1

Let $k$ be a algebraically closed field, consider $f: \mathbb A_k^2\rightarrow \mathbb A_k^4$ given by $f(u,v)=(u^3,u^2v,uv^2,v^3)$. Let $X=f(\mathbb A_k^2)$. Then how to determine $I(X)$?

If I let $s=u,t=v/u$, then it becomes $(s^3,s^3t,s^3t^2,s^3t^3)$, looking at this form, I am quite confident $I(X)$ must be $(x_2^2-x_1x_3, x^3_2-x_1^2x_4)$, but I don't know how to prove it, help please.

CYC
  • 2,493
  • Note that in $k[x_1, x_2, x_3, x_4]/(x_2^2-x_1x_3)$, your second generator, $x_2^3 - x_1^2x_4$, factors as $$ x_2^3 - x_1^2x_4 =x_2(x_2^2) - x_1^2x_4 = x_2x_1x_3 - x_1^2x_4 = x_1(x_2x_3 - x_1x_4) $$ where $x_2x_3 - x_1x_4$ is the term you're really after. – Arthur Mar 23 '16 at 23:53
  • You can use Gröbner bases to solve the implicitization problem in general. Here are two relevant threads: one and two. – Viktor Vaughn Mar 24 '16 at 01:39

2 Answers2

1

Let $X,Y,Z,W$ be the four coordinates on $A^4$. Then I claim that the ideal of your variety, call it $C$, is given by $J = (XW - YZ, Y^2 - XZ, Z^2 - YW)$.

Honestly I'm not sure how to check this without a bit of geometric machinery - I'm sure it is possible to do it just on the level of equation manipulation though.

(It makes more sense to me to think of this projectively, but I guess it doesn't matter too much.)

Outline:

  1. You can check that these polynomials vanish on the image of your map.
  2. The ideal $J$ is prime. (You can choose judicious patches, meaning inverting some of the elements $X$, $Y$, $Z$, $W$ to check this. After some simplification you should get just a localization of the polynomial ring in two variables. Really this tells us that the corresponding projective variety is integral, and a little work will show that this implies its affine cone is as well.) (Looking at some of the other answers - there are easier ways to prove that this ideal is prime. For example, what Mariano did in one of the links.)
  3. The vanishing of $J$ is 2 dimensional. (You can use the previous computation.)
  4. So $V(J)$ is a closed, irreducible algebraic set that contains $C$, and is of the same dimension. So in fact $V(J) = C$. (Why is C Zariski closed? This is where it is more natural for me to think of it projectively. But it is still true.)

BTW The projectivization of this variety, thought of as a curve in $\mathbb{P}^3$ is called the twisted cubic.

Elle Najt
  • 21,422
  • "I'm sure it is possible to do it just on the level of equation manipulation" It's not short, but it can be done. Main trick: the given relations translates to the vanishing of the $2\times 2$ minors of a certain $2\times 3$ matrix, which may therefore be taken as rank $1$ (we check the origin separately, which excludes rank $0$). See my answer. – Arthur Mar 24 '16 at 01:26
1

It is not hard to check that the image of $f$ is contained in the variety $Z(x_1x_3 - x_2^2, x_2x_4 - x_3^2, x_1x_4 - x_2x_3)$, so the real difficulty lies in showing that any point on this variety is contained in the image of $f$.

Let's take a point $a = (a_1, a_2, a_3, a_4)\in \Bbb A^4_k$ on the variety. This means that the relations above hold for the coordinates of $a$. We want to show that there are $u, v$ such that $a = f(u, v)$. If $a$ is the origin, then $u = v = 0$ works. For the following, assume that $a$ is not the origin.

If we write up the matrix $$ M = \begin{pmatrix}a_1&a_2&a_3\\a_2&a_3&a_4\end{pmatrix} $$ then we see that this matrix has rank $1$: it cannot have rank $0$ since we assumed that $a$ had components that were non-zero, and the relations show exactly that every $2\times 2$ minor vanishes, so the rank cannot be $2$.

(Note that this matrix is tailored to the given relations. This is not a "magic matrix" that solves all such problems, and there are certainly problems that cannot be solved using this approach. However, whenever you have several second-degree relations, I would suggest looking for a matrix which happens to have minors that coincide with the relations.)

From linear algebra, this tells us that there are two non-zero vectors $b = (b_1, b_2)\in k^2$ and $c = (c_1, c_2, c_3)\in k^3$ such that $M = b^Tc$, in other words $$ M = \begin{pmatrix}b_1c_1&b_1c_2&b_1c_3\\b_2c_1&b_2c_2&b_2c_3\end{pmatrix} $$If either $b_1$ or $b_2$ is zero, then that means that an entire row of $M$ is zero, which means that $a$ has only one of $a_1$ or $a_4$ as non-zero component. In either case let $u$ or $v$ be a cube root of the respective $a_i$, and we're done.

From now on, assume that neither $b_1$ nor $b_2$ is zero. The rest of the proof consists of showing that we may take $b = (u, v)$ and $c = (u^2, uv, v^2)$.

By comparing the two forms of $M$ above, we see that if one $c_i$ is zero, then all the components of $c$ are zero, and that cannot be the case (again since we assumed that $a$ is not the origin). Therefore $c$ cannot have any zero components.

Since $k$ is a field, we may scale $b$ by any non-zero factor we want, as long as we scale $c$ by the inverse of that factor. The trick is to find the right factor.

$k$ is algebraically closed, so $b_1$ and $c_1$ have cube roots. Pick one for each, and name the chosen cube root of $b_1$ $\beta$, and $\gamma$ for $c_1$. Now scale $b$ by $\frac{\gamma}{\beta^2}$ and $c$ by $\frac{\beta^2}{\gamma}$. $M$ remains unchanged, while the first element of this new $b$ is $\beta\gamma$ and the first element of $c$ is $\beta^2\gamma^2$. We take $\beta\gamma$ as our $u$.

Finally, we know that $b_1c_2 = b_2c_1$ and $b_1c_3 = b_2c_2$ from the two expressions for $M$ above, which means that $c$ truly is of the form $(b_1^2, b_1b_2, b_2^2)$, and we are finished.

Arthur
  • 204,511
  • I think at some point one still needs to show that the ideal is prime. Your computations would show that the zero set is irreducible, by comparing it to the image of an irreducible variety (also irreducible), but it may not be reduced (i.e. there may be nilpotents in the coordinate ring). (There won't be of course.) – Elle Najt Mar 24 '16 at 12:34