Home / Blog / Faisceaux Algébriques Cohérents 3 – coherent sheaves on projective space

Faisceaux Algébriques Cohérents 3 – coherent sheaves on projective space

Published on January 11, 2018
Reading time 18 minutes

As we know, cohomology of sheaves on affine schemes is boring. The next simplest case we can consider is sheaves on projective schemes.

7. The \( \operatorname{Proj} \) construction#

We can define projective space by gluing affine space together, but there is a more canonical construction. By a graded ring \( S \), we implicitly assume that \( S = \bigoplus _ {d \ge 0} S _ d \).

Definition 1. The scheme \( \operatorname{Proj} S \) is defined as the following. A point in \( \operatorname{Proj} S \) is a homogeneous prime ideal \( \mathfrak{p} \) of \( S \) such that \( \mathfrak{p} \) doesn’t contain all of \( S _ 1, S _ 2, \ldots \). The topology is generated by the sets \( D(f) = \{ \mathfrak{p} : f \notin \mathfrak{p}\} \) for homogeneous \( f \in S _ d \) with \( d \ge 1 \), and the structure sheaf on \( D(f) \) is given by the \( \mathrm{Spec} \) of the \( 0 \)-th graded piece \( (S _ f) _ 0 \).

Example 1. If \( S = A[x _ 0, \ldots, x _ n] \), we have \( \operatorname{Proj} S = \mathbb{P} _ A^n \).

So any kind of section needs to come from a \( 0 \)-th graded element in the ring. This makes sense, because on projective space, functions like \( x \) or \( y^2 / (z+1) \) don’t give well-defined values. On affine schemes, we were able to construct sheaves on \( \mathrm{Spec} A \) from \( A \)-modules. Similarly, we can construct sheaves on \( \operatorname{Proj} S \) from graded \( S \)-modules.

Definition 2. Let \( S \) be a graded ring, and let \( M = \bigoplus _ n M _ n \) be a graded \( S \)-module. We define a quasi-coherent sheaf \( \tilde{M} \) on \( \operatorname{Proj} S \), so that \( \tilde{M} \vert _ {D(f)} \cong ((M _ f) _ 0)^{\tilde{}} \).

The sheaves are glued via, where \( \deg f = a \) and \( \deg g = b \), the isomorphisms \( ((M _ f) _ 0) _ {g^a/f^b} \cong (M _ {fg}) _ 0 \cong ((M _ g) _ 0) _ {f^b / g^a} \). We immediately see that this is a quasi-coherent sheaf on \( \operatorname{Proj} S \). Of course, this construction is functorial in \( M \), and a short exact sequence \( 0 \rightarrow M _ 1 \rightarrow M _ 2 \rightarrow M _ 3 \rightarrow 0 \) induces an exact sequence \[ \displaystyle 0 \rightarrow \tilde{M} _ 1 \rightarrow \tilde{M} _ 2 \rightarrow \tilde{M} _ 3 \rightarrow 0 \] of quasi-coherent sheaves, because the construction is exact affine-locally.

But for graded modules, we have an interesting operation called "twisting". Namely, for each integer \( n \), we can shift the indices to define a new graded module \( M(n) \), so that \[ \displaystyle M(n) _ k = M _ {n+k}. \] This makes huge difference, because sections are always supposed to be degree \( 0 \) elements, and are now degree \( n \) elements of \( M \). An analogous definition can be made for quasi-coherent sheaves in general.

Definition 3. Let \( \mathscr{F} \) be a quasi-coherent sheaf on \( X = \operatorname{Proj} S \), where \( S \) is a graded ring generated by \( S _ 1 \) as an \( S _ 0 \)-algebra. Note that this condition ensures that \( X = \bigcup _ {f \in S _ 1} D(f) \). For each \( f \) homogeneous of degree \( 1 \), we let \( \mathscr{F}(n) _ f = \mathscr{F}(n) \vert _ {D(f)} \cong \mathscr{F} \vert _ {D(f)} \). But we glue them by the isomorphisms \( \mathscr{F}(n) _ f \vert _ {D(fg)} \rightarrow \mathscr{F}(n) _ g \vert _ {D(fg)} \) given by \( s \mapsto s \cdot f^n / g^n \). On triple intersections, the gluing maps commute, so this indeed defines a quasi-coherent sheaf.

