0

I'm attempting to find a series approximation to:

$$f(x)=\frac{1}{1-(1+x)^{-n}}$$

where $n\in\mathbb{N^+}$ and is a constant, and $x\in\mathbb{R}$ and $0<x<1$.

Using Wolfram Alpha, I noted the first order approximation is:

$$f(x)\approx\frac{1}{nx} + \frac{n+1}{2n}$$

With this goal post in mind, I went about trying to find the first order approximation for myself. The original formula has a singularity at $x=0$, so I figured I can't use the Taylor series formula directly. Instead, with the geometric series formula I found the following:

$$f(x)=\sum_{i=0}^\infty\frac{1}{(1+x)^{in}} $$

But I'm not sure how to proceed from here. Is there a trick I'm missing?

Looking around for a hint on how to proceed, I noticed that $f(x)$ has some similarities to mortgage monthly payments formula:

$$C(r)=\frac{r(1+r)^nP}{(1+r)^n-1}$$

where $P$ is the initial load size (principal), $r$ is the interest rate per month and $n$ is the number of payments (months). I found similar questions here and here, but it's still not clear to me what the initial steps are for finding the first order approximation given by Wolfram Alpha.

Any help would be much appreciated, even if it's just the first few steps towards the solution. Thanks.

Akyidrian
  • 267
  • Sage gives the first few terms as $-\frac{{\left(n^{4} - 20 , n^{2} + 19\right)} x^{3}}{720 , n} - \frac{{\left(n^{2} - 1\right)} x^{2}}{24 , n} + \frac{{\left(n^{2} - 1\right)} x}{12 , n} + \frac{n + 1}{2 , n} + \frac{1}{n x} + \cdots$. – A rural reader Jan 26 '23 at 01:16

2 Answers2

2

Start by multiplying both numerator and denominator by $(1+x)^n$: $$f(x)=\frac{1}{1-(1+x)^{-n}}=\frac{(1+x)^n}{(1+x)^n-1}$$ Now let's use binomial expansion $$(1+x)^n=1+nx+\frac{n(n-1)}2x^2+...$$ This yields $$f(x)=\frac{1+nx+\frac{n(n-1)}2x^2+...}{nx+\frac{n(n-1)}2x^2+...}\\=\frac1{nx}\frac{1+nx+\frac{n(n-1)}2x^2+...}{1+\frac{n-1}2x+...}$$ We can now use $$\frac1{1+\frac{n-1}2x+...}\approx1-\frac{n-1}2x$$ This yields $$f(x)\approx\frac1{nx}\left(1+nx+\frac{n(n-1)}2x^2+...\right)\left(1-\frac{n-1}2x\right)\\\approx\frac1{nx}\left(1+nx-\frac{n-1}2x\right)\\=\frac1{nx}\left(1+\frac{n+1}2x\right)=\frac1{nx}+\frac{n+1}{2n}$$ Note that at my last approximation I've dropped terms in $x^2$. If I want to keep those, I need one more term in an earlier expression.

Andrei
  • 39,869
  • Thanks for your answer. Can you elaborate on how you determined the following approximation: $$\frac1{1+\frac{n-1}2x+...}\approx1-\frac{n-1}2x$$ – Akyidrian Feb 06 '23 at 19:20
  • 1
    You want the denominator to contain only constant terms or terms with $x$ to powers greater than $2$. $$\frac{1}{1+x}=\frac{1-x}{(1+x)(1-x)}=\frac{1-x}{1-x^2}$$ Now ignore the $x^2$ – Andrei Feb 07 '23 at 02:28
2

For any positive integer $n$, $$\eqalign{f(x) &= \frac{1}{1 - (1+x)^{-n}} = \frac{(1+x)^n}{(1+x)^n - 1}\cr &= \frac{(1+x)^n}{\displaystyle\prod_{j=0}^{n-1} ((x + 1 - e^{2\pi i j/n}) }}$$ This is a rational function with limit $1$ as $x \to \infty$ and simple poles at $e^{2\pi i j/n}-1$ for $j = 0$ to $n-1$, i.e. $\omega- 1$ for the $n$'th roots of unity $\omega$. Thus it can be written as $$ \eqalign{f(x) &= 1 + \sum_{j=0}^{n-1} \frac{c_j}{x + 1 - e^{2\pi i j/n}}\cr &= 1 + \frac{c_0}{x}- \sum_{j=1}^{n-1} c_j \sum_{k=0}^\infty (-1+e^{2\pi i j/n})^{-1-k} x^k}$$ for appropriate constants $c_j$.

Robert Israel
  • 470,583
  • I suspect my complex maths has atrophied as I'm not sure I completely follow. Do you care to expand on: $(1+x)^n - 1 = \displaystyle\prod_{j=0}^{n-1} ((x + 1 - e^{2\pi i j/n})$? How do you also determine the "appropriate constants" such that you can obtain the approximation? – Akyidrian Feb 06 '23 at 20:18
  • 1
    For any monic polynomial of degree $n$, $p(z) = \prod_{k=1}^n (z-r_k)$ where $r_k$ are the roots of $p(z)$ (counted with multiplicity). The roots of $z^n - 1$ are the $n$'th roots of unity, namely $e^{2\pi i j/n}$ for $j = 0 \ldots n-1$. $c_j$ is the residue of $f(x)$ at the root $x = e^{2\pi i j/n} - 1$. – Robert Israel Feb 07 '23 at 16:34
  • Thanks - I can see where you're going. I definitely need to review my complex mathematics; it's been a while. – Akyidrian Feb 08 '23 at 05:20