Skip to the Main Content

Note:These pages make extensive use of the latest XHTML and CSS Standards. They ought to look great in any standards-compliant modern browser. Unfortunately, they will probably look horrible in older browsers, like Netscape 4.x and IE 4.x. Moreover, many posts use MathML, which is, currently only supported in Mozilla. My best suggestion (and you will thank me when surfing an ever-increasing number of sites on the web which have been crafted to use the new standards) is to upgrade to the latest version of your browser. If that's not possible, consider moving to the Standards-compliant and open-source Mozilla browser.

July 19, 2006

2-Palatini

Posted by Urs Schreiber

A few entries ago, I was claiming that other people are implicitly claiming that the field content of D=11 supergravity encodes precisely a 3-connection taking values in a certain Lie 3-algebra (). In my first attempt to make a couple of remarks on that, I ran out of time (). Here is the second attempt.

There are in particular two questions which I would like to comment on.

1) What about the 3-form really being a Chern-Simons 3-form?

2) What happens in degree 2?


1) What about the 3-form really being a Chern-Simons 3-form? We know that the 3-form C should really come from Chern-Simons 3-forms of a Lorentz and an E 8 connection. This information is not present in the classical FDA formulation of supergravity. Here is a general observation on connections on Chern-Simons 2-gerbes, which might be relevant.
Consider a Lie group G. Its Lie algebra is encoded in the fda defined by

(1)da a+12C a bca ba c=0,

where {C a bc} are the structure constant in some chosen basis. Nilpotency of d is equivalent to the Jacobi identity. Next, consider the crossed module GG. The fda corresponding to its Lie 2-algebra is given by

(2) da a+12C a bca ba c+b a=0 db a+C a bca bb c=0.

A 2-connection with values in this Lie 2-algebra is given by a 1-form A a and a 2-form B a and has curvatures

(3)F 1 =F A+B F 2 =d AB.

Without mentioning anything like 2-connections, but implicitly considering precisely this, such 2-connections have been studied for instance in hep-th/0204059. The 2-group aspect is discussed very nicely in hep-th/0206130. Suppose we demand this 2-curvature to vanish. This is equivalent to

(4)B=F A.

Hence a flat (GG)-2-connection is the same as an ordinary G-connection. In fact, a trivial flat (GG)-2-bundle is the same as an ordinary G-bundle. Next, let’s also add a generator in degree 3 to our fda, to get the Lie 3-algebra encoded by

(5) da a+12C a bca ba c+b a=0 db a+C a bca bb c=0 dc+k abb ab b=0,

where k ab is proportional to the Killing form on the Lie algebra of G. This ensures that d is still nilpotent. A 3-connection with values in this Lie 3-algebra is a 1-form A a, a 2-form B a and a 3-form C. Its curvature is

(6)F 1 =F A+B F 2 =d AB F 3 =dC+tr(BB).

Assume again that the curvature vanishes. In the first two degrees this is, as before, equivalent to B=F A. In third degree it now says that

(7)dCtr(F AF A).

But this means that C itself must be, up to a closed part, the Chern-Simons 3-form of A

(8)CCS(A).

Therefore a 3-connection of the above sort is the local connection of a Chern-Simons gerbe corresponding to some G-bundle with connection. Now take G to be the product of E 8 with the Lorentz group in D=11 and there we go. There is one problem. The known FDA description of SUGRA uses not the Lorentz group in lowest degree, but the Poincaré group. Plus, there is already something going on in degree 3. Both facts imply that applying the above construction to this case is not enirely straightforward. Partly motivated by this, in the next subsection I make some comments on the Poincaré versus Lorentz issue.


2) What happens in degree 2?

