Let $S$ be a noetherian scheme. Grothendieck [1] constructed for any projective $S$-scheme $X$ the Hilbert scheme $\operatorname{Hilb}_{X/S}$, which represents the functor $$T \mapsto \underline{\operatorname{Hilb}}_{X/S}(T) = \{Y \hookrightarrow X \times_S T \,|\, \text{family of closed subschemes, flat over } T\}.$$ Now if $X, Y$ are both projective over $S$, then the functor $$T \mapsto \underline{\operatorname{Hom}}(X, Y)(T) = \{X_T \to Y_T\}$$ can be embedded into $\underline{\operatorname{Hilb}}_{X \times_S Y/S}(T)$, by sending a morphism $f: X_T \to Y_T$ to its graph $\Gamma_f \subset X_T \times_T Y_T$. This identifies $\underline{\operatorname{Hom}}(X,Y)(T)$ with the subset of schemes $\Gamma \subset X_T \times_T Y_T$, such that the restricted projection $\Gamma \to X$ is an isomorphism. Grothendieck claims that this defines an open subscheme $$\operatorname{Hom}(X,Y) \subset \operatorname{Hilb}_{X \times Y/S}.$$ How can I see that this is the case?

Similarly, Grothendieck claims that $\operatorname{Isom}(X,Y) \subset \operatorname{Hom}(X,Y)$ is an open subset. How can I see this?

[1] Grothendieck, Techniques de construction et théorèmes d'existence en géométrie algébrique IV: les schémas des Hilbert; Séminaire Bourbaki


2 Answers 2


By definition of the Hilbert scheme, we have a "universal family", i.e. a closed subscheme $i : \mathscr{Z} \to \mathrm{Hilb}_{X\times_S Y/S} \times_S (X\times_S Y)$ such that the composite with the projection $\mathrm{Hilb}_{X\times_S Y/S} \times_S (X\times_S Y) \to \mathrm{Hilb}_{X\times_S Y/S}$ is flat.

Let $\alpha : \mathscr{Z} \to \mathrm{Hilb}_{X\times_S Y/S} \times_S X$ be the composite of $i$ with the projection onto the first two factors. Using Stacks 05XD, we have an open subscheme $\mathscr{U} \subseteq \mathrm{Hilb}_{X\times_S Y/S}$ such that any map $f : W \to \mathrm{Hilb}_{X\times_S Y/S}$ of $S$-schemes factors through $\mathscr{U}$ if and only if the basechange $\alpha_W : \mathscr{Z}_W \to (\mathrm{Hilb}_{X\times_S Y/S} \times_S X)_W$ is an isomorphism. Unwinding definitions, we find $\mathscr{U}$ represents exactly the subfunctor $\underline{\mathrm{Hom}}(X,Y)$ you described.

  • $\begingroup$ We can use a similar strategy to get Isom(X, Y). I'll leave that to you :) $\endgroup$
    – Brian Shin
    Oct 28, 2021 at 20:39
  • $\begingroup$ Thanks for your answer. Do you know if there is a similar result to 05XD for schemes only? I'm just learning about algebraic spaces, stacks etc. so I would appreciate a "more basic" approach. $\endgroup$ Oct 28, 2021 at 20:44
  • $\begingroup$ @red_trumpet I believe so. Looking through the proof at Stacks Project, it doesn't appear to use anything essentially specific to algebraic spaces. But I must confess to also being relatively new with algebraic spaces. $\endgroup$
    – Brian Shin
    Oct 28, 2021 at 20:57
  • $\begingroup$ You can delete the sentence "In particular fW is representable." from the proof of Tag 05XD when all algebraic spaces are assumed to be schemes. $\endgroup$ Nov 2, 2021 at 22:07

Another source I found is in Nitsure's Parts 2. Construction of Hilbert and Quot schemes from Funamental Algebraic Geometry - Grothendieck's FGA Explained. He proves this in detail in Theorem 5.23. This section is also available on the arXiv.


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.