Not quite a telescoping sum (1)

Calculus Level 3

n = 0 [ 1 8 n + 1 1 8 n + 7 ] \displaystyle \sum_{n=0}^{\infty} \left[\frac{1}{8n+1}-\frac{1}{8n+7} \right] can be written in the form π ( a + b c ) \pi \left(\dfrac{\sqrt{a}+b}{c}\right) where a a , b b and c c are positive integers with a a being square free.

What is the value of a + b + c a+b+c ?


The answer is 11.

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.

3 solutions

Pi Han Goh
May 5, 2020

Let us first logarithmically differentiate the Weierstrass Product : cos ( π x ) = n = 0 ( 1 4 x 2 ( 2 n + 1 ) 2 ) \cos(\pi x) = \prod_{n=0}^\infty \left(1 - \frac{4x^2}{(2n+1)^2} \right) to get n = 0 1 ( 2 n + 1 ) 2 ( 2 x ) 2 = π 8 x tan ( π x ) \sum_{n=0}^\infty \frac1{ (2n+1)^2 - (2x)^2} = \frac\pi{8x} \tan(\pi x) and noticing that 1 8 n + 1 1 8 n + 7 = 3 8 1 ( 2 n + 1 ) 2 ( 2 3 8 ) 2 \frac1{8n+1} - \frac1{8n+7} = \frac38 \cdot \frac1{(2n+1)^2 - \left(2\cdot \frac38 \right)^2 }

we can conclude that n = 0 ( 1 8 n + 1 1 8 n + 7 ) = 3 8 n = 0 1 ( 2 n + 1 ) 2 ( 3 4 ) 2 = 3 8 π 8 3 8 tan ( π 3 8 ) = π 8 ( 1 + 2 ) \sum_{n=0}^\infty \left ( \frac1{8n+1} - \frac1{8n+7} \right) = \frac38 \sum_{n=0}^\infty \frac1{(2n+1)^2 - \left(\frac34 \right)^2 } = \frac38 \cdot \frac\pi{8 \cdot \frac38} \tan\left( \pi \cdot \frac38 \right) = \frac{\pi}{8}(1 + \sqrt2)

Thank you for sharing.

Karan Chatrath - 1 year, 1 month ago
Karan Chatrath
May 4, 2020

For x < 1 \lvert x \rvert <1 consider the following geometric sequences:

1 1 x 8 = lim n k = 0 n x 8 k ( 1 ) \frac{1}{1-x^8} = \lim_{n \to \infty} \sum_{k=0}^{n} x^{8k} \ \dots \ (1) x 6 1 x 8 = lim n k = 0 n x 8 k + 6 ( 2 ) \frac{x^6}{1-x^8} = \lim_{n \to \infty} \sum_{k=0}^{n} x^{8k+6} \ \dots \ (2)

Integrating both sides with respect to x x from 0 to 1 gives:

lim t 1 0 t 1 1 x 8 d x = lim n k = 0 n 1 8 k + 1 ( 3 ) \lim_{t \to 1^{-}} \int_{0}^{t} \frac{1}{1-x^8} dx= \lim_{n \to \infty} \sum_{k=0}^{n} \frac{1}{8k+1} \ \dots \ (3) lim t 1 0 t x 6 1 x 8 d x = lim n k = 0 n 1 8 k + 7 ( 4 ) \lim_{t \to 1^{-}} \int_{0}^{t} \frac{x^6 }{1-x^8}dx = \lim_{n \to \infty} \sum_{k=0}^{n} \frac{1}{8k+7} \ \dots \ (4)

Subtracting (3) and (4) gives:

lim n k = 0 n [ 1 8 k + 1 1 8 k + 7 ] = lim t 1 0 t ( 1 x 6 1 x 8 ) d x = π ( 2 + 1 ) 8 \lim_{n \to \infty} \sum_{k=0}^{n} \left[\frac{1}{8k+1}-\frac{1}{8k+7}\right]= \lim_{t \to 1^{-}} \int_{0}^{t} \left(\frac{1-x^6}{1-x^8}\right) dx=\boxed{ \frac{\pi \left(\sqrt{2}+1\right)}{8}}

The steps of evaluation of the right-hand side limit are left out. I will add those details if requested.

You're not quite right. The series in (3) and (4) diverge.

Pi Han Goh - 1 year, 1 month ago

Log in to reply

Yes, I agree that (3) and (4) are divergent. But I don't see why subtracting two divergent series is incorrect. A linear combination of two divergent series can be convergent as is the case here.

Karan Chatrath - 1 year, 1 month ago

Log in to reply

Infinity - infinity is not a valid operation. You can't perform arithmetic operation on infinity, which is not a number, but just a concept.

Pi Han Goh - 1 year, 1 month ago

Log in to reply

@Pi Han Goh Okay, point taken. I have made a small adjustment to my solution. I think this addresses your concern.

Karan Chatrath - 1 year, 1 month ago

Log in to reply

@Karan Chatrath No, n = 0 1 8 n + 1 \displaystyle \sum_{n=0}^\infty \frac1{8n+1} diverges .

