Room isn't enough for 2

Calculus Level 1

Consider the sum of the reciprocals of every positive integer that doesn't contain the digit 2:

1 1 + 1 3 + 1 4 + + 1 10 + 1 11 + 1 13 + + 1 19 + 1 30 + 1 31 . \frac{1}{1} + \frac{1}{3} + \frac 14 + \cdots + \frac{1}{10} + \frac{1}{11} + \frac{1}{13} +\cdots+ \frac{1}{19} + \frac{1}{30} + \frac 1{31} \cdots.

Does this sum converge?

No, it diverges Yes, it converges

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

Efren Medallo
Aug 2, 2017

Let us divide the sum S S into a different sub-series we'll call b n b_n , such that b n b_n is the sum of all elements of S S of denominator k k for 1 0 n 1 k < 1 0 n 10^{n-1} \leq k < 10^n .

So from this, we can get that

b 1 = 1 + 1 3 + 1 4 + . . . + 1 9 b_1 = 1 + \frac{1}{3} + \frac{1}{4} + ... + \frac{1}{9} b 2 = 1 10 + 1 11 + . . . + 1 99 b_2 = \frac{1}{10} + \frac{1}{11} + ... + \frac{1}{99}

and so on.

Now we investigate the number of terms in each of b n b_n .

It will be easy to note that b 1 b_1 has less than 9 9 terms, or that b 1 < 9 b_1 <9 .

Similarly, we can also see that b 2 b_2 has less than 9 2 9^2 terms, and that b 2 < 9 2 10 b_2 < \frac{9^2}{10} .

In general, b n b_n will have less than 9 n 9^n terms, and b n < 9 n 1 0 n 1 b_n < \frac{9^n}{10^{n-1}} .

From here we can deduce that

S < 9 + 9 2 10 + 9 3 1 0 2 + . . . S < 9 + \frac{9^2}{10} + \frac{9^3}{10^2} + ...

with which the RHS depicts a geometric series with common ratio 9 10 \frac{9}{10} .

So, we can say that S < 90 S < 90 . Thus, the series is convergent .

*Notes:

  • S 90 S \neq 90 .

  • Fun fact : This also works for any series of the same format, but with any other digit removed (say, 6 6 , or 7 7 ).

Why doesn't this trick work when we are only removing those entries with the digit 0?

Agnishom Chattopadhyay - 3 years, 10 months ago

Log in to reply

The series with the digit 0 removed will still converge (to 23.1 \approx 23.1 ).

I'm not sure what the "fun fact" is referring to. We could prove it in a similar manner, but be careful with counting the number of terms.

Calvin Lin Staff - 3 years, 10 months ago

Oh! I'm sorry. Calvin's right; it still does converge with all those containing 0 0 removed. I somehow didn't remove them all while doing a similar pattern as above. Let me edit that.

Efren Medallo - 3 years, 10 months ago

Log in to reply

What if we were to remove a specific string of digits? Would it still converge if we were not allowed to use integers which contain (say) 314159?

Calvin Lin Staff - 3 years, 10 months ago

Log in to reply

@Calvin Lin It still will. In such case, we are bound to consider that b n < 1 0 k n b_n < 10^{kn} terms, where k k is the number of digits of the substring to be removed in the series.

Efren Medallo - 3 years, 10 months ago

Log in to reply

@Efren Medallo That's the right idea, but it think you have a typo in the inequality. Note that for k = 1 k=1 (and the digit is non-zero), we have b n < 9 n |b_n| < 9^n . We need lim b n + 1 b n < 10 \lim \frac{ b_{n+1} } { b_n } < 10 in order for this simple counting argument to work directly.

Calvin Lin Staff - 3 years, 10 months ago
Ivan Koswara
Aug 2, 2017

Fix a positive integer d d , and consider all k k satisfying the given condition having exactly d d digits. There are 8 9 d 1 8 \cdot 9^{d-1} such k k (the first digit of k k has 8 choices: all except 0 and 2; the other digits have 9 choices: anything except 2). All these k k , being having d d digits, must be at least 1 0 d 1 10^{d-1} . Thus all these k k contribute at most 8 9 d 1 1 0 d 1 \dfrac{8 \cdot 9^{d-1}}{10^{d-1}} to the sum; each of these k k 's contribute 1 k 1 1 0 d 1 \frac{1}{k} \le \frac{1}{10^{d-1}} , and there are 8 9 d 1 8 \cdot 9^{d-1} of them. Summing over all d d , we have that the sum S S satisfies

S d = 1 8 9 d 1 1 0 d 1 = 8 d = 0 9 d 1 0 d = 8 10 = 80 \displaystyle\begin{aligned} S &\le \sum_{d=1}^\infty \frac{8 \cdot 9^{d-1}}{10^{d-1}} \\ &= 8 \cdot \sum_{d'=0}^\infty \frac{9^{d'}}{10^{d'}} \\ &= 8 \cdot 10 \\ &= 80 \end{aligned}

Thus S S has an upper bound. Since all terms of S S are increasing, by monotone convergence theorem S S converges .


This can be generalized to ruling out any finite sequence of decimal digits. For example, if we take out all terms having the sequence 2017 2017 , the series will also converge.

Neat proof!

Agnishom Chattopadhyay - 3 years, 10 months ago
Jon Haussmann
Aug 2, 2017

This is called a Kempner Series .

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...