimeNov 20th 2014

    Here is a naive question: since dRBG\flat_{dR}\mathbf{B}G is a fiber of a map to BG\mathbf{B}G, it automatically comes with a “tautological” action by GG. You seem to be saying that this action is not (or not obviously) the adjoint action, since you’re trying to exhibit dRBG\flat_{dR}\mathbf{B}G as a fiber of a different map to BG\mathbf{B}G and calling that the adjoint action. What is this tautological action then? I would have thought that the adjoint action is the only available action of GG on dRBG\flat_{dR}\mathbf{B}G.

    • CommentRowNumber16.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014
    • (edited Nov 20th 2014)

    Yes, I am after just that canonical action, but I didn’t see how to produce a map

    G/ adGBG=( dRBG)/G. G/_{ad} G \longrightarrow \flat \mathbf{B}G = (\flat_{dR}\mathbf{B}G)/G \,.

    This map exists in the models at least for ordinary groups, but I didn’t see how to producte it abstractly. So instead I tried to force such kind of map. But my attempt via the pushout was no good, as you pointed out.

    I am thinking of another way to do it, but not sure how to make that work:

    Assume that GG is at least braided, i.e. that it is Ω 2B 2G\Omega^2 \mathbf{B}^2 G. Then consider the map of cospan diagrams

    BG Δ BG×BG Δ BG BG (Bθ,Bθ) B( dRBG)×B( dRBG) (Bθ,Bθ) BG \array{ \mathbf{B}G &\stackrel{\Delta}{\longrightarrow}& \mathbf{B}G \times \mathbf{B}G &\stackrel{\Delta}{\longleftarrow}& \mathbf{B}G \\ \downarrow^{} && \downarrow && \downarrow \\ \mathbf{B}G &\stackrel{(\mathbf{B}\theta, \mathbf{B}\theta)}{\longrightarrow}& \mathbf{B} (\flat_{dR}\mathbf{B}G) \times \mathbf{B} (\flat_{dR}\mathbf{B}G) & \stackrel{(\mathbf{B}\theta, \mathbf{B}\theta)}{\longleftarrow} & \mathbf{B}G }

    Forming pullbacks gives a map G/ adGsomewhereG/_{ad}G \to somewhere over BG\mathbf{B}G, where “somewhere” knows something about a GG-action on dRBG\flat_{dR}\mathbf{B}G.

    If we could go from there to

    BG Bθ B( dRBG) * \array{ \mathbf{B}G &\stackrel{\mathbf{B}\theta}{\longrightarrow}& \mathbf{B} (\flat_{dR}\mathbf{B}G) & \stackrel{}{\longleftarrow} & \ast }

    we’d be done, as the limit over that last diagram is BG\flat \mathbf{B}G.

    Hm.

    • CommentRowNumber17.
    • CommentAuthorUrs
    • CommentTimeNov 21st 2014
    • (edited Nov 21st 2014)

    Quick remark before I need to rush off:

    that diagram in #16 gives a map from G/ adGG/_{ad}G to what deserves to be written G\( dRBG× dRBG)/GG \backslash (\flat_{dR}\mathbf{B}G\times \flat_{dR}\mathbf{B}G)/G, hence an adjoint-equivariant map from GG to ( dRBG× dRBG)/G(\flat_{dR}\mathbf{B}G\times \flat_{dR}\mathbf{B}G)/G.

    In case GG is 0-truncated and in the standard models, then that GG-action on flat forms is transitive and so in that case ( dRBG× dRBG)/G dRBG(\flat_{dR}\mathbf{B}G\times \flat_{dR}\mathbf{B}G)/G \simeq \flat_{dR}\mathbf{B}G. In general I suppose this equivalence won’t hold. But maybe that extra quotient construction then is the next closest thing that exists in general.

    • CommentRowNumber18.
    • CommentAuthorMike Shulman
    • CommentTimeNov 21st 2014

    What are the two actions of GG on dRBG× dRBG\flat_{dR}\mathbf{B}G\times \flat_{dR}\mathbf{B}G?