The Poincaré-group happens to be a semidirect product of the Lorentz and the translation group, with one factor being abelian. Every such semidirect product group is already naturally a 2-group. It’s Lie 2-algebra corresponds to the fda defined by

(9)da a+12C a bca ba c=0 dt i+α i aja at j=0,

where {a a} are a basis of the dual of the Lorentz Lie algebra, {t i} a basis of the dual of the translation algebra, and α encodes the action of one on the other. Motivated by this observation, there have been attempts to formulate gravity in terms of the Poincaré 2-group (). But notice this: In the above fda, the t i live in degree 2. Hence a 2-connection with values in this guy contains a 2-form valued in the translation Lie algebra. What’s that? On the other hand, in the known fda formulation of sugra, one regards the t i as living in degree 1. Accordingly, the connection with values in this fda has a 1-form with values in the translation Lie algebra. This 1-form gets interpreted as the vielbein, which makes the fda SUGRA formulation something like an example for a Palatini-formulation of gravity (). Still, let’s assume we agree that it is unnatural to have the t i not in degree 2. But let’s also assume that we agree that we want a vielbein. There is one way out: Just like so(n,1) acts on n,1, it also acts on the abelian group 2 n,1 of bivectors, surely. Hence we can just as well consider the 2-group coming from the crossed module

(10)( 2 n,1tSO(n,1))

with trivial t. The corresponding fda looks precisley like the one above, just with t i and α suitably reinterpreted. A 2-connection with values in this Lie 2-algebra now consists of an so(n,1)-valued 1-form, together with a 2 n,1-valued 2-form. This now has a chance to encode a vielbein, since by wedging any vielbein with itself we do obtain such a 2-form. There is apparently an even better solution, using spin groups instead. This idea can be found discussed in section IV and V of hep-th/0204059 (see also gr-qc/9502037), whose authors consider precisely this, albeit without identifying n-connections explicitly (they call what they do “generalized differential calculus”). So consider D=n+1=3+1 for a moment and pick the double cover of the Lorentz group group G=SL(2,) and the translation group T 3,1. Using the 2-component spinorial language (and the Pauli matrices σ a AB), we can write a vielbein 1-form e a as e AB and wedge it with itself to form the SL(2,)-valued 2-form

(11)B AB=e AAe B A.

So we get a 2-connection for our Lie 2-algebra given by an SL(2,)-valued connection 1-form ω AB and that “surface-vielbein” 2-form B AB. In terms of these the Einstein-Hilbert action (in D=4) is of BF-theory form ()

(12) Xtr(F ωB).

Consider again the situation where the curvature of our 2-connection vanishes in degree 2. This now says that our 1-form connection is torsion free with respect to the metric encoded by the vielbein. The vielbein-2-form B should be precisely the “area vielbein” corresponding to the area metrics which have been argued to encode the geometry of string backgrounds ().

Update. Something closely related is also discussed in gr-qc/9804061.

Posted at July 19, 2006 5:20 PM UTC

TrackBack URL for this Entry:   http://golem.ph.utexas.edu/cgi-bin/MT-3.0/dxy-tb.fcgi/876

18 Comments & 1 Trackback

Re: 2-Palatini

Thanks for bringing “generalized exterior calculus” to my attention. Pretty neat.

Generalized exterior forms, geometry and space-time

Posted by: Eric on July 22, 2006 6:59 PM | Permalink | Reply to this

Re: 2-Palatini

If you like the formalism of generalized forms, you may be also interested in math-ph/0604060.

Posted by: Amitabha on July 24, 2006 7:27 AM | Permalink | Reply to this

Re: 2-Palatini

Nice. Thank you.

This might sound like a silly question, but what is the difference between a vector field and a (-1)-form? I often thought of duality mapping, i.e. transpose, as a reversal of the sign of the degree. Maybe this is not the most appropriate place to discuss it (?)

Posted by: Eric on July 24, 2006 10:46 PM | Permalink | Reply to this

Re: 2-Palatini

what is the difference between a vector field and a (-1)-form?

I believe that - apart from language - there is none, except maybe for the fact that you want to define an exterior derivative on a (-1)-form.

Of course Amitabha Lahiri in the above paper then proceeds to define duals of, in particular, (-1)-forms. Again, I would say, up to language these are just 1-forms.

The remark I would make about this formalism is that its raison d’être is, as far as I am concerned at least, Chen’s calculus of differential forms on path space.

In particular, given some n-form on X, you get a (n1) form on loop space from it essentially by contracting with tangent vectors of loops. That’s where the exterior product with “(-1)-forms” come in.

