Hendryk Pfeiffer has a new preprint on arxiv

[Kea]

But they seem to have generalized it by replacing the sphere in the definition with a general surface...

Yes, it would seem so. Replace the spheres by the two genus g surfaces. The tubular neighbourhood is something extra again. A lot of SW theory seems to use those put-a-bundle-on-it tricks that are very powerful if one has the skills to use them!

Look what I found online:
http://www.mrlonline.org/mrl/1997-004-006/1997-004-006-011.pdf

 

[selfAdjoint]

Looks interesting. I was following up spin-c structures and Taubes theorem and I found Baez's TWF weeks http://math.ucr.edu/home/baez/week44.html" , when a lot of this stuff was introduced, based on a breakthrough by Witten.
I still haven't succeeded in wrapping my head around spin-c structure in spite of reading Baez's' description and the http://en.wikipedia.org/wiki/Spin_structure" definition.

 

[Kea]

I was following up spin-c structures and Taubes theorem and I found Baez's TWF weeks http://math.ucr.edu/home/baez/week44.html" , when a lot of this stuff was introduced...

The reference from week 45 is available online
http://www.math.harvard.edu/~kronheim/thomconj.pdf

Now, I must go and read it!

 

[selfAdjoint]

The reference from week 45 is available online
http://www.math.harvard.edu/~kronheim/thomconj.pdf
Now, I must go and read it!

This is a paper I want to read as soon as I am ABLE to read it! Is there any textbook on this stuff? Not necessarily S-W invariants but the kind of concepts he uses in introducing them to the mathematicians who read the paper back when it came out?

'Nother question. In these surgery papers they talk about sewing in a fibre space, but all I see them use is cross-products. Are you aware of any use of real fibrations? Consider e.g. the Hopf fibration, a famous nontrivial fibering of the 3-sphere over the 2-sphere with fibre the 1-sphere (aka circle). Seems to me you could cut a 4-ball out of a 4-manifold and sew in an HF matching on the boundary, but maybe that would be like sewing in a crosscap on a surface - making the result non-orientable?

 

[Kea]

Is there any textbook on this stuff? Are you aware of any use of real fibrations? Seems to me you could cut a 4-ball out of a 4-manifold and sew in an HF matching on the boundary, but maybe that would be like sewing in a crosscap on a surface - making the result non-orientable?

Well, there are textbooks that go into spinC structures. Unfortunately our library doesn't have any so I forget the titles. There's Morgan's own book on SW, for instance. As for fibrations, the 'disc neighbourhood' of the surfaces is a very good non-trivial use of fibrations. It is analogous to the knot neighbourhoods essential to field theoretic constructions of the knot polnomials in 3D, where twists can be thought of as ribbon twists.

As for the Hopf bundles: I'd have to think about that. But doesn't the unknottedness of the codim 1 case mean that Hopf bundles won't do much here? I can only assume these guys thought about this (no doubt casually in their sleep). Stern is quite clear that the only known invariants of smooth structure act like the SW ones.

We can keep trying, but sorry selfAdjoint, I must be off for a long weekend...the wilderness beckons.

 

[torsten]

this appears to be very important. for the general good, we need a pedagogical post about the torus surgery
I guess the fun thing is you cut a donut out of the middle, and do something tricky with it, and glue it back in----and you think the manifold isn't changed but it is now different in an essential way.
Please somebody (selfAdjoint or volunteer) expand on this a bit at an introductory level and say why torus surgery, which one can do in 3D, is the key to why in 4D you can have such a great diversity of DeeEsses.

Hi marcus,
maybe I can illuminate you a little bit. Let M be a compact, simply-connected 4-manifold. Now you look for an embedded torus in the 4-manifold so that the torus has no self-intersections. Then cut out a neighborhood of the torus, i.e. D^2\times T^2 and sew in a twisted torus. A twisted torus is simple to describe: cut the torus along a circle to get a cyclinder, then clue both ends together but before make n rotations of one end. Such an operation is called a Dehn twist.
The whole 4-dimensional operation is called a logarithmic surgery.

I hope that helps.

Torsten

 

[marcus]

Hi marcus,
maybe I can illuminate you a little bit. Let M be a compact, simply-connected 4-manifold. Now you look for an embedded torus in the 4-manifold so that the torus has no self-intersections. Then cut out a neighborhood of the torus, i.e. D^2\times T^2 and sew in a twisted torus. A twisted torus is simple to describe: cut the torus along a circle to get a cyclinder, then glue both ends together but before make n rotations of one end. Such an operation is called a Dehn twist.
The whole 4-dimensional operation is called a logarithmic surgery.
I hope that helps.
Torsten

thanks Torsten! yes that helps some. D^2\times T^2 just gives the torus (the hollow surface of a donut) a little thickness, so that is a neighborhood of the original torus. I can picture cutting that out.

