3

I'm trying to prove the following:

Let $n\in \mathbb N$ and $$ x_n = \frac{(2n)!!}{(2n-1)!!} $$ Where: $$ (2n)!! = 2\cdot 4 \cdot 6 \cdot \dots \cdot 2n \\ (2n - 1)!! = 1\cdot 3 \cdot 5 \cdot \dots \cdot (2n-1) $$ Prove $\{x_n\}$ is not bounded.

Intuitively it feels like it is unbounded. Let's consider the following fraction:

$$ x_n = \frac{2\cdot 4 \cdot 6 \cdot \dots \cdot (2n)}{1\cdot 3 \cdot 5 \cdot \dots \cdot (2n-1) } $$

Now if we consequently take a number from nominator and denominator we get:

$$ \frac{2}{1} > 1 \\ \frac{4}{3} > 1 \\ \frac{6}{5} > 1 \\ \dots \\ \frac{2n}{2n-1} > 1 $$

So the fraction is a product of rational numbers each of which is greater than $1$ and the product of rational numbers greater than $1$ is increasing.

I've tried to formalize that by expanding $(2n)!!$ and $(2n-1)!!$:

$$ (2n)!! = (2n)(2n-2)(2n-4)\cdots(4)(2) = 2^kn!\\ (2n-1)!! = (2n-1)(2n-3)\cdots(5)(3)(1) = \\ = \frac{(2n-1)(2n-2)(2n-3)\cdots(5)(4)(3)(2)(1)}{(2n-2)(2n-4)\dots(4)(2)} = \\ \frac{(2n-1)!}{2^{n-1}(n-1)!} $$

So using the above:

$$ x_n = \frac{2^nn!2^{n-1}(n-1)!}{(2n-1)!} = \frac{2^{2n-1}n!(n-1!)}{(2n-1)!} $$

Here is where I got stuck. How do i proceed with the proof using some constant $M$ and some number $N > n$ such that $x_N > M$? Should i introduce some inequality?

I've seen a similar question, but the sequence there is proven to be divergent which i guess is more a calculus concept (yet very similar to (un)boundedness), and i'm in search of a precalculus solution.

roman
  • 5,391
  • $\frac{1}{4^n}\binom{2n}{n}\leq\frac{1}{\sqrt{\pi n}}$ has a simple proof via $\cos(x)\leq e^{-x^2/2}$ over $(0,\pi/2)$ (which on its turn follows from $\tan x\geq x$, integration and exponentiation): $$\frac{1}{4^n}\binom{2n}{n}=\frac{\pi}{2}\int_{0}^{\pi/2}(\cos x)^{2n},dx < \frac{\pi}{2}\int_{0}^{+\infty}e^{-nx^2},dx.$$ – Jack D'Aurizio Oct 29 '18 at 13:58
  • In your question you just need to show that $\int_{0}^{\pi/2}(\cos x)^{2n},dx$ converges to zero as $n\to +\infty$, which is trivial from the dominated convergence theorem, for instance. – Jack D'Aurizio Oct 29 '18 at 14:00
  • @JackD'Aurizio wouldn't that require some knowledge in calculus (particularly integration)? Unfortunately it seems like i'm missing the basis required to follow your approach for now. Anyway thank you for pointing this out! – roman Oct 29 '18 at 14:06

4 Answers4

3

As simple as it gets:

LEMMA: For $k\in N$ and $k>1$

$$\frac{2k}{2k-1}\gt\frac{\sqrt{3k+1}}{\sqrt{3k-2}}\tag{1}$$

Proof: It's trivial. Both sides are positive, square them:

$$\frac{4k^2}{4k^2-4k+1}\gt\frac{3k+1}{3k-2}$$

..which is equivalent to:

$$4k^2(3k-2)\gt(4k^2-4k+1)(3k+1)$$

$$12k^3-8k^2\gt12k^3+4k^2-12k^2-4k+3k+1$$

$$12k^3-8k^2\gt12k^3-8k^2-k+1$$

$$0\gt-k+1$$

...which is obviously true.

End of lemma proof.

