I would like to know what can be said about (global) deformations of holomorphic/algebraic rank two vector bundles on $\mathbb{P}^3$. I am particularly interested in the case of topologically trivial bundles, i.e., where the Chern classes as well as Atiyah-Rees-$\alpha$-invariant are trivial.

A precise question would be: Is it true that every topologically trivial rank two vector bundle on $\mathbb{P}^3$ can be deformed to the trivial bundle? If not, how could one define a reasonable invariant to distinguish non-deformable bundles? (The parameter space of the deformation could be anything, not necessarily irreducible or smooth.)

I guess this question could possibly be phrased asking if the moduli of vector bundles of topologically trivial rank two bundles is connected but this is a bit delicate because we are dealing with unstable bundles and so there is no coarse moduli space as such.

I only know a single paper discussing questions like this for rank two bundles on $\mathbb{P}^3$ (apart from the huge body of literature which focuses only on stable bundles):

C. Banica. Topologisch triviale Vektorbündel auf $\mathbb{P}^n(\mathbb{C})$. J. Reine Angew. Math. 344 (1983) 102-119.

The paper shows that the moduli space $M(2)$ of topologically trivial bundles with splitting type $d=2$ has two connected components, but the methods suggest that it would be fairly difficult to study $M(d)$ in general. Certainly the results in Banica's paper do not decide the deformability question. Has there been any further work on such questions?

I would greatly appreciate references to further literature dealing with unstable bundles on projective spaces (beyond what is in the book of Okonek-Schneider-Spindler) as well as hints on possible approaches to try on this problem...

  • $\begingroup$ Since a family of vector bundles on $\mathbb{P}^3$ with one member trivial has general member trivial, do you mean that any topologically trivial bundle is the limit of trivial bundles? There is at least a question by Peskine and Lazarsfeld asking that such limits are in fact trivial over complex numbers. $\endgroup$
    – Mohan
    Aug 21, 2014 at 14:52
  • $\begingroup$ @Mohan: I do not think that all limits are trivial - there are non-trivial limits of trivial bundles over $\mathbb{P}^1$ and $\mathbb{P}^2$. Moreover, once you have a non-trivial limit, it may be possible to further deform the bundle to something that is no longer a limit of trivial bundles. This happens for $\mathbb{P}^2$, and this is the reason for stating explicitly that the parameter space does not need to be irreducible. $\endgroup$ Aug 21, 2014 at 15:23
  • $\begingroup$ I did not mean on projective line or plane. Question of Peskine is just for 3-space and rank 2 bundles(and hence for all larger dimensions) $\endgroup$
    – Mohan
    Aug 21, 2014 at 15:29
  • $\begingroup$ Could you give me a reference? This seems to say that the elements in $M(2)$ of Banica are not deformable to the trivial bundle. Has this something to do with properties of local vs. global complete intersection of codimension 2 in $\mathbb{P}^3$? $\endgroup$ Aug 21, 2014 at 15:34
  • $\begingroup$ This question due to Peskine can be found in Springer Lecture Notes 1389 (1988), where he asks, given a family of smooth curves in 3-space with general member a complete intersection, is the special member a complete intersection. To the best of my knowledge, the answer is not known, but I answered it in the negative for positive characteristics. $\endgroup$
    – Mohan
    Aug 21, 2014 at 17:03

1 Answer 1


As I mentioned earlier, Peskine (and possibly Kollar too) asked whether given a family of smooth curves in 3-space with general member a complete intersection, is the special member also a complete intersection. To the best of my knowledge, the answer is not known (over complex numbers). Under the above hypothesis, it is immediate that $\omega_{C_t}=\mathcal{O}_{C_t}(d)$ for some $d$ for all $t$ in the family where $\omega_C=K_C$, the canonical bundle. Thus by Serre construction, we get a family of rank two vector bundles $E_t$ such that $E_t$ maps onto the ideal sheaf of $C_t$. Again by Serre, $E_t$ is a direct sum of line bundles if and only if $C_t$ is a complete intersection. Thus, Peskine's question is equivalent to, given a family of rank 2 vector bundles with general member a direct sum of lines, is it true also for the special member. Of course, this does not answer your question, since Peskine's question is unresolved.

  • $\begingroup$ Thanks very much, also for the discussion. My confusion was that bundles of the form $0\to\mathcal{O}(n)\to E\to\mathcal{O}(-n)\otimes\mathcal{I}_Y\to 0$ with $Y$ complete intersection are nonsplit but deformable to the split bundle for $\mathbb{P}^2$, but they are split for $\mathbb{P}^3$. Your answer seems to say that if Peskine's question has a positive answer, then there are at least two classes of rank two bundles up to deformation - the trivial bundle and the rest. $\endgroup$ Aug 22, 2014 at 13:01
  • $\begingroup$ You said that you answered Peskine's question negatively in positive characteristic. Does that give rise to interesting deformations from split bundles in characteristic $p$? $\endgroup$ Aug 22, 2014 at 13:02
  • $\begingroup$ They are `interesting', but not very explicit. It would be nice to know which bundles appear this way as a limit of decomposable bundles. $\endgroup$
    – Mohan
    Aug 22, 2014 at 22:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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