    • CommentRowNumber19.
    • CommentAuthorMike Shulman
    • CommentTimeNov 21st 2014

    In case GG is 0-truncated and in the standard models

    Do you mean by this “for a 0-truncated GG in the standard models”? Do you know whether the map G/ adGBGG/_{ad} G \to \flat \mathbf{B}G exists in the standard models when GG is not 0-truncated?

    • CommentRowNumber20.
    • CommentAuthorUrs
    • CommentTimeNov 21st 2014
    • (edited Nov 21st 2014)

    What are the two actions of GG on dRBG× dRBG\flat_{dR}\mathbf{B}G\times \flat_{dR}\mathbf{B}G?

    So generally for GHG\to H a group homomorphism then

    H * H/G BG * BH \array{ H &\longrightarrow& \ast \\ \downarrow && \downarrow \\ H/G &\longrightarrow& \mathbf{B}G \\ \downarrow && \downarrow \\ \ast &\longrightarrow& \mathbf{B}H }

    exhibits the induced GG-action on HH.

    Here the group homomorphism in question is now G dRBGG \longrightarrow \flat_{dR}\mathbf{B}G (since I am assuming GG to be braided). The action in question is hence via that homomorphism, or rather its diagonal.

    (( dRBG)×( dRBG))/G G\(( dRBG)×( dRBG))/G BG * BG B( dRBG)×B( dRBG) \array{ ((\flat_{dR} \mathbf{B}G) \times (\flat_{dR} \mathbf{B}G))/G &\longrightarrow& G \backslash ((\flat_{dR} \mathbf{B}G) \times (\flat_{dR} \mathbf{B}G))/G &\longrightarrow& \mathbf{B}G \\ \downarrow && \downarrow && \downarrow \\ \ast &\longrightarrow& \mathbf{B}G &\longrightarrow& \mathbf{B}(\flat_{dR} \mathbf{B}G)\times \mathbf{B}(\flat_{dR} \mathbf{B}G) }

    Do you mean by this “for a 0-truncated G in the standard models”?

    Yes

    Do you know whether the map G/ adGBGG/_{ad} G \to \flat \mathbf{B}G exists in the standard models when GG is not 0-truncated?

    I don’t even know that, unfortunately.

    • CommentRowNumber21.
    • CommentAuthorMike Shulman
    • CommentTimeNov 22nd 2014

    …that GG-action on flat forms is transitive and so in that case ( dRBG× dRBG)/G dRBG(\flat_{dR}\mathbf{B}G\times \flat_{dR}\mathbf{B}G)/G \simeq \flat_{dR}\mathbf{B}G

    Doesn’t that require the action to be free as well as transitive?

    • CommentRowNumber22.
    • CommentAuthorMike Shulman
    • CommentTimeNov 22nd 2014

    Wait a minute, I am confused. G/ adGG/_{ad}G is the free loop space LBGL\mathbf{B}G. But the free loop space of any object LXXL X \to X always has a section, assigning to each x:Xx:X the constant path. So if there is a map G/ adGBGG/_{ad}G \to \flat \mathbf{B}G over BG\mathbf{B} G, then BGBG\flat \mathbf{B}G \to \mathbf{B}G also has a section; but I don’t think that can happen unless BG\mathbf{B}G is discrete.

    • CommentRowNumber23.
    • CommentAuthorUrs
    • CommentTimeNov 24th 2014
    • (edited Nov 24th 2014)

    Doesn’t that require the action to be free as well as transitive?

    That’s true, right. In fact the part acting non-freely is just G\flat G again, and this is what gives ( dRBG)/G*/GBG(\flat_{dR}\mathbf{B}G)/G \simeq \ast /\flat G \simeq \flat \mathbf{B}G. So I was wrong in the last equivalence in #17. Instead of the quotient of the diagonal action being one copy of dRBG\flat_{dR}\mathbf{B}G it’s more something like the product of that with BG\flat \mathbf{B}G.

