compute $I=\lim\limits_{n\to+\infty}\sqrt[n]{\int\limits_0^1x^{n+1}(1-x)\cdots(1-x^n)dx}$

attempt: I tried to evaluate the integral $$\begin{align} \int\limits_0^1x^{n+1}(1-x)\cdots(1-x^n)dx&=\int\limits_0^1x^{n+1}\left(1-x-x^2-x^3+x^5+\cdots\right)dx\\ &=\int\limits_0^1x^{n+1}-x^{n+2}-x^{n+3}-x^{n+4}+x^{n+6}+\cdots dx\\ &=\left.\frac{x^{n+2}}{n+2}-\frac{x^{n+3}}{n+3}-\frac{x^{n+4}}{n+4}-\frac{x^{n+5}}{n+5}+\frac{x^{n+6}}{n+6}+\cdots\right|_0^1\\ &=\frac{1}{n+2}-\frac{1}{n+3}-\frac{1}{n+4}-\frac{1}{n+5}+\frac{1}{n+6}+\cdots \end{align}$$

attempt 2:

I think I can use mean value theorem for integrals

if $f$ are continuous into $[a,b]$ then exists $\xi\in[0,1]$ such that $\int\limits_a^b f(x)dx=f(\xi)(b-a)$

then using it for $[a,b]=[0,1]$ since the function inside integral are continuous, then exists $\xi\in[0,1]$ such that

$$\begin{align} \sqrt[n]{\int\limits_0^1x^{n+1}(1-x)\cdots(1-x^n)dx}&=\sqrt[n]{\xi^{n+1}(1-\xi)\cdots(1-\xi^n)(1-0)}\\ &=\sqrt[n]{\xi^{n+1}(1-\xi)\cdots(1-\xi^n)} \end{align}$$

then making for $n\in\mathbb{N}^*$

$$\begin{align} a_n&=\sqrt[n]{\xi^{n+1}(1-\xi)\cdots(1-\xi^n)}\\ a_{n+1}&=\sqrt[n+1]{\xi^{n+2}(1-\xi)\cdots(1-\xi^n)(1-\xi^{n+1})}\\ &=\sqrt[n+1]{\xi\xi^{n+1}(1-\xi)\cdots(1-\xi^n)(1-\xi^{n+1})}\\ &=\sqrt[n+1]{\xi\left[\sqrt[n]{\xi^{n+1}(1-\xi)\cdots(1-\xi^n)}\right]^n(1-\xi^{n+1})}\\ &=\sqrt[n+1]{a_n^n\xi(1-\xi^{n+1})}\\ a_{n+1}^{n+1}&=a_n^n\xi(1-\xi^{n+1}) \end{align}$$

and $\xi\in(0,1)\Rightarrow\xi^n\in(0,1)\Rightarrow 1-\xi^n\in(0,1)\Rightarrow a_n>0$ and $a_n\in(0,1)$


$$\begin{align} a_{1}&=\xi^2(1-\xi)\\ a_{2}&=\sqrt{\xi^3(1-\xi)(1-\xi^2)}\\ &=\sqrt{\xi^3(1-\xi)^2(1+\xi)}\\ &=\xi(1-\xi)\sqrt{\xi(1+\xi)} \end{align}$$

for $\xi\in(0,1)\Rightarrow \xi(1-\xi)>0$

$$\begin{align} 0<\xi&<1+\xi\\ 0<\xi^2&<\xi(1+\xi)=\xi+\xi^2\\ \xi&<\sqrt{\xi(1+\xi)}\\ \xi^2(1-\xi)&<\xi(1-\xi)\sqrt{\xi(1+\xi)}\\ a_1&<a_2 \end{align}$$

then i think if i can proof that $a_{n+1}>a_n,n\in\mathbb{N}^*$ and $\xi\in(0,1)$ then the limit would be $1$

  • 1
    $\begingroup$ Very Nice question! $\endgroup$ – Peter Jan 28 '15 at 17:30
  • $\begingroup$ the Product $(1-x)....(1-x^n)$ is known as a PochhammerQ Symbol which have some interesting asymptotic expansions... $\endgroup$ – tired Jan 28 '15 at 17:34
  • $\begingroup$ We have: $$ I_n = \int_{0}^{1}x^{n+1}(1-x)(1-x^2)\cdot\ldots\cdot(1-x^n)\,dx \geq \int_{0}^{1}x^{n+1}(1-x)^n\,dx = \frac{1}{(n+1)\binom{2n+2}{n+1}},$$ so the limit is $\geq \frac{1}{4}$, but I am having a hard time in proving a converse bound in order to show that the limit is exactly $\frac{1}{4}$. By a simple argument, the limit is $\leq 1$. $\endgroup$ – Jack D'Aurizio Jan 28 '15 at 17:39
  • $\begingroup$ In fact, the limit seems to be $1$. $\endgroup$ – Peter Jan 28 '15 at 17:43
  • $\begingroup$ Perhaps a shot in the dark, but parameterizing the integral as $$I_{a,n}=\int_0^1 (ax)^{n+1}(1-x)\cdots(1-x^n)\,dx$$ and differentiating $n+1$ times gives the differential equation ${I_a}^{(n+1)}-(n+1)!I_a=0$ with characteristic equation $r^{n+1}-(n+1)!=0$. $\endgroup$ – user170231 Jan 28 '15 at 17:47

We have: $$\begin{eqnarray*} I_n &=& \int_{0}^{1}x^{n+1}(1-x)(1-x^2)\cdot\ldots\cdot(1-x^n)\,dx \\&\leq& \int_{0}^{1}x^{n+1}(1-x^n)^n\,dx = \frac{\Gamma(n)\,\Gamma\left(1+\frac{2}{n}\right)}{\Gamma\left(2+n+\frac{2}{n}\right)}\leq\frac{1}{n^2},\tag{1}\end{eqnarray*}$$ while on the other hand, if we set: $$ A(n,m)\triangleq\int_{0}^{1} x^n \prod_{k=1}^{m}(1-x^k)\,dx,\qquad B(n,m)=\int_{0}^{1} x^n (1-x^m)^m\,dx \tag{2}$$ we have: $$\begin{eqnarray*} I_n = A(n+1,n) &=& A(n+1,n-1)-A(2n+1,n-1)\\&=&A(n+1,n-2)- A(2n,n-2)-A(2n+1,n-1)\end{eqnarray*} $$ and since $A(2n+2-i,n-i)\leq B(2n+2-i,n-i)\leq\frac{1}{(n+2)^3}$ it follows that: $$ I_n = A(n+1,1)-\sum_{k=1}^{n-1}A(2n+2-k,n-k) \geq \frac{1}{(n+2)(n+3)}-\frac{n-1}{(n+2)^3}\tag{3}$$ so $I_n\geq\frac{2}{(n+2)^3}$ and this inequality, together with $(1)$, implies: $$ I_n^{\frac{1}{n}}\xrightarrow[n\to +\infty]{}\color{red}{1}. \tag{4}$$


I can't comment because apperently just guys with super high level can help you. About this approach with the mean valued integral theorem: I've tried it.It won't work simply because $\xi$ depends on $n$ if it was a constant for any $n$ I would agree with you, this limit is 1. But the sequence $a_n$ you created is not necessarly monotonic. Because $\xi$ is a function of $n$


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.