Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Find $f(x)$, the unknown function satisfying $$f(x) = \sum\limits_{n=0}^{+\infty}\frac{x^{4n}}{(4n)!}$$

I'm looking for a direct solution which is different from mine, if possible.

share|cite|improve this question
@AndréNicolas Would you please explain further? I don't understand. Do you mean to plug $x=1,-1,i,i$ into the series? Or to plug in $x,-x,ix,-ix$? – Frenzy Li Oct 26 '12 at 14:30
@AndréNicolas Plugging in $x$, $-x$, $ix$, $-ix$ into $f$ yields $f(x)$, $f(x)$, $f(x)$, $f(x)$. – Did Oct 26 '12 at 14:40
Wrong $f$! Let $g(x)=e^x$. Write down the series for $g(x)$. Evaluate at $x$, $-x$, $ix$, $-ix$. Add up. The coefficients of $x^{4k+i}$ are $0$ for $i=1,2,3$ and $\frac{4}{(4k)!}$ for $i=0$. – André Nicolas Oct 26 '12 at 14:46
up vote 3 down vote accepted

By discrete Fourier transform.

Let $e(x)=\mathrm e^x=\displaystyle\sum\limits_{n\geqslant0}\frac{x^n}{n!}$. Note that, for every integer $n$, $\displaystyle\sum\limits_\zeta\zeta^n$ is $4$ when $n=0\pmod{4}$ and is $0$ otherwise, where the sum is over the set of roots $\{\zeta\in\mathbb C\mid \zeta^4=1\}=\{1,-1,\mathrm i,-\mathrm i\}$. Hence, $$ \sum\limits_\zeta e(\zeta x)=\sum\limits_{n\geqslant0}\frac{x^n}{n!}\cdot\sum\limits_\zeta\zeta^n=4\cdot\sum\limits_{n\geqslant0}\frac{x^{4n}}{(4n)!}. $$ In other words, $$ \sum\limits_{n\geqslant0}\frac{x^{4n}}{(4n)!}=\frac{e(x)+e(-x)+e(\mathrm ix)+e(-\mathrm ix)}4=\frac{\cosh(x)+\cos(x)}2. $$

Likewise, for every $k$ in $\{1,2,3\}$, $\displaystyle\sum\limits_\zeta\zeta^{n-k}$ is $4$ when $n=k\pmod{4}$ and is $0$ otherwise, hence $$ \sum\limits_\zeta\zeta^{-k}\cdot e(\zeta x)=\sum\limits_{n\geqslant0}\frac{x^n}{n!}\cdot\sum\limits_\zeta\zeta^{n-k}=4\cdot\sum\limits_{n\geqslant0}\frac{x^{4n+k}}{(4n+k)!}. $$ In other words, $$ \sum\limits_{n\geqslant0}\frac{x^{4n+k}}{(4n+k)!}=\frac{e(x)+(-1)^ke(-x)+(-\mathrm i)^{k}e(\mathrm ix)+\mathrm i^{k}e(-\mathrm ix)}4. $$ For example, $$ \sum\limits_{n\geqslant0}\frac{x^{4n+3}}{(4n+3)!}=\frac{\sinh(x)-\sin(x)}2. $$

share|cite|improve this answer
I'm not sure what this means: $\sum\limits_zz^n$. Is it $1^n+i^n+(-1)^n+(-i)^n ?$ – Frenzy Li Oct 26 '12 at 13:59
It is. $ $ $ $ $ $ – Did Oct 26 '12 at 14:00
Original: Why $\sum_{n\ge 0}\frac{x^n}{n!}\cdot\sum_\zeta\zeta^n=4\cdot\sum\limits_{n\ge 0}\frac{x^{4n}}{(4n)!}$? Edit: Silly me! I forgot about the modulo argument! – Frenzy Li Oct 26 '12 at 14:14
Why is this answer related to DFT? – Frenzy Li Oct 26 '12 at 14:16
Because the discrete Fourier transform of $(x_n)_{0\leqslant n\leqslant3}$ is $(y_k)_{0\leqslant k\leqslant3}$ defined by $y_k=\sum\limits_{n=0}^3x_n\zeta^{-kn}$. The beginning of the post is based on the fact that $(4,0,0,0)$ is the Fourier transform of $(1,1,1,1)$. – Did Oct 26 '12 at 14:36

This series converges on $\mathbb R$ because $\displaystyle \lim_{n\to+\infty}\frac{1}{(4n)!}^{\frac{1}{4n}} = \lim_{n\to+\infty}\frac{1}{\sqrt[n]{n!}}=0 $. By observation:

The sum of $\mathrm e^x$ and $\mathrm e^{-x}$ has no odd-powered terms: $\displaystyle \mathrm e^{x}+\mathrm e^{-x} = 2\cosh x = 2\sum\limits_{k=0}^{+\infty}\frac{x^{2n}}{(2n)!}$.

