Related information Integral of $\int^{\infty}_0 \frac{x^n}{x^s+1}dx$
This is an integral very similar to the gamma function integral:
$$R(s)=\int^{\infty}_0 (1+x^s)^{-1} e^{-x}\,dx$$
i want to find the function $R$.
I do know some values of $R$:
$$R(0)=1$$
$$R\left(\frac{1}{2}\right)=\frac{-\pi \text{erfi}(1)+\text{Ei}(1)+e \sqrt \pi}{e}$$
$$R(1)= -e\text{Ei}(-1)$$
$$R(2) = \text{Ci}(1)\sin(1)-\text{Si}(1)\cos(1)+\frac{1}{2}\pi\cos(1)$$
Can any of you provide hints or solutions?
Also, thanks to an answer by Sewer we know that:
$$\lim_{s \to \infty}R(s)=1$$
- 2,285
-
No, it doesn't reduce to "a gamma integral". If $s$ is a nonzero integer, the given integral expresses in terms of the exponential integral function (just do partial fractions for $(1+x^s)^{-1}$). – metamorphy Jun 07 '20 at 10:01
-
It does seem so, I did some calculations on paper for some values of s. – razivo Jun 07 '20 at 10:07
-
@metamorphy, s is a parameter, so I can’t do partial fractions. – razivo Jun 07 '20 at 10:08
-
Why? For an integer $s>0$ we have $(1+x^s)^{-1}=(1/s)\sum_{j=1}^{s}(1-x/x_j)^{-1}$, where $x_j=\exp\big((2j-1)\pi\mathrm{i}/s\big)$. Similarly for $s<0$. (And no, this won't work for arbitrary $s$, don't even try.) – metamorphy Jun 07 '20 at 10:30
-
@metamorphy, s is not an integer in my case. – razivo Jun 07 '20 at 10:40
-
Then forget the above. (It was only to show that it doesn't reduce to gammas.) And now the question is - what do you need exactly?.. (Please respond by editing the question itself, as it is unclear - not in comments) – metamorphy Jun 07 '20 at 10:48
-
My computer program does not simplify $R(3)$ further than a Meier-G function: $$R(3) = \displaystyle \frac{\sqrt{3} {G_{1, 4}^{4, 1}\left(\begin{matrix} \frac{2}{3} & \\frac{2}{3}, 0, \frac{1}{3}, \frac{2}{3} & \end{matrix} \middle| {\frac{1}{27}} \right)}}{6 \pi} $$ – Calvin Khor Jun 07 '20 at 12:52
-
does it get all the other 4,5,6,7.. to the Meier-G function? – razivo Jun 07 '20 at 12:54
-
Yes, for instance $R(10) = \displaystyle \frac{\sqrt{5} {G_{1, 11}^{11, 1}\left(\begin{matrix} \frac{9}{10} & \\frac{9}{10}, 0, \frac{1}{10}, \frac{1}{5}, \frac{3}{10}, \frac{2}{5}, \frac{1}{2}, \frac{3}{5}, \frac{7}{10}, \frac{4}{5}, \frac{9}{10} & \end{matrix} \middle| {\frac{1}{10000000000}} \right)}}{160 \pi^{\frac{9}{2}}}$ and note that $11=10+1$, $9=10-1$, and $10000000000=10^{10}$ – Calvin Khor Jun 07 '20 at 12:55
-
@CalvinKhor have a guass for the formula? – razivo Jun 07 '20 at 13:02
-
@razivo almost, all the terms are "obvious" except for the 160 at the bottom. There is a division by $\sqrt 2$ precisely when there isn't an integer power of $\pi$, so maybe it should be $(2\pi)^{9/2} $ instead – Calvin Khor Jun 07 '20 at 13:03
-
@razivo the above guess is correct, have just collected this data into an "answer" – Calvin Khor Jun 07 '20 at 13:25
-
3For $s \in \mathbb R^+$, the integral gives a Fox H-function: $$R(s) = H_{2, 1}^{1, 2} {\left( 1 \middle| {(0, 1), (0, s) \atop (0, 1)} \right)}.$$ For $s =l/m$, $R(s)$ can be reduced to $G_{m, l + m}^{l + m, m}$. – Maxim Jun 07 '20 at 18:31
3 Answers
I'll just keep here some values of $R$ that my computer generated. If there's a special value you want, tell me and I can try it when I'm free. I'm just using Sympy, the code is simplify(integrate(exp(-t)/t**s+1),(t,0,oo))) (simplify does nothing in everything I tried, but one can hope lol) I know almost nothing about $G$ functions, so these could be saying almost nothing. Regardless:
\begin{align} R(3) &= \displaystyle \frac{\sqrt{3} {G_{1, 4}^{4, 1}\left(\begin{matrix} \frac{2}{3} & \\\frac{2}{3}, 0, \frac{1}{3}, \frac{2}{3} & \end{matrix} \middle| {\frac{1}{27}} \right)}}{6 \pi}, \\ R(4) &= \displaystyle \frac{\sqrt{2} {G_{1, 5}^{5, 1}\left(\begin{matrix} \frac{3}{4} & \\\frac{3}{4}, 0, \frac{1}{4}, \frac{1}{2}, \frac{3}{4} & \end{matrix} \middle| {\frac{1}{256}} \right)}}{8 \pi^{\frac{3}{2}}}, \\ R(5) &= \displaystyle \frac{\sqrt{5} {G_{1, 6}^{6, 1}\left(\begin{matrix} \frac{4}{5} & \\\frac{4}{5}, 0, \frac{1}{5}, \frac{2}{5}, \frac{3}{5}, \frac{4}{5} & \end{matrix} \middle| {\frac{1}{3125}} \right)}}{20 \pi^{2}},\\ R(6) &= \displaystyle \frac{\sqrt{3} {G_{1, 7}^{7, 1}\left(\begin{matrix} \frac{5}{6} & \\\frac{5}{6}, 0, \frac{1}{6}, \frac{1}{3}, \frac{1}{2}, \frac{2}{3}, \frac{5}{6} & \end{matrix} \middle| {\frac{1}{46656}} \right)}}{24 \pi^{\frac{5}{2}}}, \\ R(7) &= \displaystyle \frac{\sqrt{7} {G_{1, 8}^{8, 1}\left(\begin{matrix} \frac{6}{7} & \\\frac{6}{7}, 0, \frac{1}{7}, \frac{2}{7}, \frac{3}{7}, \frac{4}{7}, \frac{5}{7}, \frac{6}{7} & \end{matrix} \middle| {\frac{1}{823543}} \right)}}{56 \pi^{3}}, \\ R(8) &= \displaystyle \frac{{G_{1, 9}^{9, 1}\left(\begin{matrix} \frac{7}{8} & \\\frac{7}{8}, 0, \frac{1}{8}, \frac{1}{4}, \frac{3}{8}, \frac{1}{2}, \frac{5}{8}, \frac{3}{4}, \frac{7}{8} & \end{matrix} \middle| {\frac{1}{16777216}} \right)}}{32 \pi^{\frac{7}{2}}} \end{align} This seems to follow the pattern $$ n\in\mathbb Z_{\ge 3} \implies R(n) = \displaystyle \frac{ {G_{1, n+1}^{n+1, 1}\left(\begin{matrix} \frac{n-1}{n} & \\\frac{n-1}{n}, 0, \frac{1}{n}, \frac{2}{n} , \dots , \frac{n-1}n \end{matrix} \middle| {n^{-n}} \right)}}{\sqrt n (2\pi)^{(n-1)/2}}$$ From the comment of metamurphy, $R(s) + R(-s) = 1$ so negative values don't need to be tried, but I'll record here anyway what the computer gives me for a few values:
\begin{align} R(-1) &= \displaystyle e \operatorname{E}_{2}\left(1\right), \\ R(-2) &= \displaystyle \left(- \frac{\pi}{2} + \operatorname{Si}{\left(1 \right)}\right) \cos{\left(1 \right)} - \sin{\left(1 \right)} \operatorname{Ci}{\left(1 \right)} + 1, \\ R(-3) &= \displaystyle \frac{\sqrt{3} {G_{1, 4}^{4, 1}\left(\begin{matrix} - \frac{1}{3} & \\- \frac{1}{3}, 0, \frac{1}{3}, \frac{2}{3} & \end{matrix} \middle| {\frac{1}{27}} \right)}}{6 \pi} \end{align}
Non-integer values: \begin{align}R(1/2) &= \displaystyle \frac{{G_{2, 3}^{3, 2}\left(\begin{matrix} 0, - \frac{1}{2} & \\0, - \frac{1}{2}, 0 & \end{matrix} \middle| {1} \right)}}{\pi}, \\ R(-1/2) &= \displaystyle \frac{{G_{2, 3}^{3, 2}\left(\begin{matrix} - \frac{1}{2}, -1 & \\- \frac{1}{2}, -1, 0 & \end{matrix} \middle| {1} \right)}}{\pi}, \\ R(3/2) &= \displaystyle \frac{\sqrt{3} {G_{2, 5}^{5, 2}\left(\begin{matrix} \frac{2}{3}, \frac{1}{6} & \\\frac{2}{3}, \frac{1}{6}, 0, \frac{1}{3}, \frac{2}{3} & \end{matrix} \middle| {\frac{1}{27}} \right)}}{6 \pi^{2}}\end{align} In particular, it cannot verify your solution for $s=1/2$, but it tells me that your solution is accurate to 124 decimals. Computer doesn't want to give me an answer for $s=1/3,2/3,4/3,e,\pi$.