    I don’t think that can happen unless BG is discrete.

    Right, that’s also true. In components the reason is that the conjugation action on GG goes to the inhomogeneous transformation Ag 1Ag+g 1dgA \mapsto g^{-1}A g + g^{-1}d g on differential forms, which prevents the 0-form from being the invariant section.

    So I need to think more about all this…

    • CommentRowNumber24.
    • CommentAuthorMike Shulman
    • CommentTimeNov 24th 2014

    the conjugation action on GG goes to the inhomogeneous transformation Ag 1Ag+g 1dgA \mapsto g^{-1}A g + g^{-1}d g

    Ah, right. So that means that we don’t want the “obvious” adjoint action on dRBG\flat_{dR}\mathbf{B}G, we want this sort of “twisted” one. So we do need some other fibration over BG\mathbf{B} G whose fiber is dRBG\flat_{dR}\mathbf{B}G and which admits a map from G/ adGG/_{ad}G. Is there a way to say “g 1dgg^{-1} dg” synthetically?

    • CommentRowNumber25.
    • CommentAuthorUrs
    • CommentTimeNov 24th 2014

    Is there a way to say “g 1dgg^{-1} dg” synthetically?

    Asked this way, the answer would be: sure, this is again just the MC form g *θg^\ast \theta, i.e. the image of g under θ:G dRBG\theta \colon G \to \flat_{dR}\mathbf{B}G.

    Something tautological or other needs to be done here to make all this fall into place…

    • CommentRowNumber26.
    • CommentAuthorUrs
    • CommentTimeNov 24th 2014
    • (edited Nov 24th 2014)

    Maybe we don’t actually want the fiber to be dRBG\flat_{dR}\mathbf{B}G. Maybe the variant ( dRBG× dRBG)/ diagG(\flat_{dR}\mathbf{B}G\times \flat_{dR}\mathbf{B}G)/_{diag} G is actually “right”.

    In the motivating example of #1 there is in fact an extra condition: the MC forms in arXiv:1305.4870 are produced patchwise in the fibers in (4.13); but then above (4.15) and above (4.17) further cohomological conditions are used to produce, in (4.18), (4.23) (4.24), an actual fiberwise MC form.

    • CommentRowNumber27.
    • CommentAuthorUrs
    • CommentTimeNov 25th 2014
    • (edited Nov 25th 2014)

    Ah, I think I have the answer. It’s super-easy. But my insistence above on having a group object in the slice was misled.

    Instead of using the adjoint action on GG, we should use the right action. That’s just

    G * BG \array{ G &\longrightarrow& \ast \\ && \downarrow \\ && \mathbf{B}G }

    from there we want to map to

    dRBG BG BG \array{ \flat_{dR}\mathbf{B}G &\longrightarrow& \flat \mathbf{B}G \\ && \downarrow \\ && \mathbf{B}G }

    Well, that’s easy, there is actually an essentially unique map

    * α BG BG \array{ \ast && \stackrel{\alpha}{\longrightarrow} && \flat\mathbf{B}G \\ & \searrow && \swarrow \\ && \mathbf{B}G }

    Now while that does certainly induce a map G dRBGG\to \flat_{dR}\mathbf{B}G on the homotopy fibers, one tends to worry that this will be the trivial map. But, no, it is the MC form. This follows from the pasting law (and it will be trivial to you, but let me spell it out anyway):

    the morphism on homotopy fibers which is induced from the previous diagram is that from the pullback diagram

    G * * BG \array{ G &\longrightarrow& \ast \\ \downarrow && \downarrow \\ \ast & \longrightarrow& \mathbf{B}G }

    to the pullback diagram

    dRBG BG * BG \array{ \flat_{dR}\mathbf{B}G &\longrightarrow& \flat \mathbf{B}G \\ \downarrow && \downarrow \\ \ast &\longrightarrow & \mathbf{B}G }

    via the universal map ϕ\phi in

    G * ϕ α dRBG BG * BG \array{ G &\longrightarrow& \ast \\ \downarrow^{\mathrlap{\phi}} && \downarrow^{\mathrlap{\alpha}} \\ \flat_{dR}\mathbf{B}G &\longrightarrow& \flat\mathbf{B}G \\ \downarrow && \downarrow \\ \ast &\longrightarrow& \mathbf{B}G }

    Here the bottom square and the total rectangle are homotopy pullbacks by assumption. Hence so is the top square. Hence

    ϕθ \phi \simeq \theta

    is indeed the MC form.

