Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm having trouble calculating the image of a point $P=[0:0:1]$ under a polynomial map:$$f:V \to \mathbb{P}^1 : [x:y:z] \mapsto [y:x]$$ where $$V = \left\{ [x:y:z] \in \mathbb{P}^2 ~\mid~ zy^2 = zx^2 + x^3 \right\}.$$

What is $f([0:0:1])$?

Dead end

Notice that the given polynomials don't work at this point (but they do at every other point of $V$) since $[0:0]$ is not a valid expression for a point in $\mathbb{P}^1$. I found a different expression, $$f([x:y:z]) = [zx+x^2:zy],$$ but this expression is not defined at $[0:0:1]$, $[0:1:0]$, or $[-1:0:1]$, so it is an objectively worse expression. How do I find a third expression?

Similar problem where things work:

Let $\mathbb{P}^n$ be projective space over an algebraically closed field $k$, so that $\mathbb{P}^n$ consists of all $(n+1)$-tuples $[x_0 : x_1 : \cdots : x_n ]$ with $(x_0,x_1,\ldots,x_n) \neq (0,0,\ldots,0)$ with two tuples $[x_0 : x_1 : \cdots : x_n ] = [y_0 : y_1 : \cdots : y_n ]$ considered equivalent if there is some nonzero $\lambda$ in $k$ with $\lambda x_i = y_i$ for $i=0,1,\ldots,n$.

Let $$V = \left\{ [x:y:z] \in \mathbb{P}^2 ~\mid~ x^2 + y^2 = z^2, (x,y,z) \neq(0,0,0) \right\}$$ be the (projective) circle, and let $$W = \left\{ [ x:y ] \in \mathbb{P}^1 ~\mid~ (x,y) \neq 0 \right\}$$ be the projective line, and let $$f : V \to W : [x:y:z] \mapsto \begin{cases} [ y-z : x ] & \text{ unless } [x:y:z] = [0:1:1] \\\ [ x : y+z ] & \text{ unless } [x:y:z] = [0:-1:1] \\ \end{cases}.$$

Note that this rule for $f$ is globally valid and the function is well-defined. In other words, the two definitions agree on their overlap, and every point of $V$ is covered by one of the definitions.

share|cite|improve this question
up vote 5 down vote accepted

You can stop looking for a third expression for $f$: that rational function is not defined at $P$.
Note carefully that I'm not saying that $f$ has a pole at $P$: I am saying that it has a point of indeterminacy at $P$ and that $f$ can't be extended to a morphism $V\to \mathbb P^1$.

In contrast, a rational map from a smooth projective curve to a projective variety is actually a morphism i.e. it has no such indeterminacy: this explains your success with the circle .
However the point $P$ on $V$ is singular : this explains your failure [or rather the failure of $f$ :-)] with the strophoid $V$ .

And actually all this is visible with the naked eye in your example: in the affine part $z=1$ of the curve, the rational map $f$ sends the point $Q=(x,y)=[x:y:1]$ to the slope $f(Q)=y/x$ of the chord $\overline{PQ}$.
In the limit when $Q$ tends to $P$ that slope tends to $+1$ or $-1$ according to the branch of the strophoid on which $Q$ lies.
However the Zariski topology is too coarse to distinguish between these two branches and that is the geometric reason why $f$ is not defined at $P$.

The last section above shows us the way to correct this failure of $f$.
We normalize the curve $V$ so that $P$ is replaced by two points $P_1, P_{-1}$ and $f$ will send these points respectively to $1$ and $-1$. Technically the normalization is $$n:\mathbb P^1\to V:[u:v]\mapsto [u^2v-v^3:u^3-uv^2:v^3] $$ and the composition of $n$ and $f$ is the everywhere defined morphism $$\tilde f=f\circ n:\mathbb P^1\to \mathbb P^1:[u:v]\mapsto [ u^3-uv^2:u^2v-v^3]=[u:v]$$
In other words $\tilde f$ is the identity!
Check that the points above $P=[0:0:1]$ are $P_1=[1:1], P_{-1}=[-1:1]$ and that $\tilde f(P_1)=[1:1]=\text {slope }(1/1), \;\tilde f(P_{-1})=[-1:1]=\text {slope } (-1/1)$

