2
$\begingroup$

Update:How is this conclusion made: "it follows from claim 2 that each $g_n$ is greater than $0$ if $n$ is large enough."

I know that $c$ is a natural number,so $c^n\gt 0$, I can say as n gets bigger $f_{k+n}(x)\rightarrow 1 $. But how about ${k(k+1)\cdots(k+n-1)}$ ? Since k is a rational number it can be a negative number too ( since in claim3 it's stated as $k\in\mathbb{Q}\setminus\{0,-1,-2,\ldots\} $)


I am having trouble while reading the proof of the 3rd claim that Miklós Laczkovich made in his proof. Can someone help me

Let $f_k(x) = 1 - \frac{x^2}k+\frac{x^4}{2! k(k+1)}-\frac{x^6}{3! k(k+1)(k+2)} + \cdots \qquad (k\notin\{0,-1,-2,\ldots\})$

Claim 1: The following recurrence relation holds: $(\forall x\in\mathbb{R}):\frac{x^2}{k(k+1)}f_{k+2}(x)=f_{k+1}(x)-f_k(x).$

Claim 2: For each $x\in \mathbb R$, $\lim_{k\to+\infty}f_k(x)=1.$

Claim 3: If $x\neq 0$ and if $x^2$ is rational, then

$(\forall k\in\mathbb{Q}\setminus\{0,-1,-2,\ldots\}):f_k(x)\neq0\text{ and }\frac{f_{k+1}(x)}{f_k(x)}\notin\mathbb{Q}.$

Proof: Otherwise, there would be a number $y\neq 0$ and integers $b$ such that $f_{k}(x)=ay$ and $f_{k+1}(x)=by$. In order to see why, take $y=f_{k+1}(x)$, $a=0$ and $b=1$ if $f_{k}(x)=0$; otherwise, choose integers $a$ and $b$ such that ${f_{k+1}(x)\over f_{k}(x)}={b\over a}$ and define $y= {f_{k}(x)\over a}={f_{k+1}(x)\over b}$.
In each case, $y$ cannot be $0$, because otherwise it would follow from claim 1 that each $f_{k+n}(x)(n \in \mathbb N)$ would be $0$, which would contradict claim 2. Now, take a natural number $c$ such that all three numbers ${bc\over k}$, ${ck\over x^2}$ and ${c\over x^2}$ are integers and consider the sequence

$g_n=\begin{cases}f_k(x) & \text{if }n=0\\ \dfrac{c^n}{k(k+1)\cdots(k+n-1)}f_{k+n}(x) & \text{otherwise.}\end{cases}$

Then

$g_0=f_k(x)=ay\in\mathbb{Z} y \text{ and }g_1=\frac ckf_{k+1}(x)=\frac{bc}ky\in\mathbb{Z}y.$

On the other hand, it follows from claim 1 that

$\begin{align} g_{n+2}&=\frac{c^{n+2}}{x^2k(k+1)\cdots(k+n-1)}\cdot\frac{x^2}{(k+n)(k+n+1)}f_{k+n+2}(x)\\[5pt] & =\frac{c^{n+2}}{x^2k(k+1)\cdots(k+n-1)}f_{k+n+1}(x)-\frac{c^{n+2}}{x^2k(k+1)\cdots(k+n-1)}f_{k+n}(x)\\[5pt] &=\frac{c(k+n)}{x^2}g_{n+1}-\frac{c^2}{x^2}g_n\\[5pt] &=\left(\frac{ck}{x^2}+\frac c{x^2}n\right)g_{n+1}-\frac{c^2}{x^2}g_n, \end{align}$

which is a linear combination of $g_{n+1}$and $g_n$ with integer coefficients. Therefore, each $g_n$ is an integer multiple of $y$. Besides, it follows from claim 2 that each $g_n$ is greater than $0$ (and therefore that $g_n ≥ |y|$) if $n$ is large enough and that the sequence of all $g_n$'s converges to $0$. But a sequence of numbers greater than or equal to $|y|$ cannot converge to $0.$

Since $f_{1/2}(π/4) = \cos(π/2) = 0,$ it follows from claim 3 that $π^2/16$ is irrational and therefore that $π$ is irrational.