Also that we can cancel terms with power $4n-2$ with the help of: $\displaystyle \cos x = \sum\limits_{k=0}^{+\infty}(-1)^{k}\frac{x^{2n}}{(2n)!}$.

Thus, $\displaystyle \frac{1}{4}(\mathrm e^x+\mathrm e^{-x}+2\cos{x}) = \sum\limits_{k=0}^{+\infty}\frac{x^{4n}}{(4n)!} = f(x)$.

share|cite|improve this answer
I tested the convergence in the wrong way. I think a root test should be used instead. – Frenzy Li Oct 26 '12 at 13:27
The ratio test should suffice. – Pedro Tamaroff Oct 26 '12 at 13:40
@PeterTamaroff Is it ok to ratio test even if there are missing terms? – Frenzy Li Oct 26 '12 at 13:41
sub-indices need to be "$n$" . – student forever Apr 18 at 11:32

The fourth derivative of the function is the function itself. Then apply the well-known theory for solution of a linear ODE with constant coefficients. Of course you get the same answer as the other methods. But this method (find a DE it satisfies) can be applied in many cases where you want to sum a power series.

share|cite|improve this answer

We know, $e^x= \sum\limits_{n=0}^{\infty}\frac{x^n}{n!}$

Let's consider, the equation, $y^r-1=0--->(1)$

So, the roots are $e^{\frac{2ki\pi}r}=y_k$(say) where $1\le k\le r$,

then $\sum y_k=0$

also, $\sum y_k^s=r$ if $r\mid s$, else $\sum y_k^s=0$ (find here)

As, $ e^{y_kx}= \sum\limits_{n=0}^{+\infty}\frac{(y_kx)^n}{n!}$

summing for $k=1$ to $r, \sum_{1\le k\le r} e^{y_kx}= \sum\limits_{n=0}^{+\infty}\frac{x^n}{n!}(\sum y_k^n)=n\sum\limits_{n=0}^{+\infty}\frac{x^{rn}}{(rn)!}$

$y_{r-k}=e^{\frac{2(r-k)\pi i}r}=e^{\frac{-2k\pi i}r}=y_{-k}$

If $r$ is even, $$\sum_{1\le k\le r} e^{y_kx} =e^x+e^{-x}+\sum_{1\le k\le \frac r2 -1}(e^{\frac{2kx\pi i}r}+e^{\frac{2(r-k)x\pi i}r}) =2\cosh x+\sum_{1\le k\le \frac r 2 - 1}(e^{\frac{2kx\pi i}r}+e^{\frac{-2kx\pi i}r}) =2\cosh x+2\sum_{1\le k\le \frac r 2 - 1}\cos \frac{2kx\pi }r$$

If $r$ is odd, $$\sum_{1\le k\le r} e^{y_kx} =e^x+\sum_{1\le k\le \frac {r-1} 2}(e^{\frac{2kx\pi i}r}+e^{\frac{2(r-k)x\pi i}r}) =e^x+\sum_{1\le k\le \frac {r-1} 2}(e^{\frac{2kx\pi i}r}+e^{\frac{-2kx\pi i}r}) =e^x+2\sum_{1\le k\le \frac {r-1} 2}\cos \frac{2kx\pi }r$$

$r=4\implies 4\sum\limits_{n=0}^{+\infty}\frac{x^{4n}}{(4n)!}$ $=2\cosh x+2\cos x$

share|cite|improve this answer

Related techniques: (I). Recalling the Laplace transform of a function $f$,

$$ F(s) = \int_{0}^{\infty} f(x) e^{-sx} dx \,, $$

and using the fact that the Laplace transform of $x^m$ is given by $\frac{\Gamma(m+1)}{x^{m+1}}$, we can compute the Laplace transform of $f(x)=\sum\limits_{n=0}^{+\infty}\frac{x^{4n}}{(4n)!}$ as

$$ F(s)=\sum_{k=0}^{\infty} \frac{1}{s^{4k+1}}=\frac{s^3}{s^4-1}$$

$$= \frac{1}{4(s-1)}+\frac{1}{4(s+1)}+\frac{1}{4(s+i)}+\frac{1}{4(s-i)}\rightarrow (1)\,. $$

Taking the inverse Laplace transform of $(1)$ yields,

$$ \frac{1}{2}\,\cos \left( x \right) +\frac{1}{2}\,\cosh \left( x \right) \,. $$

share|cite|improve this answer
I'm not familiar with Laplace transform. Would you please tell me how I could understand it intuitively? – Frenzy Li Oct 26 '12 at 14:50
@FrenzYDT: Laplace transform is a very useful technique and it has a lot of applications such as solving differential equations. You should practice it to learn it. – Mhenni Benghorbal Oct 26 '12 at 15:41
Mhenni: Seriously, do you consider that your comment answers @FrenzyDT's query? – Did Oct 26 '12 at 18:01

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.