share|cite|improve this answer
I suspected that the singularity is the problem, but I was missing the intuition for why the map cannot be extended. Thanks! – Andrew Aug 7 '12 at 21:02
Ah, so using the projective line for the range allows me to handle poles nicely, but there are other problems ("branches" maybe?) and my function is not single-valued at P. Is there some easy way to find a "Riemann surface" for f, where it would be single-valued? Maybe $V$ is actually the projection of $\tilde V \subset \mathbb{P}^3$ and $\tilde f$ is well-defined on $\tilde V$? – Jack Schmidt Aug 7 '12 at 21:02
I think $\tilde V$ must be the blowup of $V$ at the nodal point. Then there will be exactly $2$ points in the exceptional locus, and we can define the map there. – Andrew Aug 7 '12 at 21:09
@Andrew: Can you write down such a thing using explicit polynomials? It would be ideal if $\tilde V$ just used an extra variable $t$ that we ignored to get $V$, but anything explicit and simple enough for me to calculate with is fine. – Jack Schmidt Aug 7 '12 at 21:18
@Georges: Thanks, the slope explanation is very clear, and I suspect was the point of the exercise. (I'm still trying to make f into a single-valued function, like $f(z)=\sqrt{z}$ and its two branches.) – Jack Schmidt Aug 7 '12 at 21:20

In order for $f=(f_0:\cdots:f_n)$ to define a rational map from some $X$ to $\mathbb P^n,$ we must have not all $f_i=0.$ This is part of the definition of a rational map, since otherwise we encounter the issue of mapping to $(0:\cdots:0),$ which makes no sense. So in your second example, you are exactly modifying the expression in order to define a rational map at every point of $V.$ In the first, you should say $f:[x:y:z]\to[y:x]$ unless if $x=y=0.$ As it stands, we have not assigned a value to $f((0:0:1)),$ and in order to do so, we must find an alternative expression for $f$ in a different chart (coinciding with the first expression on an overlap), as you did in the second example.

Variation on a theme:

Over the chart $z=1$ that we've been using, the blowup $\tilde V$ of $V$ can be computed as the strict transform of $V$ under $\mathrm{Bl}_{(0,0)}\mathbb A^2\subseteq\mathbb A^2\times\mathbb P^1\to\mathbb A^2.$ Away from $(0,0),$ the point $(x,y)$ corresponds to $(x,y;x:y),$ so if e.g. $x\neq 0$ we get $(x,y;1:y/x)$ as the unique point in the fibre above $(x,y).$ (As you noticed, this is the graph of your function!) But above the origin are two points, $(0,0;1:1)$ and $(0,0;1:-1)$ corresponding to the branches of your parametrization at $(0,0).$ There is a natural map $\tilde f:\tilde V\to\mathbb P^1,$ the restriction of the projection $\mathbb A^2\times\mathbb P^1\to\mathbb P^1,$ which lifts the original $f$ (and $\tilde f:\tilde V\to\mathbb P^1$ is essentially the identity, as with the normalization).

share|cite|improve this answer
Yes, I'm asking for that alternative expression. – Jack Schmidt Aug 7 '12 at 20:42
Isn't this the rational parametrization of a node by $\mathbb P^1$? I'm not quite convinced such an expression exists. – Andrew Aug 7 '12 at 20:53
(from Georges's answer's thread) Weird, that's almost toooo easy: that's just the graph! I've got three nice answers now (a weird $f(x)=\pm\sqrt{1+x}$ answer too), so plenty to explore. If you want to add your blow-up to your answer, I'll upvote it too. – Jack Schmidt Aug 7 '12 at 21:47

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.