1

$$f(x)=\sum_{n=1}^{\infty}{\frac{x^{n-1}}{n}},$$

Prove that $f(x)+f(1-x)+\log(x)\log(1-x)=\frac{{\pi}^2}{6}$

In my mind though,I think that this is related to Basel problem$\left(\sum\limits_{n=1}^{\infty}{\frac{1}{n^2}}\right)$,but I don't know how to solve this.

Any help would be greatly appreciated :-)

Edit:

My attempt:

I cannot use latex expertly,so I post image.The circled partenter image description here

Li Taiji
  • 863
  • 1
    Hint: Write everything as series and put them together. –  Feb 15 '19 at 12:26
  • What have you tried? – mrtaurho Feb 15 '19 at 12:28
  • I have tried doing it.But the series don't include $\sum_{n=1}^{\infty}{\frac{1}{n^2}}$.Could you please give me some suggestions? – Li Taiji Feb 15 '19 at 12:29
  • I have tried doing it... Could you perhaps add in detail what you actually tried and especially where you struggled? And are you sure about your given function $f(x)$; that there is only $x^{n-1}/n$ and not $x^{n-1}/n^2$? – mrtaurho Feb 15 '19 at 12:31
  • @mrtaurho Indeed, it doesn't work without the square and like its stated by the OP, one would have to add assumptions on $x$. Thanks for pointing out! –  Feb 15 '19 at 12:37
  • I have edited my attempt. the $f(x)$ is $x^{n-1}/n$, not $x^{n-1}/n^2$. – Li Taiji Feb 15 '19 at 12:49
  • @梦里年华似烟花 Well, then the equation is false. For $x^{n-1}/n^2$ it would be close to the functional equation of the Dilogarithm function. However, as Ahmad Bazzi showed in his answer aswell as your own attempt pointed out the equation is simply wrong. – mrtaurho Feb 15 '19 at 12:52
  • @mrtaurho I'll go to ask my teacher if the question is correct.Anyway, thank you.And could you please tell me the condition that $f(x)$ is $x^{n-1}/n^2$?.emmm..Please don't mind my poor English. – Li Taiji Feb 15 '19 at 13:26
  • @梦里年华似烟花 I have added an answer by myself explainig why I think so. – mrtaurho Feb 15 '19 at 14:51

2 Answers2

1

This is not true. Counter-example $x = 0.5$. Also note that: $$f(x) = - \frac{\log(1-x)}{x}$$ $$f(1-x) = - \frac{\log(x)}{1-x}$$ $$f(x)+f(1-x)+\log(x)\log(1-x) =\frac{x(1-x)\log x \log(1-x) -(1-x)\log(1-x) - x \log x}{x(x-1)}$$ which looks like this enter image description here

Ahmad Bazzi
  • 12,076
  • 3
    I guess the OP is missing a square within the denominator of his series representation for $f(x)$. However, your answer illustrates the problem quite well (+1) – mrtaurho Feb 15 '19 at 12:34
  • I'll ask my teacher if the question is correct.Anyway, thank you. – Li Taiji Feb 15 '19 at 13:24
1

As you and Ahmad Bazzi already noticed the given equation is wrong. However, both of you already showed this issue I will not reproduce your results but going into more detail why I feel like there has to be a $x^{n-1}/n^2$ instead of the current $x^{n-1}/n~($to be precise I am also not sure about the $x^{n-1})$.

First of all let me introduce to you the Dilogarithm function $\operatorname{Li}_2(z)$ defined by a series very similiar to your one, and also to the one of the logarithm, namely

$$\operatorname{Li}_2(z)~=~\sum_{n=1}^\infty \frac{z^n}{n^2}\tag1$$

As you can see it is directly deducable from this series that we got $\operatorname{Li}_2(1)=\zeta(2)=\pi^2/6$. Anyway, there are some more nice properties about this function, e.g. an integral representation and functional equations. Right now we are especially interested in the latter one. More precise we want to show the following equation

$$\operatorname{Li}_2(z)+\operatorname{Li}_2(1-z)~=~\frac{\pi^2}6-\log(z)\log(1-z)\tag2$$

This looks familiar, doesn't it? Yes, exactly! It is your given equation by setting $f(x)=\operatorname{Li}_2(x)$. However, as you may notice your series representation of $f(x)$ does not match the one of the Dilogarithm but I suppose there was a mistake made either by proposing the task or while transfering it. The other option would be that this example was chosen to be disproved. I will just continue with a proof of $(2)$ only relying on $(1)$ aswell as on the of the natural logarithm.

