3
$\begingroup$

Let $f: X \to Y$ be a morphism of schemes over a field. I'd like to know why the sections of the sheaf $f ^* \Omega _Y$ are given by $d \phi$ for $\phi \in \mathscr O_Y$.

Does this mean that e.g. $f ^* \Omega _Y(X)=\Omega _Y(Y)$, without any additional constraints? Why can the sections of $f ^* \Omega _Y$ be seen this way?

$f^*$ is defined by $f ^* \mathscr F = f^{-1} \mathscr F \otimes_{f^{-1}\mathscr O_Y} \mathscr O_X$, and it is known that $(f^* \mathscr O_Y^{\oplus n})(U)= (\mathscr O_X^{\oplus n})(f^{-1}(U))$, so since $\Omega_Y$ is locally-free it follows that $\Omega_X$ is locally-free too. This seems "close" to what we want to show. How does the correspondence between sections of $f ^* \Omega _Y$ and elements $d\phi$ work?

$\endgroup$
4
$\begingroup$

It is true that there is a canonical morphism of sheaves of $\mathcal O_X$-Modules $\phi: f^*\Omega_Y \to \Omega_X$.
By adjunction this morphism corresponds to an easier to understand morphism of $\mathcal O_Y$-Modules $\psi:\Omega_Y\to f_*\Omega_X$ :
Given an open subset $U\subset X$ the morphism $\psi_U:\Omega_Y(U)\to f_*\Omega_X(U)=\Omega_X (f^{-1}U)$ associates to a differential form $\omega \in \Omega_Y(U)$ the lifted differential form $f^*(\omega)\in \Omega _X(f^{-1}(U)$, just as in differential topology.

Beware however that $f^*\Omega_Y$ and $\Omega_X$ are not isomorphic in general and even their spaces of global sections $f^*\Omega_Y(X)$ and $\Omega_X(X)$ need not be isomorphic:
For example if $X$ is a smooth complete curve of genus $g\neq 1$ over the field $k$ and if $f:X\to Y=\operatorname {Spec}(k)$ is the morphism to a point, then $f^*\Omega_Y=f^*\mathcal O_Y=\mathcal O_X$ so that $$f^*\Omega_Y(X)=\mathcal O_X(X)=k\neq \Omega_X(X)=k^g $$
Finally let me mention that your question "Does this mean that e.g. $f ^* \Omega _Y(X)=\Omega _Y(Y)$ ?" also has a negative answer :
If $p:P= \operatorname {Spec}(k) \to Z$ is the inclusion of a point into a smooth projective curve of degree $g\gt1$ we have $$p ^* \Omega _Z(P)=\mathcal O_P(P)=k\neq \Omega _Z(Z)=k^g$$

$\endgroup$
  • $\begingroup$ Thank you for your answer. Gathmann claims that the canonical morphism $f^* \Omega_Y \to \Omega_X$ is given by $d \psi \mapsto d(\psi \circ f)$. I wonder what does this formula mean, in particular why the elements of $f^* \Omega_Y$ where denoted by $d (\text{function in }O_Y)$. Is there an intuitive explanation for what are the sections of the pullback when e.g. $f$ is an inclusion $X \to \Bbb P^n$? $\endgroup$ – Emolga Oct 28 '16 at 6:03
  • 1
    $\begingroup$ The intuitive explanation is that a differential form on $X$ is the linear form on $TX$ obtained by restricting a linear form on $T\mathbb P^n$ to $TX$. However the definition in algebraic geometry is very formal and uses an algebraic gadget called "Kähler differentials" which has as essential virtue that $d(uv)=udv+vdu$ for elements of the base ring. Regular local functions $\psi$ thus have a differential $d\psi$ and Gathmann's formula for the morphism $f^*\Omega_Y \to \Omega_X$ is indeed correct. The reason for that formalism is the possibility of study of varieties with singularities. $\endgroup$ – Georges Elencwajg Oct 28 '16 at 6:59

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.