3
$\begingroup$

Suppose we have a finite quantity $a$, which we would like to prove to be irrational, supposing that it is indeed irrational.

Then, would it be enough to show that $$a=\lim_{n\to\infty}\frac{u_n}{v_n},$$ for some positive integers $u_n,v_n$, where $u_n,v_n\to\infty$ as $n\to\infty$. If so, then would there have to be some divisor properties between the denominator and numerator so that cancellation does not produce an integer or rational number as $n\to\infty$, e.g. suppose $(u_n,v_n)=1$ for all $n$ ?

$\endgroup$
7
  • 3
    $\begingroup$ As the rational numbers $\mathbb Q$ are dense in the real numbers $\mathbb R$, any real number (either rational or irrational) is the limit of a sequence of rational numbers. So, no, the answer is no (any $a$ can be written in that form for suitable integers $u_n$ and $v_n$) $\endgroup$ – AndreasT Aug 15 '14 at 15:24
  • 1
    $\begingroup$ The condition that $v_n \nmid u_n$ is extremely ineffective, it only rules out $a$ being an integer. Don't you mean that the gcd of $v_n$ and $u_n$ should be 1? $\endgroup$ – Erick Wong Aug 15 '14 at 15:48
  • $\begingroup$ @Erick yes you're right. I've updated the question. $\endgroup$ – Pixel Aug 15 '14 at 15:49
  • 1
    $\begingroup$ All rational numbers have finite continued-fraction representations. So if you can prove a number has an infinite continued fraction representation it must also be irrational. $\endgroup$ – NovaDenizen Aug 15 '14 at 15:53
  • $\begingroup$ Given a number, it is undecidable in general whether it is real or not (because there are more real numbers than there are Turing machines). $\endgroup$ – user1729 Aug 15 '14 at 15:55
4
$\begingroup$

Here's a similar condition that is sufficient: there exists a sequence of integers $u_n, v_n \to\infty$ such that $(u_n, v_n) = 1$ and $$\lim_{n\to\infty} v_n a - u_n = 0.$$ (Alternatively, we could just require that $a$ not be exactly equal to any $u_n/v_n$ rather than $(u_n,v_n)=1$.)

$\endgroup$
6
  • $\begingroup$ Thanks Erick. Do you have a reference for this or a resource for further details? I'd like to study more. $\endgroup$ – Pixel Aug 15 '14 at 15:53
  • 2
    $\begingroup$ I'd be shocked if the idea isn't in any book on irrationality, but I don't know a reference offhand (maybe try something basic like Niven's "Numbers: Rational and Irrational"?). The proof is very simple: if $a=p/q$, and $a \ne u_n/v_n$, then $v_n a - u_n$ is a non-zero rational with denominator (at most) $q$. Therefore it is at least $1/q$ in absolute value, and can't go to $0$. $\endgroup$ – Erick Wong Aug 15 '14 at 15:56
  • $\begingroup$ @ Erick but isn't this just the same as my original question, since $$\lim_{n\to\infty}v_n a-u_n=0 \iff \lim_{n\to\infty}a=\lim_{n\to\infty}\frac{u_n}{v_n}\iff a=\lim_{n\to\infty}\frac{u_n}{v_n},$$ and $u_n,v_n\to\infty$ as $n\to\infty$, $(u_n,v_n)=1$ ? $\endgroup$ – Pixel Aug 19 '14 at 8:53
  • 1
    $\begingroup$ @pbs Absolutely not. Compare your first iff against $a=1$, $u_n = n$, $v_n = n+1$. The $\Leftarrow$ is very much false. $\endgroup$ – Erick Wong Aug 19 '14 at 14:30
  • 1
    $\begingroup$ @ Erick wow - I see that now: while $$\lim_{n\to\infty}\frac{u_n}{v_n} = \lim_{n\to\infty}\frac{n}{n+1}=1,$$ we have $$\lim_{n\to\infty}(n+1)\times 1-n=\lim_{n\to\infty} 1 = 1\neq 0.$$ $\endgroup$ – Pixel Aug 19 '14 at 15:35
3
$\begingroup$

Consider

$$u_n = \frac{10^n - 1}{3} \hspace{1cm}v_n = 10^{n}$$

Then:

$$\frac{u_n}{v_n} = 0.\underbrace{3333\cdots 3}_{n \text{ times}}$$

What means that

$$\lim_{n\to\infty} \frac{u_n}{v_n} = 0.333\cdots = \frac{1}{3}$$

So no, that condition is not sufficient.

$\endgroup$
2
$\begingroup$

This will not suffice. Let $a$ be an irrational number then there is some sequence of rational numbers $(q_{n})_{n=1}^{\infty}$ that converges to $a$. Denote $q_{n}=\frac{u_{n}'}{v_{n}'}$ where $u_{n}',v_{n}'$ are coprime. Let $u_{n}=nu_{n}',v_{n}=nv_{n}'$ then $$ \lim_{n\to\infty}\frac{u_{n}}{v_{n}}=\lim_{n\to\infty}\frac{nu_{n}'}{nv_{n}'}=\lim_{n\to\infty}\frac{u_{n}'}{v_{n}'}=\lim_{n\to\infty}q_{n}=a $$

So every irrational number satisfy your property.

But similarly if $a$ is rational take $q_{n}\equiv\frac{na}{n\cdot1}$ which will show that every rational number also satisfy this test. So in fact all real numbers satisfy the test and so we can't get information about a number that satisfies this test - they all do

$\endgroup$

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.