8

On another site, someone asked about proving that

$$ \int_{0}^{1} \frac{\ln^{n}x}{(1-x)^{m}} \, dx = (-1)^{n+m-1} \frac{n!}{(m-1)!} \sum_{j=1}^{m-1} (-1)^{j} s (m-1,j) \zeta(n+1-j), \tag{1} $$

where $n, m \in \mathbb{N}$, $n \ge m$, $m \ge 2$, and $s(m-1,j)$ are the Stirling numbers of the first kind.

My attempt at proving $(1)$:

$$ \begin{align}\int_{0}^{1} \frac{\ln^{n}x}{(1-x)^{m}} \, dx &= \frac{1}{(m-1)!} \int_{0}^{1} \ln^{n} x \sum_{k=m-1}^{\infty} k(k-1) \cdots (k-m+2) \ x^{k-m+1} \ dx \\ &= \frac{1}{(m-1)!} \sum_{k=m-1}^{\infty} k(k-1) \cdots (k-m+2) \int_{0}^{1} x^{k-m+1} \ln^{n} x \, dx \\ &= \frac{1}{(m-1)!} \sum_{k=m-1}^{\infty} k(k-1) \cdots (k-m+2) \frac{(-1)^{n} n!}{(k-m+2)^{n+1}}\\ &= \frac{1}{(m-1)!} \sum_{k=m-1}^{\infty} \sum_{j=0}^{m-1} s(m-1,j) \ k^{j} \ \frac{(-1)^{n} n!}{(k-m+2)^{n+1}} \\ &= (-1)^{n} \frac{n!}{(m-1)!} \sum_{j=0}^{m-1} s(m-1,j) \sum_{k=m-1}^{\infty} \frac{k^{j}}{(k-m+2)^{n+1}} \end{align} $$

But I don't quite see how the last line is equivalent to the right side of $(1)$. (Wolfram Alpha does say they are equivalent for at least a few particular values of $m$ and $n$.)

  • 1
    Why do you repeatedly used \displaystyle instead of using an actual display, when your math notation is not in an inline setting? For example, just below the line where you say "On another forum", you had your integral equal to a certain sum, and that was all on a line by itself, but instead of setting it as a display, you used \displaystyle. (I changed it, as you can see.) – Michael Hardy May 27 '13 at 15:08
  • 2
    You can take help from robjhon's answer in this question: http://math.stackexchange.com/questions/75934/a-proof-of-int-01-left-frac-ln-t1-t-right2-mathrmdt-frac-pi/76032#76032 – Shobhit Bhatnagar May 27 '13 at 15:35

2 Answers2

4

Substitute $x=e^{-t}$ and get that the integral is equal to

$$(-1)^n \int_0^{\infty} dt \, e^{-t} \frac{t^n}{(1-e^{-t})^m} $$

Now use the expansion

$$(1-y)^{-m} = \sum_{k=0}^{\infty} \binom{m+k-1}{k} y^k$$

and reverse the order of summation and integration to get

$$\sum_{k=0}^{\infty} \binom{m+k-1}{k} \int_0^{\infty} dt \, t^n \, e^{-(k+1) t}$$

I then get as the value of the integral:

$$\int_0^1 dx \, \frac{\ln^n{x}}{(1-x)^m} = (-1)^n\, n!\, \sum_{k=0}^{\infty} \binom{m+k-1}{k} \frac{1}{(k+1)^{n+1}}$$

Note that when $m=0$, the sum reduces to $1$; every term in the sum save that at $k=0$ is zero.

Note also that this sum gives you the ability to express the integral in terms of a Riemann zeta function for various values of $m$, which will provide the Stirling coefficients.

Ron Gordon
  • 138,521
  • That was nice! (+1) One question: writing first $dt$ in the integral, is this common for engineers, or physics, or... I first time see that. – Cortizol May 27 '13 at 17:57
  • @Cortizol: Thanks. It's common in physics, especially in optics which is my background. – Ron Gordon May 27 '13 at 18:53
2

Here is another approach based on the beta function

$$ F = \int_{0}^{1}x^{\alpha}(1-x)^{-m} dx = \beta( \alpha+1,1-m ),\quad \alpha+1>0,\, 1-m>0. $$

Now, our integral $I$ can be evaluated as

$$ I = \lim_{\alpha \to 0} \frac{d^n}{d \alpha^n} \beta( \alpha+1,1-m ),\quad n>m-1,\,m\geq 2. $$

Here is a useful identity

$$ {\partial \over \partial x} \mathrm{B}(x, y) = \mathrm{B}(x, y) \left( {\Gamma'(x) \over \Gamma(x)} - {\Gamma'(x + y) \over \Gamma(x + y)} \right) = \mathrm{B}(x, y) (\psi(x) - \psi(x + y)), $$

where $\psi(x)$ is the digamma function.