Let $\mathbb{F}$ be a field, which for the sake of this discussion, is such that char $\mathbb{F} \neq 2$.

By Corollary 9.4 in Scharlau's Quadratic and Hermitian Forms, the Grothendieck-Witt group $GW(\mathbb{F})$ is generated by elements $\langle \alpha \rangle, \alpha \in \mathbb{F}^{\times}$, subject to the relations

  1. $\langle \alpha \rangle = \langle \alpha \beta^{2} \rangle$ for all $\alpha , \beta \in \mathbb{F}^{\times}$.
  2. $\langle \alpha \rangle + \langle \beta \rangle = \langle \alpha + \beta \rangle + \langle (\alpha + \beta)\alpha \beta \rangle$.

I understand the proof of this result as presented in Scharlau.

However, in Morel's $\mathbb{A}^{1}$ Algebraic topology over a field, lemma 2.9, he says that the second relation may be obtained from the first relation and the relation

$\langle \alpha \rangle + \langle -\alpha \rangle = 1 + \langle -1\rangle$.

I understand the motivation behind this relation (matrices of the form on the LHS are congruent to the hyperbolic plane when char $\mathbb{F} \neq 2$), but cannot formally derive the second relation using this and relation 1).

I am probably just missing a trick. Any help would be much appreciated!


1 Answer 1


I think Morel's claim is not correct.

Consider the case $F=\mathbb{Q}_3$. Let $M$ denote the free abelian group generated by $\langle\alpha\rangle$, $\alpha\in \mathbb{Q}_3^\times$. We have $\mathbb{Q}_3^\times/(\mathbb{Q}_3^\times)^2=\{[1],[-1],[3],[-3]\}$, so the quotient of $M$ by the first relation is a free abelian group with basis $\langle 1 \rangle,\langle -1\rangle, \langle 3 \rangle,\langle -3\rangle$. If Morel's claim is correct, then $\operatorname{GW}(F)$ must be the quotient of this group by the relation $$ \langle 3\rangle + \langle -3\rangle = \langle 1\rangle + \langle -1\rangle. $$ However, we have $$ \langle 3\rangle + \langle 3\rangle\neq \langle 6\rangle + \langle 54\rangle \quad( = \langle -3\rangle + \langle -3\rangle) $$ in this group, so the second relation does not hold.

On the other hand, it is true that the first and the second relation imply $\langle \alpha\rangle + \langle -\alpha\rangle=\langle 1\rangle + \langle -1\rangle$. Let $x=\dfrac{\alpha+1}{2}$ and $y=\dfrac{\alpha-1}{2}$, so that $x^2-y^2=\alpha$. By the second relation, we have $$ \langle x^2\alpha\rangle + \langle -y^2\alpha\rangle = \langle (x^2-y^2)\alpha\rangle + \langle -x^2y^2(x^2-y^2)\alpha^3\rangle. $$ Using the first relation and $x^2-y^2=\alpha$, we can rewrite this as $$ \langle \alpha\rangle + \langle -\alpha\rangle = \langle 1\rangle + \langle -1\rangle. $$ Perhaps this is what Morel wanted to note.

  • $\begingroup$ Thanks, but could you spell out why in your counterexample that $\langle 3 \rangle + \langle 3 \rangle \neq \langle 6 \rangle + \langle 54 \rangle$ ? (I haven't really worked with $\mathbb{Q}_{3}$). $\endgroup$
    – Sunny Sood
    Sep 30, 2021 at 10:09
  • $\begingroup$ It is because $(\langle 3\rangle+\langle 3\rangle)-(\langle -3\rangle+\langle -3\rangle)$ is not contained in the subgroup generated by $(\langle 3\rangle+\langle -3\rangle) -(\langle 1\rangle+\langle -1\rangle)$ in the free abelian group with basis $\langle 1\rangle,\langle -1\rangle,\langle 3\rangle,\langle -3\rangle$. $\endgroup$ Sep 30, 2021 at 10:15
  • $\begingroup$ Ahh... I see. Thanks! $\endgroup$
    – Sunny Sood
    Sep 30, 2021 at 12:49

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.