Once pulled back to the loop, it might be that we want to act with (n1) forms on it.

In particular, in the main example we have a 2-form B, contract it with a tangent vector θ of some loop to get ι θB and act on it with some 1-form A.

That’s how these pairs of 2-forms and 1-forms arise.

More formally, we may think of these pairs as local 2-connections coming to us as morphisms from the pair algebroid (infinitesimal paths in X) to the given Lie algebra.

This is nice, because it teaches us immediately how triples, quadruples etc. of forms should behave.

This, somewhat secretly, plays a major role for instance in p-form supergravity , where we are in fact dealing with triples of forms, a 1-form, a 2-form and a 3-form ().

Posted by: urs on July 27, 2006 2:27 PM | Permalink | Reply to this

Re: 2-Palatini

Sorry, I didn’t see the comments earlier.

Urs, I feel, is only partially right. I also thought that the -1-forms should be like vectors, but there is an obvious difference. Contraction by a vector is a (super-)derivative, it obeys Leibniz rule, whereas wedge product with a -1-form does not.

But the -1-forms can be more like integrals along curves, and Urs is more right when he points to loop space. Loop space would naturally take an n-form to an (n-1)-form. But then we want to add (not necessarily act with) an ordinary n-1-form to that. Then we need either to distinguish a point on the loop (where to add the n-1 form) or to consider path space instead of loop space.(Currently thinking about that …)

The dual space (vector) seems harder to interpret geometrically, and I am unable to say that I agree with Urs that these are just 1-forms…

Amitabha

Posted by: Amitabha Lahiri on July 28, 2006 1:13 PM | Permalink | Reply to this

-1 Forms

[…] obvious difference. Contraction by a vector is a (super-)derivative, […]

Right, let’s see.

Let θ be a (-1)-form.

Let ω 1 be a 1-form.

Then θω 1 is a 0-form, which depends linearly both on θ and ω 1, depending on nothing else (no metric or the like). This can only mean that wedging a 1-form with θ is the same as contracting it with some vector.

Let that vector be v. Then

(1)θω 1=ι vω 1.

But now, let ω 2 be a 2-form. For convenience, and without restriction of generality, assume that it factors into two 1-forms like

(2)ω 2=α 1β 1.

Now what is θω 2?

I guess we have

(3)θω 2 =θα 1β 1 =(θα 1)β 1 =(ι vα 1)β 1.

But I also guess we want

(4)θω 2 =θα 1β 1 =α 1θβ 1 =α 1(θβ 1) =α 1(ι vβ 1).

But this implies that α 1 is proportional to β 1, which implies that ω 2=0.

Did I make an assumption which violates the axioms of how (-1)-forms behave?

Posted by: urs on July 28, 2006 1:34 PM | Permalink | Reply to this

Re: -1 Forms

Did I make an assumption which violates the axioms of how (-1)-forms behave?

Well, yes. Because there is no such thing as an `ordinary’ -1-form, so the (generalised) -1-forms are the same space as 0-forms …

I’ll post a longer reply if this gets throguh.

Posted by: Amitabha on July 29, 2006 10:26 AM | Permalink | Reply to this

Re: -1 Forms

Ok, the previous one got through. So let me try a longer version of the reply to Urs.

Let us first do Urs’ calculation in the context of generalized forms, which is where -1-forms make sense. We should write θ=(0,θ 0), where the θ on the left hand side is a (generalized) -1-form, and θ 0 on the right hand side is a 0-form. Note that all -1-forms are like this, because there is no `ordinary’ -1 form.

Similarly, generalized 1-forms are written as α=(α 1,α 2). If we set α 2=0, we get the algebra of ordinary 1-forms. Let us do that, it will not affect the result.

Then

αβ=(α 1β 1,0),

θα=(0,θ 0α 1),

θβ=(0,θ 0β 1).

So by explicit calculation

(θα)β=(0,θ 0α 1β 1)=α(θβ).

But neither side is proportional to either α or β, both of which have non-vanishing first, and vanishing second components. This should answer Urs’ query.

