1

I'm struggling in trying to solve integral. Could you help me? Please, don't solve the problem, give some hints of what I should do.
$$\int^1_0 \frac{\log x}{1-x}dx$$

amWhy
  • 210,739

3 Answers3

8

Since the range of the integral is from $0$ to $1$, you may like to use Geometric Series

$$\frac{1}{1-x} = \sum_{k = 0}^{+\infty} x^k$$

Hence

$$ \sum_{k = 0}^{+\infty} \int_0^1 x^k\ln(x)\ dx$$

The integral is pretty standard, you can do it by parts many times or in other ways. Eventually

$$\int x^k\ln(x)\ dx = \frac{x^{k+1} ((k+1) \log (x)-1)}{(k+1)^2}$$

hence

$$\int_0^1 x^k\ln(x) = -\frac{1}{(k+1)^2}$$

This holds for $\Re(k) > -1$ which is true.

Then all together

$$- \sum_{k = 0}^{+\infty} \frac{1}{(k+1)^2} = -\zeta(2) = -\frac{\pi^2}{6}$$

6

This is a non-elementary function integral. Specifically via a simple substitution of $u=1-x \to du = -dx$, we yield the known special function integral :

$$\int \frac{\ln x}{1-x}\mathrm{d}x=-\int\frac{\ln(1-u)}{u}\mathrm{d}u = \dots = \text{Li}_2(1-x) + C$$

thus we need to find another way to work around it.

For an analytical approach, continuing Jack's idea (credits) :

$$\int x^n\ln(x)\mathrm{d}x = \frac{x^{n+1}\ln(x)}{n+1}-\frac{1}{n+1}\int x^n\mathrm{d}x= \frac{x^{n+1}\big[(n+1)\ln(x)-1\big]}{(n+1)^2} + C$$

for which we used integration by parts as :

$$\int f\mathrm{d}g = fg - \int g\mathrm{d}f$$

$$\mathrm{d}f = \frac{1}{x}\mathrm{d}x$$ $$g=\frac{x^{n+1}}{n+1}$$

Now, exploiting the well known geometric series :

$$\frac{1}{1-x} = \sum_{k = 0}^{\infty} x^k$$

will lead to the expression for our calculation, of :

$$\sum_{n = 0}^{\infty} \int_0^1 x^n\ln(x)\ dx \to \int_0^1 x^n\ln(x)= -\frac{1}{(n+1)^2} \space \text{for} \space \Re\{n\} > -1$$

which finally leads to the calculation via the usage of the Zeta Function :

$$- \sum_{n = 0}^{\infty} \frac{1}{(n+1)^2} = -\zeta(2) = -\frac{\pi^2}{6}$$

Truly, our calculations are correct, as :

$$\int_0^1 \frac{\ln x}{1-x}\mathrm{d}x = \text{Li}_2(0) - \text{Li}_2(1) = - \frac{\pi^2}{6}$$

Alternatively, if you're not familiar with such a procedure, the best possible outcome will come via numerical analysis, using Simpson's rule for example :

$$\int_0^1 \frac{\ln(x)}{1-x}\mathrm{d}x \approx \frac{1-0}{6}\bigg(f(0) + + 4f\bigg(\frac{1}{2}\bigg) + f(1)\bigg), \space \text{where} \space f(x) = \frac{\ln(x)}{1-x}$$

Rebellos
  • 21,666
  • Everything is clear except for jumping to Zeta Function. Could You do it in a bit more detail – egorponomaryov Mar 11 '18 at 16:27
  • @ЕгорПономарёв Are you familiar with the Zeta Function in your courses or not ? – Rebellos Mar 11 '18 at 16:28
  • Not really. Could You explain what $Li_2$ is? – egorponomaryov Mar 11 '18 at 16:33
  • @ЕгорПономарёв Well then the Zeta Function approach is not for you (yet). The $\text{Li}_2$ function is called the dilogaritm - polylogarithm function and is a special function for improper expressions. What lesson and part of courses was this exercise for ? – Rebellos Mar 11 '18 at 16:35
  • Mathematical Analysis, 1-st course of University – egorponomaryov Mar 11 '18 at 16:38
  • 1
    @ЕгорПономарёв For a first course of mathematical analysis this seems a rather misplaced exercise, as you need to exploit a lot of different field (classical analysis, complex analysis, a bit of special function analysis). Maybe there is a typo ? What have you been taught this far ? Also, are you familiar with the numerical approximation I listed at the end ? – Rebellos Mar 11 '18 at 16:40
  • That was an extra problem, because my teacher think that I'm pretty clever, so that wasn't a typo. I don't know the Simpson's rule. – egorponomaryov Mar 11 '18 at 16:46
  • 4
    @ЕгорПономарёв Being clever doesn't mean you can solve everything (especially if it needs important tools from higher courses). I guess the best bet you can throw at this is a Taylor approximation ! – Rebellos Mar 11 '18 at 16:47
0

Things to try:

  • Make the change of variable $x \mapsto 1-x$. This simplifies the denominator, leaves the interval of integration unchanged (not important, but sometimes useful, especially if a linear combination of the original and new integrands simplifies), and makes the numerator something whose Taylor series you should know.
  • Attempt integration by parts to allow you to differentiate the logarithm, turning it into a rational function. (I'm pretty sure this doesn't help for this problem, but is in the constellation of things to do to integrals that look like this.)
  • Try solving the differential equation $f'(x) = \frac{\log x}{1-x}$, which can be turned into a relatively simple recurrence for the Taylor coefficients.
  • Look it up in tables. For instance, in Gradshteyn & Ryzhik, 7th ed., this is 4.231.2.
Eric Towers
  • 70,953