Maths Olympiad Prep

Library / /28 of 34

Geometry Difficulty 7.4 National Olympiad, round 2 Prove it United States

Let ABCABC be an equilateral triangle with side length 11. Points A1A_1 and A2A_2 are chosen on side BCBC, points B1B_1 and B2B_2 are chosen on side CACA, and points C1C_1 and C2C_2 are chosen on side ABAB such that BA1<BA2BA_1 < BA_2, CB1<CB2CB_1 < CB_2, and AC1<AC2AC_1 < AC_2.
Suppose that the three line segments B1C2B_1C_2, C1A2C_1A_2, and A1B2A_1B_2 are concurrent, and the perimeters of triangles AB2C1AB_2C_1, BC2A1BC_2A_1, and CA2B1CA_2B_1 are all equal. Find all possible values of this common perimeter.

Solution

Claim (p=1p = 1 implies concurrence) — Suppose the six points are chosen so that triangles AB2C1AB_2C_1, BC2A1BC_2A_1, CA2B1CA_2B_1 all have perimeter 11. Then lines B1C2\overline{B_1C_2}, C1A2\overline{C_1A_2}, and A1B2\overline{A_1B_2} are concurrent.

Proof. The perimeter conditions mean that B2C1\overline{B_2C_1}, C2A1\overline{C_2A_1}, and A2B1\overline{A_2B_1} are tangent to the incircle of ABC\triangle ABC.
Figure 1

¶ Barycentric solution (by Carl, Krit, Milan) We show that, if the common perimeter is 11, then the lines concur. To do this, we use barycentric coordinates. Let A=(1:0:0)A = (1:0:0), B=(0:1:0)B = (0:1:0), and C=(0:0:1)C = (0:0:1). Let A1=(0:1a1:a1)A_1 = (0:1-a_1:a_1), A2=(0:a2:1a2)A_2 = (0:a_2:1-a_2), B1=(b1:0:1b1)B_1 = (b_1:0:1-b_1), B2=(1b2:0:b2)B_2 = (1-b_2:0:b_2), C1=(1c1:c1:0)C_1 = (1-c_1:c_1:0), and C2=(c2:1c2:0)C_2 = (c_2:1-c_2:0). The line B1C2B_1C_2 is defined by the equation
det[xyzb101b1c21c20]=0; \det \begin{bmatrix} x & y & z \\ b_1 & 0 & 1-b_1 \\ c_2 & 1-c_2 & 0 \end{bmatrix} = 0;
i.e.
x((1b1)(1c2))+y((1b1)c2)+z(b1(1c2))=0. x(-(1-b_1)(1-c_2)) + y((1-b_1)c_2) + z(b_1(1-c_2)) = 0.

Computing the equations for the other lines cyclically, we get that the lines B1C2B_1C_2, C1A2C_1A_2, and A1B2A_1B_2 concur if and only if
det[(1b1)(1c2)(1b1)c2b1(1c2)c1(1a2)(1c1)(1a2)(1c1)a2(1a1)b2a1(1b2)(1a1)(1b2)]=0. \det \begin{bmatrix} -(1-b_1)(1-c_2) & (1-b_1)c_2 & b_1(1-c_2) \\ c_1(1-a_2) & -(1-c_1)(1-a_2) & (1-c_1)a_2 \\ (1-a_1)b_2 & a_1(1-b_2) & -(1-a_1)(1-b_2) \end{bmatrix} = 0.

Let this matrix be MM. We also define the similar matrix
N=[(1b2)(1c1)(1b2)c1b2(1c1)c2(1a1)(1c2)(1a1)(1c2)a1(1a2)b1a2(1b1)(1a2)(1b1)]. N = \begin{bmatrix} -(1-b_2)(1-c_1) & (1-b_2)c_1 & b_2(1-c_1) \\ c_2(1-a_1) & -(1-c_2)(1-a_1) & (1-c_2)a_1 \\ (1-a_2)b_1 & a_2(1-b_1) & -(1-a_2)(1-b_1) \end{bmatrix}.

Claim — We have detM=detN\det M = \det N.

Proof. To show detM=detN\det M = \det N, it suffices to demonstrate that the determinant above is invariant under swapping subscripts of “1” and “2,” an operation we call Ψ\Psi.
We use the definition of the determinant as a sum over permutations. The even permutations give us the following three terms:
(1b1)(1c2)(1c1)(1a2)(1a1)(1b2)=i=12((1ai)(1bi)(1ci))(1a1)b2(1b1)c2(1c1)a2=((1a1)(1b1)(1c1))(a2b2c2)c1(1a2)a1(1b2)b1(1c2)=((1a2)(1b2)(1c2))(a1b1c1). -(1-b_1)(1-c_2)(1-c_1)(1-a_2)(1-a_1)(1-b_2) = -\prod_{i=1}^{2}((1-a_i)(1-b_i)(1-c_i)) \\ (1-a_1)b_2(1-b_1)c_2(1-c_1)a_2 = ((1-a_1)(1-b_1)(1-c_1))(a_2b_2c_2) \\ c_1(1-a_2)a_1(1-b_2)b_1(1-c_2) = ((1-a_2)(1-b_2)(1-c_2))(a_1b_1c_1).

