Olympiad Maths Prep

Track / Stage 8 / 50 of 180 #1750 of 2000

Problem 1750

IMO Shortlist mid-range; USAMO P2/P5
Combinatorics Difficulty 8.2 Prove it USAMO · United States

Let nn be a positive integer. There are n(n+1)2\frac{n(n+1)}{2} marks, each with a black side and a white side, arranged into an equilateral triangle, with the biggest row containing nn marks. Initially, each mark has the black side up. An *operation* is to choose a line parallel to one of the sides of the triangle, and flipping all the marks on that line. A configuration is called *admissible* if it can be obtained from the initial configuration by performing a finite number of operations. For each admissible configuration CC, let f(C)f(C) denote the smallest number of operations required to obtain CC from the initial configuration. Find the maximum value of f(C)f(C), where CC varies over all admissible configurations.
(This problem was suggested by Warut Suksompong.)

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

Solution (By Warut Suksompong). The answer is 6n4+n4n4=n+2n46\lfloor\frac{n}{4}\rfloor + n - 4\lfloor\frac{n}{4}\rfloor = n + 2\lfloor\frac{n}{4}\rfloor.
For n=1n=1 the answer is clearly 1, since there is only one configuration other than the initial one, and that configuration takes 1 step to get to. From now on we will consider n2n \ge 2.
Note that there are 3n3n possible operations in total, since we can select 3n3n lines to perform an operation on (nn lines parallel to each side of the triangle.) Performing an operation twice on the same line is equivalent to doing nothing. Hence, we will describe any combination of operations as a triple of nn-tuples ((a1,a2,,an),(b1,b2,,bn),(c1,c2,,cn))((a_1, a_2, \dots, a_n), (b_1, b_2, \dots, b_n), (c_1, c_2, \dots, c_n)), where each element ai,bi,cia_i, b_i, c_i is either 0 or 1 (00 means no operation, 11 means the opposite), each tuple of the triple denotes operating on a line parallel to one of the sides, and the indices denote the number of marks in the row of operation. Let AA denote the set of all such 3n3n-tuples, so that A=23n|A| = 2^{3n}.
Let BB denote the set of all admissible configurations. Let N=n(n+1)2N = \frac{n(n+1)}{2}. We will describe each element of BB by an NN-tuple (z1,z2,,zN)(z_1, z_2, \dots, z_N), where zi=0z_i = 0 if mark ii is black and zi=1z_i = 1 otherwise.
For each element aAa \in A, let b=f(a)b = f(a) be the element of BB that is the result of applying the operations in aa. Then f(a+a)=f(a)+f(a)f(a + a') = f(a) + f(a') for all a,aAa, a' \in A, where addition is considered in modulo 2. Let KK be the set of all aAa \in A such that f(a)f(a) is the all-black configuration. The following eight elements are easily seen to be in KK.

* ((0,0,,0),(0,0,,0),(0,0,,0))=id((0, 0, \dots, 0), (0, 0, \dots, 0), (0, 0, \dots, 0)) = \text{id}
* ((0,0,,0),(1,1,,1),(1,1,,1))=x((0, 0, \dots, 0), (1, 1, \dots, 1), (1, 1, \dots, 1)) = x
* ((1,1,,1),(1,1,,1),(0,0,,0))=y((1, 1, \dots, 1), (1, 1, \dots, 1), (0, 0, \dots, 0)) = y
* ((1,1,,1),(0,0,,0),(1,1,,1))=x+y((1, 1, \dots, 1), (0, 0, \dots, 0), (1, 1, \dots, 1)) = x + y
* ((0,1,0,1,),(0,1,0,1,),(0,1,0,1,))=z((0, 1, 0, 1, \dots), (0, 1, 0, 1, \dots), (0, 1, 0, 1, \dots)) = z
* ((0,1,0,1,),(1,0,1,0,),(1,0,1,0,))=x+z((0, 1, 0, 1, \dots), (1, 0, 1, 0, \dots), (1, 0, 1, 0, \dots)) = x + z
* ((1,0,1,0,),(1,0,1,0,),(0,1,0,1,))=y+z((1, 0, 1, 0, \dots), (1, 0, 1, 0, \dots), (0, 1, 0, 1, \dots)) = y + z
* ((1,0,1,0,),(0,1,0,1,),(1,0,1,0,))=x+y+z((1, 0, 1, 0, \dots), (0, 1, 0, 1, \dots), (1, 0, 1, 0, \dots)) = x + y + z
We will show that they are the only elements of KK.
Suppose L=((a1,a2,,an),(b1,b2,,bn),(c1,c2,,cn))L = ((a_1, a_2, \dots, a_n), (b_1, b_2, \dots, b_n), (c_1, c_2, \dots, c_n)) is in KK. For i+j+k=2n+1i+j+k = 2n+1, there is a unique mark contained in a row of length ii in the aa direction, a row of length jj in the bb direction, and a row of length kk in the cc direction. Operations aia_i, bjb_j, and ckc_k are the only operations affecting the mark, so ai+bj+ck=0a_i + b_j + c_k = 0. By adding one or both of xx and yy if necessary, we will assume that bn=cn=0b_n = c_n = 0. Since a2+bn1+cn=a2+bn+cn1=0a_2 + b_{n-1} + c_n = a_2 + b_n + c_{n-1} = 0, we have that bn1=cn1b_{n-1} = c_{n-1}. We now have two cases.

