Let $f:X \rightarrow T$ be a proper morphism from a projective scheme $X$ to a smooth projective curve $T$ over $\mathbb{C}$. I know that fiber dimension is upper-semicontinuous, but is the degree of fibers also upper-semicontinuous? Here I fix the embedding of $X$ into some projective space (and so all the fibers of $f$ over each closed point of $T$) and then define the degree inside this fixed projective space in an usual way.

Actually, what I have in mind is the following situation: Let $\mathbb{P}^N$ be a projective space over $\mathbb{C}$ and let $Z$ and $Z_t$ be two irreducible varieties, where $Z$ is fixed and $Z_t$ moves with $t \in T$ as a parameter on a smooth curve $T$. I want to prove upper semicontinuity of degrees of the intersection scheme of $Z$ and $Z_t$ in $\mathbb{P}^N$, when dimension of $Z$ and $Z_t$ are smaller than $N \over 2$ but they intersect. In other words, the intersection is not proper in $\mathbb{P}^N$ and the expected dimension of the intersection is negative. For example, you could consider a line L in $\mathbb{P}^N$ and a family of curves $C_t$ with t as its parameter such that $L \cap C_t$ is nonempty, when $N>2$.


There's a slight stupidity here which is that if the dimension of the fiber jumps up, the degree can go down. E.g. let $X = (T \times C) \cup (p \times \mathbb P^2) \subseteq T \times \mathbb P^2$, where $C$ is a plane conic and $p\in T$ is a point. Then over most points of $T$, the fiber is $C$ of degree $2$, but over $p$ it's $\mathbb P^2$ of degree $1$. But this is just because studying the leading coefficient of a polynomial, as the polynomial changes degree (careful! this means "dimension" here), is a silly thing to do.

Assuming that the dimension is constant not just semicontinuous, then the degree is indeed semicontinuous. Let $X' \subseteq X$ be the closure of the fiber over the generic point, i.e. the union of the components of $X$ that dominate $T$. Then since $C$ is $1$-dimensional and regular, $X' \to X$ is flat, so the degree of fibers is actually constant. Adding back any remaining components of $X$, each of which maps to a point, we increase the degree at those points.

| cite | improve this answer | |
  • $\begingroup$ Thank you for your answer. I think your counterexample works for my main question, but it does not seem to occur in my situation of 2nd paragraph. What I am thinking for my situation is the following: Consider a quadric surface $Q$ in $\mathbb{P}^3$ and cosider lines $L_t$ in $\mathbb{P}^3$ intersecting with $Q$ at 2 points when $t \neq 0$ and $L_0$ is in $Q$. My question is, is it right that $L_0 \cap Q$ is just $L_0$, not $2L_0$? (If it's $L_0$ then this can be a counterexample for my situation although it is proper intersection for general $t$, but I am not sure that this is right.) $\endgroup$ – user55992 Jul 14 '14 at 9:40

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.