1 Introduction
It has been pointed out to us by Professor Y. Matzusawa that the proof of Theorem 2.3 in [Reference Barrios1] contains a mistake. Furthermore, he found a counterexample to the statement of the theorem using the strict transform of a blow-up (see [Reference Matsuzawa5, Example 1.10]). Fortunately, such examples (negatively embedded varieties) are essentially the only thing that can fail. In fact, in this note, we show that all the results of our work [Reference Barrios1] remain correct under the assumption that the subvariety under consideration is positively embedded: that is, its normal bundle is ample. As in our original argument, the proof goes by induction on the dimension. However, this time, the argument is more algebraic rather than geometric. We remark that Professor Y. Matzusawa found an alternative way to prove a weaker Riemann–Roch inequality which nonetheless does not require positivity assumptions (see [Reference Matsuzawa5, Lemma 2.1]).
Theorem 3.2 below replaces [Reference Barrios1, Theorem 2.3]. The only difference is that now we require that the normal bundle be ample. Importantly, [Reference Barrios1, Theorem 3.2] about height bounds for rational points remains correct under the same additional assumption. Examples of subvarieties with ample normal bundle are abundant; for instance, complete intersections of ample hypersurfaces.
We thank Professor Y. Matzusawa for carefully reading a previous draft of this note and for providing valuable feedback.
2 Algebra preliminaries
Lemma 2.1. Let A be a ring and
$\mathfrak {a},\mathfrak {b}$
be ideals of A. Then, for any positive integer r,
Proof. Consider an element
$x{\kern-1pt}\in{\kern-1pt} {(\mathfrak {a}{\kern-1pt}+{\kern-1pt}\mathfrak {b})^r}/{\mathfrak {b}}$
. Then,
$x{\kern-1pt}={\kern-1pt}\sum x_i\cdots x_r{\kern-1pt}+{\kern-1pt}b$
, where
${x_i{\kern-1pt}\in{\kern-1pt} \mathfrak {a}+\mathfrak {b}}$
and
$b\in \mathfrak {b}$
. Let
$x_i=a_i+b_i$
and
$x_j=a_j+b_j$
be elements in
$\mathfrak {a}+\mathfrak {b}$
. We note that
$x_i\cdot x_j=~a_1a_2+b_{ij}$
, so that
$x_ix_j\in \mathfrak {a}^r+\mathfrak {b}$
. Thus,
$\sum x_1\cdots x_r\in \mathfrak {a}^r+\mathfrak {b}$
and, therefore,
$x\in ~{(\mathfrak {a}^r+\mathfrak {b})}/{\mathfrak {b}}$
.
On the other hand, if
$x\in {(\mathfrak {a}^r+\mathfrak {b})}/{\mathfrak {b}}$
, then
$x=(\sum a_1\cdots a_r+b)+\overline {b}$
, where
$a_i\in \mathfrak {a}$
and
$b,\overline {b}\in \mathfrak {b}$
. Since
$\mathfrak {a}\subset \mathfrak {a}+\mathfrak {b}$
, the
$a_i$
are elements in
$\mathfrak {a}+\mathfrak {b}$
. Therefore,
$\sum a_i\cdots a_r\in (\mathfrak {a}+\mathfrak {b})^r$
giving the other inclusion.
Lemma 2.2. Let A be a ring and
$\mathfrak {a},\mathfrak {b}$
be ideals of A. Then, for any positive integer r,
Proof. Consider the exact sequence
Tensoring by
$A/\mathfrak {b}$
,
The kernel of the first morphism is
$(\mathfrak {a}^r+\mathfrak {b})/\mathfrak {b}$
, so Lemma 2.1 gives the result.
Lemma 2.3. Let A be a Noetherian integral domain. Let h be a prime element in A and
$\mathfrak {p}\subset A$
be a prime ideal generated by a regular sequence. If
$h\not \in \mathfrak {p}$
, then
$h\mathfrak {p}^r=(h)\cap \mathfrak {p}^r$
for any positive integer r.
Proof. The inclusion
$h\mathfrak {p}^r\subseteq (h)\cap \mathfrak {p}^r$
is always true. We need to show the other inclusion.
