PDA

View Full Version : questions on nonabelian p-form gauge theories and self-duality in 6 dimensions


Jake Mannix
Apr9-04, 02:21 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>Hello all,\nThis post is partly directed to Prof. Baez, in reference to his\npaper (see the talk and links at the bottom of this page:\nhttp://math.ucr.edu/home/baez/gauge/ ), but also to people familiar\nwith the general physics folklore that "there exist no lagrangian\ndescriptions of self-dual 2-forms in 6 dimensions" (c.f. Witten\'s\npaper "The Effective Action of a 5-brane").\n\nIn Prof. Baez\'s paper, he describes (in more categorical language\nthan I\'ll use, as I\'m not sure what the field-theoretic relevance this\nhas for the time being) an action for a coupled system of a\nlie-algebra valued one form A, and a lie-algebra valued 2-form B (I\'ll\nbe dealing specifically with the case Baez calls the "automorphism\n2-group", so these two lie algebras can be identified).\nIn particular, to review, we have field strengths:\n\nF = dA + 1/2 A^A - B\nG = dB + A^B\n\nAnd he writes down a lagrangian density of the simplest possible\nform in these variables:\n\nL = |F|^2 + |G|^2\n\nMy first question is this: what are the gauge symmetries of this\nsystem?\n\nFrom my understanding of the origins of these things as\nconnections on gerbes, and from Baez\' description as being 2-group\nrelated - in this case there\'s a left Aut(G) action and compatible\nright G action on the fibers, and hence on the connection. But how\ndoes it show up as even infinitessimal variations on A, B?\n\nSecond, Baez mentions self-duality relations in *five*\ndimensions:\n\n*F = G\n\nWhich seems like it should be a string-particle duality given\nthat magnetic sources for strings in D=5 are particles.\n\nWhich brings me to my third question: string theorists like to\ntalk about 2-forms with self-dual field strength in D=6, which in this\nnotation would be just:\n\n*G = G\n\nBut what will this mean? The bianchi identities are, after a\nlittle rearrangement in more classical terms:\n\nD^2(A) = 0, DG = F^B,\n\nWhile the field equations are:\n\nD*F = (*DB)^B\nD*G = -*F,\n\nSo self-duality is requiring:\n\nF^B = - *F.\n\nAre there any other things that it requires? Are these\nconstraints solvable nontrivially? How many remaining degrees of\nfreedom are there once imposed?\n\nSome of this I\'m sure is obvious, but I haven\'t played around\nwith these enough yet to know what is going on here...\n\n-Jake Mannix\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>Hello all,
This post is partly directed to Prof. Baez, in reference to his
paper (see the talk and links at the bottom of this page:
http://math.ucr.edu/home/baez/gauge/ ), but also to people familiar
with the general physics folklore that "there exist no lagrangian
descriptions of self-dual 2-forms in 6 dimensions" (c.f. Witten's
paper "The Effective Action of a 5-brane").

In Prof. Baez's paper, he describes (in more categorical language
than I'll use, as I'm not sure what the field-theoretic relevance this
has for the time being) an action for a coupled system of a
lie-algebra valued one form A, and a lie-algebra valued 2-form B (I'll
be dealing specifically with the case Baez calls the "automorphism
2-group", so these two lie algebras can be identified).
In particular, to review, we have field strengths:

F = dA + 1/2 A^A - B[/itex]
G = dB + A^B

And he writes down a lagrangian density of the simplest possible
form in these variables:

L = |F|^2 + |G|^2

My first question is this: what are the gauge symmetries of this
system?

From my understanding of the origins of these things as
connections on gerbes, and from Baez' description as being 2-group
related - in this case there's a left Aut(G) action and compatible
right G action on the fibers, and hence on the connection. But how
does it show up as even infinitessimal variations on A, B?

Second, Baez mentions self-duality relations in *five*
dimensions:

*F = G

Which seems like it should be a string-particle duality given
that magnetic sources for strings in D=5 are particles.

Which brings me to my third question: string theorists like to
talk about 2-forms with self-dual field strength in D=6, which in this
notation would be just:

*G = G

But what will this mean? The bianchi identities are, after a
little rearrangement in more classical terms:

D^2(A) = 0, DG = F^B,

While the field equations are:

D*F = (*DB)^B
[itex]D*G = -*F,

So self-duality is requiring:

F^B = - *F.

Are there any other things that it requires? Are these
constraints solvable nontrivially? How many remaining degrees of
freedom are there once imposed?

Some of this I'm sure is obvious, but I haven't played around
with these enough yet to know what is going on here...