Let \( S \) be a graded ring generated by \( S _ 1 \) as an \( S _ 0 \)-algebra, and let \( X = \operatorname{Proj} S \). Here are some properties that can be easily verified: (Hartshorne [Har77] III.5.12)

  • For any graded \( S \)-module \( M \), we have \( \tilde{M}(n) \cong (M(n))^{\tilde{}} \).
  • The sheaf \( \mathscr{O} _ X(n) \) is invertible, i.e. locally free, for all \( n \).
  • We have \( \mathscr{O} _ X(n) \otimes _ {\mathscr{O} _ X} \mathscr{F} \cong \mathscr{F}(n) \) for every quasi-coherent sheaf \( \mathscr{F} \).

One important remark is that to define \( \mathscr{F}(n) \), we need to specify what \( S \) is. If \( \operatorname{Proj} S \cong \operatorname{Proj} T \) for two graded rings \( S \) and \( T \), the sheaves \( \mathscr{F}(n) \) can be defined with either of \( S \) and \( T \), and there is no reason for them to be isomorphic.

Over a general scheme, we can also define all these twisted sheaves \( \mathscr{F}(n) \), as long as we have chosen an invertible sheaf \( \mathscr{O} _ X(1) \), because we can define \( \mathscr{F}(n) = \mathscr{F} \otimes \mathscr{O} _ X(1)^{\otimes n} \). If \( X \) is defined over \( Y \), an invertible sheaf \( \mathscr{L} \) is called very ample if there is a locally closed immersion \( i : X \rightarrow \mathbb{P} _ Y^n \) such that \( \mathscr{L} = i^\ast(\mathscr{O}(1)) \). We can develop the theory of twisted sheaves in this setting, but I prefer not to.

Example 2. Let us take the projective space \( S = A[x _ 0, \ldots, x _ n] \) and \( X = \mathbb{P}^n = \operatorname{Proj} S \), and try to compute cohomology of \( \mathcal{O} _ X(m) \). On each open \( D(x _ i) \cong \mathrm{Spec} A[x _ 0 / x _ i, \ldots, x _ n / x _ i] \), the sheaf \( \mathscr{O} _ X(m) \) looks like the homogeneous degree \( m \) polynomials in \( A[x _ 0, \ldots, x _ n, x _ i^{-1}] \). Then the Čech complex looks like the degree \( m \) components of \[ \displaystyle 0 \rightarrow \prod _ {i}^{} A[x _ 0, \ldots, x _ n, x _ i^{-1}] \rightarrow \prod _ {i < j}^{} A[x _ 0, \ldots, x _ n, x _ i^{-1}, x _ j^{-1}] \rightarrow \cdots. \] Because each is a direct sum of a copy of \( A \) corresponding to each (Laurent) monomial, we can compute them separately and arrive at \( H^k(\mathbb{P} _ A^n, \mathscr{O} _ X(\bullet)) = 0 \) for \( 0 < k < n \) and \[ \displaystyle \begin{aligned} H^0(\mathbb{P} _ A^n, \mathscr{O} _ X(\bullet)) & \displaystyle\cong A[x _ 0, \ldots, x _ n], \\ H^n(\mathbb{P} _ A^n, \mathscr{O} _ X(\bullet)) & \displaystyle\cong (x _ 0 \cdots x _ n)^{-1} A[x _ 0^{-1}, \ldots, x _ n^{-1}]. \end{aligned} \] So \( H^0(\mathbb{P} _ A^n, \mathscr{O} _ X(m)) \) is a free \( A \)-module of rank \( \binom{m+n}{n} \) for \( m \ge 0 \) and \( H^n(\mathbb{P} _ A^n, \mathscr{O} _ X(m)) \) is free of rank \( \binom{-m-1}{n} \) for \( m \le -n-1 \).

8. Between graded modules and quasi-coherent sheaves#

We have seen how to construct a quasi-coherent sheaf from a module. In the affine case, there was an equivalence of categories between modules and quasi-coherent sheaves. To go from quasi-coherent sheaves to modules, we only needed to take the global sections. Will something similar happen over \( \operatorname{Proj} S \) as well?

Definition 4. Let \( \mathscr{F} \) be a quasi-coherent sheaf over \( X = \operatorname{Proj} S \), where \( S \) is generated by \( S _ 1 \) as an \( S _ 0 \)-algebra. Note that any element \( f \in S _ n \) can be thought of as a global section in \( \mathscr{O} _ X(n) \). Then we define a graded \( S \)-module \( \Gamma(\mathscr{F}) \) by \[ \displaystyle \Gamma(\mathscr{F}) _ m = \Gamma(X, \mathscr{F}(m)). \] The multiplication structure is given by \( (f, s) \mapsto f \otimes s \in \Gamma(X, \mathscr{O} _ X(m) \otimes \mathscr{F}(n)) = \Gamma(\mathscr{F}) _ {m+n} \).

