Maths Olympiad Prep

Library / /224 of 397

Combinatorics Difficulty 6.0 National Olympiad Prove it Taiwan

Let nn be a positive integer such that 33 does not divide nn. An nn-th order triangular array is formed by dividing a large equilateral triangle into n2n^2 smaller equilateral triangles of equal size, by drawing lines through all the nn-equal-division points on each side, parallel to the other sides. An orange is placed at each vertex of the small triangles in the nn-th order triangular array, giving a total of
(n+1)(n+2)2 \frac{(n+1)(n+2)}{2}
oranges. A set of three oranges A,B,CA, B, C is called good if and only if AB\overline{AB} and AC\overline{AC} are each a side of some small triangle, and BAC=120\angle BAC = 120^{\circ}.

Yen-Yen may each time choose a good set of three oranges and remove them. Find the maximum number of oranges that Yen-Yen can remove.

Solution

Answer: (n+1)(n+2)23\frac{(n+1)(n+2)}{2} - 3.

Denote by an (n,k)(n, k)-trapezoidal array the figure obtained by removing from an nn-th order triangular array an nkn-k-th order triangular array sharing a common vertex with it. Let A00,A01,A02,,A0nA_{00}, A_{01}, A_{02}, \dots, A_{0n} denote in order the oranges on one base edge of the nn-th order triangular array, and recursively define the orange labeled AijA_{ij} so that Aij,A(i1)j,A(i1)(j+1)A_{ij}, A_{(i-1)j}, A_{(i-1)(j+1)} form a small equilateral triangle (AijA(i2)(j+1)A_{ij} \neq A_{(i-2)(j+1)}); we say that the orange AijA_{ij} is the jj-th orange in the ii-th row.

(1) First we prove the upper bound. It suffices to prove the following:
Among A00,A01,A02,,A0nA_{00}, A_{01}, A_{02}, \dots, A_{0n}, there must be at least one orange that is never selected.
This shows that the least upper bound must be strictly less than (n+1)(n+2)2\frac{(n+1)(n+2)}{2}, and since the least upper bound must be a multiple of 33, the result follows.

Proof of the Lemma: Suppose otherwise, i.e., assume that all these n+1n+1 oranges can be removed.

* Let kk be the smallest integer such that (A0(k+1),A0k,A1(k1))(A_{0(k+1)}, A_{0k}, A_{1(k-1)}) are simultaneously selected; if no such integer exists, let k=n+1k = n + 1;
* Let mm be the largest integer less than kk such that (A0(m1),A0m,A1m)(A_{0(m-1)}, A_{0m}, A_{1m}) are simultaneously selected; if no such integer exists, let m=1m = -1;
* Since A1(k1)A_{1(k-1)} and A1mA_{1m} must be selected in different rounds, we must have k1m+1k - 1 \ge m + 1. This guarantees the existence of m<i<km < i < k, and note that these A0iA_{0i} are not selected in either of the two selections mentioned above.

* Also note that, by the minimality of kk and the maximality of mm, the selection that includes A0iA_{0i} cannot simultaneously include another A0jA_{0j}, and must include at least one A1jA_{1j}. But note that there are a total of km1k-m-1 such A0iA_{0i}, while between A1(k1)A_{1(k-1)} and A1mA_{1m} there are only (k1)m1(k-1)-m-1 possible A1jA_{1j}'s to choose, so this is impossible. Contradiction!

(2) Next we construct a selection method that achieves the upper bound. We proceed by induction:
* n=1n=1: no orange needs to be removed
* n=2n=2: remove (A20,A10,A11)(A_{20}, A_{10}, A_{11})
* n=3n=3: remove (A30,A20,A11)(A_{30}, A_{20}, A_{11}); (A21,A12,A02)(A_{21}, A_{12}, A_{02})
* Suppose the case n=k3n=k-3 holds; consider n=kn=k: first apply the induction hypothesis to the first k3k-3 rows. Consider the parity of kk:
* If k=2m+1k=2m+1, we may select:
(A30_{30}, A21_{21}, A22_{22}); (A20_{20}, A11_{11}, A12_{12}); (A10_{10}, A01_{01}, A02_{02}); and
(A0(2i+1)_{0(2i+1)}, A0(2i+2)_{0(2i+2)}, A1(2i+2)_{1(2i+2)}); (A1(2i+1)_{1(2i+1)}, A2(2i+1)_{2(2i+1)}, A2(2i+2)_{2(2i+2)}), for 1im21 \le i \le m-2, and
(A3(k3)_{3(k-3)}, A2(k2)_{2(k-2)}, A1(k2)_{1(k-2)}); (A0(k2)_{0(k-2)}, A0(k1)_{0(k-1)}, A1(k1)_{1(k-1)})
* If k=2mk=2m, we may select: (A30_{30}, A21_{21}, A11_{11}); (A20_{20}, A10_{10}, A01_{01}) and
(A0(2i+1)_{0(2i+1)}, A0(2i+2)_{0(2i+2)}, A1(2i+2)_{1(2i+2)}); (A1(2i+1)_{1(2i+1)}, A2(2i+1)_{2(2i+1)}, A2(2i+2)_{2(2i+2)}), for 1im21 \le i \le m-2, and
(A3(k3)_{3(k-3)}, A2(k2)_{2(k-2)}, A1(k2)_{1(k-2)}); (A0(k2)_{0(k-2)}, A0(k1)_{0(k-1)}, A1(k1)_{1(k-1)})

By induction, the result follows.

Want a route through all this instead of an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement translated into English from zh; metadata (topic, difficulty) added by this project.