Let $k$ be a field. Let $X$ be a (non empty) connected proper $k$-scheme. I would like to prove the maximum principle, that is for any $k$-morphism $\varphi \colon X \to \mathbb A_k^1$, the image of $\varphi$ is a closed point.

From a previous result, as $\mathbb A_k^1$ is a separated $k$-scheme, I already know that $\varphi(X)$ is closed in the affine line. It remains to show that $\varphi(X)$ is a single point. I have no clue…

I tried to exploit the fact that a morphism $\varphi \colon X \to \mathbb A_k^1$ is given by a morphism $k[T] \to \mathscr O_X(X)$ (via the left adjoint of the inclusion of affine schemes into the locally ringed spaces) and it lead me to exhibit $\varphi$ as follow : there is an global section $f \in \mathscr O_X(X)$ such that for all $x \in X$, $$\varphi(x) = \ker (k[T] \to \kappa(x) \colon Q \mapsto Q(f)) .$$ But I don't see why this prime ideal of $k[T]$ wouldn't change with respect to $x$.


First we claim that the image of $\varphi$ cannot be all of $\Bbb{A}^1$. If it were all of $\Bbb{A}^1$, by composing with $\Bbb{A}^1 \hookrightarrow \Bbb{P}^1$ we get a dominant map $X \to \Bbb{P}^1$ that is furthermore proper and thus has closed image. But now this means the map $X \to \Bbb{P}^1$ is surjective, a contradiction.

So we know the image of $\varphi$ is a proper closed subset of $\Bbb{A}^1$. It is now enough to observe that the image of a connected set under a continuous function is also connected. For then the only connected proper closed subset of $\Bbb{A}^1_k$ is a one point set.

Added: Why is $X \to \Bbb{P}^1$ proper?

Well this comes from the famous "Property $\mathscr{P}$" exercise:

Let $\mathscr{P}$ be a property of morphisms that is stable under composition and base extension. Then suppose $f : X\to Z$ which factors as $X \stackrel{g}{\to} Y \stackrel{h}{\to} Z$. If $f$ and the diagonal $\Delta{h}$ have property $\mathscr{P}$, then so does $g$.

| cite | improve this answer | |
  • $\begingroup$ I think you meant $\mathbb A_k^1$ instead of $k$. Just to be sure I'm getting it right : a closed subset of $\mathbb A_k^1$ is (topologically) $\mathrm{Spec}\,(k[T]/I)$ for some ideal $I$ of $k[T]$ ; but a scheme (in fact, a locally ringed space) is connected if and only if its global section ring does not have non trivial idempotent, which is in turn equivalent to be unwritable as a product of two non trivial rings ; in conclusion, a closed subset $V(I=(Q)) \simeq \mathrm{Spec}\, (k[T]/(Q))$ is connected iff $Q$ irreducible or null (by the Chinese theorem). Is that it ? $\endgroup$ – Pece Jan 4 '14 at 14:21
  • $\begingroup$ … And if it is, why can't $\varphi(X)$ be the connected closed subset $\mathbb A_k^1$ (that is $Q=0$) ? $\endgroup$ – Pece Jan 4 '14 at 14:31
  • $\begingroup$ @Pece For your first statement, you're really overcomplicating it: All it is saying is that a polynomial of degree $n$ can have at most $n$ number of roots! $\endgroup$ – user38268 Jan 4 '14 at 14:53
  • $\begingroup$ @Pece For your second comment- I overlooked that part, and have now edited my answer. $\endgroup$ – user38268 Jan 4 '14 at 15:03
  • $\begingroup$ I get the first part, embedding $\mathbb A^1$ into $\mathbb P^1$. For the second part, I don't get why you mean by "All it is saying etc." : from where do you see (without my overcomplicating explanation) that $V((Q))$ is connected iff $Q$ is irreducible ? $\endgroup$ – Pece Jan 4 '14 at 15:10

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.