In general, if \( s \in M _ m \) is any homogeneous element, it can be regarded as a global section in \( \Gamma(X, \mathscr{F}(m)) \) because it can be viewed locally and can be glued well. Conversely, if \( \mathscr{F} \) is a quasi-coherent sheaf, each global section of \( \mathscr{F}(n) \) restricts a section \( \Gamma(D(f), \mathscr{F}) \). This gives a map \( (\Gamma(\mathscr{F}))^{\tilde{}} \rightarrow \mathscr{F} \) over \( D(f) \) for each \( f \in S _ 1 \), and they glue well. So there are canonical maps \[ \displaystyle \alpha : M \rightarrow \Gamma(\tilde{M}), \quad \beta : (\Gamma(\mathscr{F}))^{\tilde{}} \rightarrow \mathscr{F}. \]

Example 3. We have seen that \( \Gamma(\mathbb{P} _ A^n, \mathcal{O}(m)) = H^0(\mathbb{P} _ A^n, \mathcal{O}(m)) \) is isomorphic to the space of degree \( m \) homogeneous polynomials in \( S = A[x _ 0, \ldots, x _ n] \). So \( \alpha : S \rightarrow \Gamma(\mathcal{O} _ X) \) is an isomorphism.

We expect these two operations \( \Gamma \) and \( \tilde{} \) to be somewhat inverses to each other.

Proposition 5 (Hartshorne II.5.15). Let \( S \) be a graded ring that is finitely generated by \( S _ 1 \) over \( S _ 0 \), (this just means that \( S \) is a quotient of \( S _ 0[x _ 0, \ldots, x _ n] \)) and let \( X = \operatorname{Proj} S \). If \( \mathscr{F} \) is a quasi-coherent sheaf over \( X \), then \( \beta : \Gamma(\mathscr{F})^{\tilde{}} \rightarrow \mathscr{F} \) is an isomorphism.

Proof. Let \( f _ 1, \ldots, f _ n \in S _ 1 \) generate \( S \) over \( S _ 0 \). Working affine-locally, we only need to check that the map of modules are bijective. To show injectivity is to show that if a global section \( s \in \mathscr{F}(m) \) vanishes over \( D(f _ i) \), then \( s \otimes f _ i^c = 0 \) in \( \Gamma(X, \mathscr{F}(m+c)) \) for a sufficiently large \( c \). But on each chart \( D(f _ j) \), the section vanishing on \( D(f _ i) \) means that \( s \in \Gamma(D(f _ j), \mathscr{F}) \) will localize to \( 0 \) in \( \Gamma(D(f _ j), \mathscr{F}) _ {f _ i/f _ j} \). Then \( s \cdot (f _ i / f _ j)^c = 0 \) for all \( j \) and large enough \( c \), and this just means that \( s \otimes f _ i^c \) vanishes in \( \mathscr{F}(m+c) \). To show surjectivity is to show that every section of \( D(f _ i) \) actually comes from a global section of \( \mathscr{F}(m+c) \). This is just because \( s \in \Gamma(D(f _ i), \mathscr{F}(m)) \) restricts to an element of \( \Gamma(D(f _ i f _ j), \mathscr{F}(m)) = \Gamma(D(f _ j), \mathscr{F}(m)) _ {f _ i/f _ j} \), and so we can find \( s _ j \) such that \( s _ j \) and \( s \) glue well in \( \mathscr{F}(m+c) \) for large \( c \). We might worry about \( s _ j \) and \( s _ {j^\prime} \) not gluing well, but injectivity tells us that they will agree once we set \( c \) higher. ▨

What about \( \alpha : M \rightarrow \Gamma(\tilde{M}) \)? It can’t be that \( \alpha \) is also an isomorphism in most cases. Consider the class \( \mathcal{C} \subseteq \mathsf{GrMod} _ {S} \) consisting of modules \( M \) such that \( M _ n = 0 \) for all sufficiently large \( n \). This is a Serre class and so we can consider the "localization" \( \mathsf{GrMod} _ S / \mathcal{C} \). It is not difficult to see that \( \tilde{} : \mathsf{GrMod} _ S \rightarrow \mathsf{QCoh} _ {\operatorname{Proj} S} \) factors through \( \mathsf{GrMod} _ S / \mathcal{C} \), that is, \( \mathcal{C} \)-bijective graded modules induce isomorphic quasi-coherent sheaves.

