1

I have recently recalled a problem from my school days, where I was asked if for every $\varepsilon > 0$ there exist positive integers $m, n$ such that $|m\sqrt 2 - n|< \varepsilon$. I have just realized, that I do not really know of a simple way to solve this problem. Furthermore, would be good if that solution could be applied also to any $q\in \Bbb R_+$ instead of $\sqrt2$.

SBF
  • 36,041

1 Answers1

1

Let $0<q\in\Bbb R\setminus\Bbb Q$, $M=\{mq-n:m,n\in\Bbb Z\}$ and $a=\inf\{x:x\in M\land x>0\}$. Assume on contrary $a>0$.

If $a\in M$, then for every $0<x\in M$ we have $0\leq x-\lfloor x/a\rfloor a<a$ and since $x-\lfloor x/a\rfloor a\in M$ this implies $x=\lfloor x/a\rfloor a\in a\Bbb Z$. Then $q=au$ and $1=av$ for some $u,v\in\Bbb Z$, and we get the contradiction $q=u/v\in\Bbb Q$.

If $a\notin M$, by definition of infimum, then there exists $m',n'\in\Bbb Z$ such that $a<m'q-n'\lt 2a$ and, again by definition of infimum, there exists $m,n\in\Bbb Z$ such that $a<mq-n\lt m'q-n'$. Then $0\lt(m'-m)q-(n'-n)\lt a$ - a contradiction.

  • It might be worth explaining why there exist such $m,m'.n.n'$ in the last paragraph? (The pigeonhole principle argument is certainly the standard way of handling this; I just haven't seen it phrased quite this way before.) – Steven Stadnicki Jul 21 '20 at 21:25
  • @StevenStadnicki: I add an explanation in the answer. – Fabio Lucchini Jul 21 '20 at 21:28