A Putnam Problem

Calculus Level 5

k = 1 n = 0 ( 1 ) k 1 k ( 1 + 2 n k ) \large \sum_{k = 1}^{\infty} \sum_{n = 0}^{\infty} \frac{(-1)^{k - 1}}{k(1 + 2^nk)}

Find the value (to 2 decimal places) of the closed form of the expression above.


Source: This is problem B6 of the Putnam 2016 competition.


The answer is 1.00.

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.

4 solutions

Mark Hennings
Dec 7, 2016

Let A = k = 1 n = 0 1 k ( 2 n k + 1 ) B = k = 1 n = 0 1 2 k ( 2 n + 1 k + 1 ) A \; = \; \sum_{k=1}^\infty \sum_{n=0}^\infty \frac{1}{k(2^nk+1)} \hspace{2cm} B \; = \; \sum_{k=1}^\infty \sum_{n=0}^\infty \frac{1}{2k(2^{n+1}k+1)} Note that while A A is the sum of 1 k ( 2 n k + 1 ) \tfrac{1}{k(2^nk+1)} over all n 0 n \ge 0 and k 1 k \ge 1 , the term B B is the sum of 1 k ( 2 n k + 1 ) \tfrac{1}{k(2^nk+1)} over all n 0 n \ge 0 and even k 1 k \ge 1 . Now B = 1 2 k = 1 n = 1 1 k ( 2 n k + 1 ) = 1 2 [ k = 1 n = 0 1 k ( 2 n k + 1 ) 1 k ( k + 1 ) ] = 1 2 ( A 1 ) \begin{aligned} B &= \tfrac12\sum_{k=1}^\infty \sum_{n=1}^\infty \frac{1}{k(2^nk+1)} \\ &= \tfrac12\left[\sum_{k=1}^\infty \sum_{n=0}^\infty \frac{1}{k(2^nk+1)} - \frac{1}{k(k+1)}\right] \\ &= \tfrac12(A-1) \end{aligned} But the sum we are interested in is k = 1 n = 0 ( 1 ) k 1 k ( 2 n k + 1 ) = A 2 B = 1 \sum_{k=1}^\infty \sum_{n=0}^\infty \frac{(-1)^{k-1}}{k(2^nk+1)} \; = \; A - 2B \; = \; \boxed{1}

Oh wow, that's a really innovative approach. What made you think of A A and B B , and to find the identity between them?

Calvin Lin Staff - 4 years, 6 months ago

Log in to reply

Writing an alternating sum as "the sum of the lot, minus twice the sum over the even-indexed terms" is a fairly standard ploy. B B is the sum over all even k k .

Mark Hennings - 4 years, 6 months ago

Log in to reply

Ah, it didn't occur to me immediately that B was the "negative portion". Could you add this line to the top of the solution to provide the motication of what you did? Thanks!

Calvin Lin Staff - 4 years, 6 months ago

Note that the sum (say S S ) converges absolutely as 1 k ( 1 + k 2 n ) < 1 k 2 2 n \dfrac{1}{k(1 + k2^n)} < \dfrac{1}{k^22^n} . Hence, we can interchange the order of summations. Interchanging and writing 1 1 + k 2 n \dfrac{1}{1+k2^n} as 0 1 x k 2 n d x \displaystyle \int_0^1 x^{k2^n} \, \text{d}x , we get the following :

S = n = 0 k = 1 ( 1 ) k 1 k 0 1 x k 2 n d x = n = 0 0 1 ln ( 1 + x 2 n ) d x ( by uniform convergence of power series ) = 0 1 ln ( 1 1 x ) d x ( by Lebesgue’s monotone convergence theorem ) = 0 1 ln x d x = 1 \begin{aligned} S &= \sum_{n = 0}^{\infty} \sum_{k = 1}^{\infty} \frac{(-1)^{k - 1}}{k}\int_0^1 x^{k2^n} \, \text{d}x \\ &= \sum_{n = 0}^{\infty} \int_0^1 \ln (1 + x^{2^n}) \, \text{d}x \quad (\text{by uniform convergence of power series})\\ &= \int_0^1 \ln \left( \frac{1}{1 - x} \right) \, \text{d}x \quad (\text{by Lebesgue's monotone convergence theorem}) \\ &= - \int_0^1 \ln x \, \text{d}x \\ &= \boxed{1} \end{aligned}

Note: The following equation is also used to get the third equality : n = 0 N 1 ( 1 + x 2 n ) = 1 x 2 N 1 x \prod_{n = 0}^{N - 1} (1 + x^{2^n}) = \frac{1 - x^{2^N}}{1 - x}

You say that the first identity is due to the uniform convergence of power series. Yes and no - you need to be more explicit. While it is true that power series converge uniformly on compact subsets of their (open) interval of convergence (i.e for x S |x| \le S for any S < R S < R , the radius of convergence), we cannot use this result here, since we are integrating on [ 0 , 1 ] [0,1] , where R = 1 R=1 .