    • CommentRowNumber28.
    • CommentAuthorUrs
    • CommentTimeNov 25th 2014

    And so to come back to #1, the answer is that given a GG-principal \infty-bundle on X 1X_1 (classified by some g:X 1BGg \colon X_1\to \mathbf{B}G) with a trivialization of its pullback to some correspondence space ZX 1Z \to X_1, then this induces the fiberwise flat 𝔤\mathfrak{g}-valued differential form that is denoted x:X 1θ x\underset{x \colon X_1}{\sum} \theta_x in this pasting diagram of homotopy pullbacks:

    G Z * θ x x:X 1θ x θ/G dRBG BG * x X 1 g BG \array{ G && \longrightarrow && Z && \longrightarrow && \ast \\ & \searrow^{\mathrlap{\theta_x}} && && \searrow^{\mathrlap{\underset{x\colon X_1}{\sum} \theta_x}} && && \searrow^{\mathrlap{\theta/G}} \\ \downarrow && \flat_{dR}\mathbf{B}G && \longrightarrow && && \longrightarrow && \flat \mathbf{B}G \\ & \swarrow && && \swarrow && && \swarrow \\ \ast && \stackrel{x}{\longrightarrow} && X_1 && \stackrel{g}{\longrightarrow} && \mathbf{B}G }

    Or rather, this is half the answer for #1. Next I need to sum up these fiberwise flat forms to a single non-flat form on the total space…

    • CommentRowNumber29.
    • CommentAuthorMike Shulman
    • CommentTimeNov 25th 2014

    Re: #27, that’s basically just the definition of ϕ\phi, right? I actually thought of that, but you really seemed to want the adjoint action on GG so I assumed it wasn’t right. (-:

    • CommentRowNumber30.
    • CommentAuthorUrs
    • CommentTimeNov 25th 2014
    • (edited Nov 25th 2014)

    Yeah, sorry, I was misled. For the absolute MC form I needed a group object, so I thought I should look for one in the slice, too. My bad, sorry for the wild goose chase.

    • CommentRowNumber31.
    • CommentAuthorMike Shulman
    • CommentTimeNov 25th 2014

    No problem, I learned some things. (-:

    • CommentRowNumber32.
    • CommentAuthorDavidRoberts
    • CommentTimeNov 26th 2014

    @Urs, what’s missing in the middle of the diagram in #28?

    • CommentRowNumber33.
    • CommentAuthorUrs
    • CommentTimeNov 26th 2014

    Whatever name you want to give to the pullback given by the bottom right square.

    • CommentRowNumber34.
    • CommentAuthorDavidRoberts
    • CommentTimeNov 26th 2014

    Ah, I see. I thought it might be something that already had a name.

    • CommentRowNumber35.
    • CommentAuthorUrs
    • CommentTimeNov 26th 2014

    Right, I should have said more about it, sorry. So let’s write PP for the GG-principal \infty-bundle which is modulated by that map g:X 1BGg\colon X_1 \to \mathbf{B}G. Then one name of that thing without name is the associated dRBG\flat_{dR}\mathbf{B}G-bundle

    P×G( dRBG) P \underset{G}{\times} (\flat_{dR}\mathbf{B}G)

    Moreover, a more evocative notaton for dRBG\flat_{dR}\mathbf{B}G is Ω flat(,𝔤)\Omega_{flat}(-,\mathfrak{g}), the sheaf of L L_\infty-algebra valued differential forms, so this is

    P×GΩ flat(,𝔤) P \underset{G}{\times} \Omega_{flat}(-,\mathfrak{g})

    and that makes most manifest how this is a fiberwise varying collection of such differential form coefficients. One might consider the étalification of this associated bundle to a concrete bundle, and that would be then something like a 𝔤\mathfrak{g}-valued cotangent bundle of X 1X_1.

