Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Is there any simple way of computing the following sum?

$$\sum_{k=1}^\infty \frac1{k\space k!}$$

share|improve this question
5  
It is equal to $\displaystyle{\int_0^1\dfrac{e^t-1}{t}\,dt}$. –  Jonas Meyer Jun 1 '12 at 20:51
1  
The value is Ei$(1) -\gamma = \int_0^1 \frac{e^t-1}{t} dt$ if that's any help. –  Logan Maingi Jun 1 '12 at 20:51
    
@Chris: I just used Steven Stadnicki's method to get that, but I didn't know whether it answered the question. –  Jonas Meyer Jun 1 '12 at 20:59
    
@JonasMeyer Agreed - I'm not really sure that it does, since it only boils it down to the integral and then some magic happens. I'd add a proof of the evaluation to my answer, but a bit of digging about the web isn't actually turning up any details on it, and none of the usual spots seem to say anything about it. –  Steven Stadnicki Jun 1 '12 at 21:12
add comment

3 Answers

up vote 12 down vote accepted

First of all, consider the power series for $e^x$, $\displaystyle\sum_{k=0}^{\infty}\frac{x^k}{k!}$. Now subtract off the constant term and divide by $x$: $\displaystyle{\frac{e^x-1}{x} = \sum_{k=1}^{\infty}\frac{x^{k-1}}{k!}}$. Now integrate: $\displaystyle{\int_0^x \frac{e^t-1}{t} dt = \sum_{k=1}^{\infty}\frac{x^k}{k\cdot k!}}$ (note that the lower limit is dictated by the constant term). Finally, evaluate at $x=1$; the value of your sum is the value of the definite integral $\displaystyle{\int_0^1 \frac{e^t-1}{t} dt }$. Wolfram Alpha evaluates this to $\mathrm{Ei}(1)-\gamma$, so there's probably no better closed form than that.

share|improve this answer
add comment

The exponential integral function can be written as:

$$ \mathrm{Ei}(x) = \gamma + \log|x| + \sum_{k=1}^{\infty} \frac{x^k}{k\; k!} $$

Plug $x = 1$ to get:

$$ \sum_{k=1}^{\infty} \frac{1}{k\; k!} = \mathrm{Ei}(1) - \gamma $$

Where $\gamma$ is Euler–Mascheroni constant and $\mathrm{Ei}(1)$ is given by A091725.

share|improve this answer
add comment

$\def\d{\delta} \def\e{\epsilon} \def\g{\gamma} \def\pv{\mathrm{PV}} \def\pv{\mathcal{P}} \def\pv{\mathrm{P}}$We show another way to get the integral representation of the sum and explain its relation to the exponential integral.

Let $$S(x) = \sum_{k=1}^\infty \frac{x^k}{k k!}.$$ The sum we are interested in is $S(1)$, but, as is often the case, it is easier to get the sum for any $x>0$. (There is a straightforward extension to $x<0$.) Notice that $$\begin{eqnarray*} S'(x) &=& \sum_{k=1}^\infty \frac{x^{k-1}}{k!} \\ &=& \frac{1}{x}\left( \sum_{k=0}^\infty \frac{x^{k}}{k!} - 1\right) \\ &=& \frac{e^x-1}{x}. \end{eqnarray*}$$ Therefore, $\displaystyle S(x) = \int_a^x dt\, \frac{e^t-1}{t}.$ To find $a$ just notice that $S(0) = 0$, so $a=0$, $$S(x) = \int_0^x dt\, \frac{e^t-1}{t}.$$

The argument of the integral is perfectly well-behaved at $t=0$, so $$\begin{eqnarray*} S(x) &=& \lim_{\e\to 0} \int_\e^x dt\, \frac{e^t-1}{t} \\ &=& \lim_{\e\to 0} \left( \int_\e^x dt\, \frac{e^t}{t} - \int_\e^x dt\,\frac{1}{t} \right) \\ &=& \lim_{\e\to 0} \left( \pv \int_{-\infty}^x dt\,\frac{e^t}{t} - \pv \int_{-\infty}^\e dt\,\frac{e^t}{t} -\log x + \log \e \right) \\ &=& \lim_{\e\to 0} \left( \mathrm{Ei}(x) - \mathrm{Ei}(\e) - \log x + \log \e \right) \\ &=& \lim_{\e\to 0} \left( \mathrm{Ei}(x) - (\g + \log \e) - \log x + \log \e \right) \\ &=& \mathrm{Ei}(x) - \g - \log x. \end{eqnarray*}$$ (See below for a derivation of $\mathrm{Ei}(\e) = \g + \log \e + O(\e)$.) Therefore, $$\sum_{k=1}^\infty \frac{x^k}{k k!} = \mathrm{Ei}(x) - \g - \log x$$ and so $$\sum_{k=1}^\infty \frac{1}{k k!} = \mathrm{Ei}(1) - \g.$$

Some details

Above we use the definition of the exponential integral $$\mathrm{Ei}(x) = \pv \int_{-\infty}^x dt\,\frac{e^t}{t},$$ where $\pv\int$ stands for the Cauchy principal value, and the series expansion for $\mathrm{Ei}(x)$ for small $x$, which we derive now. Split the integral, $$\begin{eqnarray*} \mathrm{Ei}(x) &=& \lim_{\d\to0}\left[ \underbrace{\int_{-\infty}^{-\d} dt\,\frac{e^t}{t}}_{I_1} + \underbrace{\int_{\d}^{x} dt\,\frac{e^t}{t}}_{I_2} \right]. \end{eqnarray*}$$ For $I_1$, let $t=-s$ and integrate by parts,
$$I_1 = \log\d - \int_\d^\infty ds\, e^{-s}\log s.$$ For $I_2$, Taylor expand $e^t$ and integrate, $$I_2 = \log x - \log\d + O(x).$$ Thus, $$\begin{eqnarray*} \mathrm{Ei}(x) &=& \lim_{\d\to0}\left[ \left(\log\d - \int_\d^\infty ds\, e^{-s}\log s\right) +\left(\log x - \log\d + O(x)\right) \right] \\ &=& \g + \log x + O(x), \end{eqnarray*}$$ where we recognize the integral representation of the Euler-Mascheroni constant, $\g = -\int_0^\infty ds\,e^{-s}\log s$.

Notice that if we kept the higher order terms in the expansion for $I_2$ we would find $$\mathrm{Ei}(x) = \g + \log x + \sum_{k=1}^\infty \frac{x^k}{k k!},$$ the correct expansion for the exponential integral for $x>0$. In fact, this immediately gives our sum, $$\sum_{k=1}^\infty \frac{1}{k k!} = \mathrm{Ei}(1) - \g.$$ This is the approach of @Ayman Hourieh.

share|improve this answer
add comment

Your Answer

 
discard

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.