4
$\begingroup$

Let $k$ be a field and let $Q$ be a quaternion algebra over $k$.

It is well known that, if $\mathrm{char}\,k\neq 2$, one can define $Q$ as the $k$-algebra of dimension $4$ generated by elements $x,y$ and relations $x^2 =a$, $y^2 =b$ and $xy = - xy$ where $a,b\in k$ and $b\neq 0$. Call this $Q=(a,b)$.

One can also define a "quaternion" algebra if $\mathrm{char}\,k = 2$, taking symbols $x,y$ and imposing relation $x^2 +x =a , y^2 =b, xy = yx + y$. Call this $Q=[a,b)$.

It is clear from the definition that, in case $\mathrm{char}\,k\neq 2$, the isomorphism class of $(a,b)$ depends only on the classes of $a,b$ in $k^*/(k^*)^2$ or, more precisely, letting $a = u^2 \alpha,b = v^2 \beta$ we get that $(a,b) = (u^2\alpha ,v^2 \beta ) \simeq (\alpha, \beta) $ and the isomorphism is given by sending $x\mapsto u^2 x$ and $y\mapsto v^2 y$.

The question is: do we have a similar description for $[a,b)$, namely with $\mathrm{char}\,k=2$? It seemed natural to me to look at the classes of $a,b$ in $k/\wp(k)$ but I don't get the right relations.

Any help would be appreciated, especially welcome is a good reference for dyadic quaternion algebras.


Edit.

So far I have been able to see that, given $k$ of $\mathrm{char}\,k=2$ and $Q=[a,b)$ a quaternion algebra, then we can replace $a$ with any of its representatives in the class $[a]\in k/\wp{k}$, namely we can replace $a\mapsto \alpha + (u^2 + u)$ and we obtain an isomorphic quaternion algebra.

However, I fail to see what we can do with $b$ (if anything can be said). Looking at the book suggested in the comments, it looks like in general a quaternion algebra is not equivalent to give a pair of elements $a,b$ and some relations, unless the characteristic is not 2.

So in the end I would say that $[a,b)$ depends on the class of $a$ in $k/\wp{k}$ as well as on the choice of $b\in k$. Does that make sense?

$\endgroup$
5
  • 1
    $\begingroup$ The standard reference is still, I believe, the book my Marie-France Vigneras. $\endgroup$
    – Vincent
    Commented May 30, 2018 at 10:34
  • 2
    $\begingroup$ Apparently someone made an English translation and LaTeX-ed it: see here maths.nju.edu.cn/~guoxj/notes/qa.pdf $\endgroup$
    – Vincent
    Commented May 30, 2018 at 10:37
  • 2
    $\begingroup$ Thanks! This book looks wonderful! By the way, just reading there I realised that I had made a mistake when trying to mimick the char $\neq 2$ isomorphism, so it looks like the isomorphism class of $[a,b)$ actually depends on the classes in $k/\wp(k)$. $\endgroup$
    – Caligula
    Commented May 30, 2018 at 10:50
  • 1
    $\begingroup$ There is no need for quotation marks in the characteristic $2$ case: what you describe there really are the quaternion algebras over $k$. In all characteristics, a quaternion algebra over $k$ is a 4-dimensional central simple $k$-algebra. These have different concrete descriptions in characteristic not $2$ and in characteristic $2$, just like classifying quadratic Galois extensions of $k$ uses different concrete descriptions in characteristic not $2$ (nontrivial elements of $k^\times/(k^\times)^2$) and in characteristic $2$ (nonzero elements of $k/\wp(k)$). $\endgroup$
    – KConrad
    Commented May 31, 2018 at 14:14
  • 1
    $\begingroup$ As much as I like Vigneras's book, I think the standard reference should become Voight's book: math.dartmouth.edu/~jvoight/quat.html $\endgroup$
    – Aurel
    Commented May 31, 2018 at 20:06

0

You must log in to answer this question.