Highly Nested Summation!

Calculus Level 5

S = i = 1 j = 1 k = 1 1 i j k ( i + j + k ) \large{S = \sum_{i=1}^\infty \sum_{j=1}^\infty \sum_{k=1}^\infty \dfrac{1}{ijk(i+j+k)}}

If S S can be expressed as A B π C \large \dfrac{A}{B} \pi^C for positive integers A , B , C A,B,C with gcd ( A , B ) = 1 \gcd(A,B)=1 , submit the value of A + B + C A+B+C as your answer.


Bonus: Generalize the expression below in terms of n n :

P ( n ) = i 1 = 1 i 2 = 1 i n = 1 1 i 1 i 2 i n ( i 1 + i 2 + + i n ) = ? \large{P(n) = \sum_{i_1=1}^\infty \sum_{i_2=1}^\infty \cdots \sum_{i_n=1}^\infty \dfrac{1}{i_1 i_2 \dotsm i_n(i_1+i_2+\ldots+i_n)}=\ ?}


The answer is 20.

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.

2 solutions

Satyajit Mohanty
Sep 20, 2015

We'll start generalizing our Bonus Problem to calculate S S by substituting n = 3 n=3 .

We first notice that:

P ( n ) = i 1 = 1 i n = 1 1 i 1 i 2 i n 0 1 x i 1 + i 2 + + i n 1 d x = 0 1 1 x ( i 1 = 1 x i 1 i 1 ) ( i n = 1 x i n i n ) d x = 0 1 ( 1 ) n ( ln ( 1 x ) ) n x d x \begin{aligned} \displaystyle P(n) &= \sum_{i_1=1}^\infty \cdots \sum_{i_n=1}^\infty \dfrac{1}{i_1i_2 \dotsm i_n} \int_0^1 x^{i_1 + i_2 + \ldots + i_n-1} dx \\ &= \int_0^1 \dfrac{1}{x} \left( \sum_{i_1=1}^\infty \dfrac{x^{i_1}}{i_1} \right) \cdots \left( \sum_{i_n=1}^\infty \dfrac{x^{i_n}}{i_n} \right) dx \\ &= \int_0^1 (-1)^n \dfrac{(\ln(1-x))^n}{x} dx \end{aligned}

We now use the substitution t = 1 x t=1-x in the above integral to get:

P ( n ) = ( 1 ) n 0 1 ( ln ( t ) ) n 1 t d t = ( 1 ) n 0 1 ( ln ( t ) ) n ( r = 0 t r ) d t = ( 1 ) n r = 0 0 1 t r ( ln ( t ) ) n d t \begin{aligned} \displaystyle P(n) &= (-1)^n \int_0^1 \dfrac{(\ln(t))^n}{1-t} dt = (-1)^n \int_0^1 (\ln(t))^n \left( \sum_{r=0}^\infty t^r \right) dt \\ &= (-1)^n \sum_{r=0}^\infty \int_0^1 t^r (\ln(t))^n dt \end{aligned}

In view of the integration by parts formulaes, we see that:

0 1 t r ( ln ( t ) ) n d t = ( 1 ) n n ! ( r + 1 ) n + 1 \int_0^1 t^r (\ln(t))^n dt = (-1)^n \dfrac{n!}{(r+1)^{n+1}}

Therefore:

P ( n ) = ( 1 ) n r = 0 ( 1 ) n n ! ( r + 1 ) n + 1 = n ! r = 0 1 ( r + 1 ) n + 1 P(n) = (-1)^n \sum_{r=0}^\infty (-1)^n \dfrac{n!}{(r+1)^{n+1}} = n! \sum_{r=0}^\infty \dfrac{1}{(r+1)^{n+1}}

P ( n ) = n ! ζ ( n + 1 ) \large \Longrightarrow \boxed{P(n) = n! \cdot \zeta(n+1)}

where ζ ( s ) \zeta(s) denotes the Riemann-Zeta Function. To calculate the value of S S , substitute n = 3 n=3 to obtain:

S = 3 ! ζ ( 4 ) = 6 π 4 90 = π 4 15 \large S = 3! \cdot \zeta(4) = 6 \cdot \dfrac{\pi^4}{90} = \dfrac{\pi^4}{15}

So A = 1 , B = 15 , C = 4 , A + B + C = 20 A=1, \ B=15, \ C=4, \quad A+B+C=\boxed{20}

Did the exact same! Nice easy problem!

Kartik Sharma - 5 years, 8 months ago

Log in to reply

Except I did it without IBP. I did it using Feynman's Method.

0 1 x n 1 x d x \displaystyle \int_{0}^{1}{\frac{x^n}{1-x} \ dx} will be our "base" integral.

0 1 x n k = 0 n x k d x \displaystyle \int_{0}^{1}{x^n \sum_{k=0}^{n}{x^k} \ dx}

k = 0 n 0 1 x n + k d x \displaystyle \sum_{k=0}^{n}{\int_{0}^{1}{{x}^{n+k} \ dx}}

k = 0 n 1 n + k + 1 \displaystyle \sum_{k=0}^{n}{\frac{1}{n+k+1}} [I would have directly used Harmonic or digamma but just to show, I have used this long method]

I ( n ) = k = 0 n 1 n + k + 1 \displaystyle I(n) = \sum_{k=0}^{n}{\frac{1}{n+k+1}}

I guess it's very easy to guess what's I ( m ) ( 0 ) {I}^{(m)}(0) ? Right?

I ( m ) ( 0 ) = ( 1 ) m Γ ( m + 1 ) ζ ( m + 1 ) \displaystyle {I}^{(m)}(0) = (-1)^m \Gamma(m+1) \zeta(m+1)

Kartik Sharma - 5 years, 8 months ago

Log in to reply

Wait. How is Digamma applied here? I don't see any possible instance of using it here.

Pi Han Goh - 5 years, 8 months ago

Log in to reply

@Pi Han Goh Through this definition of digamma -

ψ ( n + 1 ) + γ = 0 1 1 x n 1 x d x \displaystyle \psi(n+1) +\gamma = \int_{0}^{1}{\frac{1-x^n}{1-x} \ dx} or what you call a harmonic number.

Kartik Sharma - 5 years, 8 months ago

These kinds of problems are obviously easy for you. I always refrain from using higher topics of calculus like Digamma and all as I myself know nothing about them. By the way, I'm designing a set of correlated problems on Calculus for my 400 followers celebration.. hope you'll like them :D

Satyajit Mohanty - 5 years, 8 months ago

Log in to reply

Could you tell me what is the answer for n = 2 with (2.40411380630063+) related to π \pi ? Probably no similar relation.

Lu Chee Ket - 5 years, 7 months ago

Log in to reply

@Lu Chee Ket When n = 2 n=2 , we arrive at 2 ζ ( 3 ) 2 \zeta(3) but since we don't know the exact values of ζ ( odd numbers ) \zeta(\text{odd numbers}) in form of π , e , \pi,e, etc., thus we can only show the answer in form of ζ \zeta function or approximated decimals.

Satyajit Mohanty - 5 years, 7 months ago

Log in to reply

@Satyajit Mohanty Thanks very much! I am trying to find if i = 1 1 i 6 \sum_{i=1}^\infty \frac{1}{i^6} is related to π 6 \displaystyle \pi^6 . Seem only up to π 4 \displaystyle \pi^4 .

https://en.wikipedia.org/wiki/Particular values of Riemann zeta function#Even positive_integers

Lu Chee Ket - 5 years, 7 months ago

Log in to reply

@Lu Chee Ket As of now, mathematicians are only able to find the exact form of ζ ( n ) \zeta(n) where n n is a positive even number.

