Browse · MathNet
PrintInternational Mathematical Olympiad Shortlisted Problems
counting and probability
Problem
Let be a positive integer, and let be a subset of . An -partition of into parts is a representation of as a sum , where the parts belong to and are not necessarily distinct. The number of different parts in such a partition is the number of (distinct) elements in the set . We say that an -partition of into parts is optimal if there is no -partition of into parts with . Prove that any optimal -partition of contains at most different parts.

Solution
If there are no -partitions of , the result is vacuously true. Otherwise, let be the minimum number of parts in an -partition of , and let be an optimal partition. Denote by the number of different parts in this partition, so we can write for some pairwise different numbers in . If , we will prove that there exist subsets and of such that and . Then, deleting the elements of from our partition and adding the elements of to it, we obtain an -partition of into less than parts, which is the desired contradiction. For each positive integer , we consider the -element subset as well as the following -element subsets of : Pictorially, if we represent the elements of by a sequence of dots in increasing order, and represent a subset of by shading in the appropriate dots, we have: Denote by the sum of elements in . Clearly, is the minimum sum of a -element subset of . Next, for all appropriate indices and we have Therefore To see this in the picture, we start with the leftmost points marked. At each step, we look for the rightmost point which can move to the right, and move it one unit to the right. We continue until the rightmost points are marked. As we do this, the corresponding sums clearly increase. For each we have found different integers of the form between 1 and . As we vary , the total number of integers we are considering is Since they are between 1 and , at least two of these integers are equal. Consequently, there exist and as well as such that as required. The result follows.
---
Alternative solution.
Assume, to the contrary, that the statement is false, and choose the minimum number for which it fails. So there exists a set together with an optimal partition of refuting our statement, where, of course, is the minimum number of parts in an -partition of . Again, we define with ; by our assumption we have . Without loss of generality we assume that . Let us distinguish two cases.
Case 1. . Consider the partition , which is clearly a minimum -partition of with at least different parts. Now, from we obtain so , which contradicts the choice of .
Case 2. . Set , and for . There are such sums; so at least two of them, say and , are congruent modulo (where ). This means that for some integer . Notice that for we have so the indices and are distinct, and we may assume that . Next, we observe that and imply so . Thus, we may remove the terms of in our -partition, and replace them by the terms of and terms equal to , for a total of terms. The result is an -partition of into a smaller number of parts, a contradiction.
---
Alternative solution.
Assume, to the contrary, that the statement is false, and choose the minimum number for which it fails. So there exists a set together with an optimal partition of refuting our statement, where, of course, is the minimum number of parts in an -partition of . Again, we define with ; by our assumption we have . Without loss of generality we assume that . Let us distinguish two cases.
Case 1. . Consider the partition , which is clearly a minimum -partition of with at least different parts. Now, from we obtain so , which contradicts the choice of .
Case 2. . Set , and for . There are such sums; so at least two of them, say and , are congruent modulo (where ). This means that for some integer . Notice that for we have so the indices and are distinct, and we may assume that . Next, we observe that and imply so . Thus, we may remove the terms of in our -partition, and replace them by the terms of and terms equal to , for a total of terms. The result is an -partition of into a smaller number of parts, a contradiction.
Techniques
Pigeonhole principleColoring schemes, extremal arguments