Given a positive integer $n$, determine all functions $f$ from the first $n$ positive integers to the positive integers, satisfying the following two conditions: (1) $\sum_{k=1}^{n}{f(k)}=2n$; and (2) $\sum_{k\in K}{f(k)}=n$ for no subset $K$ of the first $n$ positive integers.
Problem
Source: Romania TST 2016 Day 3 P1
Tags: algebra, number theory
01.11.2017 07:06
I feel like I've seen this before on WOOT practice olympiads, with $n = 50$. Maybe I'm imagining things. Anyways, we just want to characterize the $n$-multisets $\{a_1, \dots, a_n\}$ summing to $2n$ but with no subset summing to $n$. If $n = 1$ then the only possible multiset is $\{2\}$, so consider $n > 1$. Let $s_k = a_1 + \dots + a_k$ for $k = 1, 2, \dots, n$. If two of them, say $s_i, s_j$ with $i < j$, share the same residue modulo $n$, then $a_{i + 1} + \dots + a_j$ is less than $2n$ and is a multiple of $n$, so it is $n$; contradiction. Thus $s_1, \dots, s_n$ form a complete residue system modulo $n$. However, $s_1, \dots, s_n$ must form a complete residue system modulo $n$ for any arrangement of $a_1, \dots, a_n$. Then, it's easy to show that $a_1 \equiv \dots \equiv a_n \equiv r \pmod{n}$ for some $r \in \{0, \dots, n - 1\}$. It's easy to see $r = 0$ is impossible for $n > 1$. Now we consider some cases: If $r = 1$, then $a_1, \dots, a_n \in \{1, n + 1\}$ with sum $2n$; hence, $\{a_1, \dots, a_n\} = \{\underbrace{1, \dots, 1}_{n - 1 \; 1\text{s}}, n + 1\}$. This works. If $r = 2$ and $n \ge 3$, then $a_1, \dots, a_n \ge 2$ with sum $2n$; hence $\{a_1, \dots, a_n\} = \{\underbrace{2, \dots, 2}_{n \; 2\text{s}}\}$. This works when $n$ is odd. If $r = 3$ and $n \ge 4$, then $a_1, \dots, a_n \ge 3$ with sum $2n$, which is impossible. Conclusion: the only such $f$ are the ones whose ranges are a permutation of $\underbrace{1, \dots, 1}_{n - 1 \; 1\text{s}}, n + 1$, or if $n$ is odd, a range of $\underbrace{2, \dots, 2}_{n \; 2\text{s}}$.
01.11.2017 16:42
CantonMathGuy wrote: However, $s_1, \dots, s_n$ must form a complete residue system modulo $n$ for any arrangement of $a_1, \dots, a_n$. Then, it's easy to show that $a_1 \equiv \dots \equiv a_n \equiv r \pmod{n}$ for some $r \in \{0, \dots, n - 1\}$. Can you eloborate , more please?
01.11.2017 16:49
This problem also appeared on Problems of Number Theory in Mathematical Competitions (Vol. 2 of Mathematical Olympiad Series) page 84-85.
01.11.2017 16:51
Can anyone explain why all of $a_i $'s are congruent modulo $n $?
02.11.2017 13:43
tenplusten wrote: Can anyone explain why all of $a_i $'s are congruent modulo $n $?
17.07.2021 19:45