Let $q = \frac ab$ be any rational number such that $a < b$. What is the smallest positive integer $n$ such that $\frac ab \times \left(2^n-1\right)$ is an integer?
-
1By definition, that the order of $2$ $ \pmod b$. Note that if $b$ is even, then there is no such $n$. – lulu Jun 16 '19 at 13:37
-
1If $a,b$ have no common divisor, then search for the smallest $n$ s.t. $b$ divides $2^n-1$. – Wuestenfux Jun 16 '19 at 13:40
-
Is there an expression that might (approximately) relate $n$ to $b$? – jII Jun 16 '19 at 13:45
-
1There is no easy way to compute $n$, unfortunately. General theory tells us that $n$ is a divisor of $\varphi(b)$, where $\varphi$ denotes Euler's totient function In particular, $\varphi(b)$ always works as an exponent $n$, but it won't, in general, be minimal. – lulu Jun 16 '19 at 13:48
-
Is there a particular reason for the binary tag? Are you trying to write a program for this? – Simply Beautiful Art Jun 16 '19 at 14:03
-
@SimplyBeautifulArt Re: binary, see the first sentence in my answer. – Bill Dubuque Jun 16 '19 at 14:55
-
I'm not particularly interested in a small tangent in an answer. I'm asking the OP what interest they have in putting up the binary tag, since it may affect answers. – Simply Beautiful Art Jun 16 '19 at 15:01
-
@SimplyBeautifulArt Yuur response is quite puzzling. Do you seriously think that the reaosn is other than I have explained? If so please do explain since that would be rather surprising. Btw, you need to use "
user" if you wish the user to be notified of your reply (I noticed it only by chance). – Bill Dubuque Jun 16 '19 at 15:16 -
I'm not saying it probably isn't, but I'm not going to assume the reasoning behind the OP's tagging. That is why I am asking them. Also for some reason mobile is being buggy for me on the @. – Simply Beautiful Art Jun 16 '19 at 15:20
-
1@Sim But it seemed you were by calling the question's raison d'etre a "small tangent" . That's why I was puzzled by your remark. – Bill Dubuque Jun 16 '19 at 15:26
-
The majority of your answer is not specific to binary, hence my remark. – Simply Beautiful Art Jun 16 '19 at 15:29
-
2@SimplyBeautifulArt and Bill Dubuque, thanks for the responses---I used the "binary" tag when posting the question for the same connection identified in Bill's answer. – jII Jun 16 '19 at 16:42
1 Answers
It's simply the binary analog of computing the decimal period of a fraction.
W.l.o.g. generality we may assume that $\,a/b\,$ is reduced, i.e. $\,d = \gcd(a,b) = 1\ $ (else cancel $d).$
Then $\, (2^{\large n}-1)a/b = k\in\Bbb Z\iff bk = (2^{\large n}-1)a\ $ so $\,b\mid a(2^{\large n}-1).\,$ Since $\,\gcd(b,a) = 1\,$ we infer by Euclid's Lemma that $\,b\mid 2^{\large n}-1,\,$ i.e. $\,2^{\large n}\equiv 1\pmod{\!b}.\,$
Since $\,2^{\large n}-1\,$ is odd its factor $b$ must be odd, and it is easy to show that such an $n$ must exist, e.g. by a pigeonhole argument, or by Euler's phi theorem (or Lagrange) we can choose $\,n = \phi(b).$
The least such $\,n > 0\,$ is known as the order of $\,2\,$ modulo $\,b.\,$ It can be verified using the Order Test and computed by various algorithms.

- 272,048