7
$\begingroup$

Can every rational number be represented as a finite sum of reciprocal numbers? You are only allowed to use each reciprocal number one time per expression (So for example 3/2 cannot be 1/2+1/2+1/2).

You could then express these numbers in a kind of "binary-reciprocal" where a 1 in the nth place from the right denotes adding 1/n. E.g 17/10 could be 10011 as it is 1+1/2+1/5.

Thanks in advance!

$\endgroup$
7
  • 3
    $\begingroup$ Googling "Egyptian fractions" should turn out to be quite fruitful for you. $\endgroup$
    – Wojowu
    Commented Sep 10, 2016 at 16:57
  • $\begingroup$ Clearly irrational numbers cannot be written as a sum of fractions. $\endgroup$
    – Arthur
    Commented Sep 10, 2016 at 16:59
  • $\begingroup$ Ok. Thanks. I cleared that up. Cleary I meant rationals. $\endgroup$ Commented Sep 10, 2016 at 17:00
  • $\begingroup$ The thing about Egyptian fractions is that they may not be distinct and if I have a number like 100 I must not use more than a single 1. Meaning that it is not obvious to me that Egyptian fractions always work. $\endgroup$ Commented Sep 10, 2016 at 17:03
  • $\begingroup$ I moved that "rational" clarification to the beginning of the Question. Rollback the edit, mtheorylord, if I unintentionally changed your meaning. $\endgroup$
    – hardmath
    Commented Sep 10, 2016 at 17:05

2 Answers 2

9
$\begingroup$

Yes, and there is a simple algorithm to achieve this. First, we subtract successive reciprocal numbers $(1/1,\ 1/2,\ 1/3,\ \ldots)$ until the remainder is less than the next reciprocal number. This is always possible because the harmonic series is divergent.

If the remainder is zero then we are done, so let us assume that $$0 < \frac{a}{b} < 1.$$ There exists a unique positive integer $m$, greater than any of the preceding denominators, so that $$\frac{1}{m} \le \frac{a}{b} < \frac{1}{m-1}.$$ Therefore, $$\frac{a}{b} - \frac{1}{m} = \frac{am-b}{bm} \ge 0.$$

But $$\frac{a}{b} < \frac{1}{m-1} \implies am-a < b \implies am-b < a,$$ so the numerator has decreased.

If the process is repeated, then the remainder will eventually be zero, since the numerators cannot decrease forever. Therefore, every positive rational number can be represented as a sum of distinct reciprocal numbers.

Example: To write 11/4 as a sum of distinct unit fractions, we subtract as many successive unit fractions as possible, starting with 1.

$$\frac{11}{4} - \frac11 - \frac12 - \frac13 - \frac14 - \frac15 - \frac16 - \frac17 - \frac18 = \frac9{280}.$$

The remainder lies between $\frac1{32}$ and $\frac1{31}$, so we subtract $\frac1{32}$.

$$\frac{9}{280} - \frac{1}{32} = \frac{1}{1120}.$$

Since the result is a unit fraction, we are done. $$\frac{11}4 = \frac11 + \frac12 + \frac13 + \frac14 + \frac15 + \frac16 + \frac17 + \frac18 + \frac1{32} + \frac1{1120}.$$ Thanks to Barry Cipra for pointing out an error in a previous version of this post.

$\endgroup$
2
  • $\begingroup$ I don't see how you are guaranteed there'll be an $m$ available after you've gotten the number down to less than $1$. For example, suppose you start at $11/4$. Then ${a\over b}={11\over4}-1-{1\over2}-{1\over3}={11\over12}$. You've already used ${1\over 2}$, but that's the unique reciprocal sandwich for ${1\over m}\le{11\over12}\lt{1\over m-1}$. $\endgroup$ Commented Sep 10, 2016 at 21:51
  • 1
    $\begingroup$ You are correct. I believe that it works if we subtract successive reciprocals as long as we can until the difference is less than the next reciprocal. $\endgroup$ Commented Sep 10, 2016 at 23:15
5
$\begingroup$

Every positive number is a subseries of the harmonic series (maybe infinite). You do the following: if $x$ is your number - add numbers till the next step is over $x$, Then don't add the following till adding again will give you something less or equal. Proceed in this manner.

For positive rationals there is a method to achive them in a finite manner - using the following replacement $1/k+1/k=1/k+1/(k+1)+1/k(k+1)$ - Botts, Truman (1967), "A chain reaction process in number theory", Mathematics Magazine, 40 (2): 55–65

$\endgroup$
11
  • $\begingroup$ But can you prove this process isn't infinite? $\endgroup$ Commented Sep 10, 2016 at 17:04
  • $\begingroup$ It can be infinite... I wrote it in the brackets $\endgroup$ Commented Sep 10, 2016 at 17:06
  • $\begingroup$ But I asked for a finite representation. Cleary if its infinite you can use a variant of Riemann's Rearrangement like you did. $\endgroup$ Commented Sep 10, 2016 at 17:07
  • $\begingroup$ You can manipulate my solution in a way that will work in a finite manner for rationals $\endgroup$ Commented Sep 10, 2016 at 17:09
  • 1
    $\begingroup$ Sure the denominators can get big, but the numerators start shrinking. See Dave Radcliffe's answer for a proof. $\endgroup$
    – Nate
    Commented Sep 10, 2016 at 19:06

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .