4
$\begingroup$

Mathcad software gives me the answer as: $$ \frac{1}{1+e^x} = \frac{1}{2} -\frac{x}{4} +\frac{x^3}{48} -\frac{x^5}{480} +\cdots$$ I have no idea how it found that and i don't understand. What i did is expand $(1+e^x)^{-1} $ as in the binomial MacLaurin expansion. I want until the $x^4$ term. So what i found is the following:

$$ (1+e^x)^{-1} = 1 - e^x + e^{2x} + e^{3x} + e^{4x}+\cdots\tag1 $$ Then by expanding each $e^{nx}$ term individually: $$ e^x = 1 + x + \frac{x^2}{2} + \frac{x^3}{6} + \frac{x^4}{24}+\cdots\tag2 $$ $$ e^{2x} = 1 + 2x + 2x^2 + \frac{3x^3}{2} + \frac{2x^4}{3}+\cdots\tag3 $$ $$ e^{3x} = 1 + 3x + \frac{9x^2}{2} + \frac{9x^3}{2} + \frac{27x^4}{8}+\cdots\tag4 $$ $$ e^{4x} = 1 + 4x + 8x^2 + \frac{32x^3}{3} + \frac{32x^4}{3}+\cdots\tag5 $$ So substituting $(2),(3),(4),(5)$ into $(1)$ i get: $$ (1+e^x)^{-1} = 1 +2x+5x^2+ \frac{22x^3}{3} + \frac{95x^4}{12} +\cdots$$ Which isn't the correct result. Am i not allowed to expand this series binomially? I've seen on this site that this is an asymptotic expansion. However, i don't know about these and i haven't been able to find much information on this matter to solve this. If someone could help me understand how to solve it and why my approach isn't correct, i would be VERY grateful. Thanks in advance.

Improve this question
$\endgroup$
  • 3
    $\begingroup$ We have $(1+e^x)^{-1}=1-e^x+e^{2x}-\cdots$ only if $x<0$. If $x>0$, you need to factor $e^x$ out of the denominator. $\endgroup$ – Clayton Aug 21 '15 at 12:55
  • 1
    $\begingroup$ You forgot the series for $(1+u)^{-1}$ converges if and only if $\lvert u\rvert <1$, which is not the case of $\mathrm e^x$. $\endgroup$ – Bernard Aug 21 '15 at 13:05
  • 2
    $\begingroup$ Bernard is correct, @RestlessC0bra: If $e^x<1$, we have $x<0$. We need it for the geometric series to converge. Otherwise, you can do what I (and mwomath) suggested, which is to factor $e^x$ out of the denominator. $\endgroup$ – Clayton Aug 21 '15 at 13:06
  • 1
    $\begingroup$ Expanding $1/(1+a)$, you get $1-a+a^2-a^3+a^4-\cdots$, with alternating signs, not $1-a+a^2+a^4+ \cdots$, with the same sign in each term after the $1$st-degree term. It's not clear why you stopped after $(5)$. ${}\qquad{}$ $\endgroup$ – Michael Hardy Aug 21 '15 at 13:07
  • 1
    $\begingroup$ @RestlessC0bra : If $x$ is real, the $e^x>0$. But one needs $e^x<1$, which means $x<0$. ${}\qquad{}$ $\endgroup$ – Michael Hardy Aug 21 '15 at 13:08
2
$\begingroup$

Since you're just looking for the first few terms: $$f(x)=\frac{1}{1+e^x} \implies f(0)=\frac 12$$ $$f'(x)=\frac{-e^x}{(1+e^x)^2} \implies f'(0)=-\frac 14$$ $$f''(x)=\frac{-e^x}{(1+e^x)^2}+2\frac{e^{2x}}{(1+e^x)^3}\implies f''(0)=0$$ $$f'''(x)=f''(x)+4\frac{e^{2x}}{(1+e^x)^3}-3\times 2\frac{e^x}{(1+e^x)^4}\implies f'''(0)=\frac 18$$

You get: $$\frac{1}{1+e^x}=\frac 1{2\times 0!} -\frac 1{4\times 1!} x+\frac 1{8\times 3!}x^3+\mathcal O (x^5)$$ $$\frac{1}{1+e^x}=\frac 1{2} -\frac 1{4} x+\frac 1{48}x^3+ \mathcal O (x^5)$$

Improve this answer
$\endgroup$
  • 1
    $\begingroup$ @RestlessC0bra Since this is a truncated Taylor series we need this "Big-O" notation to describe how closely the series approximates the given function. $\endgroup$ – GeorgSaliba Aug 21 '15 at 15:15
  • $\begingroup$ I mark this as the accepted answer because it was the most simple for me. But i'll study all the other answers as well. Thank you all. $\endgroup$ – Nikos Aug 21 '15 at 20:23
  • $\begingroup$ @RestlessC0bra - how come your username is spelt with a lower case "o" but when typing it with an "@" it appears with a zero or "0"? maybe now you can also consider using $\Theta$ instead... :) $\endgroup$ – Hypergeometricx Aug 21 '15 at 21:12
  • $\begingroup$ @GeorgSaliba - what's the difference between $O(.)$ and $\Theta (.)$ here? $\endgroup$ – Hypergeometricx Aug 21 '15 at 21:13
  • $\begingroup$ @hypergeometric Sorry, it should be $\mathcal O (.)$ because it's an upper bound not a tight bound (where $\Theta (.)$ would've been correct). $\endgroup$ – GeorgSaliba Aug 21 '15 at 21:21
4
$\begingroup$

Let $\dfrac1{1+e^x}=a_0+a_1x+a_2x^2+\cdots$

$$\implies(1+e^x)\left(a_0+a_1x+a_2x^2+\cdots\right)=1$$

As $e^x=\sum_{r=0}^\infty\dfrac{x^r}{r!}$

$$\implies\left(2+\dfrac x1+\dfrac{x^2}2+\dfrac{x^3}{3!}+\cdots\right)\left(a_0+a_1x+a_2x^2+\cdots\right)=1$$

Comparing the coefficients of different powers of $x,$

$\implies1=2a_0\iff a_0=?$

$0=2a_1+a_0\iff a_1=?$

$0=2a_2+a_1+\dfrac{a_0}2$

and so on

Improve this answer
$\endgroup$
  • $\begingroup$ This is a nice, simple approach. $\endgroup$ – user84413 Aug 21 '15 at 17:33
  • $\begingroup$ And what is the radius of convergence? $\endgroup$ – BeesaFangirl DOTO Dec 27 '19 at 21:59
  • $\begingroup$ @BeesaFangirlDOTO, The Series should converge when the series of $e^x$ converges. $\endgroup$ – lab bhattacharjee Dec 28 '19 at 10:46
3
$\begingroup$

Just a sketch of a proof to be expanded later (sorry, I am in a hurry): by taking the logarithmic derivative of the Weierstrass product of the $\cosh$ function we have the Taylor series of the $\tanh$ function, with the coefficients depending on values of the $\zeta$ function at even integers. With a little maquillage it is not difficult to turn that into the Taylor series of $\frac{1}{1+e^x}$. Or we may just play a bit with the generating function of Bernoulli numbers.

So, I was saying: $$\cosh z = \prod_{n\geq 0}\left(1+\frac{4z^2}{(2n+1)^2 \pi^2}\right)\tag{1}$$ hence: $$ \tanh z = \sum_{n\geq 0}\frac{8z}{(2n+1)^2 \pi^2 + 4z^2}\tag{2} $$ leads to: $$ \tanh z = \sum_{m\geq 0}(-1)^m z^{2m+1}\sum_{n\geq 0}\frac{2^{2m+3}}{\pi^{2m+2}(2n+1)^{2m+2}}\tag{3}$$ by expanding the general term of the RHS of $(2)$ as a geometric series. Computing the innermost sum in terms of the $\zeta$ function it follows that: $$\tanh z = \sum_{m\geq 0}\frac{2(-1)^m\left(2^{2m+2}-1\right) }{\pi^{2m+2}}\zeta(2m+2)\, z^{2m+1}\tag{4}$$ but since $\frac{1}{1+e^{z}}=\frac{1-\tanh(z/2)}{2}$ it follows that: $$\frac{1}{1+e^{z}}=\frac{1}{2}-\sum_{m\geq 0}\frac{(-1)^m\left(2^{2m+2}-1\right) }{2^{2m+1}\pi^{2m+2}}\zeta(2m+2)\, z^{2m+1}\tag{5}$$ so to solve the original problem it is enough to recall that $\zeta(2)=\frac{\pi^2}{6}$ and $\zeta(4)=\frac{\pi^4}{90}$.

Moreover, through $(5)$ it is straightforward to compute the magnitude of the coefficients by simply approximating $\zeta(2m+2)$ with $1$. On the other hand, that is trivial also by considering that the radius of convergence of the Taylor series of $f(z)=\frac{1}{1+e^z}$ at $z=0$ is $\pi$, since $f(z)$ is a meromorphic function with two simple poles at $z=\pm \pi i$.

Improve this answer
$\endgroup$
  • $\begingroup$ I haven't delved that deep into mathematics yet, but i'd certainly want to see the proof. $\endgroup$ – Nikos Aug 21 '15 at 13:39
  • $\begingroup$ @RestlessC0bra: please just wait a little, I'll type it later. $\endgroup$ – Jack D'Aurizio Aug 21 '15 at 13:40
  • $\begingroup$ @RestlessC0bra: done ;) $\endgroup$ – Jack D'Aurizio Aug 21 '15 at 15:46
2
$\begingroup$

Here's another way to derive it term by term: $$y^{-1}=1+e^x\Rightarrow -y^{-2}y'=e^x=y^{-1}-1\Rightarrow y'=y^2-y$$

Now you can continue to differentiate with the minimum of fuss:

$$y''=2yy'-y'$$ $$y'''=2yy''+2y'y'-y''$$ $$y^{(4)}=2yy'''+6y''y'-y'''$$

And so on, evaluating at each step.

Improve this answer
$\endgroup$
2
$\begingroup$

$\begin{array}\\ f(x) &=\frac{1}{1+e^x}\\ &=\frac{1}{2+(e^x-1)}\\ &=\frac12\frac{1}{1+(e^x-1)/2}\\ &=\frac12\sum_{n=0}^{\infty}(-1)^n\frac{(e^x-1)^n}{2^n}\\ &=\frac12\sum_{n=0}^{\infty}(-1)^n\frac{(x+x^2/2+x^3/6+...)^n}{2^n}\\ &=\frac12\sum_{n=0}^{\infty}(-1)^n(x/2)^n(1+x/2+x^2/6+...)^n\\ &=\frac12\sum_{n=0}^{\infty}(-1)^n(x/2)^n(1+nx/2+...)\\ &=\frac12\left(1-\frac{x}{2}(1+\frac{x}{2}+...)+\frac{x^2}{4}(1+...)+...\right)\\ &=\frac12\left(1-\frac{x}{2}-\frac{x^2}{4} +\frac{x^2}{4}(1+...)+...\right)\\ &=\frac12\left(1-\frac{x}{2}+O(x^3)\right)\\ &=\frac12-\frac{x}{4}+O(x^3)\\ \end{array} $

Improve this answer
$\endgroup$

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.