What Urs has shown seems to be precisely that wedging with the -1-form cannot be the same as contracting with a vector.

Do you agree, Urs?

Posted by: Amitabha on July 29, 2006 11:13 AM | Permalink | Reply to this

Re: -1 Forms

Thanks for the explanation. That makes sense to me.

What I wrote was based on what I had seen on p. 3 of math-ph/0604060.

I haven’t looked at Sparling’s text, which is referred to there, but from what it says on that p. 3 I got the impression you were assuming the existence of an object of the sort I was talking about in my last comment.

If this isn’t the case, never mind.

Posted by: urs on July 29, 2006 2:02 PM | Permalink | Reply to this

Re: -1 Forms

I remember doing a calculation like this many ages ago when I was still thinking about this stuff and I seem to recall that it actually does work for noncommutative forms.

Perhaps generalized forms and (discrete) noncommutative forms are more closely related than the former and commutative forms?

Posted by: Eric on July 29, 2006 4:23 PM | Permalink | Reply to this

Re: -1 Forms

Trying to brush off some cobwebs…

Wedging with a (-1)-form somehow makes me think of cap product of a 1-cochain.

Posted by: Eric on July 29, 2006 7:09 PM | Permalink | Reply to this

Re: -1 Forms

Wedging with a (-1)-form somehow makes me think of cap product of a 1-cochain.

Given what Amitabha has explained above, this looks improbable to me.

The terminology is maybe not be the most suggestive one. A “generalized p form” in this business is defined to be a pair consisting of an ordinary p form and an ordinary (p+1)-form.

Given the sort of applications this is bound to arise in, we should think of the higher degree form as a connection form, and the lower degree one as an “auxialiary” form. This is at least what happens for connections on gerbes, hence for p=1. There we have a 2-form and a 1-form, and it is the 2-form part which is the genuine 2-connection component, whereas the 1-form part is sort of “auxiliary”.

A priori, these generalized forms are defined only for p0. But since we are dealing with pairs there is a natural option to extend this definition to p=1, because this will involve a 0-form in highest degree. And one simple takes the lower degree to be “zero”. More precisely, we extend the rules for wedging generalized forms, which are pairs of forms, to this (1)-case, which is not really a pair anymore, but just a 0-form.

This makes good sense, as Amitabha Lahiri has described in his latest comment.

So this (1)-form is, as a piece of data, a 0-form, equipped with a special rule for how to wedge it with other generalized forms.

In the above sense, we can regard it as the connection on a (-1)-gerbe, which has a 0-form connection and no auxiliary stuff.

In any case, it is important that, as data, these (-1)-forms are 0-forms.

I am not precisely sure how natural it is to consider pairs of forms in every degree. Because we know that the connection on some nonabelian p-gerbe is a (p+1)-tuple of forms of all degrees up to p+1.

And, at least in the abelian case, we know the natural product structure on these (p+1)-tuples: it’s the Chern-Simons-like cup product ().

But that’s just my point of view.

I should add that I am on vacation and probably won’t be able to send any messages across the web for the next week. So if you don’t see any further replies from me, it’s because of that.

Posted by: urs on July 29, 2006 9:37 PM | Permalink | Reply to this

Re: -1 Forms

Have a great vacation! You deserve it :)

I guess I am trying to understand things from the alternative point of view where you have a “true” (-1)-form ζ such that

(1)ap=(α p,α p+1)=α p+α p+1ζ.

and

(2)dζ=k.

But then maybe I am hopeless confused because Amitabha said

Note that all -1-forms are like this, because there is no `ordinary’ -1 form.

I don’t understand how this statement relates to the statement in the paper

We note here that the generalized p-form can be alternatively defined by assuming the existence of a (−1)-form field, i.e., a form of degree −1, due to Sparling[1].

This seems really clean to me. I would probably prefer to understand the meaning of the (-1)-form field. Aparently, based on both of your comments, we should think of it as an integral along a curve.

I really don’t know what I’m talking about, but it almost seems like a generalized form is a globbing together of a manifold and its path space and defining a form on that (or something).