Pi Han Goh - 5 years, 7 months ago

Log in to reply

@Pi Han Goh n ! r = 0 1 ( r + 1 ) n + 1 n! \sum_{r=0}^\infty \dfrac{1}{(r+1)^{n+1}} is the exact sum for all but no simple expression generally; I read a writing and find that only π 2 6 \frac {\pi^2}{6} and π 4 90 \frac {\pi^4}{90} are available. For odd n = 3, people just evaluate it up to many significant figures.

Lu Chee Ket - 5 years, 7 months ago

Log in to reply

@Lu Chee Ket That's what I said. As of now, there's no exact form for ζ ( 3 ) \zeta(3) . Please familiarize yourself with Riemann zeta function .

Pi Han Goh - 5 years, 7 months ago

Log in to reply

@Pi Han Goh I have already read the website you introduced when I discussed with Satyajit Mohanty. The question is I don't find exact form for n = 6, 8 and 10 neither.

Lu Chee Ket - 5 years, 7 months ago

Log in to reply

@Lu Chee Ket Here .

Pi Han Goh - 5 years, 7 months ago

Log in to reply

@Pi Han Goh Thanks a lot! He didn't tell me these as I only told him that I am searching for these.

Lu Chee Ket - 5 years, 7 months ago
Lu Chee Ket
Nov 8, 2015

The problem is set as A B π C \dfrac{A}{B} \pi^C to satisfy Brilliant's methods of answering. In General, such problems would come directly. Your method is not accurate. It's just a guess which fortunately worked well with this problem.

Satyajit Mohanty - 5 years, 7 months ago

Log in to reply

You may appreciate how a correct guess can be arrived with plenty of of observations from sufficient extends. 6.49351257882126152+ with extreme 1,133,000 is continuing in my evaluation. Yet, I also described as followed.

Lu Chee Ket - 5 years, 7 months ago

Log in to reply

Still, this is a wrong method. Like most (or all?) of your other solutions, you have arrived at the right answer using the wrong method. The question is not phrased like "This summation appears to be approximately equal to A B π C \frac AB \pi^C for integers A , B , C A,B,C . What is the most likely reasonable values of A , B , C A,B,C ? Submit your answer as A + B + C A+B+C ."

Pi Han Goh - 5 years, 7 months ago

Log in to reply

@Pi Han Goh Brilliant means having ways particularly alternatives even when we are in a situation of helpless resources. Not because of knowledge but the Tao that makes these sensible. The evidence is almost all correct for guesses attempted. If not once then only because of laziness.

Lu Chee Ket - 5 years, 7 months ago

i = 1 1 i ( i ) = π 2 6 , f o r \displaystyle \sum_{i=1}^\infty \frac {1}{i (i)} = \frac {\pi^2}{6}, for n = 1 n = 1

i = 1 1 i 2 + 2 = π 4 90 , f o r \displaystyle \sum_{i=1}^\infty \frac {1}{i^{2+2}} = \frac {\pi^4}{90}, for n = 3 n = 3

i = 1 j = 1 k = 1 1 i j k ( i + j + k ) = 6.493 + , f o r \displaystyle \sum_{i=1}^\infty \displaystyle \sum_{j=1}^\infty \displaystyle \sum_{k=1}^\infty \frac {1}{i j k (i + j + k)} = 6.493+, for n = 3 n = 3

Having found that ( 6.493 + ) / π 4 90 6 , (6.493+)/ \frac {\pi^4}{90} \rightarrow 6, I guess exact relation is determined.

Whether we believe a relation of ( π 2 ) 2 (\pi^2)^2 of π 4 \pi^4 is crucial. Once we believe in this way, I think an exact relation of 6 times to very good approximation to 6.493+ can help to determine an exact answer without knowing the mathematics concerned! So far, 6.49347873699081405+ for extreme of 1,038,000 has been obtained from real evaluation. 6 × π 4 90 = π 4 15 \displaystyle 6 \times \frac {\pi^4}{90} = \frac {\pi^4}{15} is therefore possible to be determined once we believe we can with fortunate trials!

Lu Chee Ket - 5 years, 7 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...