Let consider the complex projective space in two ways: Topologically by setting $$\mathbb{P}_n =\mathbb{P}(\mathbb{C}^{n+1}) = \mathbb{C}^{n+1}/{\sim}$$

as space of complex lines and via

$$\mathbb{P}^n = \operatorname{Proj}(\mathbb{C}[T_0, T_1, \ldots, T_n])$$

as $\mathbb{C}$-scheme.

We have the (tautological) invertible sheaf $\mathcal{O}_{\mathbb{P}_n}(1)$ defined locally by $\mathcal{O}_{\mathbb{P}_n}(1) \vert _{D_+(T_i)} = \operatorname{Spec}(\mathbb{C}[T_0, T_1, \ldots, T_n]_{(T_i),1})= \operatorname{Spec}(\mathbb{C}[T_0, T_1, \ldots, T_n])$ where $\mathbb{C}[T_0, T_1, \ldots, T_n]_{(T_i),1}$ means the $1$-graded part of canonically graded ring $\bigoplus _i \mathbb{C}[T_0, T_1, \ldots, T_n]_{(T_i),i}$. Let $\mathcal{O}_{\mathbb{P}_n}(-1)$ be the dual sheaf of $\mathcal{O}_{\mathbb{P}_n}(1)$.

It's easy to see that $\{(l,u)\in \mathbb{P}_n \times \mathbb{C}^{n+1} \vert v \in l\}$ is a line bundle over $\mathbb{P}_n$.

My question is why holds

$$\mathcal{O}_{\mathbb{P}_n}(-1) = \{(l,u)\in \mathbb{P}_n \times \mathbb{C}^{n+1} \mid v \in l\}$$

Remark: I know that we can identify locally free sheaves with vector bundles if global sections of the structure sheaf are given as regular functions $U \to \mathbb{C}$, but I don't see how to conclude the desired equation.

  • $\begingroup$ HINT: Remember that sections $\mathscr O(k)(U)$ are homogeneous polynomials of degree $k$ on $U$. Can you see that $\mathscr O(1)$ is the dual of the tautological line bundle? $\endgroup$ – Ted Shifrin Apr 7 '18 at 3:02
  • $\begingroup$ @TedShifrin Which command do you use to produce that calligraphic $O$? Thanks in advance. $\endgroup$ – Batominovski Apr 7 '18 at 4:48
  • $\begingroup$ $$ \begin{align} & \mathbb{C}^{n+1}/\sim \\ & \mathbb{C}^{n+1}/{\sim} \end{align} $$ I changed the first line above to the second, and did some other MathJax corrections. There's a common-sense reason why you should expect to need to do it as on the second line above. $\endgroup$ – Michael Hardy Apr 7 '18 at 6:10

Let $V$ be a vector space, and $\Bbb P(V) = X$ be the space of its lines. Write $\mathcal O_X(-1)'$ for the line bundle $\{(l,v) \in X \times V : v \in l\}$.

We will build an isomorphism $\mathcal O_X(-1)' \times \mathcal O_X(1) \to \mathcal O_X$. If $s'\in \Gamma(U,O(-1)')$ we remark that $s'$ is the same as an algebraic function $v : U \to V$ where $v(u) \in u$ for each $u \in U$. A section $s$ $\in \Gamma(U,\mathcal O(1))$ is just a regular section on $U$ and of degree $1$, say $\varphi$.

Warning : it does not define a function, e.g the section $x_0$ has no well defined value at $[1:0:0]$. There are two ways to avoid this problem : first, working with quotients of sections gives a well defined map (but this is not regular anymore), or you can also consider $\varphi$ as a regular function on $W = \{v \in V : [v] \in U\}$.

In particular we obtain an "evaluation map" $$\text{ev} : \mathcal O_X(-1) \otimes \mathcal O_X(1) \to \mathcal O_X, (v, \varphi) \mapsto \varphi(v) : (u \mapsto \varphi(v(u)))$$ which is easily checked to be an isomorphism.

  • $\begingroup$ Hi. Thank you for the answer. One step stays unclear: Why are the elements $s' \in \Gamma(U,O(-1)')$ identical to algebraic functions $v: U \to V$ with $v(u) \in u$ property? $\endgroup$ – KarlPeter Apr 11 '18 at 21:35
  • $\begingroup$ @KarlPeter : Hi ! It's because the total space of $\mathcal O(-1)'$ is $E' =: \{(v,l) \in V \times \Bbb P(V) : v \in l \} $ and the projection $p : E' \to \Bbb P(V)$ is exactly the restriction of the projection $V \times \Bbb P(V) \to \Bbb P(V)$. In particular, $s' \in \Gamma(U, O(-1)')$ is just a map $s' : U \to E'$ with $p \circ s' = id$. This is exactly algebraic functions $v : U \to V$ with the property that $v(u) \in u$. $\endgroup$ – Nicolas Hemelsoet Apr 12 '18 at 9:43
  • $\begingroup$ I noticed that in your definition of $ev$ there occurs following problem: By your explanations above $v \in \Gamma(U, O(-1)')$ is essentially a map $v: U \to V$ and $\varphi \in \Gamma(U, O(1))$ a regular function, especially a map $U \to F$ for $F= \mathbb{R}, \mathbb{C}$. What does then the expression $\varphi(v(u))$ here mean? Since $v(u) \in V$ we get a domain problem,right? $\endgroup$ – KarlPeter Nov 15 '18 at 0:37
  • $\begingroup$ I edited my answer. $\endgroup$ – Nicolas Hemelsoet Nov 15 '18 at 5:53

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.