Let $k \in \mathbb{Z}$ and consider the field extension $K := \mathbb{Q}[\sqrt{k}]$. Define a norm on $K$ given by $\|p+q\sqrt{k}\| := \sqrt{p^2+q^2}$. For any $z \in K$, I was interested to know when $\|z^2\| = \|z\|^2$.

It can be shown that $z=p+q\sqrt{k}$ satisfies $\|z^2\|=\|z\|^2$ if, and only if, $$q^2(k+1)(2p^2+(k-1)q^2)=0.$$

Clearly, $q=0$ means that $z \in \mathbb{Q}$ and $\sqrt{p^4}=\left(\sqrt{p^2}\right)^{\! 2}$ for all $p \in \mathbb{Q}$. When $k=-1$ we have $\mathbb{Q}[\operatorname{i}]$ which is a subset of $\mathbb{C}$, and we already know that the modulus satisfies $|wz|=|w||z|$.

My Questions:

  • What is the significance of the condition $2p^2+(k-1)q^2=0$?
  • Does it give any information about the norm $\|p+q\sqrt{k}\| = \sqrt{p^2+q^2}$?
  • For a fixed $k$, what is so special about the $p+q\sqrt{k}$ which satisfy the condition?

Let's focus on solutions of $2p^2+(k-1)q^2=0$ in $\mathbb{Q}$ for a fixed $k\in\mathbb{Z}$.

  • If $k>1$, there is clearly only one solution $p=q=0$.

  • If $k=1$, then $p=0$, $q\in\mathbb{Q}$ (even though I assume you don't want to consider cases where $k$ is a square).

  • If $k<1$, then as before $p=0$, $q=0$ is a trivial solution. Assume $p^2+q^2>0$. $$2p^2=(1-k)q^2$$

    If $k$ is odd, then $k=1-2t$ for $t\ge1$. We get $$p^2=\frac{1-k}{2}q^2$$ $$p^2 = tq^2$$ Clearly $p\ne0\ne q$ so $t=\frac{p^2}{q^2}$ and therefore $\sqrt{t}=\left|\frac{p}{q}\right|\in\mathbb{Q}$. It's known that a square root of a (positive) integer is either an integer or irrational. Therefore $\sqrt{t}=s^2$, $s\in\mathbb{N}$. So $k=1-2s^2$ and $p=\pm sq$. If odd $k<1$ is not of this form, the only solution is the trivial one.

    If $k$ is even, let $p=\frac{u_1}{u_2}$ and $q=\frac{v_1}{v_2}$ where $u_1,u_2,v_1,v_2\in\mathbb{Z}-\{0\}$ and $u_1,u_2$ and $v_1,v_2$ are relatively prime. Then $$2\frac{u_1^2}{u_2^2}=(1-k)\frac{v_1^2}{v_2^2}$$ $$2u_1^2 v_2^2=(1-k)v_1^2 u_2^2$$ $$2u^2=(1-k)v^2$$ for $u,v\in\mathbb{Z}$, $u\ne 0\ne v$. Since $k$ is even, multiplicity of the prime factor $2$ of LHS is odd and multiplicity of the prime factor $2$ of RHS is even. Therefore the last equation does not have any solutions.

Let's summarize all results (not considering squares for $k$):

  • $k>1$, $k$ is not a square
    • $(p,0)$ for $p\in\mathbb{Q}$
  • $k=-1$
    • $(p,q)$ for $p,q\in\mathbb{Q}$
  • $k<-1$, $k=1-2s^2$ for $s\in\mathbb{N}$
    • $(p,0)$ for $p\in\mathbb{Q}$
    • $(\pm sq,q)$ for $q\in\mathbb{Q}$
  • $k<-1$, $k\ne 1-2s^2$ for $s\in\mathbb{N}$
    • $(p,0)$ for $p\in\mathbb{Q}$
  • 1
    $\begingroup$ +1 Thank you for the detailed, and well thought-out, calculations. I would very much like to see how your findings relate to the three questions that I posted at the end of my question. I am interested in the significance more than the identity. Do you have any ideas? $\endgroup$ – Fly by Night Oct 1 '13 at 19:15

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.