Proposition 6 (Serre n63 Proposition 3). Let \( S = k[x _ 0, \ldots, x _ n] \) with \( k \) a field, and let \( M \) be a finitely generated graded module (or \( \mathcal{C} \)-isomorphic to one). Then \( \alpha : M \rightarrow \Gamma(\tilde{M}) \) is an \( \mathcal{C} \)-isomorphism. Moreover, \( H^i(\mathbb{P} _ k^n, \tilde{M}(\bullet)) \rightarrow 0 \) is an \( \mathcal{C} \)-isomorphism for \( i > 0 \).

Proof. By Hilbert’s syzygy theorem (here’s a nice proof written by Tom Lovering), there exists a finite graded resolution of \( M \) \[ \displaystyle 0 \rightarrow F^s \rightarrow F^{s-1} \rightarrow \cdots \rightarrow F^0 \rightarrow M \rightarrow 0 \] by finite rank free modules. (By free, I mean that each is isomorphic to \( \bigoplus _ i S(n _ i) \). All maps should preserve grading.) We’re going to prove this by induction on the minimal length of such a resolution. If \( M \) is free, we have already checked this in Example 2. For longer lengths, consider a short exact \( 0 \rightarrow N \rightarrow F \rightarrow M \rightarrow 0 \) where \( F \) is finite rank free and \( N \) has a shorter resolution than \( M \). We have a short exact sequence \( 0 \rightarrow \tilde{N}(\bullet) \rightarrow \tilde{F}(\bullet) \rightarrow \tilde{M}(\bullet) \rightarrow 0 \) of sheaves (note that twisting is exact), and then a long exact sequence \[ \displaystyle \cdots \rightarrow H^i(X, \tilde{N}(\bullet)) \rightarrow H^i(X, \tilde{F}(\bullet)) \rightarrow H^i(M, \tilde{M}(\bullet)) \rightarrow \cdots. \] But all \( H^i(X, \tilde{N}(\bullet)) \) and \( H^i(X, \tilde{F}(\bullet)) \) are \( \mathcal{C} \)-isomorphic to \( 0 \) by the induction hypothesis, except for \( i = 0 \) in which case they are \( \mathcal{C} \)-isomorphic to \( N \) and \( F \). This shows that \( 0 \rightarrow N \rightarrow F \rightarrow \Gamma(X, \tilde{M}) \rightarrow 0 \) is a \( \mathcal{C} \)-exact sequence and \( H^i(X, \tilde{M}(\bullet)) \) is \( \mathcal{C} \)-isomorphic to \( 0 \). Therefore \( \alpha \) is an \( \mathcal{C} \)-isomorphism. ▨

For a quasi-coherent sheaf \( \mathscr{F} \) on \( \mathbb{P} _ k^n \) (or any locally Noetherian scheme in general), let us say that it is coherent if on each affine, the sheaf corresponds to a finitely generated module. Denote by \( \mathsf{Coh} _ X \) the full subcategory of coherent sheaves. It is not hard to see that a finitely generated graded module \( S \) induces a coherent sheaf. Conversely, a coherent sheaf induces a \( \mathcal{C} \)-finitely generated graded module.

Proposition 7 (Serre n60 Theorem 2). If \( \mathscr{F} \) be a coherent sheaf on \( \mathbb{P} _ k^n \), then \( \Gamma(\mathscr{F}) \) is \( \mathcal{C} \)-isomorphic to a finitely generated \( k[x _ 0, \ldots, k _ n] \)-module.

