Infinite Sum with Base Representations

Calculus Level 5

Let s ( n ) s(n) be the sum of the digits of n n 's base- 2017 2017 representation. Compute the sum n = 1 s ( n ) n ( n + 1 ) . \sum_{n = 1}^{\infty} \frac{s(n)}{n(n+1)}.

Round your answer to the nearest thousandth.


The answer is 7.613.

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

Alan Yan
Dec 12, 2017

Let p = 2017 p = 2017 . We will use the following lemmas:

Lemma 1 v p ( n ! ) = n s ( n ) p 1 v_p(n!) = \frac{n - s(n)}{p-1} Lemma 2 For s > 1 s > 1 , s s 1 > ζ ( s ) > 1 s 1 \frac{s}{s-1} > \zeta(s) > \frac{1}{s-1} The lemma follows from comparing ζ ( s ) \zeta(s) to the area under the curve f ( t ) = t s f(t) = t^{-s} , giving us 1 + 1 t s d t > ζ ( s ) > 1 t s . 1 + \int_1^\infty t^{-s} \, dt > \zeta(s) > \int_1^\infty t^{-s}. Corollary lim s 1 + ( s 1 ) ζ ( s ) = 1 \lim_{s \to 1^+} (s-1)\zeta(s) = 1 .

We come back to the main problem. Observe that n 1 s ( n ) n ( n + 1 ) = n 1 s ( n ) n s ( n ) n + 1 = n = 1 s ( n ) s ( n 1 ) n = n = 1 1 ( p 1 ) v p ( n ) n . \begin{aligned} \sum_{n \geq 1} \frac{s(n)}{n(n+1)} & = \sum_{n \geq 1} \frac{s(n)}{n} - \frac{s(n)}{n+1}\\ & = \sum_{n = 1}^\infty \frac{s(n) - s(n-1)}{n} \\ & = \sum_{n = 1}^\infty \frac{1 - (p-1)v_p(n)}{n}. \end{aligned}

We define the function F ( s ) = n = 1 1 ( p 1 ) v p ( n ) n s F(s) = \sum_{n = 1}^\infty \frac{1 - (p-1)v_p(n)}{n^s} for s > 1 s > 1 . This simplifies to F ( s ) = ζ ( s ) ( p 1 ) n = 1 v p ( n ) n s = ζ ( s ) ( p 1 ) k = 1 m = 1 1 ( p k m ) s = ζ ( s ) ( p 1 ) ζ ( s ) ( p s 1 p s ) = ζ ( s ) ( p s p p s 1 ) . \begin{aligned} F(s) & = \zeta(s) - (p-1)\sum_{n = 1}^\infty \frac{v_p(n)}{n^s} \\ & = \zeta(s) - (p-1) \sum_{k = 1}^\infty \sum_{m = 1}^\infty \frac{1}{(p^km)^s} \\ & = \zeta(s) - (p-1) \zeta(s) \left ( \frac{p^{-s}}{1 - p^{-s}} \right ) \\ & = \zeta(s) \left ( \frac{p^s - p}{p^s - 1} \right ). \end{aligned} Now, we take the limit as s 1 + s \to 1^+ , lim s 1 + F ( s ) = 1 p 1 lim s 1 + ζ ( s ) ( s 1 ) p s p s 1 = p p 1 lim s 1 + p s 1 1 s 1 = p p 1 log p \lim_{s \to 1^+} F(s) = \frac{1}{p-1}\lim_{s \to 1^+} \zeta(s)(s-1) \frac{p^s - p}{s-1} = \frac{p}{p-1} \lim_{s \to 1^+} \frac{p^{s-1} - 1}{s-1} = \frac{p}{p-1} \log p where the last equality follows from L'Hospital's Rule.

That's some significant math. I didn't follow all of it, but I was impressed.

I ended up using a Python program, but it was taking a looooonnnnnggg time to converge.

Steven Perkins - 3 years, 6 months ago
Mark Hennings
Dec 18, 2017

Let s p ( n ) s_p(n) be the sum of the digits of n n in the base p p representation of n n . Note that s p ( n ) = O ( ln n ) s_p(n) = O(\ln n) as m m \to \infty , and hence it is clear that the infinite serie S p = n 1 s p ( n ) n ( n + 1 ) S_p \; = \; \sum_{n \ge 1}\frac{s_p(n)}{n(n+1)} converges.