I can also imagine a twisted torus. I suppose that before one sews that back in one must also thicken it some by Cartesian product with
a disk D^2 , as before. Or is that even necessary?

These must be very basic elementary things in differential topology and i am a little embarrassed not to know this already! I think I had a course about differential geometry a while ago that should have taught me this, but did not


In any case, thanks for the kind explanation.

 

[selfAdjoint]

just gives the torus (the hollow surface of a donut) a little thickness, so that is a neighborhood of the original torus. I can picture cutting that out

I think you are picturing the unit interval cross the torus. Think about the factors: the torus is a 2-surface like an inner tube, and the 2-ball is like a disk. The cross product of those two is a four dimensional manifold with boundary the 3-torus. I can work with that, but I can't visualize it, can you? The point is that the original torus lies inside the neighborhood, and only the boundary of the neighborhood contacts the manifold. So you have room to do the twist without messing up the reattachment. It's the neighborhood that gets warped by having to match the twisted torus on the inside and the untwisted manifold on the outside.

 

[marcus]

The point is that the original torus lies inside the neighborhood, and only the boundary of the neighborhood contacts the manifold. So you have room to do the twist without messing up the reattachment...

Yay!

the boundary of the disk D² is the circle factor that ups the 2-torus to a 3-torus-----which is the boundary of the neighborhood
and the radial distance of the disk represents leeway
(for screwing the original 2-torus around in)

OK a tiny increment of intuition----OK a crumb from your blanketyblank differential topology banquet.

 

[Kea]

I still haven't succeeded in wrapping my head around spin-c structure...

Well, the weekend was nice. A naughty mouse at the bivy ate half my Snickers bar.

Twistor theory to the rescue! Reference (around page 215):

Twistor Geometry and Field Theory
R.S. Ward, R.O. Wells jnr
Cambridge Mon. Math. Phys. (1990)

Recall that SL(2,C) is the double cover of the Lorentz group, so another name for it is Spin(1,3)_{0}. Analogously, whenever there is a two-to-one covering one talks about spin structures.

What about spinC structures? Given a complex structure on a bundle of rank 2n one gets such a structure. This weakening of spin structures is one motivation. As well as the short exact sequence

1 \rightarrow Z_{2} \rightarrow \textrm{Spin}(2n) \rightarrow<br /> \textrm{Spin}(2n) \backslash Z_{2} \rightarrow 1

just alluded to, there is now also a sequence

1 \rightarrow \textrm{Spin}(2n) \backslash Z_{2} \rightarrow<br /> \textrm{Spin}^{C}(2n) \rightarrow U(1) \rightarrow 1

which amounts to the fact that if H^{2}(M,Z) = 0 then spinC structures can be reduced to spin ones.

In terms of Clifford algebra: \textrm{Cl}(V) is the quotient of the tensor algebra by the ideal generated by v \otimes v + Q(v) for some quadratic form Q. Define \textrm{Cl}(V)^{C} = \textrm{Cl}(V) \otimes C. Let (\textrm{Cl}(V)^{C})* denote the invertible elements. Let \alpha denote the involution on \textrm{Cl}(V) coming from v goes to -v. Define the group

\Gamma^{C}(V) = \{ v \in (\textrm{Cl}(V)^{C})* : \alpha (v) w<br /> v^{-1} \in V \hspace{10mm} \textrm{for} \hspace{10mm} w \in V \}

and then define the group \textrm{Pin}^{C}(V) to be the elements of \Gamma^{C}(V) of norm 1. Then there is an exact sequence

1 \rightarrow U(1) \rightarrow \textrm{Pin}^{C}(V) \rightarrow O(V) \rightarrow 1

which hopefully makes the magical appearance of the U(1) a little clearer. Call the second arrow from the right \rho. Then

\textrm{Spin}^{C}(V) = \{ v \in \textrm{Pin}^{C}(V) :<br /> \textrm{det} \rho (v) = 1 \}

Sigh.

 

[Chronos]

Kea, I was going to ask you to pass the ice bucket so I could freshen up my drink. Now I need a towel so I can apply it to my forehead.

 

[Pietjuh]

When I saw the title of this thread I immediately conjured up an image of marcus having the Pfeiffer disease. Luckily it wasn't the case! :)

 

[Kea]

reference

The reference given by the Nair paper being discussed in the other thread is

Spin Geometry
H.B. Lawson, M. Michelsohn
Princeton U. P. (1989)

Haven't seen it myself, so can't say what it's like.