Let x be an element in
$(h)\cap \mathfrak {p}^r$
. Since
$x\in (h)$
, we have
$x=ha$
for some
$a\in A$
. As
$h\not \in \mathfrak {p}$
, then
$h\not \in \mathfrak {p}^r$
. Now,
$\mathfrak {p}^r$
is primary by [Reference Hochster3, Theorem 2.1] and as
$x=ha\in \mathfrak {p}^r$
, we have
$a\in \mathfrak {p}^r$
. Therefore,
$x=ha\in h\mathfrak {p}^r$
and we obtain the result.
3 Riemann–Roch
As in [Reference Barrios1], let X be a projective variety over a field K,
$Y\subset X$
a subvariety of X also defined over K and
$\mathcal {A}$
a very ample line sheaf on X. Let
$r \gg 1$
.
Lemma 3.1. Assume
$X,Z,\mathcal {A}$
and r are as above. Denote by
$i:X\to \mathbb {P}^n$
the inclusion defined by
$\mathcal {A}$
and let
$H\subseteq \mathbb {P}^n$
be a general hyperplane. Consider
$X'=X\cap H$
and
$Z'=~Z\cap H$
. Then,
Proof. The assertion follows from the isomorphisms
where the second isomorphism follows from Lemma 2.2.
Theorem 3.2. Let X be a smooth projective variety defined over a number field k and let Z be a smooth irreducible subvariety of dimension
$d\geq 1$
also defined over k such that the normal bundle
$\mathcal {N}_{Z/X}$
is ample on Z. Consider
$\mathcal {I}_Z$
, the ideal sheaf defined by Z. Then, for any positive integer r and any very ample line sheaf
$\mathcal {A}$
on X,
where
$c=\dim X-d$
.
Proof. By [Reference Hartshorne2, Theorem 8.18, II], we can choose an element
$f\in H^0(X,\mathcal {A})$
such that the zero locus of f, denoted by
$X'$
, is a smooth and irreducible subvariety of X, and such that if
$d\geq 2$
, then
$Z'=Z\cap X'$
is irreducible.
Consider now the exact sequence
Tensoring by
$\mathcal {O}_X/\mathcal {I}_Z^r$
, we obtain
Lemma 2.3 implies
$\mathrm {Tor}^{\mathcal {O}_X}(\mathcal {O}_X/\mathcal {I}_{X'}~;\mathcal {O}_X/\mathcal {I}_Z^r)$
is trivial, so the sequence is
Since
$\mathcal {A}$
is flat,
Finally, by the choice of f, we have
$\mathcal {I}_{X'}=\mathcal {A}^\vee $
. Thus, the final sequence is
Passing to global sections, we have the equality
Now, consider
$f'\in H^0(X',\mathcal {A}|_{X'})$
with the same properties as f and denote by
$X^{\prime }_1$
the zero locus of
$f'$
and
$Z^{\prime }_1=Z'\cap X^{\prime }_1$
. Then,
Sequence (3.1) implies
$h^0(X,\mathcal {O}_X/\mathcal {I}_Z^r)\geq h^0(X',\mathcal {O}_{X'}/\mathcal {I}_{Z_1'}^r)$
.
We can repeat this argument until we obtain a curve Y and a set of points W obtained by the intersection of X and Z, respectively, with hyperplanes that come from
$\mathcal {A}$
as described before. Then, we need to compute
$h^0(Y,\mathcal {O}_Y/\mathcal {I}_W^r)$
. This dimension is equal to
$\#W\cdot \binom {c+r-1}{c}$
, that is,
$(\mathcal {A}^d\cdot Z)\cdot \binom {c+r-1}{c}$
. For r sufficiently large, the last binomial coefficient is equal to
${r^c}/{c!}+O(r^{c-1})$
. Thus, inductively,
By the assumption on
$\mathcal {N}_{Z/X}$
, [Reference Lazarsfeld4, Proposition 6.3.13] implies
$h^0(X,\mathcal {O}_X/\mathcal {I}_Z^r)\leq 1$
, which leads to the desired inequality.