a. First, suppose that bn1=cn1=0b_{n-1} = c_{n-1} = 0. Then from a3+bn2+cn=a3+bn1+cn1=a3+bn+cn2a_3+b_{n-2}+c_n = a_3+b_{n-1}+c_{n-1} = a_3+b_n+c_{n-2}, we have that bn2=cn2=0b_{n-2} = c_{n-2} = 0. Continuing in this manner (considering equalities with a4,a5,a_4, a_5, \dots), we find that all the bib_i's and cic_i's are 0, from which we deduce that L=idL = \text{id}.

b. Second, suppose that bn1=cn1=1b_{n-1} = c_{n-1} = 1. Then from a3+bn2+cn=a3+bn1+cn1=a3+bn+cn2a_3 + b_{n-2} + c_n = a_3 + b_{n-1} + c_{n-1} = a_3 + b_n + c_{n-2}, we have that bn2=cn2=0b_{n-2} = c_{n-2} = 0. Continuing in this manner (considering equalities with a4,a5,a_4, a_5, \dots), we find that (b1,b2,,bn)=(c1,c2,,cn)=(,1,0,1,0)(b_1, b_2, \dots, b_n) = (c_1, c_2, \dots, c_n) = (\dots, 1, 0, 1, 0), from which we deduce that either L=zL = z or L=x+zL = x + z.

Hence LL is one of the eight elements listed above. It follows that the 23n2^{3n} elements of AA form 23n32^{3n-3} sets consisting of elements differing by an element of KK. Each aAa \in A corresponds to an element of BB which is the result of applying the operations in aa. For each element aAa \in A, let x1x_1 be the number of aia_i with odd indices which are equal to 1, and let x2x_2 be the number of aia_i with even indices which are equal to 1. Define y1,y2,z1y_1, y_2, z_1, and z2z_2 similarly for the bib_i's and cic_i's. Note that for each choice of (x1,x2,y1,y2,z1,z2)(x_1, x_2, y_1, y_2, z_1, z_2), there is at least one element of AA which corresponds to this choice.
Let T(a)T(a) be the minimum value of TT over the set containing aa. The maximum of this value over all the sets is the desired answer. For an element aAa \in A with minimal value of TT in its set, adding an element of KK to aa must increase the value of TT. Conversely if adding any element of KK to aa decreases TT, then aa has the minimal value of TT within its set. Applying this observation for x,y,x+y,z,x+z,y+z,x+y+zKx, y, x+y, z, x+z, y+z, x+y+z \in K, we see that aAa \in A has the minimal value of TT within its set if and only if the following inequalities hold.

(a) x1+x2+y1+y2nx_1 + x_2 + y_1 + y_2 \le n
(b) x1+x2+z1+z2nx_1 + x_2 + z_1 + z_2 \le n
(c) y1+y2+z1+z2ny_1 + y_2 + z_1 + z_2 \le n

(d) x2+y2+z23n/22=Vx_2 + y_2 + z_2 \le \left\lfloor \frac{3\lfloor n/2 \rfloor}{2} \right\rfloor = V
(e) x1+y1+z22n/2+n/22=Wx_1 + y_1 + z_2 \le \left\lfloor \frac{2\lfloor n/2 \rfloor + \lfloor n/2 \rfloor}{2} \right\rfloor = W

(f)x2+y1+z12n/2+n/22=W (f) \quad x_2 + y_1 + z_1 \le \left\lfloor \frac{2\lfloor n/2 \rfloor + \lfloor n/2 \rfloor}{2} \right\rfloor = W
(g)x1+y2+z12n/2+n/22=W (g) \quad x_1 + y_2 + z_1 \le \left\lfloor \frac{2\lfloor n/2 \rfloor + \lfloor n/2 \rfloor}{2} \right\rfloor = W
Adding the last four inequalities and dividing by 4, we obtain
T(a)V+3W2. T(a) \le \left\lfloor \frac{V + 3W}{2} \right\rfloor.
We analyze this in four separate cases:

a. n=4kn = 4k. Then V=W=3kV = W = 3k, and so T(a)6kT(a) \le 6k, with equality for x1=x2=y1=y2=z1=z2=kx_1 = x_2 = y_1 = y_2 = z_1 = z_2 = k.
b. n=4k+1n = 4k + 1. Then V=3kV = 3k and W=3k+1W = 3k + 1, and so T(a)6k+1T(a) \le 6k + 1 with equality for x1=x2=y1=y2=z2=kx_1 = x_2 = y_1 = y_2 = z_2 = k and z1=k+1z_1 = k + 1.
c. n=4k+2n = 4k + 2. Then V=3k+1V = 3k + 1 and W=3k+1W = 3k + 1, and so T(a)6k+2T(a) \le 6k + 2 with equality for x1=x2=y1=y2=kx_1 = x_2 = y_1 = y_2 = k and z1=z2=k+1z_1 = z_2 = k + 1.
d. n=4k+3n = 4k + 3. Then V=3k+1V = 3k + 1 and W=3k+2W = 3k + 2, and so T(a)6k+3T(a) \le 6k + 3 with equality for x1=x2=y2=kx_1 = x_2 = y_2 = k and y1=z1=z2=k+1y_1 = z_1 = z_2 = k + 1.
In each case, observe that T(a)n+2n4T(a) \le n + 2 \lfloor \frac{n}{4} \rfloor and that equality is attained for some configuration in AA, hence the maximum value of T(a)T(a) is as claimed, concluding our proof.

Source: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.