    • CommentRowNumber36.
    • CommentAuthorDavidRoberts
    • CommentTimeNov 26th 2014
    • (edited Nov 26th 2014)

    Or sections of the adjoint bundle P×_G g, such as appears in the left of the algebroid Atiyah sequence? But you said that we shouldn’t be thinking of the adjoint action (though I may be thinking of the wrong bit, I can’t read back over the thread to unravel it now)

    • CommentRowNumber37.
    • CommentAuthorUrs
    • CommentTimeNov 26th 2014

    Right, the action is that by gauge transformations. If GG is a Lie 1-group, then

    gG(U)=C (U,G)g \in G(U) = C^\infty(U,G) takes AΩ flat(U,𝔤)A \in \Omega_{flat}(U,\mathfrak{g}) to

    Ag 1Ag+g 1dg. A \mapsto g^{-1}A g + g^{-1} d g \,.

    In the dcct pdf this is prop. 5.4.44, see also example 3.9.37.

    (And the statement about the GG-equivariance of θ\theta is now also on the nnLab here).

    • CommentRowNumber38.
    • CommentAuthorUrs
    • CommentTimeNov 28th 2014
    • (edited Nov 28th 2014)

    I may still be confused. We have the construction in #28

    G Z * θ x x:X 1θ x θ/G dRBG P×G( dRBG) BG * x X 1 g BG \array{ G && \longrightarrow && Z && \longrightarrow && \ast \\ & \searrow^{\mathrlap{\theta_x}} && && \searrow^{\mathrlap{\underset{x\colon X_1}{\sum} \theta_x}} && && \searrow^{\mathrlap{\theta/G}} \\ \downarrow && \flat_{dR}\mathbf{B}G && \longrightarrow && P \underset{G}{\times}(\flat_{dR}\mathbf{B}G) && \longrightarrow && \flat \mathbf{B}G \\ & \swarrow && && \swarrow && && \swarrow \\ \ast && \stackrel{x}{\longrightarrow} && X_1 && \stackrel{g}{\longrightarrow} && \mathbf{B}G }

    and my next task would be to find from these fiberwise flat differential forms a globally defined differential form on ZZ. From looking at the model one expects this not to be necessarily flat anymore, as the horizontal part of the differential will see how the fiberwise flat forms change as one moves along the fibers. My expectation was that I’d have to first pass from fiberwise flat to fiberwise general 𝔤\mathfrak{g}-valued forms Ω(,𝔤)\Omega(-,\mathfrak{g}) and then observe the vanishing of some cocycle in order to find a section of a trivial Ω(,𝔤)\Omega(-,\mathfrak{g})-bundle over ZZ which would be the globally defined form in question.

    But what confuses me now is that I seem to get all this for free and for flat forms: namely the map x:X 1θ x\sum_{x\colon X_1}\theta_x in the above diagram is equivalently a section of the pullback of the bundle P×G( dRBG)P \underset{G}{\times}(\flat_{dR}\mathbf{B}G) to ZZ. But by the commutativity of the diagram, that is equivalently the pullback to BG\flat \mathbf{B}G along Z*BGZ\to \ast \to \mathbf{B}G, which is the trivial flat form bundle Z×( dRBG)Z \times (\flat_{dR}\mathbf{B}G). Hence the above gives a section of that hence a flat form

    Z dRBG Z \to \flat_{dR}\mathbf{B}G

    That worries me a bit. It would be good, but somehow it seems too good to be true. Maybe it is an indication that I am missing something and my intended formalization of #1 still does not work.