5

Let $0<x\leq 1$. Prove that there exists a unique sequence $(p_n)$ of integers such that $1<p_1\leq p_2 \leq\cdots$ and $$x=\sum_{n=1}^\infty\frac{1}{p_1\dots p_n}.$$

For the existence part I completely lost, can someone give an idea?

For the uniqueness part I'm also stuck but I tried the following. Let $$x=\sum_{n=1}^\infty\frac{1}{q_1\dots q_n}$$ be another representation of $x$ such that, for $n_0\in\mathbb{N}$ such that for $n< n_0$, $q_n=p_n$ but $p_{n_0}<q_{n_0}$. Then it must be

$$ \sum_{n_0}^\infty \frac{1}{p_1\dots p_{n_0}\dots p_n} = \sum_{n_0}^\infty \frac{1}{p_1\dots p_{n_0-1}q_{n_0}q_{n_0+1}\dots q_n}. \tag{$*$} $$

With this I want to deduce the contradiction $x<x$. To do this I compute

\begin{equation} \begin{split} x &= \sum_{n=1}^{n_0-1}\frac{1}{p_1\dots p_n} + \frac{1}{p_1\dots q_{n_0}} + \sum_{n=n_0+1}^\infty\frac{1}{p_1\dots q_{n_0}\dots q_n}\\ &< \sum_{n=1}^{n_0-1}\frac{1}{p_1\dots p_n} + \frac{1}{p_1\dots p_{n_0}}+ \sum_{n=n_0+1}^\infty\frac{1}{p_1\dots q_{n_0}\dots q_n}\\ &< \sum_{n=1}^{n_0-1}\frac{1}{p_1\dots p_n} + \frac{1}{p_1\dots p_{n_0}} -\frac{1}{p_1\dots q_{n_0}} +\sum_{n=n_0}^\infty\frac{1}{p_1\dots p_{n_0}\dots p_n}\\ &= x + \frac{1}{p_1\dots p_{n_0}} -\frac{1}{p_1\dots q_{n_0}}, \end{split} \end{equation}

where in the second inequality I used $(*)$. Obviously the last part is a number greater than zero $\alpha$, so $x<x+\alpha$. This is the farthest I can get.

user2820579
  • 2,389

2 Answers2

1

Let $p_1$ be the smallest positive integer for which $1/p_1<x.$ Let $x_1 = x,$ and this will be the first term in a sequence of $x\text{s.}$ Then $$ \require{cancel} \xcancel{\begin{align} \frac 1 {p_1} < {} & x \le \frac 1 {p_1-1} \\ & \text{so} \\ 0 < x - \frac 1 {p_1} \le {} & \frac 1 {p_1-1} - \frac 1 {p_1} = \frac 1 {p_1(p_1-1)}. \end{align}} $$ The above probably needs further work. See the comments below.

And so we have $$ 0 < (p_1-1)(p_1x-1) \le 1. $$ Let $x_2 = \text{this latest number}=(p_1-1)(p_1x-1).$

And now $p_2$ is to $x_2$ as $p_1$ to $x_1$ (and here I'm not talking about a ratio; I just mean $p_2$ is found in the same way).

Just keep going like that. That shows existence.

  • How do you ensure $p_1\leq p_2\leq \dots$? Also I get $x = \frac{x_2}{p_1(p_1-1)} + \frac{1}{p_1}$ and since $x_3 = (p_2-1)(x_2p_2-1)$ this implies $x_2=\frac{x_3}{p_2(p_2-1)}+\frac{1}{p_2}$ and $x=\frac{1}{p_1} + \frac{1}{p_2 p_1 (p_1-1)} + \frac{x_3}{p_1 p_2(p_2-1)(p_1-1)}$. I don't get the series. – user2820579 Jul 02 '20 at 20:39
  • Shouldn't be $0<xp_1-1\leq 1/(p_1-1)\leq 1$ with $x_2 = xp_1 -1$? – user2820579 Jul 02 '20 at 20:52
1

Idea for uniqueness: Suppose $q_1>p_1$ and

$$\tag 1 \frac{1}{p_1}+ \frac{1}{p_1p_2}+\cdots = \frac{1}{q_1}+ \frac{1}{q_1q_2}+\cdots.$$

The right side of $(1)$ is then no more than

$$\frac{1}{q_1}+ \frac{1}{q_1^2}+\frac{1}{q_1^3}+\cdots = \frac{1}{q_1-1} \le \frac{1}{p_1}$$

But the left side of $(1)$ is greater than $\dfrac{1}{p_1},$ contradiction. So $q_1>p_1$ is impossible, and by symmetry so is $p_1>q_1.$ Thus $p_1=q_1.$

Possible lather, rinse, repeat scenario.

zhw.
  • 105,693