5
$\begingroup$

I'm interested in $c_1^2(L)$ where $L$ is the determinant line bundle associated to a $Spin^c$ structure on a compact oriented four-manifold $X^4$.

Specifically I'd like to know the numerical value of this characteristic number for some simple examples; let's say $\mathbb{CP}^2\#\overline{\mathbb{CP}^2}$ and $\mathbb{CP}^2\#\overline{\mathbb{CP}^2}\#\overline{\mathbb{CP}^2}$. I've seen this number show up while reading about Seiberg-Witten theory but I can never find any sources that give it's value on an actual manifold. Any reference to a source which has calculated some of this would be excellent.

As an alternative or add-on, what methods are/would be used to calculate such a thing? My knowledge about characteristic classes is pretty weak. (Speaking of which, one misc. question: Is this class sensitive to orientation? i.e. does it change sign if one changes the orientation of $X$?)

Thank you for any information.

$\endgroup$
5
$\begingroup$

Last question first. Recall that a $\text{Spin}^c$-structure on a 4-manifold is a lift of the structure group of the tangent bundle from $O(4)$ to $\text{Spin}^c(4)$. The map $\text{Spin}^c(4) \to O(4)$ factors through $SO(4)$, so a $\text{Spin}^c$-structure comes equipped with an orientation. When we talk about $\text{Spin}^c$ structures, we're usually doing so on manifolds that are oriented in the first place, in a way compatible with the $\text{Spin}^c$-structure; if these two orientations disagreed, there's no reason to believe stuff coming from the orientation (eg the Hodge star) would play well with stuff coming from the $\text{Spin}^c$-structure (eg Clifford algebras).

Onto your actual question. Any Kahler manifold carries a $\text{Spin}^c$-structure; often this is defined in terms of spinor bundles (eg in Morgan's book), but I'll keep with the language of structure groups above. Because in 4D $\text{Spin}^c(4) = (SU(2) \times SU(2)) \times_{\Bbb Z/2} U(1)$, writing an element of $U(2)$ as $\lambda g$, where $\lambda \in U(1)$ and $g \in SU(2)$, we have a homomorphism $f: U(2) \to \text{Spin}^c(4)$ given by $f(\lambda g) = (g, \lambda, \lambda)$. You can verify that this is well-defined. The map $\text{Spin}^c(4) \to SO(4)$ is given by $(g,h,\lambda) \mapsto \pm(g,h) \in (SU(2) \times SU(2))/\pm I = SO(4)$. This means that our maps $U(2) \to \text{Spin}^c(4) \to SO(4)$ are compatible, the composition being the natural homomorphism $U(2) \to SO(4)$; so given a Kahler structure, we inherit a $\text{Spin}^c$-structure.

The map $U(2) \to \text{Spin}^c(4)$ is compatible with the determinant map (of both) to $U(1)$. This tells me that the determinant line bundle of the Kahler manifold is the same as the determinant line bundle of the $\text{Spin}^c$-structure, and hence they have the same Chern class. So it suffices to see what $c_1(L)^2$ is for the determinant line bundle of the complex structure; but it is a general fact of complex vector bundles that $c_1(E) = c_1(\Lambda^{top}(E))$, so that $c_1(L) = c_1(TM)$ here.

To calculate it, I'm just going to invoke some theorems. First, we have the relationship $c_1(E)^2 - 2c_2(E) = p_1(E)$ for any complex vector bundle $E$; second, we have the fact that for a 4-manifold, $p_1(M) = 3\sigma(M)$; and last, for a complex manifold of complex dimension $n$, $c_2(M) = \chi(M)$. Putting this all together, we get $c_1(E)^2 = 2\chi(M) + 3\sigma(M)$. If you know the index formula for the linearized Seiberg-Witten operator, you'll realize this gives us an exciting fact: the Seiberg-Witten moduli space has expected dimension zero on a Kahler manifold!

If I twisted the Spin-c structure from the "standard one", I would get $c_1(\mathfrak s \otimes L) = c_1(\mathfrak s) + 2c_1(L)$ (the point being that the determinant map is really a $z \mapsto z^2$-thing, because of that $_{\Bbb Z/2}$ in the definition of $\text{Spin}^c$-structure), so we can in principle then calculate $c_1^2$ for every $\text{Spin}^c$ structure on a Kahler manifold.

As for your specific examples: $\chi(\Bbb{CP}^2 \# n \overline{\Bbb{CP}^2}) = 3+n$, and $\sigma(\Bbb{CP}^2 \# n \overline{\Bbb{CP}^2}) = 1-n$, so we get $c_1(\Bbb{CP}^2 \# n \overline{\Bbb{CP}^2})^2 = 9-n$ from the previous computation.

$\endgroup$
  • $\begingroup$ Thank you very much for your answer. I am quite ignorant of complex/Kähler geometry. Is it the case that $\mathbb{CP}^2\#n\overline{\mathbb{CP}^2}$ is Kähler for every n? What happens if we consider $k\mathbb{CP}^2\#l\overline{\mathbb{CP}^2}$ for arbitrary $k$ and $l$? These aren't all Kähler manifolds, are they? $\endgroup$ – Brian Klatt Aug 2 '16 at 17:26
  • $\begingroup$ @BrianKlatt Yeah, they're Kahler manifolds for $k=1$ - you took the Kahler manifold $\Bbb{CP}^2$ and "blew it up" at n points, which gives you another Kahler manifold. With $k>1$, those are never Kahler (or even symplectic) manifolds, though that's actually a theorem of Seiberg-Witten theory. Actually, everything in this answer is valid for almost complex manifolds, with the same results, though you have to be a little more careful with your phrasing. It's true that if $k$ is odd, your manifolds admit almost complex structures, which then have naturally associated spin-c structures. $\endgroup$ – user98602 Aug 2 '16 at 17:30
  • $\begingroup$ So if $k$ is odd, is the basic point that $c_1^2=2\chi + 3\sigma$ still holds? $\endgroup$ – Brian Klatt Aug 2 '16 at 17:35
  • $\begingroup$ There's a more fundamental point to be made - manifolds do not come equipped with a $\text{Spin}^c$ structure; you have to give them one. If you equip your manifold with an almost complex structure, then the induced $\text{Spin}^c$ structure indeed satisfied the relation you gave. If you don't, it doesn't make sense to ask for a calculation of $c_1(\mathfrak s)^2$. In the above cases, where you say "$c_1$ of..." etc, I assume you're equipping these manifolds with the $\text{Spin}^c$ structures coming from their (almost) complex structures. $\endgroup$ – user98602 Aug 2 '16 at 17:38
  • $\begingroup$ Ah, thanks for pointing that out. I think I was assuming there was a canonical choice if the manifold was simply-connected (as in the spin case). $\endgroup$ – Brian Klatt Aug 2 '16 at 17:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.