1
$\begingroup$

I’m trying to solve the following problem given in a textbook:

Let $k$ be an algebraically closed field and $Q=V(F)$ a quadric in $\mathbb{P}^3(k)$, where $F$ is an irreducible polynomial in $X,Y,Z,T$, and hence gives rise to a quadratic form on $k^4$ which we assume is non-degenerate.

Show that after some change of coordinates we can write $Q=V(XT-YZ)$.

I’ve solved the case where $\text{char}(k)\neq2$ by diagonalising the quadratic form and then making a suitable change of coordinates. However this process involves using the correspondence between quadratic forms and symmetric bilinear forms which is not valid in characteristic $2$ (and if we could diagonalise $F$ then it would become reducible).

My first issue is that I can’t find a reference defining what it means for a quadratic form to be non-degenerate in characteristic $2$, so if I tried to come up with a counterexample I wouldn’t know if it was valid or not.

Beyond this, I’m not even sure that the result is true, I can’t seem to find anywhere claiming that it is. Has the textbook simply forgotten to specify that $\text{char}(k)\neq2$, or am I missing something?

Any help would be much appreciated.

$\endgroup$
0
$\begingroup$

Having spent more time on this, I think I have solved the problem, and the result is in fact true.

Arf defines non-singularity for quadratic forms of characteristic $2$ here, explained in English here. From these papers, we see that if $Q$ is non-singular over a field of characteristic $2$, we can write $$Q=(aX^2+XT+bT^2)+(cY^2+YZ+dZ^2)$$ for some $a,b,c,d\in k$ and an appropriate choice of coordinates. Let’s consider $aX^2+XT+bT^2$.

If $a=b=0$ then we have $XT$ already, if say $b=0$ then we have $X(aX+T)$ so taking the inverse of the transformation sending $X\mapsto X$ and $T\mapsto aX+T$ we have $XT$.

Then we assume $a,b\neq0$. Sending $X\mapsto\frac{1}{\sqrt{a}}X$ and $T\mapsto\frac{1}{\sqrt{b}}T$ we have $X^2+\alpha XT+T^2$ for $\alpha=\frac{1}{\sqrt{ab}}$. All square roots exist since $k$ is algebraically closed, and for the same reason we can also find a root $\beta$ of the polynomial $x^2+\alpha x+1$. Then sending $X\mapsto X+\frac{1}{\alpha^2}T$ and $T\mapsto\beta X+\frac{\alpha+\beta}{\alpha^2}T$ we have $XT$. This transformation is invertible since $$\begin{vmatrix}1 & \beta \\\frac{1}{\alpha^2} & \frac{\alpha+\beta}{\alpha^2}\end{vmatrix}=\frac{1}{\alpha}\neq0$$

We can repeat the same process for $Y$ and $Z$, and so we can write $Q=XT+YZ=XT-YZ$.

$\endgroup$

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.