Olympiad Maths Prep

Track / Stage 6 / 102 of 400 #1102 of 2000

Problem 1102

National olympiad, first round
Combinatorics Difficulty 6.1 Prove it

Given 35 positive integers, the sum of which is 100 and none of them is greater than 50. Prove that there are some of them whose sum is 50.

This one wants a proof. Work it on paper, read the official solution, then mark yourself honestly — the ladder only means something if the record is true.

Official solution

The question is related to problem 2518 and exercise 2304. (Their solutions can be found in the January and May issues of 1986, respectively.) From the note added to the solution of the latter, it is clear that the "balancing procedure" applied in both cases does not necessarily lead to a result. That is, if we assign weights to the given numbers and place them in the pans of a two-pan balance in decreasing order of magnitude, placing them in the left pan if the balance is even, and otherwise in the lighter pan, the pans will not necessarily be in balance when the weights run out.

If the procedure "gets stuck," that is, placing a weight of mass mm in one of the pans would exceed half the total mass, 50 grams, then the total mass of the weights in both pans up to this point is at least 50(m1)=51m50-(m-1)=51-m, so at most 2m22m-2 mass of weights remain off the balance. The mass in the pans is at most 100m100-m, and thus up to the point of getting stuck, we could have placed at most

[100mm]=[100m]1 \left[\frac{100-m}{m}\right]=\left[\frac{100}{m}\right]-1

weights on the balance, since each of the weights used so far is at least mm in mass due to the decreasing order.

Thus, we still need to place at least 35([100m]1)=36[100m]36100m35-\left(\left[\frac{100}{m}\right]-1\right)=36-\left[\frac{100}{m}\right] \geqq 36-\frac{100}{m} weights, and their total mass is at most 2m22m-2 according to the first estimate derived from the fact of getting stuck. More precisely, excluding the "critical" weight of mass mm, at least 35100m35-\frac{100}{m} weights are left, each with a mass of at most m2m-2 grams.

Since the mass of each weight is at least 1, we have 35100mm235-\frac{100}{m} \leqq m-2, or m237m+1000m^2 - 37m + 100 \geqq 0. Given that mm is an integer, this inequality holds if and only if 35m35 \leqq m or 2m2 \geqq m.

The case m35m \geqq 35 is not possible, because in the decreasing order, there would already be at least one weight of at least 35 grams in each pan before getting stuck, and together with the "critical" weight of mass mm, their total would be at least 3353 \cdot 35, which is greater than 100.

With a weight of mass m=1m=1, the procedure cannot get stuck, as in this case, proceeding one by one, the 50-gram mass in either pan cannot be exceeded.

Thus, the procedure can only get stuck in the case m=2m=2, which can happen if 49 grams of weight have accumulated in both pans, and a 2-gram weight remains. (In this case, we have no 1-gram weights.) This can indeed occur, for example, if we have 30 weights of 3 grams and 5 weights of 2 grams. It can be shown that in this case, there exists a subset of the weights in each pan such that the difference in their total mass is exactly 1 gram. By swapping these two subsets, the difference between the pans becomes 2, which can be balanced with the remaining weight. However, we will choose a different path now.

If there are no 1-gram weights, then the number of 2-gram weights is clearly at least five (42+313>100)(4 \cdot 2 + 31 \cdot 3 > 100). Let's set aside five of the 2-gram weights. We will show that among the remaining thirty weights, there exists a subset whose total mass is an even number between 40 and 50, and thus can be completed to 50 grams using the set-aside 2-gram weights.

The total mass (90 grams) and the number (30) of the remaining weights are both even, so the number of odd-mass and even-mass weights among them is also even. Therefore, we can pair the odd-mass weights and the even-mass weights separately and consider the sum of the weights in each pair. This way, we get 15 even numbers, each at least 4, and their total is 90. We will show that among these, there are some whose sum SS (which is clearly even) satisfies 40S5040 \leqq S \leqq 50.

Let the 15 numbers in decreasing order be b1,b2,,b15b_1, b_2, \ldots, b_{15}, and let SkS_k be the sum of the first kk numbers. Since b14b_1 \geqq 4, we have Sk=90(bk+1++b15)90(15k)4=30+4kS_k = 90 - (b_{k+1} + \ldots + b_{15}) \leqq 90 - (15-k) \cdot 4 = 30 + 4k, so S550S_5 \leqq 50. If S540S_5 \geqq 40, we are done.

If S538S_5 \leqq 38, then S55b5S_5 \geqq 5 \cdot b_5 implies b5385<8b_5 \leqq \frac{38}{5} < 8, so if i5i \geqq 5, then bi6b_i \leqq 6. Therefore, if we start from a number less than 40 and increase it by at most 6 at each step until we reach 90, we cannot "skip" the interval [40, 50], so there will be some kk for which 40Sk5040 \leqq S_k \leqq 50. This completes the proof.

Source: NuminaMath-1.5, licensed Apache-2.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.