A set of numbers is called sum-free if whenever and are (not necessarily distinct) elements of the set, is not an element of the set. For example, and the empty set are sum-free, but is not. What is the greatest possible number of elements in a sum-free subset of ?
- A)
- B)
- C)
- D)
- E)
Answer
C
Key insight
Look at the largest number in the set. Pair up the smaller numbers so each pair adds to it; the set can hold at most one number from each pair.
Solution
This is a "greatest possible" question, so it has two halves: find a set that works, then show nothing bigger can work.
Part 1: ten numbers is possible.
- Take the top half, . The smallest sum you can make is , which is bigger than . So no sum lands back in the set. That is numbers.
- (Another set that works: the odd numbers . Odd plus odd is always even, so a sum is never one of them.)
Part 2: eleven numbers is impossible.
- The largest number in the set is the one that limits everything, so look at it first. Suppose the largest number is .
- Pair up the numbers below so that each pair adds to :
The number is left over without a partner.
- Because is in the set, the set cannot contain both numbers of any pair. For example, if it had and , then would be in the set, which breaks the rule.
- The set cannot contain either, because . (The problem says the two numbers may be the same.)
- So below the set has at most one number from each of the pairs. Adding itself gives at most numbers.
- If the largest number is smaller than , the same pairing works, with even fewer numbers below it to choose from. For example, if the largest is , the pairs give at most numbers below it, and again at most in total.
Ten is possible and eleven is not, so the greatest size is .
The answer is .
Why this works
When a question asks how many numbers you can pick without breaking a rule, look at one special number (here the largest) and see what it forbids. Pairing the other numbers so that each pair "fights" over one spot turns the question into counting pairs. The same idea shows that from to you can never pick more than about half the numbers, and the top half always works.
Alternative approach
Count forbidden numbers instead of pairs. Say the set has numbers and the largest is . For each of the other numbers , the number cannot be in the set (otherwise ). These "forbidden" numbers are all different from each other, because different values of give different values of . So there are numbers in the set, plus different numbers that are not in the set, all between and . That is , so .
The trap
Choosing 10 through 20 (11 numbers) and forgetting that a number may be added to itself: 10 + 10 = 20 breaks the rule. That gives the wrong choice (D).
Concise solution
Ten is attainable. Two different sets of size work.
- The top half : any two of its elements (equal or not) sum to at least , which is outside altogether.
- The odd numbers : a sum of two odd numbers is even, so it is never one of them.
Eleven is impossible. Let be sum-free with and let be its largest element, so .
For any we cannot have , since then would be an element of . So the "mirror" map sends into the complement of .
Apply this to the elements of . Each such satisfies , so its mirror also lies in , and the map is injective. We therefore have two sets inside :
each of size , and by the previous paragraph they are disjoint. Counting,
so , and since is an integer, .
The greatest possible size is .
The answer is .
Review in the AoPS books
- Introduction to Number Theory, Chapter 1: Integers: The Basics
- Intermediate Counting & Probability, Chapter 2: Sets and Logic (§2.2, §2.3)
- Intermediate Counting & Probability, Chapter 5: The Pigeonhole Principle
- Introduction to Algebra, Chapter 20: Special Functions (§20.3)
Techniques
Consider the largest/smallest element or boundary case · Pigeonhole principle