5

What is the Taylor series expansion of the lower incomplete gamma function $\gamma(s,-x)$?

The standard answer from wikipedia is interms of $x^k$: $$\gamma(s,-x)=\sum_{k=0}^{\infty}\frac{(-x)^se^{x}(-x)^k}{s(s+1)\cdots(s+k)}=(-x)^se^{x}\Gamma(s)\sum_{k=0}^{\infty}\frac{(-1)^k x^k}{\Gamma(s+k+1)}\tag{1}$$

Mathematica 10.0 gives a series in terms of $x^{-k}$:

$$\gamma(s,-x)=\Gamma(s)+e^x x^s (-1)^s\left(\frac{1}{x}+\frac{s-1}{x^2}+\frac{(s-1)(s-2)}{x^3}+\cdots\right)\tag{2}$$

The series in $(2)$ suits our needs because it is in terms of $1/x$ so we can truncate it at $1/x^3$ for example.

The series in $(1)$ does not suits our need because we have to keep large enough terms, say $K$ terms, to make $\frac{x^K}{|\Gamma(s+K+1)|}<1$ as $x\to+\infty$.

The expression $(2)$ is actually from the asymptotic series for $\Gamma(s,-x)$ because $$\gamma(s,-x)=\Gamma(s)-\Gamma(s,-x)\tag{3}$$

It is not obvious to me how one can derive from $(1)$ to $(2)$. That is why I ask this question.

Gary
  • 31,845
mike
  • 5,604
  • 1
    https://en.wikipedia.org/wiki/Incomplete_gamma_function#Lower_incomplete_Gamma_function – marty cohen Jun 07 '16 at 16:53
  • 2
    Is the Question intended to explore a notion of power series expansion "at infinity"? Generally a terse problem statement of this kind is at risk of being placed on hold for lack of context. There are several ways to add context. You can explain what about the problem interests you, for example, or what approaches you tried or at least considered and where you ran into difficulty. – hardmath Jun 07 '16 at 17:00
  • @hardmath. Thanks for the advice. – mike Jun 07 '16 at 20:19
  • 1
    Thanks, that improves the Question! – hardmath Jun 07 '16 at 21:26
  • 1
    Just for the record, the expression said to come from Wikipedia also appears in Olver et al. (2010) as equation 8.11.4. See an online version here: http://dlmf.nist.gov/8.11#E4

    [ Olver, F. W. J., Lozier, D. W., Boisvert, R. F., and Clark, C. W., eds. (2010). NIST Handbook of Mathematical Functions. Cambridge University Press. ]

    – Vicent Oct 27 '17 at 09:20

1 Answers1

4

Both (1) and (2) can be derived from the integral definition of the lower gamma function. By definition: $$ \gamma(s,-x) = \int_0^{-x} t^{s-1} e^{-t} dt $$ To see either, it suffices to perform repeated integration by parts on this definition. Letting: $$ u = e^{-t} \qquad dv=t^{s-1}dt \\ du = -e^{-t}dt \qquad v=\frac{t^s}{s} $$ One obtains: $$ \gamma(s,-x) = \left [\frac{e^{-t}t^s}{s} \right|_0^{-x}+\frac{1}{s} \int_0^{-x} t^s e^{-t}dt \\= \frac{e^{x}(-x)^s}{s}+\frac{1}{s} \gamma(s+1,-x) \\= e^{x}(-x)^s \left ( \frac{1}{s} + \frac{x}{s(s+1)} + \frac{x^2}{s(s+1)(s+2)} + \cdots \right) $$ This relation may be repeated to obtain (1). To obtain a similar version of (2). Simply invert the IBP and let: $$ u = t^{s-1} \qquad dv=e^{-t}dt \\ du = (s-1) t^{s-2}dt \qquad v=-e^{-t} $$ Then: $$ \gamma(s,-x) = \left [-e^{-t}t^{s-1} \right|_0^{-x}+(s-1) \int_0^{-x} t^{s-2} e^{-t}dt \\= -e^{x}(-x)^{s-1}+(s-1) \gamma(s-1,-x) \\= -e^{x}(-x)^s \left(\frac{1}{x} + \frac{(s-1)}{x^2} + \frac{(s-1)(s-2)}{x^3} + \cdots \right) $$ Your form (2) is simply then from applying a similar analysis on the integral for $\Gamma(s,-x)$ through (3).

Alternatively, one can avoid all of the integral definitions by simply using the recurrences thus derived, as in here. To truly go from (1) to (2) though, one would have to effectively work back to a more general definition, before progressing.

Paul LeVan
  • 1,046
  • Thanks again. By the way, what dose IBP stand for? – mike Jun 11 '16 at 15:28
  • $u = e^{-t},dv=t^{s-1},du = -e^{-t}, v=\frac{t^s}{s}$ should be $u = e^{-t},dv=t^{s-1}\color{red}{dt},du = -e^{-t}\color{red}{dt},v=\frac{t^s}{s}$, right? – mike Jun 11 '16 at 15:57
  • 1
    @mike Integration by Parts, and yes, those all can be placed on there. – Paul LeVan Jun 11 '16 at 17:12