A real number is given. The sequence consists of all the positive integral such that . Prove that there are at most three different numbers among the numbers , , , .
[i]A corollary of a theorem from ergodic theory[/i]
A real number is given. The sequence consists of all the positive integral such that . Prove that there are at most three different numbers among the numbers , , , .
[i]A corollary of a theorem from ergodic theory[/i]
1. Consider the sequence and the fractional part:
Let be a real number. The sequence consists of all positive integers such that . Here, denotes the fractional part of .
2. **Reduction to interval :**
We can assume without loss of generality because adding an integer to does not change the fractional parts .
3. Interpretation on a unit circle:
Imagine the numbers as points on a circle of unit length. We lay off arcs of length successively from point 0. We are interested in the arrangement of points on a given arc of length .
4. Define shifts:
- A shift to the right occurs if .
- A shift to the left occurs if .
- The magnitude of the shift is .
- The number of steps of the shift is .
5. Consecutive shifts in the same direction:
Two consecutive shifts in the same direction are always equal. For example, if , then and , implying and .
6. Define smallest shifts:
- Let be the smallest number of steps for a right shift .
- Let be the smallest number of steps for a left shift .
- Assume .
7. Lemma 1:
If and for some , lies in the interval , then there exists a shift of with number of steps .
8. Lemma 2:
There are no shifts with a number of steps no more than , except for the three (or possibly two) mentioned above.
Proof of Lemma 2:
- Suppose there exists a shift of with number of steps .
- If it is a shift to the right, then . Find such that . Then . Since cannot be less than , the transition from to contains a shift to the left and less than steps, which is impossible.
- If the shift by is a shift to the left, then and . Taking such that , we obtain that the transition from to contains a shift to the right and less than steps, which is again a contradiction.
9. Proof of Lemma 1:
- Take a natural for which lies in the half-interval .
- Then .
- At no natural does the fractional part lie in the interval —at and by the assumption for the number , and at the other by Lemma 2, which has already been proved.
10. Conclusion:
The statement of the problem follows from Lemmas 1 and 2. If and for some the fractional part lies in the interval , then:
- From points with fractional parts smaller than , we have a shift by .
- From points with fractional parts greater than , we have a shift by .
- The other points are shifted by (both shifts with fewer steps are impossible).
If no such exists, there are no points of the last type.
We have proved that if the differences take three different values, then one of these values is equal to the sum of the other two.