I want to integrate $$ \int_0^{\infty}dx\,e^{-ax}\frac{1-(2x)^b}{1-2x} $$ where $a,b>0$. My The naive approach was to consider $b$ to be an integer, in which case you get a truncated geometric series above that gives you a sum of gamma functions. Nonetheless, by doing this you get a function that when tried for $b$ non integer gives you a number with an imaginary part, so the naive approach of solving for integer $b$ and hoping that the result can be generalized does not work. Any suggestions?

Any ideas about how to proceed?

  • 4
    $\begingroup$ Mathematica says $\frac{1}{2} e^{-\frac{a}{2}-i \pi b} \left(\Gamma (b+1) \Gamma \left(-b,-\frac{a}{2}\right)-e^{i \pi b} \Gamma \left(0,-\frac{a}{2}\right)\right)$, which is a hint. Note that Mathematica gets clumsy integrating around branch cuts, and sometimes ends up with a nonzero imaginary part erroneously. $\endgroup$ Feb 24, 2017 at 18:41
  • 1
    $\begingroup$ This integral diverges because of a pole at $x=1/2$. If you want a "principal value integral" you should say so. $\endgroup$ Feb 24, 2017 at 18:50
  • 2
    $\begingroup$ @GeraldEdgar no pole, it is an avoidable singularity $\endgroup$ Feb 24, 2017 at 18:54
  • 1
    $\begingroup$ You are right, no ingularity. I had $1-2x^b$ in error. $\endgroup$ Feb 24, 2017 at 22:55

2 Answers 2


The Mathematica 11 code

i = Integrate[Exp[-a*x]*(1 - (2*x)^b)/(1 - 2*x), {x, 0, Infinity}, 
  PrincipalValue -> True, Assumptions -> a > 0 && b > 0]

outputs $$\frac{1}{2} e^{-\frac{a}{2}} \left((-1)^b \Gamma (b+1) \Gamma \left(-b,-\frac{a}{2}\right)+\text{Ei}\left(\frac{a}{2}\right)+i \pi \right) $$ Let us check it:

N[i /. {a -> 2, b -> 1}]

$ 0.5\, +0. i $


N[i /. {a -> 2.1, b -> 3/2}]

$0.68219\, +7.770178924974121\,\,10^{-17} i $

  • $\begingroup$ Try it for b non integer and you will see an imaginary part $\endgroup$ Feb 24, 2017 at 19:27
  • $\begingroup$ You are not right. See the addition in my answer. The very small imaginary part appears because of errors when calculating. $\endgroup$
    – user64494
    Feb 24, 2017 at 19:39

Using a certain exponential integral from Maple, it comes out as $$ {\frac {-b!\,{2}^{b+1}+ \left( -{a}^{b+1}{\rm Ei}_1 \left(-a/2 \right) + \left( b+1 \right) !\,{2}^{b+1}{\rm Ei}_{b+2} \left(-a/2 \right) \right) {{\rm e}^{-a/2}}}{2{a}^{b+1}}} $$ It works for fractional $b$, interpreting $b! = \Gamma(b+1)$ as usual.


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.