The literature about K3 surfaces is extensive. Let us consider the fact that the zero locus of any smooth homogeneous degree 4 equation is a K3 surface.

Let $R$ the quotient ring of a homogeneous polynomial in $4$ variables and degree $4$.

I am not being succesfull to find anything related to the study of derivations and locally nilpotent derivations on these sort of ring. What could be a reference for

  1. a description of the derivations on $R$?
  2. a description of the locally nilpotent derivations on $R$?
  3. a description of the automorphism groups of $R$?
  4. any condition for the group of automorphisms of $R$ to be finite/finitely generated?
  • 2
    $\begingroup$ What do you call a derivation on a surface? If you mean a tangent vector field, they do not exist ($\neq 0$) on a K3. $\endgroup$
    – abx
    Dec 27, 2019 at 17:20
  • 1
    $\begingroup$ @abx: "do not exist" means "=0". But the question is, indeed, quite unclear. $\endgroup$
    – Sasha
    Dec 27, 2019 at 19:38
  • $\begingroup$ It was really bad written and it was missing an essential part of the data. Sorry for my mistake in not double checking it. I tried to clarify now... $\endgroup$
    – Binai
    Dec 28, 2019 at 3:52
  • $\begingroup$ 1. is just given by the Euler sequence. They are just derivations of the polynomial ring (which are obvious) which takes the equation of the hypersurface to a multiple of the same equation. $\endgroup$
    – Mohan
    Dec 29, 2019 at 15:57
  • $\begingroup$ @Mohan It is pretty clear that the derivations should have this property related to the polynomial ideal that we are taking the quotient, but I want to know a characterization (if it exists). $\endgroup$
    – Binai
    Dec 31, 2019 at 5:14

1 Answer 1


I can try to say a little about your first question, on derivations of $R$, in characteristic 0. As Mohan said, the derivations of $R$ are just derivations of the polynomial ring taking the defining equation to itself, but I'd add that it's a bit subtle what the collection of such derivations looks like. It's clear we have derivations of every positive degree $k$, for example by first applying the Euler operator and then multiplying by a polynomial of degree $k$. So, one natural question you can ask is whether there's a derivation of $R$ of negative degree. Note that such a derivation is locally nilpotent (though not all locally nilpotent derivations have negative degree).

The general context for your question is that we're given a ring $R=S(X,L):=\bigoplus H^0(X,L^m)$, which is naturally the section ring of a (smooth) projective variety with a chosen ample line bundle. By a theorem of J.M. Wahl ("A cohomological characterization of $\mathbb P^n$") if $R$ admits a derivation of negative degree, then $X$ must in fact be $\mathbb P^n$. Thus, in your example there will be no derivations of negative degree. (This crucially uses the characteristic 0 hypothesis.)

In fact, one can rule out differential operators of negative degree of any order on your $R$, not just derivations: one can show that the existence of a differential operator of order $m$ and degree $-e$ gives rise to an element of $H^0(X,(\mathrm{Sym}^m T_X)(-e))$, and in turn to an ample subsheaf of $\mathrm{Sym}^m T_X$. By powerful results of Miyaoka (Corollary 8.6 of "Deformations of a Morphism along a Foliation and Applications") this in turn forces $X$ to be uniruled (again, in characteristic 0!), which your K3 surface is not.

  • $\begingroup$ It is pretty clear that the derivations should have this property related to the polynomial ideal that we are taking the quotient, but I want to know a description (if it exists), i.e, what "the collection of such derivations looks like". $\endgroup$
    – Binai
    Dec 31, 2019 at 5:12
  • $\begingroup$ Ah, for sure that characteristic 0 is a crucial step. I would be happy with answers in this case! $\endgroup$
    – Binai
    Dec 31, 2019 at 5:16

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.