PS I have sympy installed but you can use it online here https://live.sympy.org
- 36,192
- 6
- 47
- 102
-
1Thanks, from the Wikipedia article I get that using the melin transform might be useful in our case. – razivo Jun 07 '20 at 13:28
-
Did you mean the Laplace transform? Actually we are looking for $\mathcal{L} \left{ \frac{1}{x^s+1} \right}(1) $ – Sewer Keeper Jun 07 '20 at 13:38
-
@SewerKeeper I think he was starting from the G function, which is an inverse Mellin transform of products of Gammas and 1/Gammas evaluated at various points https://en.wikipedia.org/wiki/Meijer_G-function – Calvin Khor Jun 07 '20 at 13:39
-
Also, in your general case, the first element of the lower part is completely separate from the rest by the function’s definition. – razivo Jun 07 '20 at 14:00
-
@CalvinKhor , can you try $R(-1)$? In general, can you try negatives in general? – razivo Jun 07 '20 at 14:06
-
1
-
@razivo metamorphy is correct, but I have edited my post with what sympy gives me anyway – Calvin Khor Jun 08 '20 at 06:52
-
@CalvinKhor as you have a general solution in terms of the Fox-H function, what are the steps to evaluating the integral? – Тyma Gaidash Sep 06 '21 at 13:22
-
@Tyma what does ‘evaluate’ mean? Also the solution here is a G function, which is simpler – Calvin Khor Sep 06 '21 at 15:20
-
Well in principle the answer would be to look at the source code of sympy...but I do not have a pen and paper method @Tyma – Calvin Khor Sep 06 '21 at 15:39
From your related question we get
$$ \begin{split} R(s) &= \int^{\infty}_0 \frac{\operatorname e^{-x}}{x^s+1}\,\operatorname dx \\ &= \int^{\infty}_0 \sum_{n=0}^{+\infty}(-1)^n\frac{x^n}{n!}\frac{1}{x^s+1}\,\operatorname dx \\ &= \sum_{n=0}^{+\infty} \frac{(-1)^n}{n!}\int^{\infty}_0\frac{x^n}{x^s+1} \operatorname d x \\ &= \sum_{n=0}^{+\infty} \frac{(-1)^n}{n!}\frac1{n+1}R\left(\frac{s}{n+1};0\right) \\ &= \sum_{n=0}^{+\infty} \frac{(-1)^n}{n!}\frac1{n+1}\frac{n+1}{s}\varGamma\left(\frac{n+1}{s}\right) \varGamma\left(1-\frac{n+1}{s}\right) \\ &=\frac{1}{s} \sum_{n=0}^{+\infty} \frac{(-1)^n}{n!} \varGamma\left(\frac{n+1}{s}\right) \varGamma\left(1-\frac{n+1}{s}\right)\\ \end{split} $$
Moreover, if $\frac{n+1}{s} \not \in \mathbb Z$, we can rewrite $R$ as
$$R(s) =\frac{\pi}{s} \sum_{n=0}^{+\infty} \frac{(-1)^n}{n!} \frac{1}{\sin\left( \frac{\pi(n+1)}{s}\right)}$$
When we used the property:
$$ \varGamma(1-z) \varGamma(z) = \frac{\pi}{\sin(\pi z)} \qquad \forall \, z\not\in\mathbb Z $$
- 36,192
- 6
- 47
- 102
- 1,510
-
unsure of how we continue from here, do you know of any sums containing $\varGamma$? – razivo Jun 07 '20 at 11:42
-
-
1Given we can make a constant $S=\frac{\pi}{s}$ it is simplified to $$S \sum^{+\infty}_{n=0} \frac{(-1)^n}{n!} \csc(S(n+1))$$ – razivo Jun 07 '20 at 11:47
-
the integral from the other post fails to make sense for $n$ sufficiently large ($s-n\le 1$) – Calvin Khor Jun 07 '20 at 11:51
-
