AMM 12181-05 issue

Calculus Level 5

k = 2 1 k 0 1 { 1 x k } d x = α γ \sum_{k=2}^{\infty}\frac1k\int_0^1\left\{\frac{1}{\sqrt[k]{x}}\right\} dx =\alpha\gamma

If the equation above holds true, find the value of α 2 \alpha^2 .

Notations:


This problem appeared in American Mathematical Monthly, 2020 April as problem 12181-05 which was proposed by Shivam Sharma, India where the original problem was to prove the closed form .


The answer is 1.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

1 solution

Naren Bhandari
Oct 10, 2020

The original proposed issue was to prove k = 2 1 k 0 { 1 x k } d x = γ \sum_{k=2}^{\infty}\frac{1}{k}\int_0^{\infty}\left\{\frac{1}{\sqrt[k]{x}}\right\}dx=\gamma

Here are two proposed solutions of mine

Solution 1 . Call the fractional integral J k = 0 1 { 1 x k } d x \displaystyle J_k= \int_0^1\left\{\frac{1}{\sqrt[k]{x}}\right\}dx for all k 2 k\geq 2 . We shall prove that the integral J k = k k 1 ζ ( k ) \displaystyle J_k = \frac{k}{k-1}-\zeta(k) . So we substitute x = u k x=u^{-k} then we have the integral above as k 1 { u } u k + 1 d u = k n 1 n n + 1 u n u k + 1 d u = k n 1 n n + 1 ( 1 u n u k + 1 ) d u k\int_1^{\infty}\frac{\left\{u\right\}}{u^{k+1}}du=k\sum_{n\geq 1 }\int_n^{n+1}\frac{u-n}{u^{k+1}}du=k\sum_{n\geq 1}\int_{n}^{n+1}\left(\frac{1}{u}-\frac{n}{u^{k+1}}\right)du Evaluating the integrals and simplification yields J k = k k 1 n 1 ( 1 ( n + 1 ) k 1 1 n k 1 ) -1 + n 1 ( 1 ( n + 1 ) k 1 n k ) n J_k=-\frac{k}{k-1}\underbrace {\sum_{n\geq 1}\left(\frac{1}{(n+1)^{k-1}}--\frac{1}{n^{k-1}}\right)}_{\textbf{-1}}+ {\sum_{n\geq 1}\left(\frac{1}{(n+1)^{k}}-\frac{1}{n^{k}}\right)n} Further note that the former sum (underbraced) is telescoping series that converges to 1 -1 this we have former sum is equal to k k 1 ( 1 ) \frac{k}{k-1}\cdots (1) and latter sum we write as n 1 ( 1 ( n + 1 ) k 1 1 n k 1 1 ( n + 1 ) k ) = n 1 ( 1 1 ( n + 1 ) k ) = ζ ( k ) ( 2 ) \sum_{n\geq 1}\left(\frac{1}{(n+1)^{k-1}}-\frac{1}{n^{k-1}}-\frac{1}{(n+1)^{k}}\right)=\sum_{n\geq 