By using the inequality (1) we have:

$$\frac{2}{1}=\frac{\sqrt{4}}{\sqrt{1}}$$

$$\frac{4}{3}\gt\frac{\sqrt{7}}{\sqrt{4}}$$

$$\frac{6}{5}\gt\frac{\sqrt{10}}{\sqrt{7}}$$

$$...$$

$$\frac{2n-2}{2n-3}\gt\frac{\sqrt{3n-2}}{\sqrt{3n-5}}\tag{1}$$

$$\frac{2n}{2n-1}\gt\frac{\sqrt{3n+1}}{\sqrt{3n-2}}\tag{1}$$

Multiply all this and you get:

$$a_n=\frac{(2n)!!}{(2n-1)!!} > \sqrt{3n+1}$$

So $a_n$ is obviously unbounded. Cute, isn't it? :)

Saša
  • 15,906
  • Introducing this lemma is very smart. How would you come up with that idea? – roman Oct 29 '18 at 13:21
  • 1
    @roman I saw the same trick used to estimate the upper bound of the central binomial coefficient: ${2n \choose n}\le\frac{4^n}{\sqrt{3n+1}}$. And I noticed that $\frac{(2n)!!}{(2n-1)!!}=\frac{4^n}{2n \choose n}$. I have just played the same old trick in a slightly different context. – Saša Oct 29 '18 at 13:43
2

This answer uses calculus-based language. It is likely not quite what the OP was looking for (requesting a pre-calculus solution), but I'm posting so that someone else will suggest a way to avoid the use of l'Hopital's rule.

Observe that $x_n$ is unbounded if and only if $\ln(x_n)$ is unbounded. Since $x_n$ is a product, we can write $$ \ln(x_n)=\sum_{k=1}^{n}\ln\left(\frac{2k}{2k-1}\right). $$

If we show that $$ \sum_{k=1}^{\infty}\ln\left(\frac{2k}{2k-1}\right) $$ exists (i.e., converges), then the $\ln(x_n)$'s must be bounded as they are the partial sums of this sum. We will see, however, that the opposite is true.

We will use the limit comparison test and compare this series to the (divergent) harmonic series with entries $\frac{1}{k}$. Therefore, we compute $$ \lim_{k\rightarrow\infty}\frac{\ln\left(\frac{2k}{2k-1}\right)}{\frac{1}{k}}. $$ This limit is an indeterminate form $\frac{0}{0}$, so we can use l'Hopital's rule (the only Calculus part of this answer) to get that this limit equals $$ \lim_{k\rightarrow\infty}\frac{\frac{2}{2k}-\frac{2}{2k-1}}{-\frac{1}{k^2}}= \lim_{k\rightarrow\infty}\frac{\frac{-2}{2k(2k-1)}}{-\frac{1}{k^2}}= \lim_{k\rightarrow\infty}\frac{2k^2}{2k(2k-1)}=1. $$ Therefore, both series diverge by the limit comparison test. Since $\sum\frac{1}{k}$ diverges to infinity, so does $\sum\ln\left(\frac{2k}{2k-1}\right)$, i.e., $\ln(x_n)$ is unbounded.

Michael Burr
  • 32,867
  • Even though it uses a bit of calculus, it is still a nice way to attack the problem. Thank you. +1 for sharing your ideas – roman Oct 29 '18 at 13:26
2

If you take

$$\log(1-x)\le-x$$ for granted*,

$$\log x_n=\sum_{k=1}^n\log\left(\frac{2k}{2k-1}\right)=-\sum_{k=1}^n\log\left(1-\frac1{2k}\right)>\frac12\sum_{k=1}^n\frac1k$$

which diverges.


You can draw it from

$$\log(1+0)\le-0$$ and

$$(\log(1+x))'=\frac1{1+x}\le(-x)'=-1.$$

2

Without integrals or derivatives: $$ x_n=\frac{(2n)!!}{(2n-1)!!}= \prod_{k=0}^{n-1}\left(1+\frac{1}{2k+1}\right)=\prod_{k=1}^{n}\left(1-\frac{1}{2k}\right)^{-1}$$ implies $$ x_n^2 = \prod_{k=1}^{n}\left(1-\frac{1}{k}+\frac{1}{4k^2}\right)^{-1}=\frac{1}{4}\prod_{k=2}^{n}\left(1-\frac{1}{k}\right)^{-1}\prod_{k=2}^{n}\left(1-\frac{1}{(2k-1)^2}\right) $$ or $$ x_n^2 = \frac{n}{4}\prod_{k=1}^{n-1}\left(1-\frac{1}{(2k+1)^2}\right). $$ On the other hand the infinite product $\prod_{k\geq 1}\left(1-\frac{1}{(2k+1)^2}\right)$ is convergent to a positive constant since $\sum_{k\geq 1}\frac{1}{(2k+1)^2}$ is convergent to a positive constant, hence $x_n\geq \sqrt{K n}$ for some mysterious constant $K>0$, which actually equals $\pi$.

Jack D'Aurizio
  • 353,855