2
$\begingroup$

I am familiar with the following:

let $X$ be a smooth projective complex variety, $D$ a smooth divisor in $X$ and $U=X \setminus Z$. Then there is on the one hand a residue map $$ \mathrm{Res}_D \colon H^n_{dR}(U) \longrightarrow H^{n-1}_{dR}(D) $$ induced from the exact sequence of sheaves $0 \to \Omega^p_X \to \Omega^p_X(\log D) \to \Omega^{p-1}_D \to 0,$ and on the other hand the tubular map in homology $$ T \colon H_{n-1}(D) \to H_n(U). $$ They are related by $$ \int_\sigma \mathrm{Res}_D(\omega)=\frac{1}{2\pi i} \int_{T(\sigma)} \omega $$ for all $\sigma \in H_{n-1}(D)$ and $\omega \in H^n_{dR}(U)$.

Question: I guess this is still true when $X$ is not supposed projective, because everything seems very local. Do you know a good reference?

$\endgroup$
1
  • $\begingroup$ I would look at the paper of Herrera-Lieberman, "Residues and principal values on complex spaces". Although I'd bet it's already in Leray's work in this form. $\endgroup$ – Dan Petersen Oct 5 '18 at 12:02
2
$\begingroup$

The Gysin sequence in the form you need is a purely topological based on the long exact cohomological sequence of a pair, the Thom isomophism and Poincare duality.

The Gysin sequence is obtained from the long exact sequence of the pair $(T(D), \partial T(D))$ where $T(D)$ is a tubular neighborhood of $D$ in $X$. Note that we have a natural projection $\pi: T(D)\to D$. Moreover, by excision

$$ H^\bullet(T(D),\partial T(D)\;)\cong H^\bullet (X, X\setminus D). $$ $\newcommand{\pa}{\partial}$ If $D$ is compact then it carries a homology class $[D]$ and its Poincare dual is the Thom class $\tau\in H^2(T(D),\partial T(D))$.

The Thom isomorphism theorem states that the map $\Psi: H^\bullet (T(D)D)\to H^{\bullet +2}(T(D),\pa T(D))$ given by $$ H^kT(D)\cong H^k(D)\ni \alpha \mapsto \tau \cup \pi^*\alpha\in H^{k+2}(T(D),\pa T(D)) $$ is an isomorphism.

The long exact sequence

$$ \cdots \to H^k(T(D),\pa T(D))\to H^k(T(D))\to H^k (\pa T(D))\to H^{k+1}(T(D),\pa T(D))\to\cdots $$

can be rewritten using the Thom isomorphism in the form $$ \cdots \to H^{k-2}(T(D))\to H^k(T(D))\to H^k (\pa T(D))\to H^{k-1}(T(D))\to\cdots $$ Note that $H^\bullet(T(D))\cong H^\bullet (D)$ so we obtain $$ \cdots \to H^{k-2}(D)\to H^k(D)\to H^k (\pa T(D))\to H^{k-1}(D)\to\cdots $$

We set $N:=\dim_{\mathbb{C}} X$ and $m:=2N-k$.

Using the Poincare dualities $$ H^j(D)\cong H_{2N-2-j}(D),\;\; H^j(\pa T(D))\cong H_{2N-1-j}( \pa T(D)) $$ we obtain a long exact sequence

$$\cdots \to H_m(D)\to H_{m-2}( D) \stackrel{T}{\to} H_{m-1}(\pa T(D))\to H_{m-1}(D)\to\cdots .$$

Above, $T$ is the so called tube morphism. Its Kronecker dual is the Residue morphism. See these notes for more details.

If $D$ is non-compact, then you will have to use Borel-Moore homology $H_\bullet^{BM}$ instead of the usual homology,

$$ H_\bullet^{BM}(X,\mathbb{C})={\rm Hom}\;(\; H^\bullet_{cpt}(X,\mathbb{C}),\mathbb{C}\;). $$

$\endgroup$

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.