Let $f(X,Y)\in \mathbb C[X,Y]$ be an irreducible polynomial. I know that the zero set of $f$ , $V(f):=\{(a,b)\in \mathbb C^2 : f(a,b)=0\}$ is connected in the usual Euclidean topology of $\mathbb C^2$ . Is there a proof of this result without using too much algebraic geometry ? What are the most elementary proof of this result ?

If we assume that $V(f)$ is smooth i.e. $$\forall p \in \mathbb C^2\colon\quad(\partial f/\partial X , \partial f/\partial Y)|_p\ne (0,0),$$ can we make a proof substantially more elementary ?

  • $\begingroup$ Will you share a reference for an advanced proof of this result? $\endgroup$ – Matt F. Apr 2 '18 at 1:36
  • $\begingroup$ @MattF.: You may take a look here math.stackexchange.com/questions/242535/… ... zero set of irreducible polynomial over algebraically closed field is an irreducible algebraic set in Zariski topology, hence connected in Zariski topology. $\endgroup$ – user Apr 2 '18 at 15:31
  • $\begingroup$ Coincidentally Terry Tao recently posted notes on this at the end of terrytao.wordpress.com/2018/03/28/… $\endgroup$ – Dap Apr 4 '18 at 4:56

How about this? By Noether normalization, after a generic change of coordinates, our curve $C$ is of the form $$y^d + f_1(x) y^{d-1} + \cdots + f_d(x)=0$$ for various polynomials $f_j$.

Let $\Delta(x)$ be the discriminant of this polynomial as a polynomial in $y$. Since $f$ is squarefree as a polynomial in $\mathbb{C}[x,y]$, we know that $\Delta(x)$ is not identically zero. So it has finitely many zeroes $z_1$, $z_2$, ..., $z_N$. Let $U = \mathbb{C} \setminus \{ z_1, z_2, \ldots, z_N \}$. For $x \in U$, there are $d$-distinct roots $y$ of the above polynomial.

Let $\pi$ be projection onto the $x$-coordinate. Then the implicit function theorem shows that $\pi^{-1}(U) \to U$ is a $d$-fold covering map. We want to show $\pi^{-1}(U)$ is connected. Suppose not. Then it breaks into components $Y_1$, $Y_2$, ..., $Y_r$ which cover $U$ with degrees $k_1$, $k_2$, ..., $k_r$ where $\sum k_i = d$.

Choose a component $Y$ of $\pi^{-1}(U)$, with $Y \to U$ of degree $k$. For $1 \leq j \leq k$ and $x \in U$, let $e_j(x)$ be the $j$-th elementary symmetric function in the $y$-coordinates of the $k$ points of $Y \cap \pi^{-1}(x)$.

The function $e_j$ is holomorphic on $U$. At the punctures $z_1$, ..., $z_N$, the function $e_j$ is bounded, since $C \to \mathbb{C}$ is proper. So, by the Riemann extension theorem, $e_j$ extends to a holomorphic function on all of $\mathbb{C}$. Also, $e_j$ is a polynomial in algebraic (multivalued) functions, so its growth rate as $|x| \to \infty$ is bounded by $|x|^N$ for some $N$. We see that $e_j(x)$ is an entire function bounded by a polynomial, so it is a polynomial. Thus, $y^k - e_1(x) y^{k-1} + \cdots \pm e_k(x)$ is a polynomial of degree $k$ vanishing on $Y$.

We can repeat this for each component $Y_i$ and get a polynomial $g_i(x,y)$ with $y$-degree $k_i$ vanishing on the corresponding points of the cover $\pi^{-1}(U) \to U$. Then $\prod g_i(x,y)$ vanishes on all of $\pi^{-1}(U)$, and hence is divisible by $f$. Since $f$ and $\prod g_i(x,y)$ have the same $y$-degree, they are proportional. This gives a nontrivial factorization of $f$, giving our contradiction.

  • $\begingroup$ Very nice answer! I am trying to understand your proof: $U$ is connected. The connected components of $\pi^{-1}(U)$ are covers of $U$. Each produces a polynomial and their product is the original one. It must be only one, so $\pi^{-1}(U)$ is connected. $C$ is the closure of $\pi^{-1}(U)$ so connected. $\endgroup$ – Orest Bucicovschi Apr 2 '18 at 18:45
  • $\begingroup$ @orangeskid That's right. I'm thinking I introduced a red herring by bringing in the symmetric group explicitly; I might edit to make it more geometric later if I get a chance. $\endgroup$ – David E Speyer Apr 2 '18 at 22:30
  • $\begingroup$ Removed references to covering groups and orbits in place of more geometric language. $\endgroup$ – David E Speyer Apr 3 '18 at 2:58

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.