Is it true that if and are bounded subsets of the real line, then can be decomposed into pairwise disjoint translated copies of in at most one way? (We allow for infinitely many translated copies.)
Problem 1268
Official solution
I. Solution. We will show that the questioned conclusion is not true. We will recursively construct the sets , and the translation sets (all non-empty subsets of ), such that
and all the translations and are pairwise disjoint.
First, let , and . From here, we proceed recursively. Suppose we have already constructed the finite sets such that , and and cover each element they cover exactly once (i.e., the sets are pairwise disjoint, and the sets are also pairwise disjoint). Let . Now, for each , we do the following: if has not yet been in , we add an element to and an element to such that their sum is exactly , and to ensure that the previous properties do not break, should not be of the form (where these are previous elements: ), and should not be of the form (where ), and the new should not be in . Each of these conditions represents a finite number of forbidden elements. We proceed similarly for . Thus, we obtain the sets , clearly , .
It is clear that satisfy the conditions, provided that the boundedness also holds. However, we can easily ensure the boundedness by choosing each new element from the interval as follows: in a typical step, we want to write a given in the form , where , and (or ). It is clear that since only a finite number of forbidden elements exist, there exist suitable numbers.
II. Solution (based on the work of Matolcsi Dávid). Let be the set of rational numbers with 3-power denominators that fall into the open interval . Let be the set of numbers of the form , where is a non-negative integer. We will show that can be decomposed into pairwise disjoint translates of in more than one way.
Let be the set of rational numbers with 3-power denominators that fall into the closed interval . Then (in fact, equality holds). Since and both contain the number , these two translates of are not disjoint. It suffices to show that both can be extended to a decomposition of into pairwise disjoint translates of . Since is countable, it suffices to show that if the translates for some do not contain the number , then there exists a number such that the translate contains the number and is disjoint from each of .
Choose an integer large enough so that is an integer for all , and . Then , so we can choose a sign such that with and , we have , i.e., . Then .
We only need to show that for and , , i.e., , or . Since , we have , so we are done if or , since in these cases . (In the latter case, we use the fact that the set is symmetric about 0.) In other cases, we show that is not an integer, from which the desired inequality immediately follows.
Let and , then , where is an integer divisible by 9, and is not, because if then it is either or not an integer, and if then it is either or not divisible by 3. Thus, cannot be an integer divisible by 9.