0
$\begingroup$

I have been reading a book on approximating $e$ and there is a couple lines that I am stuck on. Here they are:

$x$ln$(1+\displaystyle\frac{1}{x}) = 1 - \displaystyle\frac{1}{2x} + \displaystyle\frac{1}{3x^2} - \displaystyle\frac{1}{4x^3} + ...$

Exponentiating each side and collecting the same powers of $\displaystyle\frac{1}{x}$ with the help of the Maclaurin series of $e^x$ we find that,

$(1 + \displaystyle\frac{1}{x})^x = e(1 - \displaystyle\frac{1}{2x} + \displaystyle\frac{11}{24x^2} - \displaystyle\frac{7}{16x^3} + \displaystyle\frac{2447}{5760x^4} - \displaystyle\frac{959}{2304x^5} + ...)$

I am not sure how they went about going from the first term which makes sense to me to the second term with some pretty random numbers. I would appreciate the help and filling in the details that I do not get.

$\endgroup$
  • $\begingroup$ Is it correct that on the right side you multiply $e$ instead of raising $e$ to the rhs? $\endgroup$ – flawr Sep 18 '14 at 16:45
  • $\begingroup$ @flawr Yes, that is not an error. $\endgroup$ – heropup Sep 18 '14 at 16:47
1
$\begingroup$

The idea is this: we know that $x \log (1 + x^{-1}) = \log (1+x^{-1})^x$. Now exponentiating both sides gives $$(1+x^{-1})^x = \exp \left(1 - \frac{1}{2x} + \frac{1}{3x^2} - \frac{1}{4x^3} + \cdots \right) = e \cdot e^{-1/2x} \cdot e^{1/3x^2} \cdot e^{-1/4x^3} \cdots.$$ Now consider each factor separately: $$\exp \frac{1}{(k+1)(-x)^k} = \sum_{n=0}^\infty \frac{1}{n!} \left( (k+1)(-x)^k \right)^{-n}. $$ Excluding the first constant factor $e$ in the infinite product, they expand $e^{-1/2x}$, $e^{1/3x^2}$, etc. and multiply out all those expansions. That's computationally tedious but it is not too bad to get the first few terms this way, because if you want to get the $m^{\rm th}$ term of the original series expansion, then you only need $n_{\rm max} = \lfloor m/k \rfloor$ terms for the $k^{\rm th}$ exponential series expansion. So for instance, if I want up to the $x^{-3}$ term for the original series, the first few exponential expansions are $$\begin{align*} e^{-1/2x} &= 1 - \frac{1}{2x} + \frac{1}{8x^2} - \frac{1}{48x^3} + \cdots, \\ e^{1/3x^2} &= 1 + \frac{1}{3x^2} + \cdots, \\ e^{-1/4x^3} &= 1 - \frac{1}{4x^3} + \cdots. \end{align*}$$ I don't need any more terms because all of them will be smaller than $x^{-3}$. Then I take the product, ignoring any subproducts smaller than $x^{-3}$, which is easily done by hand. Of course, if you want higher order terms, this is best left to a computer.


Per the OP's request. Pay careful attention to how we discard any intermediate products that are of higher order than $x^{-3}$:

$$\begin{align*} e^{-1/2x}e^{1/3x^2}e^{-1/4x^3} &= \left(1 - \tfrac{1}{2x} + \tfrac{1}{8x^2} - \tfrac{1}{48x^3} + \cdots \right)\left(1 + \tfrac{1}{3x^2} + \cdots \right)\left(1 - \tfrac{1}{4x^3} + \cdots \right) \\ &= \left(1 - \tfrac{1}{2x} + \tfrac{1}{8x^2} - \tfrac{1}{48x^3} + \cdots \right)\left(1 + \tfrac{1}{3x^2} - \tfrac{1}{4x^3} + \cdots \right) \\ &= 1-\tfrac{1}{2x}+\tfrac{1}{8x^2}+\tfrac{1}{3x^2}-\tfrac{1}{48x^3}-\tfrac{1}{6x^3}-\tfrac{1}{4x^3}+\cdots \\ &= 1 - \tfrac{1}{2x} + \tfrac{11}{24x^2} - \tfrac{7}{16x^3} + \cdots. \end{align*}$$

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Property of logs. Makes sense. You can put the $x$ back up there. Your second line makes sense. If we thought about it using a simple example like $x^{(a+b)}$ then this is just $x^a \times x^b$. Makes sense. The only thing I am unsure of is where the coefficients are coming from. I dont see where they are coming from. By coefficients I mean the second series I presented in my question. $\endgroup$ – Nick Freeman Sep 18 '14 at 17:05
  • $\begingroup$ Take the three exponential expansions I wrote in my answer, and multiply them together. The result will match the series of $(1+x^{-1})^x$ up to the $x^{-3}$ term. $\endgroup$ – heropup Sep 18 '14 at 17:19
  • $\begingroup$ So what your saying is take all the series for each exponential term and multiply them together? If thats the case, I would find that quite tedious on paper. Lots of foiling. However just so I am on the same page with you, could you show me the process in which we were get up to the $x^{-3}$ term? I just want to make sure I have it and then I think I will understand this perfectly. $\endgroup$ – Nick Freeman Sep 18 '14 at 17:28

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.