Proof. If \( \mathscr{F} \) is a coherent sheaf on \( \mathbb{P} _ k^n \), it is generated by a finite number of sections over \( D(f _ i) \), and hence finitely generated by global sections of \( \mathscr{F}(n) \) for \( n \) sufficiently large. This gives a surjection \( \mathscr{L}^0 \rightarrow \mathscr{F} \rightarrow 0 \) of sheaves, where \( \mathscr{L}^0 \) is a finite direct sum of \( \mathscr{O} _ X(m) \). The kernel is also coherent, and hence we can continue it to an exact \( \mathscr{L}^1 \rightarrow \mathscr{L}^0 \rightarrow \mathscr{F} \) where \( \mathscr{L}^i \) are finite direct sums of \( \mathscr{O} _ X(m) \). If we let \( M \) to be the cokernel \( \Gamma(\mathscr{L}^1) \rightarrow \Gamma(\mathscr{L}^0) \rightarrow M \rightarrow 0 \), then \( M \) is clearly finitely generated, because \( \Gamma(\mathscr{L}^0) \) is. If we apply the \( \tilde{} \) construction, we get an exact \( \mathscr{L}^1 \rightarrow \mathscr{L}^0 \rightarrow \tilde{M} \rightarrow 0 \), and this shows that \( \tilde{M} \cong \mathscr{F} \). Applying \( \Gamma \) shows that \( \Gamma(\mathscr{F}) \cong \Gamma(\tilde{M}) \), and \( \Gamma(\tilde{M}) \) is \( \mathcal{C} \)-isomorphic to \( M \), which is finitely generated. ▨

Corollary 8. For \( S = k[x _ 0, \ldots, x _ n] \) and \( X = \operatorname{Proj} S \), the functors \( \tilde{} : \mathsf{FinGrMod} _ S / \mathcal{C} \rightarrow \mathsf{Coh} _ {X} \) and \( \Gamma : \mathsf{Coh} _ {X} \rightarrow \mathsf{FinGrMod} _ S / \mathcal{C} \) exhibit an equivalence of the two categories.

If you’re really interested in what the essential image of \( \Gamma : \mathsf{Coh} _ X \rightarrow \mathsf{GrMod} _ S \) is, Serre has the following proposition.

Proposition 9 (Serre n67 Proposition 9). Let \( S = k[x _ 0, \ldots, x _ n] \). The essential image of \( \Gamma : \mathsf{Coh} _ X \rightarrow \mathsf{GrMod} _ S \) consists of \( \mathcal{C} \)-finitely generated modules \( M \) satisfying:

  • If \( m \in M \) such that \( x _ i m = 0 \) for all \( i \), then \( m = 0 \).
  • If homogeneous elements \( m _ i \in M \) of the same degree satisfy \( x _ j m _ i = x _ i m _ j \) for all \( i, j \), then there exists an \( m \in M \) such that \( m _ i = x _ i m \).

9. The Hilbert polynomial and Euler characteristic#

We can now discuss one application of all the theory we have developed so far. Let \( k \) be a field, and consider a graded ring \( S \) such that \( S _ 0 = k \) and \( S \) is finitely generated by \( S _ 1 \) over \( S _ 0 = k \). As we have said, this corresponds to a closed subscheme of \( \mathbb{P} _ k^n \) for some \( n \). Let us write \( X = \operatorname{Proj} S \) as usual.

Proposition 10. If \( X = \operatorname{Proj} S \) as above, and \( \mathscr{F} \) is a coherent sheaf over \( X \), then each \( H^i(X, \mathscr{F}) \) is a finite-dimensional \( k \)-vector space.

Proof. For \( \mathscr{F} \) a coherent sheaf over \( X \), we know that \( \Gamma(\mathscr{F}) \) is \( \mathcal{C} \)-isomorphic to a finitely generated \( M \). Then \( \mathscr{F} \cong \tilde{M} \). If we have a short exact sequence \( 0 \rightarrow N \rightarrow F \rightarrow M \rightarrow 0 \) where \( F \) is a finitely rank free module, then we get an exact \( 0 \rightarrow \tilde{N} \rightarrow \tilde{F} \rightarrow \tilde{M} \) and thus a long exact \[ \displaystyle \cdots \rightarrow H^i(X, \tilde{N}) \rightarrow H^i(X, \tilde{F}) \rightarrow H^i(X, \tilde{M}) \rightarrow H^{i+1}(X, \tilde{N}) \rightarrow \cdots. \] We know that each \( H^i(X, \tilde{F}) \) is a finite-dimensional \( k \)-vector space, because we exactly know the dimensions from Example 2. So if each \( H^i(X, \tilde{N}) \) has finite dimension, then each \( H^i(X, \tilde{M}) \) also has finite dimension. Thus we can prove this by induction on the minimal length of a finite rank free resolution. (Again, we’re using Hilbert’s syzygy theorem.) ▨

