Maths Olympiad Prep

Library / China Mathematical Competition (Complementary Test) /16 of 17

China Mathematical Competition (Complementary Test)

Geometry Difficulty 7.1 National Olympiad, round 2 Prove it China

As shown in Fig. 1, ABCDABCD is a convex quadrilateral with ∠B+∠D<180∘\angle B + \angle D < 180^\circ, and PP is a moving point on the plane. Let
f(P)=PA×BC+PD×CA+PC×AB. f(P) = PA \times BC + PD \times CA + PC \times AB.
(1) Prove that PP, AA, BB, CC are concyclic when f(P)f(P) reaches the minimum.

(2) Suppose that point EE is on the arc AB\text{AB} of the circumscribed circle OO of △ABC\triangle ABC, satisfying
AEAB=32,BCEC=3−1,∠ECB=12∠ECA; \frac{AE}{AB} = \frac{\sqrt{3}}{2}, \quad \frac{BC}{EC} = \sqrt{3} - 1, \quad \angle ECB = \frac{1}{2} \angle ECA;
furthermore, DADA, DCDC are tangent to ⊙O\odot O, AC=2AC = \sqrt{2}. Find the minimum of f(P)f(P).

Figure 1
Fig. 1

Solutions — 3

Solution 1

(1) As shown in Fig. 1, by the Ptolemy inequality we have
PA×BC+PC×AB≥PB×AC. PA \times BC + PC \times AB \ge PB \times AC.
Therefore,
f(P)=PA×BC+PC×AB+PD×CA≥PB×CA+PD×CA=(PB+PD)×CA. \begin{aligned} f(P) &= PA \times BC + PC \times AB + PD \times CA \\ &\ge PB \times CA + PD \times CA \\ &= (PB + PD) \times CA. \end{aligned}
The equality holds if and only if PP, AA, BB, CC lie on ⊙O\odot O and PP on AC^\widehat{AC}. Furthermore, PB+PD≥BDPB + PD \ge BD, and the equality holds if and only if PP lies on line BDBD. Combining the results above, we have
f(P)min⁡=BD×CA. f(P)_{\min} = BD \times CA.

(2) Denote ∠ECB=α\angle ECB = \alpha. Then ∠ECA=2α\angle ECA = 2\alpha. By the sine theorem we have
AEAB=sin⁡2αsin⁡3α=32. \frac{AE}{AB} = \frac{\sin 2\alpha}{\sin 3\alpha} = \frac{\sqrt{3}}{2}.
That is to say, 3sin⁡3α=2sin⁡2α\sqrt{3} \sin 3\alpha = 2\sin 2\alpha. By using trigonometric identities,
3(3sin⁡α−4sin⁡3α)=4sin⁡αcos⁡α. \sqrt{3}(3\sin \alpha - 4\sin^3 \alpha) = 4\sin \alpha \cos \alpha.
By simplification,
33−43(1−cos⁡2α)−4cos⁡α=0. 3\sqrt{3} - 4\sqrt{3}(1 - \cos^2\alpha) - 4\cos\alpha = 0.
That is to say, 43cos⁡2α−4cos⁡α−3=0. \text{That is to say, } 4\sqrt{3}\cos^2\alpha - 4\cos\alpha - \sqrt{3} = 0.
The solutions are cos⁡α=32 and cos⁡α=−123 (discarded). \text{The solutions are } \cos \alpha = \frac{\sqrt{3}}{2} \text{ and } \cos \alpha = -\frac{1}{2\sqrt{3}} \text{ (discarded).}
Therefore, α=30∘ and ∠ECA=60∘. \text{Therefore, } \alpha = 30^\circ \text{ and } \angle ECA = 60^\circ.
On the other hand,
BCEC=3−1=sin⁡(∠EAC−30∘)sin⁡∠EAC. \frac{BC}{EC} = \sqrt{3} - 1 = \frac{\sin(\angle EAC - 30^\circ)}{\sin \angle EAC}.
That is to say,
32sin⁡∠EAC−12cos⁡∠EAC=(3−1)sin⁡∠EAC. \frac{\sqrt{3}}{2}\sin\angle EAC - \frac{1}{2}\cos\angle EAC = (\sqrt{3} - 1)\sin\angle EAC.
Then
2−32sin⁡∠EAC=12cos⁡∠EAC. \frac{2 - \sqrt{3}}{2} \sin \angle EAC = \frac{1}{2} \cos \angle EAC.
Therefore,
tan⁡∠EAC=12−3=2+3. \tan \angle EAC = \frac{1}{2 - \sqrt{3}} = 2 + \sqrt{3}.
We obtain ∠EAC=75∘ and ∠AEC=45∘. \text{We obtain } \angle EAC = 75^\circ \text{ and } \angle AEC = 45^\circ.
Since △ADC\triangle ADC is an isosceles triangle and AC=2AC = \sqrt{2}, we have CD=1CD = 1. Furthermore, △ABC\triangle ABC is an isosceles triangle, so BC=2BC = \sqrt{2} and AB=2AB = 2. Then
BD2=AB2+AD2=4+1=5. BD^2 = AB^2 + AD^2 = 4 + 1 = 5.
We have BD=5. Therefore, \text{We have } BD = \sqrt{5}. \text{ Therefore,}

