1
$\begingroup$

If $0\longrightarrow M\longrightarrow E\longrightarrow N\longrightarrow 0$ is a short exact sequence of vector bundles on a surface. Here $M$ and $N$ are line bundles, and so rank $ E$=2. Also, if $L$ is an ample line bundle, we have that $\mu_L(M)>\mu_L(E)>\mu_L(N)$.

My question is this : Why does $\textrm{Ext}^1(N,M)/\textrm{Hom}(N,M)$ make sense?

From the initial exact sequence, by applying the functor Hom$(N,\_)$, we get the sequence $0\longrightarrow \textrm{Hom}(N,M)\longrightarrow \textrm{Hom}(N,E)\longrightarrow \textrm{Hom}(N,N)\longrightarrow \textrm{Ext}^1(N,M)\longrightarrow..$.

Similarly in the case if we apply Hom$(\_,M)$, we get $0\longrightarrow \textrm{Hom}(N,M)\longrightarrow \textrm{Hom}(E,M)\longrightarrow \textrm{Hom}(M,M)\longrightarrow \textrm{Ext}^1(N,M)\longrightarrow..$.

So how do we get this quotient?

$\endgroup$
  • $\begingroup$ Where did you read that the quotient makes sense? $\endgroup$ – Mariano Suárez-Álvarez Apr 22 '15 at 14:41
  • $\begingroup$ I am reading a paper - Stability of rank-3 Lazarsfeld-Mukai bundles on K3 surfaces. It is mentioned in that. $\endgroup$ – gradstudent Apr 22 '15 at 14:43
  • $\begingroup$ There is no sensible map from Hom to Ext in a general situation. $\endgroup$ – Mariano Suárez-Álvarez Apr 22 '15 at 14:58
  • $\begingroup$ Is there a natural action of Hom on Ext? $\endgroup$ – gradstudent Apr 22 '15 at 15:03
  • 2
    $\begingroup$ There is an action of $\hom(M,M)$ and of $\hom(N,N)$, but not of $\hom(M,N)$ $\endgroup$ – Mariano Suárez-Álvarez Apr 22 '15 at 15:06
2
$\begingroup$

As stated in the comments, this does not answer the question at all, but rather one that arose there.

Suppose our category had enough projectives. Choose a projective resolution $d_{\bullet}\!: P_{\bullet} \to N$. Using projectivity of $P_0$, resp. projectivity of $P_1$ and $\text{im}(f_0 \circ d_0) \subseteq \ker(E \to N) = M$, we get lifts of the identity $$\require{AMScd} \begin{CD} P_2 @>d_1>> P_1 @>d_0>> P_0 @>>> N @>>> 0 \\ @VVV @VVf_1V @VVf_0V @VV\text{id}V \\ 0 @>>> M @>>> E @>>> N @>>> 0. \end{CD} $$ It follows that $E \cong M \amalg_{P_1} P_0$ is the pushout. On the other hand, consider the complex $$0 \to \text{Hom}(P_0,M) \xrightarrow{d_0^*} \text{Hom}(P_1,M) \xrightarrow{d_1^*} \text{Hom}(P_2,M) \to \ldots$$ By construction, $f_1 \in \ker(d_1^*)$. This defines the class of our extension in $\text{Ext}^1(N,M) = H^1(\text{Hom}(P_{\bullet},M))$. Accordingly, it splits iff $f_1 \in \text{im}(d_0^*)$, and in fact, we have a bijection between $(d_0^*)^{-1}f_1$ and splittings of our extension*. Then $\text{Hom}(N,M) \cong \ker(d_0^*)$ acts on this simply transitively by addition.

*The splittings $M \amalg_{P_1} P_0 \cong E \to M$ are precisely given by $(\text{id}_M,\varphi)$ with $\varphi \circ d_0 = f_1$, that is, such that $\varphi \in (d_0^*)^{-1}f_1 \subseteq \text{Hom}(P_0,M)$.

$\endgroup$
  • $\begingroup$ Of course I was being silly in the comments to the question. The argument does apply to coherent sheaves, just by dualizing everything. Sorry. $\endgroup$ – Thomas Poguntke Apr 28 '15 at 18:51

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.