Three positive integers $a,b,c$ have the property that $b > 2a$ and $c > 2b$.
Is it always possible to find a real number $r$, such that each fractional part of the numbers $r\cdot a$, $r\cdot b$ and $r\cdot c$ lies within the interval $(\frac{1}{3}, \frac{2}{3}]$ ?

  • 1
    $\begingroup$ This might be an acceptable question on this site, but why not posting it on a more suitable site, such as the mathematics site? $\endgroup$
    – WhatsUp
    Nov 12, 2021 at 14:24
  • $\begingroup$ @WhatsUp, right, it is sometimes hard to decide which site fits best for such questions. I keep it here and see, if somebody wants to give it a try. $\endgroup$
    – ThomasL
    Nov 12, 2021 at 19:09

1 Answer 1


Yes, it is possible.


Let us think of r as a time, the fractional parts of the other variables can then be understood as a circular track of unit length on which a,b,c travel each at their own, constant speed. Two thirds of the track are "neutral" and one third is the target.

Because b travels at more than twice the speed of a it will move more than double the distance in the same time. In particular, while a traverses the target, b will cover more ground than fits in the neutral zone. In other words b will for some of the time be in the target, too.

We can make the same argument w.r.t. b and c, but that does not directly imply that all three will be in the target at the same time. To demonstrate this it will suffice to show that (*) during some of b's cycles a will be in the target for the entire time b is. There are two cases:

Case 1) b is a multiple of a: if b/a>=4 (*) is clearly satisfied because one passage of a through the target translates to at least 4/3 cycles for b which are guaranteed to contain at least one contigouous sweep through the target. But if b/a=3 then we can directly verify that a's very first passage trough the target fully contains one of b's.

Case 2) b is not a multiple of a. Then because their ratio is rational, we will encounter a finite set of phase shifts (i.e. offsets of b's cycles relative to the beginnings of each of a's cycles) which will be of the form 1/n 2/n 3/n etc. Each offset of b contains a shifted period of length >2/3 (at a units) when b is in the target zone. a's target of width 1/3 will be fully inside for a stretch of offsets of width >1/3. If n>=3 these must cover the entire loop. If n=2 the ratio b/a must be 5/2 or more. Now each offset of b contains a target zone of length >=5/6. Therefore a stretch of length >=1/2 of offsets will contain the entire a target. As n=2 this is just enough to cover the full loop.

  • $\begingroup$ Is there any way to make Case 2 more rigorous? I can see in my mind's eye how the phase shift for b/a >2 will guarantee (*) but I always like to be able to back these things up with an algebraic argument. +1 from me for the formulation of (*) $\endgroup$
    – hexomino
    Nov 14, 2021 at 17:18
  • $\begingroup$ @hexomino I was a bit sloppy with that part of the argument, hope it is better now. $\endgroup$
    – loopy walt
    Nov 14, 2021 at 18:11
  • $\begingroup$ Nice solution using speed travel. $\endgroup$
    – ThomasL
    Nov 15, 2021 at 21:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.