0

Demonstrate recursively that

$$\prod_{k = 0}^\infty (1 + x^{2^k}) = \frac{1}{1-x}$$

My work:

Define

$$a_n = \prod_{k = 0}^n (1 + x^{2^k}) = (1 + x^{2^n})a_{n - 1} \iff a_n - (1 + x^{2^n})a_{n - 1} = 0$$ $$A(x) = \sum_{n = 0}^\infty a_nx^n = a_0 + \sum_{n = 1}^\infty a_{n}x^n = 1 + x + \sum_{n = 1}^\infty a_{n}x^n$$ $$\implies xA(x) = \sum_{n = 1}^\infty a_{n-1}x^n$$

Therefore,

\begin{align} A(x) - (1 + x^{2^n})xA(x) &= 1 + x + \sum_{n = 1}^\infty a_{n}x^n - \sum_{n = 1}^\infty (1 + x^{2^n})a_{n-1}x^n\\ A(x) \left(1 - x(1 + x^{2^n})\right) &=1 + x + \sum_{n = 1}^{\infty} (a_n - (1 + x^{2^k})a_{n - 1})x^n\\ A(x) \left(1 - x(1 + x^{2^n})\right) &=1 + x\\ A(x) &= \frac{1 + x}{1 - x(1 + x^{2^n})}\\ &= \frac{1 + x}{1 - x - x^{2^n + 1}} \end{align}

To find $a_n$, I now want to transform $A(x)$ in the form $\sum_{k = 0}^\infty a_nx^n$ and then take the limit as $n \to \infty$. However, I’m unsure how to do this since I don’t think the expression can be decomposed into partial fractions.

My question:

  • Is my approach correct?
  • How do I continue?

Note: I know that there are other solutions, but I specifically want to see if defining a recurrence can yield a solution.

  • 1
    You should probably work with $A(y)$ instead. – J.G. Jul 06 '22 at 21:26
  • For a recursion see the answers of this post. – Dietrich Burde Jul 06 '22 at 21:29
  • The series $\sum_{n=0}^{\infty}a_n x^n$ cannot uniquely determine $a_n$'s since $a_n$'s themselves are polynomials in $x$. You should use different variables for the generating function so it can uniquely determine $a_n$'s; for example, you may follow J.G.'s advice and consider $$\sum_{n=0}^{\infty} a_n y^n$$ instead. – Sangchul Lee Jul 06 '22 at 21:33
  • If you expand the products, you get the geometric series $1+x+x^2+\dots$, which also converges to $\frac1{1-x}$ for $|x|<1$. – PC1 Jul 06 '22 at 22:01
  • You can write $\prod_{k=0}^{2^{n-1}}\dots=\sum_{k=0}^{2^n-1}x^n$ recursively. – PC1 Jul 06 '22 at 22:04
  • @PC1 I’m not sure I see the recursion, could you elaborate? –  Jul 06 '22 at 22:07
  • Will write an answer. – PC1 Jul 06 '22 at 22:09
  • @PC1 Okay, sure. I just hope that you aren’t forming a recursion from the sum, but rather from the product. (Because transforming the product into the sum is already a big algebraic step that pretty much kills the essence of a recursive solution). –  Jul 06 '22 at 22:10
  • I'm not sure if I understand your comment but I have written what I think is a recursive solution... – PC1 Jul 06 '22 at 22:16
  • 2
    Side-note, this is equivalent to the fact that nonnegative integers have unique binary expansions. Cheers – Jair Taylor Jul 06 '22 at 22:19
  • 1
    @JairTaylor Yeah! I've actually implied that here. –  Jul 06 '22 at 22:20
  • This doesn't make much sense to me. I believe you totally misapplied the concept of generating functions. The way you've done it, you'll find that an are polynomials themself. W – tryst with freedom Jul 06 '22 at 23:20

2 Answers2

1

(This may be a duplicate answer, but I worked it out myself, so, whatever.)

As in other answer answers, let $a_n = \prod_{k = 0}^n (1 + x^{2^k}) $ and $b_n=(1-x)a_n$.

Then, if $n \ge 2$,

$\begin{array}\\ b_n &= (1-x)\prod_{k = 0}^n (1 + x^{2^k})\\ &= (1-x)(1+x)\prod_{k = 1}^n (1 + x^{2^k})\\ &= (1-x^2)\prod_{k = 1}^n (1 + x^{2^k})\\ &= (1-x^2)(1+x^2)\prod_{k = 2}^n (1 + x^{2^k})\\ &= (1-x^4)\prod_{k = 2}^n (1 + x^{2^k})\\ \end{array} $

I claim that, for $m \le n$, $b_n = (1-x^{2^m})\prod_{k = m}^n (1 + x^{2^k}) $.

Proof.

This is true for $m=0, 1$.

If true for $m<n$ then

$\begin{array}\\ b_n &= (1-x^{2^m})\prod_{k = m}^n (1 + x^{2^k})\\ &= (1-x^{2^m})(1+x^{2^m})\prod_{k = m+1}^n (1 + x^{2^k})\\ &= (1-x^{2^{m+1}})\prod_{k = m+1}^n (1 + x^{2^k})\\ \end{array} $

Setting $m=n-1$,

$\begin{array}\\ b_n &= (1-x^{2^n})\prod_{k = n}^n (1 + x^{2^k})\\ &= (1-x^{2^n})(1 + x^{2^n})\\ &= (1-x^{2^{n+1}})\\ \end{array} $

Therefore, if $|x| < 1$, $\lim_{n \to \infty} b_n =1 $ so $\lim_{n \to \infty} a_n =\dfrac1{1-x} $.

marty cohen
  • 107,799
0

Let's say that $A_n=\prod_{k=0}^n(1+x^{2^k})$. We have $A_0=1+x$. We will prove that $B_n=\sum_{k=0}^{2^n-1}x^k$ is equal to $A_n$. It is clear that $B_0=1+x=A_0$.

So we assume that $A_i=B_i$ and will prove that this implies $B_{i+1}=A_{i+1}$.

\begin{align} A_{i+1}&=\prod_{k=0}^{i+1}(1+x^{2^k})\\ &=\left(1+x^{2^i}\right)\prod_{k=0}^{i}(1+x^{2^k})\\ &=A_i+x^{2^i}A_i\\ &=B_i+x^{2^i}B_i\\ &=\sum_{k=0}^{2^i-1}x^k+x^{2^i}\sum_{k=0}^{2^i-1}x^k\\ &=\sum_{k=0}^{2^{i+1}-1}x^k\\ &=B_{i+1} \end{align}

PC1
  • 2,175
  • Thanks for the solution, (and I should probably have made this clearer) but I'm trying to approach this from the perspective of not knowing what the product evaluates to. Further, this feels more inductive than recursive at its essence, which is not exactly what I am looking for. –  Jul 06 '22 at 22:17
  • You could try to prove that $A_i=\frac{1-x^{2^i+1}}{1-x}$ but I don't know if this would be recursive enough... – PC1 Jul 06 '22 at 22:21
  • How is induction different from recursion..? @BeKind See here for instance – tryst with freedom Jul 06 '22 at 22:49
  • I couldn't really grasp what's the problem, the proof is using a recurrence... – PC1 Jul 06 '22 at 23:00
  • @PC1 My main issue is that I am looking for a derivation, rather than a proof. –  Jul 06 '22 at 23:01