1}\left(-1-\frac{1}{(n+1)^{k}}\right)=-\zeta(k)\cdots(2) Thus from ( 1 ) (1) and ( 2 ) (2) we have J k = k k 1 ζ ( k ) \displaystyle J_k=\frac{k}{k-1}-\zeta(k) . Now we are left to evaluate k 2 J k = k 2 1 k ( k k 1 ζ ( k ) ) = k 2 ( 1 k 1 1 k ( ζ ( k ) 1 k ) ) \sum_{k\geq 2}J_k= \sum_{k\geq 2}\frac{1}{k}\left(\frac{k}{k-1}-\zeta(k)\right)=\sum_{k\geq 2}\left(\frac{1}{k-1}-\frac{1}{k}-\left(\frac{\zeta(k)-1}{k}\right)\right) here the sum k 2 ( 1 k 1 1 k ) = lim m k = 1 m ( 1 k 1 1 k ) = lim m ( 1 1 m + 1 ) = 1 \displaystyle \sum_{k\geq 2}\left(\frac{1}{k-1}-\frac{1}{k}\right)=\lim_{m\to\infty}\sum_{k=1}^m\left(\frac{1}{k-1}-\frac{1}{k}\right)=\lim_{m\to\infty}\left(1-\frac{1}{m+1}\right)=1 and to evaluate the sum k 2 ζ ( k ) 1 k ( 3 ) \displaystyle \sum_{k\geq 2}\frac{\zeta(k)-1}{k}\cdots(3) recall that ζ ( k ) 1 = m = 2 1 m k \displaystyle \zeta(k)-1=\sum_{m=2}^{\infty}\frac{1}{m^k} . plugging in ( 3 ) (3) gives us k = 2 ζ ( k ) 1 k = k = 2 1 k ( m = 2 1 m k ) = m = 2 ( k 2 1 k m k ) = m 2 ( 1 m ln ( m 1 m ) ) \sum_{k=2}^{\infty}\frac{\zeta(k)-1}{k}=\sum_{k=2}^{\infty}\frac{1}{k}\left(\sum_{m=2}^{\infty}\frac{1}{m^k}\right)=\sum_{m=2}^{\infty}\left(\sum_{k\geq 2}\frac{1}{k m^k}\right)=\sum_{m\geq 2}\left(-\frac{1}{m}-\ln\left(\frac{m-1}{m}\right)\right) the latter sum further can be written as m 2 ( 1 m ln m + ln ( m 1 ) ) = lim m j = 2 m ( 1 j + ln ( j 1 ) ln j ) -\sum_{m\geq 2}\left(\frac{1}{m}-\ln m+\ln(m-1)\right)=-\lim_{m\to\infty}\sum_{j= 2}^m\left(\frac{1}{j}+\ln(j-1)-\ln j\right) Note that j = 2 m 1 j = H m 1 \displaystyle \sum_{j=2}^{m} \frac{1}{j}= H_m-1 and j = 2 m ( ln ( j 1 ) ln j ) = 0 ln m = ln m \displaystyle \sum_{j=2}^m (\ln(j-1)-\ln j)= 0-\ln m=-\ln m therefore we yield lim m j = 2 m ( 1 j + ln ( j 1 ) ln j ) = lim m ( H m ln m 1 ) = 1 γ ( 4 ) -\lim_{m\to\infty}\sum_{j=2}^m\left(\frac{1}{j}+\ln(j-1)-\ln j\right)=-\lim_{m\to\infty} (H_m-\ln m-1)=1-\gamma\cdots(4) subtracting 3 3 from 4 4 we have the desired result k 2 J k = γ \displaystyle \sum_{k\geq 2}J_k=\gamma .


