A user on stack exchange suggested to think of the following problem as a good way to distinguish between algebraic mindset and an analysis mindset:

a) Prove $\sqrt{2}$ is irrational by expressing it is equal to $\frac{m}{n}$ and then come up with some irreducibility argument (lowest terms). (We did this in my introduction to maths course)

b) Prove that whenever some rational $a^2 <2$, there is some other rational $b>a$, such that $b^2 <2$

I found this second way of thinking about it extremely interesting and have been messing around with inequalities like $0< a^2 -1<1$, but I think I'm completely missing the point on the second one, to be quite frank, I'm kind of lost. I do get that the point is that you will never reach $\sqrt2$ since it's a limit point of the set of integers leading up to it, but it's not contained within $\mathbb{Q}$. This is the feature that makes $\mathbb{Q}$ not closed (does not contain all of its limit points).

  • 3
    $\begingroup$ Do you mean $a^2≤2$ maybe? I mean, it is also true that whenever we are given some rational $a$ with $a^2<4$ we can find a rational $b>a$ with $a^2<b^2<4$. $\endgroup$ – lulu May 15 '18 at 19:53
  • 2
    $\begingroup$ It's not clear how (b) indicates that $\sqrt{2}$ is irrational. If $a^2<1$ then there is a $b>a$ so that $b^2<1.$ $\endgroup$ – Thomas Andrews May 15 '18 at 19:54
  • $\begingroup$ Yeah, you need $a^2\le2\implies\exists b>a,b^2\le2$ $\endgroup$ – Akiva Weinberger May 15 '18 at 19:55
  • $\begingroup$ This answer shows how to prove by descent, without using unique factorization, that if $m^2<n<(m+1)^2$ for $n,m$ non-negative integers, then $\sqrt{n}$ is not rational. math.stackexchange.com/questions/5/… $\endgroup$ – Thomas Andrews May 15 '18 at 19:59
  • $\begingroup$ See also math.stackexchange.com/questions/2778191/… $\endgroup$ – Przemysław Scherwentke May 15 '18 at 20:00

If $a>0$ and $a^2<2$ then $b^2>2$ when $b=2/a$. If $a$ is close to $\sqrt2$ we'd expect $b$ to be roughly as close to $\sqrt2$ so $c=\frac12(a+b)$ should be closer. But $c^2>2$. So try $d=2/c$. Your task: prove $d^2<2$ and $d>a$.

  • $\begingroup$ Now notice that $d=\frac{4}{a+b}>\frac{4}{b+b}=\frac{2}{b}=a$, So indeed $d>a$, also $d^2=\frac{4}{c^2}<\frac{4}{2}=2$ $\blacksquare$ $\endgroup$ – Wesley Strik May 16 '18 at 21:04
  • $\begingroup$ We can also express our final number explicitly in terms of a: for every $a>0$, with $a^2 <2$, there exists a unique number $d=\frac{4a}{a^2+2}$, such that $d>a$ and $d^2<2$ $\endgroup$ – Wesley Strik May 16 '18 at 21:10

You can take $$b=a+\frac {1}{n} $$ with $n $ such that

$$a <a+\frac {1}{n}<\sqrt {2}$$


$$n=\lfloor \frac {1}{\sqrt {2}-a} \rfloor +1.$$

  • $\begingroup$ Downvoting. What does this answer mean? How can $b = a + \frac1n$ but also $a + \frac1n < b$? $\endgroup$ – BallBoy May 15 '18 at 20:11
  • $\begingroup$ @Y.Forman Thanks. i fixed it. $\endgroup$ – hamam_Abdallah May 15 '18 at 20:44
  • $\begingroup$ Thanks for editing! $\endgroup$ – BallBoy May 15 '18 at 20:45
  • $\begingroup$ If $a^2 < 2$ then we want $n$ such that $2 > (a+1/n)^2=a^2+2a/n+1/n^2$ which is true if $2 > a^2+3a/n$ since $1/n^2 < a/n$ is the same as $na > 1$ which is true for large enough $n$. Therefore $n$ works if $3a/n < 2-a^2$ or $n > 3a/(2-a^2)$. So this answer is fine. Just needs more elaboration. $\endgroup$ – marty cohen May 15 '18 at 20:52

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.