The first term is invariant under Ψ\Psi, while the second and third terms are swapped under Ψ\Psi. For the odd permutations, we have a contribution to the determinant of
cyc(1b1)(1c2)(1c1)a2a1(1b2); \sum_{\text{cyc}} (1-b_1)(1-c_2)(1-c_1)a_2a_1(1-b_2);
each summand is invariant under Ψ\Psi. This finishes the proof of our claim. \square

Now, it suffices to show that, if AB2C1AB_2C_1, BC2A1BC_2A_1, and CA2B1CA_2B_1 each have perimeter 11, then
det[(1b2)(1c1)(1b2)c1b2(1c1)c2(1a1)(1c2)(1a1)(1c2)a1(1a2)b1a2(1b1)(1a2)(1b1)]=0. \det \begin{bmatrix} -(1-b_2)(1-c_1) & (1-b_2)c_1 & b_2(1-c_1) \\ c_2(1-a_1) & -(1-c_2)(1-a_1) & (1-c_2)a_1 \\ (1-a_2)b_1 & a_2(1-b_1) & -(1-a_2)(1-b_1) \end{bmatrix} = 0.

Indeed, we have AB2=b2AB_2 = b_2 and AC1=c1AC_1 = c_1, so by the law of cosines,
1b2c1=1AB2AC1=B2C1=b22+c12b2c1. 1 - b_2 - c_1 = 1 - AB_2 - AC_1 = B_2C_1 = \sqrt{b_2^2 + c_1^2 - b_2c_1}.
This gives
(1b2c1)2=b22+c12b2c1    12b22c1+3b2c1=0. (1 - b_2 - c_1)^2 = b_2^2 + c_1^2 - b_2c_1 \implies 1 - 2b_2 - 2c_1 + 3b_2c_1 = 0.
Similarly,
12c22a1+3c2a1=0 and 12a22b1+3a2b1=0. 1 - 2c_2 - 2a_1 + 3c_2a_1 = 0 \text{ and } 1 - 2a_2 - 2b_1 + 3a_2b_1 = 0.

Now,
N[111]=[(1b2)(1c1)+(1b2)c1+b2(1c1)(1c2)(1a1)+(1c2)a1+c2(1a1)(1a2)(1b1)+(1a2)b1+a2(1b1)]=[1+2b2+2c13b2c11+2c2+2a13c2a11+2a2+2b12a2b1]=[000]. N \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix} = \begin{bmatrix} -(1-b_2)(1-c_1) + (1-b_2)c_1 + b_2(1-c_1) \\ -(1-c_2)(1-a_1) + (1-c_2)a_1 + c_2(1-a_1) \\ -(1-a_2)(1-b_1) + (1-a_2)b_1 + a_2(1-b_1) \end{bmatrix} \\ = \begin{bmatrix} -1 + 2b_2 + 2c_1 - 3b_2c_1 \\ -1 + 2c_2 + 2a_1 - 3c_2a_1 \\ -1 + 2a_2 + 2b_1 - 2a_2b_1 \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \\ 0 \end{bmatrix}.

So it follows detN=0\det N = 0, as desired.

---

Synthetic approach (from author) We prove the converse of the problem first:

Claim (p=1p = 1 implies concurrence) — Suppose the six points are chosen so that triangles AB2C1AB_2C_1, BC2A1BC_2A_1, CA2B1CA_2B_1 all have perimeter 11. Then lines B1C2\overline{B_1C_2}, C1A2\overline{C_1A_2}, and A1B2\overline{A_1B_2} are concurrent.

Proof. The perimeter conditions mean that B2C1\overline{B_2C_1}, C2A1\overline{C_2A_1}, and A2B1\overline{A_2B_1} are tangent to the incircle of ABC\triangle ABC.
Figure 1
Hence the result follows by Brianchon's theorem. \Box

Now suppose p1p \neq 1. Let B2C1\overline{B'_2C'_1} be the dilation of B2C1\overline{B_2C_1} with ratio 1p\frac{1}{p} at center AA, and define C2C'_2, A1A'_1, A2A'_2, B1B'_1 similarly. The following diagram showcases the situation p<1p < 1.

By the reasoning in the p=1p=1 case, note that B1C2\overline{B'_1C'_2}, C1A2\overline{C'_1A'_2}, and A1B2\overline{A'_1B'_2} are concurrent. However, B1C2\overline{B_1C_2}, C1A2\overline{C_1A_2}, A1B2\overline{A_1B_2} lie in the interior of quadrilaterals BCB1C2BCB'_1C'_2, CAC1A2CAC'_1A'_2, and ABA1B2ABA'_1B'_2, and these quadrilaterals do not share an interior point, a contradiction.
Thus p1p \ge 1. Similarly, we can show p1p \le 1, and so p=1p=1 is forced (and achieved if, for example, the three triangles are equilateral with side length 1/31/3).

Therefore, the only possible value of the common perimeter is 1\boxed{1}.

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 reproduced verbatim; metadata (topic, difficulty) added by this project.