So let us define the number \[ \displaystyle h^i(X, \mathscr{F}) = \dim _ k H^i(X, \mathscr{F}) \] for a coherent sheaf \( \mathscr{F} \). If \( 0 \rightarrow \mathscr{F} \rightarrow \mathscr{G} \rightarrow \mathscr{H} \rightarrow 0 \) is exact, then we get a long exact \[ \displaystyle \cdots \rightarrow H^i(X, \mathscr{F}) \rightarrow H^i(X, \mathscr{G}) \rightarrow H^i(X, \mathscr{H}) \rightarrow H^{i+1}(X, \mathscr{F}) \rightarrow \cdots. \] If we define the Euler characteristic of \( \mathscr{F} \) as \[ \displaystyle \chi(X, \mathscr{F}) = \sum _ {i=0}^{\infty} (-1)^i h^i(X, \mathscr{F}), \] (note that this is a finite sum because cohomology vanishes for \( i > n \)) it follows that \( \chi(X, \mathscr{F}) + \chi(X, \mathscr{H}) = \chi(X, \mathscr{G}) \).

Proposition 11. For any coherent \( \mathscr{F} \), the Euler characteristics \( \chi(X, \mathscr{F}(m)) \) is a polynomial in \( m \).

Proof. We again use induction on the length of the resolution. We first check that if \( \mathscr{F} = \mathscr{O} _ X \), then \[ \displaystyle \chi(X, \mathscr{O} _ X(m)) = \frac{(m+1) (m+2) \cdots (m+n)}{n!}. \] So \( \chi(X, \mathscr{L}(m)) \) is a polynomial in \( m \) if \( \mathscr{L} \) is a direct sum of a finite number of \( \mathscr{O} _ X(m _ i) \). Also, if \( 0 \rightarrow \mathscr{G} \rightarrow \mathscr{L} \rightarrow \mathscr{F} \rightarrow 0 \) is exact and \( \chi(X, \mathscr{G}(m)) \) is a polynomial, then \( \chi(X, \mathscr{F}(m)) = \chi(X, \mathscr{L}(m)) - \chi(X, \mathscr{G})) \) is also a polynomial. ▨

Definition 12. This polynomial \( p _ X(m) = \chi(X, \mathscr{O} _ X(m)) \) is called the Hilbert polynomial of \( m \).

Corollary 13. The dimension \( \dim _ k S _ m \) is equal to a polynomial in \( m \), namely \( p _ X(m) \), for \( m \) sufficiently large.

Proof. Note that we have proven that \( H^i(X, \tilde{M}(m)) = 0 \) and \( H^0(X, \tilde{M}(m)) \cong M _ m \) for \( m \) sufficiently large and \( i > 0 \). Let \( M = S \). ▨

Again, the Hilbert polynomial is not well-defined if we’re only given the scheme \( X \). Because twisting is defined when there is an embedding \( X \hookrightarrow \mathbb{P} _ k^n \) (or just a very ample line bundle), this data is needed to define the polynomial. But \( \mathscr{F}(0) = \mathscr{F} \) always, and so we can evaluate \( p _ X(0) \) without any trouble.

Definition 14. We define the Euler characteristic of \( X \) as \[ \displaystyle \chi(X) = \chi(X, \mathscr{O} _ X) = p _ X(0) = \sum _ {i = 0}^{\infty} (-1)^i h^i(X, \mathscr{O} _ X). \]

Corollary 15. The value of \( p _ X(0) \) doesn’t depend on the embedding \( X \hookrightarrow \mathbb{P} _ k^n \) (or, the very ample line bundle \( \mathscr{O} _ X(1) \)).

This can be used to define the arithmetic genus of a curve. If \( X \) is a curve and \( g \) is its genus, they will satisfy the formula \( \chi = 1 - g \). For instance, the Hilbert polynomial for \( \mathbb{P} _ k^1 \) is \( \chi _ {\mathbb{P} _ k^1}(m) = \frac{1}{2} (m+1) (m+2) \) and so \( \chi(\mathbb{P} _ k^1) = 1 \). For an elliptic curve \( E \hookrightarrow \mathbb{P} _ k^2 \), we have \( \chi _ E(m) = 3m \) and so \( \chi(E) = 0 \).

References#

[Har77] Robin Hartshorne, Algebraic geometry, Springer-Verlag, New York-Heidelberg, 1997, Graduate Texts in Mathematics, No. 52.

[Ser55] Jean-Pierre Serre, Faisceaux Algébriques Cohérents, Ann. of Math. (2) 61 (1955), 197–278.