What happens if you pullback an ordinary 0-form to the path space? Would that be a (-1)-form on the path space? We probably discussed this a million times already.

Getting old sucks and its worse when you weren’t too bright when you were younger :)

Cheers,
Eric

Posted by: Eric on July 29, 2006 10:18 PM | Permalink | Reply to this

Re: -1 Forms

I don’t understand how this statement relates to the statement in the paper

Yes, that’s what caused some misunderstanding in the above discussion. As far as I understood Amitabha’s latest comment, this statement in the paper is a little misleading. I assume from now on that what he explained in his latest comment is the way to think about it. Because that’s what makes sense, while otherwise we would run into the problem I pointed out.

What happens if you pullback an ordinary 0-form to the path space?

It disappears. It becomes the zero 0-form on path space.

Because what we really do is that we pull back along the evaluation map

(1)ev:PX×[0,1]X

which takes a path and a parameter and sends it to the path’s position at that parameter value, and then we integrate the result over the [0,1]-factor (called “integration over the fiber”).

For 0-forms on X this always produces the zero 0-form on PX.

In more pedestrian words, what you do is you take one of the indices of your p-form on X and contract it with the derivative of your path and integrate the result over the interval. If there is no index to contract with, you get 0.

[…] it almost seems like a generalized form is […]

Well, I made some sceptical comments on that, already. While it is true that a large and important class of 1-forms on path space come from a pair of a 2-fom and a 1-form on X, I am not sure yet that the concept of “generalized forms” is the best destillation of the information contained here.

As I said before, I think the full answer is to look at chain maps from some dg-algebras to the deRham complex. This contains all the information there is (linearized, of course), and it contains these pairs of differential forms as special cases.

But that’s just my point of view. Amitabha Lahiri is welcome to correct me. (But I might not be able to reply before next week.)

Posted by: urs on July 29, 2006 10:52 PM | Permalink | Reply to this

Re: -1 Forms

This probably makes no sense, but if we consider the map F:PMM from path space PM to the underlying manifold M, then I would think the usual rules should apply

(1) S p1F *(α p)= F *(S p1)α p

except the degree changes by one when you pull back to path space.

If we let αΩ 0(M) such that F *(α)=ζ be the (-1)-form on path space satisfying dζ=k, then we have

(2) pdF *(α)=k

for every pPM. However, if regular rules apply we also have

(3) pdF *(α)= pF *(α)= F *()α=k.

This likely makes no sense because it would imply that integrating over the emptyset on path space is nonzero and the pushforward of the emptyset is non-empty on the base manifold.

Anyway, this probably serves nothing more than to demonstrate once again how clueless I am :)

Cheers,
Eric

Posted by: Eric on July 29, 2006 10:54 PM | Permalink | Reply to this

Re: -1 Forms

This likely makes no sense because it would imply that integrating over the emptyset on path space is nonzero and the pushforward of the emptyset is non-empty on the base manifold.

Or maybe the boundary of a point in path space is not the emptyset?

I’ll try to stop rambling now…

Posted by: Eric on July 29, 2006 10:57 PM | Permalink | Reply to this

Re: -1 Forms

Eric,

I have to run now, sorry.

The thing is, the map we are looking at is not PXX but PX×[0,1]X, and the degree shift occurs because we integrate over the [0,1]-factor, while leaving the PX-factor alone.

I can try to say more about this – next week. :-)

All the best, Urs

Posted by: urs on July 29, 2006 11:01 PM | Permalink | Reply to this

Re: -1 Forms

Just saw your post…

Didn’t see it before writing the previous two whacky comments. I’ll think about what you said. I was trying to imagine what a (-1)-form would be like on path space by pulling back a 0-form on base space, but apparently that will not get me too far. A part of me still thinks there should be a way to pull back a 0-form to a (-1)-form on path space. Hmmm…

Posted by: Eric on July 29, 2006 11:17 PM | Permalink | Reply to this
Read the post SuGra 3-Connection Reloaded
Weblog: The n-Category Café
Excerpt: Discussion of n-category description of supergravity continued.
Tracked: August 17, 2006 7:53 PM

Post a New Comment