-Jake Mannix

Jake Mannix
Apr14-04, 03:17 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>I wrote in message news:&lt;12c1bbee.0404081648.95a7105@posting.google.c om&gt;...\n&lt;snip&gt;\n&gt; In particular, to review, we have field strengths:\n&gt;\n&gt; F = dA + 1/2 A^A - B\n&gt; G = dB + A^B\n&gt;\n&gt; And he writes down a lagrangian density of the simplest possible\n&gt; form in these variables:\n&gt;\n&gt; L = |F|^2 + |G|^2\n&gt;\n&gt; My first question is this: what are the gauge symmetries of this\n&gt; system?\n\nSo here\'s my current guess on some gauge symmetries, and my thoughts\non the meaning:\n\nIf we let:\n\na be a lie-algebra valued function, and\nb be a lie-algebra valued 1-form (unrelated to a),\n\nthen let the following infinitessimal variations of the connection\n(A,B) be:\n\nA -&gt; A + [A,a] + da + b\nB -&gt; B + [B,a] + db + b ^ A\n\nThen in the case where b = 0, this is just the usual set of\n(infinitessimal) gauge transformations, where A transforms like the\nconnection as usual (an adjoint action, plus an inhomogeneous term),\nand B transforms covariantly.\n\nOn the other hand, when we consider the abelian case, where\ncommutators (and hence wedge products of lie-algebra valued forms)\nvanish, then this reduces to shifting A and B by exact forms, and so\nthe on F = dA - B is to send it to itself (since dA -&gt; dA + db, while\nB -&gt; B + db), and certainly G = dB is invariant.\n\nLooking at the full nonabelian action, F = dA + 1/2 A ^ A - B, this\nstill transforms covariantly w.r.t. the \'a\' parameter, and is\ninvariant under the b-action, as the A and B terms still cancel. The\n3-form term still transforms covariantly w.r.t. the a-action, but now\nw.r.t. the b-shift, we have\n\nG = dB + A ^ B -&gt; d(b ^ A) + b ^ B + A ^ (db + b ^ A),\n\nand since the variation is into a commutator (on the lie-algebra\npiece), the term tr(G*G) should (to 1st order in the b-parameter) have\nvanishing variation.\n\nIf this is right, it\'s good for physics because the reason why we need\ngauge-invariance for vector particles is to shift away the negative\nnorm time-component (for which we need one degree of freedom per gauge\ngroup generator). For the two-form, the B_{0,i} terms are all\nnegative norm, and so we need a (co-)vector of degrees of freedom to\ncancel these out: which is taken care of by the b-shifts above: one\nfull 1-form of freedom.\n\nSo I still can\'t tell what the finite (non-infinitessimal) form of the\ngauge transformations are for these 2-forms (and their partner\n1-forms), but at least I think these make sense for small variations.\nAnd additionally, it\'s apparent that the 1-form connection and it\'s\npart of the curvature is somehow auxilliary - as A -&gt; A + b is part of\nit\'s transformation, it can be *completely* shifted away, and its part\nof the curvature can be absorbed into B.\n\nOf course, so doing uses up the big part of the gauge freedom which\nwas useful for cancelling out the unphysical degrees of freedom of B,\nbut at least knowing that those terms *can* be eliminated should be\nenough... so *either* A can be shifted away, *or* the unphysical parts\nof B... so maybe it\'s not auxilliary? Now I\'m a little confused\nagain.\n\n-Jake Mannix\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>I wrote in message news:<12c1bbee.0404081648.95a7105@posting.google.com>...
<snip>
> In particular, to review, we have field strengths:
>
> F = dA + 1/2 A^A - B
> G = dB + A^B
>
> And he writes down a lagrangian density of the simplest possible
> form in these variables:
>
> L = |F|^2 + |G|^2
>
> My first question is this: what are the gauge symmetries of this
> system?

So here's my current guess on some gauge symmetries, and my thoughts
on the meaning:

If we let:

a be a lie-algebra valued function, and
b be a lie-algebra valued 1-form (unrelated to a),

then let the following infinitessimal variations of the connection
(A,B) be:

A -> A +[/itex] [A,a] + da + b
B -> B + [B,a] [itex]+ db + b ^ A

Then in the case where b = 0, this is just the usual set of
(infinitessimal) gauge transformations, where A transforms like the
connection as usual (an adjoint action, plus an inhomogeneous term),
and B transforms covariantly.

On the other hand, when we consider the abelian case, where
commutators (and hence wedge products of lie-algebra valued forms)
vanish, then this reduces to shifting A and B by exact forms, and so
the on F = dA - B is to send it to itself (since dA -> dA + db, while
B -> B + db), and certainly G = dB is invariant.

Looking at the full nonabelian action, F = dA + 1/2 A ^ A - B, this
still transforms covariantly w.r.t. the 'a' parameter, and is
invariant under the b-action, as the A and B terms still cancel. The
3-form term still transforms covariantly w.r.t. the a-action, but now
w.r.t. the b-shift, we have

G = dB + A ^ B -> d(b ^ A) + b ^ B + A ^ (db + b ^ A),

and since the variation is into a commutator (on the lie-algebra
piece), the term tr(G*G) should (to 1st order in the b-parameter) have
vanishing variation.

If this is right, it's good for physics because the reason why we need
gauge-invariance for vector particles is to shift away the negative
norm time-component (for which we need one degree of freedom per gauge
group generator). For the two-form, the B_{0,i} terms are all
negative norm, and so we need a (co-)vector of degrees of freedom to
cancel these out: which is taken care of by the b-shifts above: one
full 1-form of freedom.

So I still can't tell what the finite (non-infinitessimal) form of the
gauge transformations are for these 2-forms (and their partner
1-forms), but at least I think these make sense for small variations.
And additionally, it's apparent that the 1-form connection and it's
part of the curvature is somehow auxilliary - as A -> A + b is part of
it's transformation, it can be *completely* shifted away, and its part
of the curvature can be absorbed into B.

Of course, so doing uses up the big part of the gauge freedom which
was useful for cancelling out the unphysical degrees of freedom of B,
but at least knowing that those terms *can* be eliminated should be
enough... so *either* A can be shifted away, *or* the unphysical parts
of B... so maybe it's not auxilliary? Now I'm a little confused
again.

-Jake Mannix