Solution 2 Let f ( x ) = { 1 x k } f(x)=\left\{\frac{1}{\sqrt[k]{x}}\right\} for all x ( 0 , 1 ] x\in (0,1] and k 2 k\geq 2

0 1 f ( x ) d x = 0 1 ( 1 x k 1 x k ) d x = k k 1 I ( k ) \int_0^1f(x)dx = \int_0^1\left(\frac{1}{\sqrt[k]{x}}-\left\lfloor \frac{1}{\sqrt[k]{x}}\right\rfloor\right)dx=\frac{k}{k-1}-I(k) We note that n x < n + 1 n\leq x < n+1 and hence n k x k < n + 1 k \sqrt[k]{n} \leq \sqrt[k]{x} < \sqrt[k]{n+1} which implies 1 n + 1 k < 1 x k 1 n k \frac{1}{\sqrt[k]{n+1}} < \frac{1}{\sqrt[k]{x}}\leq \frac{1}{\sqrt[k]{n}} . also on the interval ( ( n + 1 ) 1 / k , n 1 / k ] , x 1 / k = n ( (n+1)^{-1/k}, n^{-1/k}], \lfloor x^{-1/k}\rfloor = n . Therefore I k = n n 1 1 n k 1 ( n + 1 ) k d x = n 1 ( 1 ( n + 1 ) k 1 n k ) n = ζ ( k ) I_k = -n\sum_{n\geq 1}\int_{\frac{1}{n^k}}^{\frac{1}{(n+1)^k}}dx=-\sum_{n\geq 1} \left(\frac{1}{(n+1)^k}-\frac{1}{n^k}\right)n=\zeta(k) giving us 0 1 f ( x ) d x = k k 1 ζ ( k ) \int_0^1 f(x)dx =\frac{k}{k-1} -\zeta(k) To prove k 2 0 1 f ( x ) d x = k 2 ( 1 k 1 1 k ζ ( k ) 1 k ) = γ \displaystyle \sum_{k\geq 2} \int_0^1 f(x) dx = \sum_{k\geq2} \left(\frac{1}{k-1}-\frac{1}{k}-\frac{\zeta(k)-1}{k}\right)=\gamma As we have in solution 1 k 2 ( 1 k 1 1 k ) = 1 ( 5 ) \displaystyle \sum_{k\geq 2} \left(\frac{1}{k-1}-\frac{1}{k}\right)=1\cdots (5) and to show that k ζ ( k ) 1 k = 1 γ \displaystyle\sum_{k\geq}\frac{\zeta(k)-1}{k}=1-\gamma we make an alternative approach as follows. We have that ζ ( k ) 1 k = 0 x k k ! d x x e x ( e x 1 ) d x \frac{\zeta(k)-1}{k}=\int_0^{\infty}\frac{x^k}{k !}\cdot \frac{dx}{xe^x(e^x-1)}dx and hence on summing we have k 2 ζ ( k ) 1 k = 0 e x x 1 x e x ( e x 1 ) d x = 0 d x e x 0 ( 1 e x 1 1 x e x ) d x = 1 A \sum_{k\geq 2}\frac{\zeta(k)-1}{k}=\int_0^{\infty}\frac{e^x-x-1}{xe^x(e^x-1)}dx=\int_0^{\infty}\frac{dx}{e^x}-\int_{0}^{\infty}\left(\frac{1}{e^x-1}-\frac{1}{xe^x}\right)dx=1-A since A = 0 ( 1 e x 1 1 x e x ) d x = γ A=\displaystyle -\int_{0}^{\infty}\left(\frac{1}{e^x-1}-\frac{1}{xe^x}\right)dx=-\gamma and and hence k 2 ζ ( k ) 1 k = 1 γ ( 6 ) \displaystyle\sum_{k\geq 2}\frac{\zeta(k)-1}{k}=1-\gamma\cdots(6) and hence from ( 5 ) (5) and ( 6 ) (6) we proved k 2 1 k 0 1 f ( x ) d x = γ \displaystyle \sum_{k\geq 2}\frac{1}{k}\int_0^1 f(x)dx=\gamma .

@Naren Bhandari Hello sir. Your Mathematics skills are higher than mountain Everest.
Can you help me in this problem. 1 sin x d x \int \frac{1}{\sqrt{\sin x}}dx
Thanks in advance.

Talulah Riley - 8 months ago

Log in to reply

Yes sir .......

SRIJAN Singh - 8 months ago

If you're satisfied with using a non-elementary function: elliptic integrals, then the answer is pretty easy to obtain.

If you're looking for an elementary solution, don't bother. It's not possible.

If you added the limits π / 6 , π / 3 \pi/6, \pi/3 , I'm still pretty confident that the definite integral does not have an elementary closed-form either.

Pi Han Goh - 8 months ago

The problem doesn't have any elementary antiderivatives. However, if you are interested to know antiderivatives of it then here you go in terms of Elliptic integral.

Naren Bhandari - 8 months ago

Log in to reply

@Naren Bhandari However if I want to put limits like π 6 \frac{π}{6} to π 3 \frac{π}{3} . Then can we evaluate?

Talulah Riley - 8 months ago

Log in to reply

Yes we can. just need to set the limits. The limit you mention yield 0 answer.

Naren Bhandari - 8 months ago

Log in to reply

@Naren Bhandari @Naren Bhandari i have changed the limits. Sorry.
But how I will find the answer of that integral to put the upper and lower limts?

Talulah Riley - 8 months ago

Log in to reply

@Talulah Riley If you dont want to use the result of it then simply set t = sin x t=\sin x then find the limits and then you will have beta function however, it will in the form incomplete beta function if you wish to work then use gauss hypergeometric relation with respect of IBF. The answer still have ugly closed form.

Naren Bhandari - 8 months ago

Log in to reply

@Naren Bhandari @Naren Bhandari After using hypergeometric relation, and then putting the limits , will I will get the same answer , which the same answer is showing in Wolfram alpha.?

Talulah Riley - 8 months ago

Log in to reply

@Talulah Riley It's all upto you, if you solve it that would be nice.

Naren Bhandari - 8 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...