Pi Han Goh - 1 year, 1 month ago

Log in to reply

@Pi Han Goh I have made a second adjustment to my solution.

Karan Chatrath - 1 year, 1 month ago

@Pi Han Goh By your argument, the problem statement can be subjected to the same kind of scrutiny.

n = 0 [ 1 8 n + 1 1 8 n + 7 ] = n = 0 1 8 n + 1 n = 0 1 8 n + 7 \sum_{n=0}^{\infty} \left[\frac{1}{8n+1} - \frac{1}{8n+7} \right] = \sum_{n=0}^{\infty} \frac{1}{8n+1} -\sum_{n=0}^{\infty} \frac{1}{8n+7}

Or are you telling me that:

n = 0 [ 1 8 n + 1 1 8 n + 7 ] n = 0 1 8 n + 1 n = 0 1 8 n + 7 \sum_{n=0}^{\infty} \left[\frac{1}{8n+1} - \frac{1}{8n+7} \right] \ne \sum_{n=0}^{\infty} \frac{1}{8n+1} -\sum_{n=0}^{\infty} \frac{1}{8n+7}

The first statement amounts to \infty-\infty . I frankly do not see a problem with my earlier arguments because like you said, infinity is a concept and not a number. So an infinite summation essentially means summing up a large number of terms of a sequence. This, to me, is logical and it does not amount to \infty-\infty .

Karan Chatrath - 1 year, 1 month ago

Log in to reply

@Karan Chatrath Yes, the second ( LaTeX \LaTeX line) is the one I'm alluding to. It's a common misconception to express a convergent series as the sum of divergent series. See Riemann rearrangement theorem .

Pi Han Goh - 1 year, 1 month ago

Log in to reply

@Pi Han Goh What you're basically saying is that ( a 1 b 1 ) + ( a 2 b 2 ) + = ( a 1 + a 2 + ) ( b 1 + b 2 + ) (a_1 - b_1) + (a_2 - b_2) + \cdots = (a_1 + a_2 + \cdots) - (b_1 + b_2 + \cdots ) must hold.

But this isn't true!

Take the simpler telescoping series:

n = 1 ( 1 n 1 n + 1 ) = 1 \sum_{n=1}^\infty \left( \frac1n - \frac1{n+1} \right) = 1

By your reasoning, you would have concluded that 1 + 1 2 + 1 3 + 1 + \frac12 + \frac13 + \cdots is finite and is greater than 1 2 + 1 3 + 1 4 + \frac12 + \frac13 + \frac14 + \cdots by exactly 1. But that's not right, because the both of these series diverge! See Is + 1 > \infty + 1>\infty ? .

Pi Han Goh - 1 year, 1 month ago

This isnot correct solution.

Naren Bhandari - 1 year, 1 month ago

Log in to reply

@Naren Bhandari Hi, thanks for the feedback. Could you provide some specific details about my errors so that I may learn from them?

Karan Chatrath - 1 year, 1 month ago

Log in to reply

@Karan Chatrath Your very first line started with x < 1 |x| < 1 , so your equations (3) and (4) does not hold because x = 1 x=1 does not satisfy this inequality.

You could just cut out the working for (3) and (4) and show your last step: The sum converts to an integral.

But to do justify this step, you need to use Dominated convergence theorem.

Alternatively, you could first show that each term of the series is positive, and that the series converge by p p -test.

Pi Han Goh - 1 year, 1 month ago

Log in to reply

@Pi Han Goh Okay, your feedback is much appreciated. Also, thanks for sharing those couple of links in your previous comments.

I thought that introducing the limits as I have, obeys the property that x < 1 \lvert x \rvert <1 . I could be more specific by saying that it is a left-hand limit. I am adding that detail to my solution as well.

Also, I am unfamiliar with the dominated divergence theorem as well as the p-test. I will read about these two subjects at my leisure and subject my solution to further scrutiny. In the meantime, you are welcome to mark my solution as confusing.

Karan Chatrath - 1 year, 1 month ago

Log in to reply

@Karan Chatrath I've submitted an alternative solution. =D

Pi Han Goh - 1 year, 1 month ago

@Karan Chatrath Also, there's a one-line solution via digamma function . Do try it out!

Pi Han Goh - 1 year, 1 month ago
Aruna Yumlembam
May 18, 2020

We can rewrite the above sum as (1-1/7)+sigma from n= 1 to infinity (1/8n-1/(8n+7))-sigma from n=1 to infinity (1/8n-1/(8n+1)) ,then recognizing that there are digamma functions hiding in both the sum we have (8/7+ digamma(1-1/8)-digamma (1/8)-8) all divided by 8 . Simplifying by using the reflection equation 1/7-1+πcot (π/8) ,since cot(π/8)=2^1/2+1. Hence we have by inputting it to the original sum, 1-1/7+1/7-1+(π(2^1/2+1)/8)=(π(2^1/2+1))/8,a=2,b=1 and c=8.Adding it gives 11 as our result.(sorry for the messy calculation.)

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...