@CalvinKhor Ok, I didn't seet it, that's correct. So we cannot use Taylor expansion of $\operatorname e^{-x}$ I guess. Should I delete my answer? – Sewer Keeper Jun 07 '20 at 11:55
-
well, if you can find special values of $s$ where your answer for $R(s)$ is finite, it may end up being correct, or perhaps the only "interesting" answer that is not infinity anyway – Calvin Khor Jun 07 '20 at 11:57
-
so, this approach fails to yield anything except in the $s \to \infty$? – razivo Jun 07 '20 at 12:08
-
if we do go to infinity, the csc and s outside can be conjoined to make it $$ \pi \sum^{+\infty}{n=0} \frac{(-1)^n}{n!} \lim{s \to \infty} \csc \left( \frac{\pi (n+1)}{s}\right) \frac{1}{s} $$ – razivo Jun 07 '20 at 12:11
-
If you want to computer $\lim_{s \to +\infty} R(s)$, you can move the limit inside the integral (because of dominated convergent theorem) and, after splitting the integral $\int_0^1$ and $\int_1^{+\infty}$, you get $$\lim_{s \to +\infty} \int_0^{+\infty} \frac{\operatorname e^{-x}}{x^s+1} , \operatorname d x =\int_0^{1} \lim_{s \to +\infty} \frac{\operatorname e^{-x}}{x^s+1} , \operatorname d x =\int_0^{1} \operatorname e^{-x}, \operatorname d x= 1$$ – Sewer Keeper Jun 07 '20 at 12:19
-
when doing the limit we reach $$\pi \sum^{+\infty}_{n=0} \frac{(-1)^n}{n!} \frac{1}{\pi (n+1)}$$ – razivo Jun 07 '20 at 12:19
-
more simply, $$-\sum^{+\infty}_{n=0} \frac{(-1)^{n+1}}{(n+1)!} = 1-\frac{1}{e}$$ – razivo Jun 07 '20 at 12:23
-
-
As Calvin Khor said, my formula is not correct, since $$\int_0^{+\infty} \frac{x^n}{x^{s}+1}, \operatorname d x $$ does not make sense if $n \ge s$. Moreover, since this integral is inside a series, there are infinite numbers such that $n \ge s$. – Sewer Keeper Jun 07 '20 at 12:28
-
1
Since the Fox-H function may be hard to understand, here is a solution using the Geometric Series. For simplicity, lets do the general integral. Notice the Incomplete Gamma function definition:
$$\int \frac{e^{-x}}{x^s+1}dx= \int \frac{e^{-x}}{1-\left(-{x^s}\right)}dx =\int e^{-x} \sum_{n=0}^\infty \left(-x^s\right)^n dx=\sum_{n=0}^\infty (-1)^n \int x^{ns}e^{-x} dx=C-\sum_{n=0}^\infty (-1)^nΓ(ns+1,x) ,|x^s|<1$$
Even though this may not converge for$ |x^s|\not<1$, this still demonstrates a generalized version of the integral. I thought it might be nice to notice being able to use the linked geometric series. Maybe I will add more possible expansions. Please correct me and give me feedback!
It also may be possible to use the Abel-Plana formula with a complicated real and imaginary part. Note that the sum should have a tricky closed form.
$$\int_0^\infty \frac{e^{-x}}{x^s+1}dx=\sum_{x=0}^\infty \frac{e^{-x}}{x^s+1} -\frac12 \frac{e^{-0}}{0^s+1}+i\int_0^\infty \frac{1}{e^{2\pi x}-1} \frac{e^{-ix}}{(ix)^s+1}-\frac{1}{e^{\pi x}-1}\frac{e^{ix}}{(-ix)^s+1} dx= \sum_{x=0}^\infty \frac{e^{-x}}{x^s+1} -\frac12 +\int_0^\infty \frac{1}{e^{\pi x}-1} \frac{i\cos(x)-\sin(x)}{i^sx^s+1}-\frac{1}{e^{\pi x}-1} \frac{i\cos(x)+\sin(x)}{(-1)^s i^s x^s+1}dx$$
- 13,576