f(P)min⁡f(P)_{\min} = BD ×\times CA = 5×2=10.\sqrt{5} \times \sqrt{2} = \sqrt{10}.

Solution 2

(1) As shown in Fig. 2, the line BDBD intercepts the circumscribed circle OO of △ABC\triangle ABC at point P0P_0, which lies on BDBD because DD is outside of ⊙O\odot O. Through AA, CC, DD draw lines perpendicular to P0AP_0A, P0CP_0C, P0DP_0D, respectively, and they constitute △A1B1C1\triangle A_1B_1C_1 by intercepting with each other. It is easy to see that P0P_0 lies in △ABC\triangle ABC, as well as in △A1B1C1\triangle A_1B_1C_1. Denoting the three inner angles of △ACD\triangle ACD by xx, yy, zz, respectively, we have
Figure 2
Fig. 2
∠AP0C=180∘−y=z+x. \angle AP_0C = 180^\circ - y = z + x.
Furthermore, from B1C1⊥P0AB_1C_1 \perp P_0A and B1A1⊥P0CB_1A_1 \perp P_0C, we have ∠B1=y\angle B_1 = y. In a similar way, we have ∠A1=x\angle A_1 = x and ∠C1=z\angle C_1 = z. It follows that △A1B1C1∼△ABC\triangle A_1B_1C_1 \sim \triangle ABC. Now let
B1C1=λBC,C1A1=λCA,A1B1=λAB. B_1C_1 = \lambda BC, \quad C_1A_1 = \lambda CA, \quad A_1B_1 = \lambda AB.
Then for any point MM on the plane, we have
λf(P0)=λ(P0A×BC+P0D×CA+P0C×AB)=P0A×B1C1+P0D×C1A1+P0C×A1B1=2S△A1B1C1≤MA×B1C1+MD×C1A1+MC×A1B1=λ(MA×BC+MD×CA+MC×AB)=λf(M). \begin{aligned} \lambda f(P_0) &= \lambda (P_0A \times BC + P_0D \times CA + P_0C \times AB) \\ &= P_0A \times B_1C_1 + P_0D \times C_1A_1 + P_0C \times A_1B_1 \\ &= 2S_{\triangle A_1B_1C_1} \\ &\le MA \times B_1C_1 + MD \times C_1A_1 + MC \times A_1B_1 \\ &= \lambda (MA \times BC + MD \times CA + MC \times AB) \\ &= \lambda f(M). \end{aligned}
That is to say, f(P0)≤f(M)f(P_0) \le f(M). Since MM is an arbitrary point, we know that f(P)f(P) reaches the minimum at P0P_0, and at the same time P0AP_0A, P0CP_0C, P0DP_0D are concyclic. This completes the proof.