Lets begin by finding the derivative of $\operatorname{Li}_2(z)$ which can be found by differentiating the series representation $(1)$ termwise. Thus, we get

\begin{align*} \frac{\mathrm d}{\mathrm dz}\operatorname{Li}_2(z)&=\sum_{n=1}^\infty \frac{\mathrm d}{\mathrm dz}\frac{z^n}{n^2}=\sum_{n=1}^\infty \frac{z^{n-1}}n=-\frac{\log(1-z)}z\\ \frac{\mathrm d}{\mathrm dz}\operatorname{Li}_2(1-z)&=\sum_{n=1}^\infty \frac{\mathrm d}{\mathrm dz}\frac{(1-z)^n}{n^2}=-\sum_{n=1}^\infty \frac{(1-z)^{n-1}}n=\frac{\log(z)}{1-z} \end{align*}

On the other hand note that we can use the product rule to differentiate $\log(z)\log(1-z)$ which yields to

$$\frac{\mathrm d}{\mathrm dz}\log(z)\log(1-z)=\frac{\log(1-z)}z-\frac{\log(z)}{1-z}$$

In other words the derivatives of both sides alone equals each other which conversely implies that their anti-derivatives are in fact the same up to a constant. Reformulating this relation we get the following

\begin{align*} \int -\frac{\log(1-z)}z\mathrm dz+\int\frac{\log(z)}{1-z}\mathrm dz&=\int-\frac{\log(1-z)}z+\frac{\log(z)}{1-z}\mathrm dz\\ \operatorname{Li}_2(z)+\operatorname{Li}_2(1-z)&=-\log(z)\log(1-z)+c \end{align*}

To determine the constant $c$ we may just plug in $z=1$ to find that

\begin{align*} \operatorname{Li}_2(1)+\operatorname{Li}_2(1-1)&=-\log(1)\log(1-1)+c\\ \operatorname{Li}_2(1)&=c\\ \therefore~c&=\frac{\pi^2}6 \end{align*}

Hence the identity $(2)$ has been proved. The here given proof is more or less the same as Lewin's in his Polylogarithms and Associated Functions. Now moreover note two things: firstly the identity only holds for the interval $[0,1]$ otherwise we will run into problems regarding the natural logarithm and secondly the series expansion $(1)$ only holds for $|z|<1$ $($the value at $z=1$ can be deduced nevertheless$)$.

Anyway, the question remains if it was intended by your teacher to give you a wrong equation or if there happend a mistake of transfering the task inbetween. In order to match equation $(2)$ $f(x)$ the terms should be given by $x^n/n^2$ rather than by $x^{n-1}/n$.

mrtaurho
  • 16,103
  • 1
    So good!Thank you.I found a similar question.link – Li Taiji Feb 15 '19 at 15:21
  • @梦里年华似烟花 As you can see the Dilogarithm is a great to tool while evaluating definite integrals containing logarithms. If you are satisfied with either mine or Ahmad Bazzi's answer you could leave an upvote and accept the one which answers your question the best :) – mrtaurho Feb 15 '19 at 15:24
  • But I think even if the $f(x)$ is $x^{n-1}/n^2$,the question is also wrong.$f(x)$ should be $x^n/n^2$. – Li Taiji Feb 15 '19 at 15:27
  • @梦里年华似烟花 I know, that is why I think the the equation given was intended to be false even though it is odd that it is almost right, i.e. that the given series is so close the one of the Dilogarithm. – mrtaurho Feb 15 '19 at 15:29
  • So if $f(x)$ is $x^n/n^2$,the question is true? – Li Taiji Feb 15 '19 at 15:31
  • @梦里年华似烟花 Precisely! Hence then we are dealing with the functional equation $(2)$ because in this case we get that $f(x)=\operatorname{Li}_2(x)$ as you can see within $(1)$ aswell $($just substitute $z=x)$. – mrtaurho Feb 15 '19 at 15:33
  • @ mrtaurho .This is a Graduate Entrance Examination question.I will confirm the question to the teacher,and edit it. – Li Taiji Feb 15 '19 at 15:39