Let j p ( n ) j_p(n) be the index of p p in n n . It is easy to see that s p ( n ) s p ( n 1 ) = 1 ( p 1 ) j p ( n ) n 1 s_p(n) - s_p(n-1) \; = \; 1 - (p-1)j_p(n) \hspace{2cm} n \ge 1 and so, since n = 1 s p ( n ) n ( n + 1 ) = n 1 s ( n ) s ( n 1 ) n = n 1 1 ( p 1 ) j p ( n ) n \sum_{n=1}^\infty \frac{s_p(n)}{n(n+1)} \; = \; \sum_{n \ge 1} \frac{s(n) - s(n-1)}{n} \; = \; \sum_{n \ge 1} \frac{1 - (p-1)j_p(n)}{n} we see that n 1 n^{-1} gets added to the sum once for each n n , but ( p 1 ) n 1 (p-1)n^{-1} is subtracted from the sum once each time p p divides n n . Thus n = 1 p K s p ( n ) n ( n + 1 ) = n = 1 p K 1 n ( p 1 ) n = 1 p K 1 1 p n ( p 1 ) n = 1 p K 2 1 p 2 n ( p 1 ) n = 1 1 1 p K n = H p K ( p 1 ) j = 1 K H p K j p j \sum_{n=1}^{p^K} \frac{s_p(n)}{n(n+1)} \; = \; \sum_{n=1}^{p^K} \frac{1}{n} - (p-1)\sum_{n=1}^{p^{K-1}}\frac{1}{pn} - (p-1)\sum_{n=1}^{p^{K-2}}\frac{1}{p^2n} - \cdots - (p-1)\sum_{n=1}^1 \frac{1}{p^Kn} \; = \; H_{p^K} - (p-1)\sum_{j=1}^K\frac{H_{p^{K-j}}}{p^j} Now we use the fact that X n = H n ln n γ = O ( n 1 ) n X_n \; = \; H_n - \ln n - \gamma \; = \; O\big(n^{-1}\big) \hspace{2cm} n \to \infty and calculate ln ( p K ) ( p 1 ) j = 1 K ln ( p K j ) p j = p p 1 ( 1 p K ) ln p 1 ( p 1 ) j = 1 K p j = p K \begin{aligned} \ln(p^K) - (p-1)\sum_{j=1}^K \frac{\ln(p^{K-j})}{p^j} & = \; \tfrac{p}{p-1}(1 - p^{-K})\ln p \\ 1 - (p-1)\sum_{j=1}^Kp^{-j} & = \; p^{-K} \end{aligned} so that n = 1 p K s p ( n ) n ( n + 1 ) = p p 1 ( 1 p K ) ln p + γ p K + [ X p K ( p 1 ) j = 1 K X p K j p j ] \sum_{n=1}^{p^K} \frac{s_p(n)}{n(n+1)} \; = \; \tfrac{p}{p-1}(1 - p^{-K})\ln p + \gamma p^{-K} + \left[ X_{p^K} - (p-1)\sum_{j=1}^K\frac{X_{p^{K-j}}}{p^j}\right] Since X n A n 1 |X_n| \le An^{-1} for some constant A A , we deduce that n = 1 p K s p ( n ) n ( n + 1 ) p p 1 ln p [ p p 1 ln p + γ + A ] p K + ( p 1 ) j = 1 K A p K j × p j [ p p 1 ln p + γ + A + A K ( p 1 ) ] p K \begin{aligned} \left| \sum_{n=1}^{p^K} \frac{s_p(n)}{n(n+1)} - \tfrac{p}{p-1}\ln p\right| & \le \; \left[\tfrac{p}{p-1}\ln p + \gamma + A\right]p^{-K} + (p-1)\sum_{j=1}^K\frac{A}{p^{K-j}\times p^j} \\ & \le \; \left[\tfrac{p}{p-1}\ln p + \gamma + A + AK(p-1)\right]p^{-K} \end{aligned} for all integers K K . Thus we deduce that S p = p p 1 ln p S_p \; = \; \tfrac{p}{p-1}\ln p . With p = 2017 p=2017 , we obtain the answer 7.613141025 \boxed{7.613141025} .

I couldn't decide which solution I liked better, so I upvoted them both. This one is pretty much exactly what I did, except it's far more scrupulous about convergence issues; the OP's solution seems to avoid those issues by the nice trick of replacing n n by n s n^s and later taking the limit as s 1 + . s\to 1^+.

Patrick Corn - 3 years, 5 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...