$\endgroup$
  • $\begingroup$ What's the first step that you fail to understand? $\endgroup$ – José Carlos Santos Jan 3 '18 at 22:55
  • $\begingroup$ First of all, against what he says otherwise there would be a number y? $\endgroup$ – Leyla Alkan Jan 3 '18 at 22:57
  • $\begingroup$ He explains why such a $y$ must exist in the second and third sentences of the proof. I am sorry, but I am unable to see where is it the obstacle. $\endgroup$ – José Carlos Santos Jan 3 '18 at 23:05
  • $\begingroup$ I think I got that part now @JoséCarlosSantos $\endgroup$ – Leyla Alkan Jan 3 '18 at 23:14
  • 1
    $\begingroup$ Since $\lim_{n\to\infty}f_n(x)=1$, all you need to prove is that the seqeunce$$\left(\frac{c^n}{l(k+1)\cdots(k+n-1)}\right)_{n\in\mathbb{N}}$$tends to $0$, which is easy: just compute the quotioent of two consecutive terms of the sequence. $\endgroup$ – José Carlos Santos Jan 4 '18 at 16:49
1
$\begingroup$

I think this might clarify.

Claim 2 asserts that $\lim_{m\to\infty}f_m(x)=1$. Using the definition of limit with $\epsilon=1$, there is some integer $M$ for which $m>M$ implies $f_m(x)>1-1=0$. Thus for any fixed $k$, $f_{k+n}(x)>0$ if $n$ is large enough to ensure that $k+n>M$.

Suppose $n$ is “large enough.” The number $g_n$ is a positive multiple of $f_{k+n}(x)$, so $g_n$ is also greater than zero for such large enough $n$.

The number $g_n$ was shown to be an integer multiple of $y$, and the smallest positive integer multiple of $y$ is $|y|$, so if $g_n>0$, it must be at least $|y|$. (Any smaller positive value could not be an integer multiple of $y$.)

Now we have that $g_n>|y|>0$ for all “large enough” $n$, and so the sequence $g_n=f_{k+n}\cdot \dfrac{c^n}{k(k+1)\cdots(k+n-1)}$ cannot approach zero in the limit as $n\to\infty$. On the other hand, $\lim_{n\to\infty} f_{k+n} = 1$, and $\lim_{n\to\infty} \dfrac{c^n}{k(k+1)\cdots(k+n-1)} = 0$ (the new terms in the denominator eventually become and stay larger than $c$). Therefore $\lim_{n\to\infty} g_{n} = 0$, since $g_n$ is the termwise product of two sequences, one of which approaches $1$ and the other of which approaches zero.

It is here that a contradiction arises, proving Claim 3. The $g_n$ cannot approach zero, yet they do, so the presumption that Claim 3 was false is mistaken.

$\endgroup$
  • $\begingroup$ Thanks so much! Are you here accepting that k is a positive rational number? $\endgroup$ – Leyla Alkan Jan 10 '18 at 1:51
  • $\begingroup$ I’m not sure what you’re asking. I think the claims hold for any $k$, but Claim 3 is only applied for $k=\frac12$. When $x>0$ and $x^2\in\mathbb Q$, it follows that for any rational $k$ that is not a non-positive integer (such as $k=\frac12$), both $f_k(x)\neq0\text{ and }\frac{f_{k+1}(x)}{f_k(x)}\notin\mathbb{Q}$. This claim is considered for $x=\pi>0$. If $\pi^2\in\mathbb Q$, the claim implies that $\frac{f_{k+1}(x)}{f_k(x)}$ is irrational, but it’s not. So $\pi^2$ is not rational. Note: You can read Laczkovich’s paper if you register with Jstor: jstor.org/stable/2974737 $\endgroup$ – Steve Kass Jan 10 '18 at 2:24
  • 1
    $\begingroup$ First, what does $1000$ have to do with anything, since the proof is about limits? As I just outlined, the $g_n$’s avoid a limit of $0$. They might all be positive after a certain point or they might all be negative after a certain point, but the assumptions of the proof by contradiction (with the correction I added to handle $g_n$ that are negative and less than $-|y|$ for large enough $n$) imply that the limit of the $g_n$ is not zero. That is all the proof needs, because the assumptions also imply that the limit of the $g_n$ is zero, hence a contradiction. $\endgroup$ – Steve Kass Jan 10 '18 at 14:29
  • 1
    $\begingroup$ To be more direct in reply to your latest comment, although the proof you quoted says each $g_n$ (for large enough $n$) is greater than zero, I think that is a mistake. Fortunately, all that matters is that for large enough $n$, each $|g_n|>0$, and so each $|g_n|\ge |y|>0$, so the limit of the $g_n$ cannot be zero. $\endgroup$ – Steve Kass Jan 10 '18 at 14:31
  • 1
    $\begingroup$ Yes exactly, all I was thinking was that since this is a published work and if I think there is a uncertainty about a particular thing in the article then it must have a good explanation which I haven't come up with yet. That's why I insisted to get a clear explanation. And I totally agree with why such a contradiction arises. After all taking absolute value of $g_n$ solved the issue about this uncertainty for me. Thanks a lot, again. Now I am ready to present the proof :)@SteveKass $\endgroup$ – Leyla Alkan Jan 10 '18 at 15:22

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.