We have to use the alternating nature of this series, plus the fact that the terms in the sequence have decreasing modulus, so that ln ( 1 + x 2 n ) k = 1 K ( 1 ) k 1 k x k 2 n 1 k + 1 x ( k + 1 ) 2 n 1 k + 1 0 x 1 \left| \ln(1+x^{2^n}) - \sum_{k=1}^K \frac{(-1)^{k-1}}{k} x^{k2^n}\right| \; \le \; \frac{1}{k+1}x^{(k+1)2^n} \; \le \; \frac{1}{k+1} \hspace{2cm} 0 \le x \le 1 to show that the series converges uniformly on [ 0 , 1 ] [0,1] .

Mark Hennings - 4 years, 6 months ago

Log in to reply

(Sorry for the really late reply)

We could interchange the sum and integral over [ 0 , r ] [0,r] for r < 1 r < 1 and then use the continuity of F ( x ) = 0 x f ( t ) d t F(x) = \int_0^x f(t) \, \text{d}t to conclude that the same can be done over [ 0 , 1 ] [0,1] right?

P.S.: Mostly this is wrong, but I want to confirm.

A Former Brilliant Member - 4 years, 6 months ago

Log in to reply

That should work, but it's a bit fiddly...

Mark Hennings - 4 years, 6 months ago

Log in to reply

@Mark Hennings Hmm... Okay.

Thanks for the reply.

A Former Brilliant Member - 4 years, 6 months ago

Thanks for justifying all of those steps. At the Putnam, you have to explain and justify your work carefully.

Calvin Lin Staff - 4 years, 6 months ago
Ishan Singh
Dec 7, 2016

Let S = k = 1 n = 0 ( 1 ) k 1 k ( k 2 n + 1 ) \displaystyle \text{S} = \sum_{k=1}^{\infty} \sum_{n=0}^{\infty} \dfrac{(-1)^{k-1}}{k(k\cdot2^n+1)}

Since S \text{S} is absolutely convergent, we can swap summations to get,

S = n = 0 k = 1 ( 1 ) k 1 k ( k 2 n + 1 ) \displaystyle \text{S} = \sum_{n=0}^{\infty} \sum_{k=1}^{\infty} \dfrac{(-1)^{k-1}}{k(k\cdot2^n+1)}

= n = 0 k = 1 [ ( 1 ) k 1 k ( 1 ) k 1 k + 1 2 n ] = \sum_{n=0}^{\infty} \sum_{k=1}^{\infty} \left[\dfrac{(-1)^{k-1}}{k} - \dfrac{(-1)^{k-1}}{k+\frac{1}{2^n}}\right]

Splitting into even and odd terms, we have,

= n = 0 k = 1 [ ( 1 ) k 1 k ( 1 k + 1 2 n 1 k + 1 2 n + 1 ) ] = \sum_{n=0}^{\infty} \sum_{k=1}^{\infty}\left[\dfrac{(-1)^{k-1}}{k} - \left(\dfrac{1}{k+\frac{1}{2^n}} - \dfrac{1}{k+\frac{1}{2^{n+1}}} \right)\right]

Converting to Digamma Function , we have,

= n = 0 [ log 2 + ψ ( 1 2 n ) ψ ( 1 2 n + 1 ) 2 n ] = \sum_{n=0}^{\infty} \left[\log 2 + \psi \left(\dfrac{1}{2^n} \right) - \psi \left( \dfrac{1} {2^{n+1}} \right) - 2^{n}\right]

= lim N n = 0 N [ log 2 + ψ ( 1 2 n ) ψ ( 1 2 n + 1 ) 2 n ] ( ) = \lim_{N \to \infty}\sum_{n=0}^{N} \left[\log 2 + \psi \left(\dfrac{1}{2^n} \right) - \psi \left( \dfrac{1} {2^{n+1}} \right) - 2^{n}\right] \quad \quad (*)

Clearly, ( ) (*) telescopes. Evaluating it, we have,

S = 1 \text{S} = \boxed{1}

After splitting, shouldn't the exponent of 2 in the end be n + 1 n+1 rather than n 1 n - 1 ?

A Former Brilliant Member - 4 years, 6 months ago

Log in to reply

Typo, corrected.

Ishan Singh - 4 years, 6 months ago

Let S0 and S1 be the sums ∑k 1 k ∑ ∞ n=0 1 k2 n+1 with k running over all odd and all even positive integers, respectively, so that S = S1 −S0 = (S0 +S1)−2S0. In S1, we may write k = 2 to obtain S1 = ∞ ∑ =1 1 2 ∞ ∑ n=0 1 2

n+1 +1

1 2 (S0 +S1) + ∞ ∑ `=1 1

2 ( +1)

1 2 (S0 +S1) + 1 2 because the last sum telescopes; this immediately yields S = 1.

Nice solution.I solved it the harder way.

A Former Brilliant Member - 4 years, 3 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...