(2) From (1) we know that the minimum of f(P)f(P) is
f(P0)=2λS△A1R1C1=2λS△ABC. f(P_0) = \frac{2}{\lambda} S_{\triangle A_1 R_1 C_1} = 2\lambda S_{\triangle ABC}.
In the same way as is shown in the proof of (2) in Solution I, we find that both △ADC\triangle ADC and △ABC\triangle ABC are right-angled isosceles triangles; so we have CD=AD=AC2=1CD = AD = \frac{AC}{\sqrt{2}} = 1, AB=2AC=2AB = \sqrt{2}AC = 2, S△ABC=1S_{\triangle ABC} = 1,
BD=AB2+AD2=5. BD = \sqrt{AB^2 + AD^2} = \sqrt{5}.
Furthermore, since △A1B1C1∼△ABC\triangle A_1B_1C_1 \sim \triangle ABC, ∠AB1B=∠AB1C+∠BB1C=90∘\angle AB_1B = \angle AB_1C + \angle BB_1C = 90^\circ, AB1BDAB_1BD is a rectangle. It follows that B1C1=BD=5B_1C_1 = BD = \sqrt{5}. Therefore λ=52\lambda = \frac{\sqrt{5}}{\sqrt{2}} and

f(P)min⁡f(P)_{\min} = 2 ×52×\times \frac{\sqrt{5}}{\sqrt{2}} \times 1 = 10.\sqrt{10}.

Solution 3

(1) We discuss the problem on the complex plane, and regard points AA, BB, CC as complex numbers. Then by the triangle inequality we have
∣PA⃗⋅BC⃗∣+∣PC⃗⋅AB⃗∣≥∣PA⃗⋅BC⃗+PC⃗⋅AB⃗∣. | \vec{PA} \cdot \vec{BC} | + | \vec{PC} \cdot \vec{AB} | \ge | \vec{PA} \cdot \vec{BC} + \vec{PC} \cdot \vec{AB} |.
That is to say,
∣(A−P)(C−B)∣+∣(C−P)(B−A)∣≥∣(A−P)(C−B)+(C−P)(B−A)∣=∣−P×C−A×B+C×B+P×A∣=∣(B−P)(C−A)∣=∣PB⃗⋅AC⃗∣.(1) \begin{aligned} & |(A-P)(C-B)| + |(C-P)(B-A)| \\ \ge & |(A-P)(C-B) + (C-P)(B-A)| \\ = & |-P \times C - A \times B + C \times B + P \times A| \\ = & |(B-P)(C-A)| \\ = & |\vec{PB} \cdot \vec{AC}|. \end{aligned} \tag{1}
Then
∣PA⃗⋅BC⃗∣+∣PC⃗⋅AB⃗∣+∣PD⃗⋅AC⃗∣≥∣PB⃗⋅AC⃗∣+∣PD⃗⋅AC⃗∣=(PB⃗+PD⃗)⋅AC⃗=∣BD⃗∣⋅AC⃗∣.(2) \begin{aligned} & |\vec{PA} \cdot \vec{BC}| + |\vec{PC} \cdot \vec{AB}| + |\vec{PD} \cdot \vec{AC}| \\ \ge & |\vec{PB} \cdot \vec{AC}| + |\vec{PD} \cdot \vec{AC}| \\ = & (\vec{PB} + \vec{PD}) \cdot \vec{AC} \\ = & |\vec{BD}| \cdot \vec{AC}|. \end{aligned} \tag{2}
The equality in (1) holds only when the complex numbers (A−P)(C−B)(A-P)(C-B) and (C−P)(B−A)(C-P)(B-A) are in the same direction. This means that there is a real λ>0\lambda > 0 such that
(A−P)(C−B)=λ(C−P)(B−A). (A-P)(C-B) = \lambda(C-P)(B-A).
That is to say,
A−PC−P=λB−AC−B. \frac{A-P}{C-P} = \lambda \frac{B-A}{C-B}.
Therefore,
arg⁡(A−PC−P)=arg⁡(B−AC−B), \arg\left(\frac{A-P}{C-P}\right) = \arg\left(\frac{B-A}{C-B}\right),
which means that the angle of rotation from PC⃗\vec{PC} to PA⃗\vec{PA} is equal to that from BC⃗\vec{BC} to AB⃗\vec{AB}. It follows that PP, AA, BB, CC are concyclic.
The equality in (2) holds only when BB, PP, DD are collinear and PP lies on the segment BDBD. This means that f(P)f(P) reaches the minimum when PP lies on the circumscribed circle of △ABC\triangle ABC and PP, AA, BB, CC are concyclic.

(2) From (1) we know that f(P)min=BD×CAf(P)_{min} = BD \times CA. The following steps are the same as those in the proof of (2) in Solution I.

Want a route through all this instead of an archive? The track puts 2,604 problems in a working order, from Junior Challenge level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.