Maths Olympiad Prep

Library / IMO Shortlist /194 of 365

IMO Shortlist

Geometry Difficulty 8.6 Shortlist Prove it IMO

Let ABCDEFA B C D E F be a convex hexagon with AB=DEA B = D E, BC=EFB C = E F, CD=FAC D = F A, and ∠A−∠D=∠C−∠F=∠E−∠B\angle A - \angle D = \angle C - \angle F = \angle E - \angle B. Prove that the diagonals ADA D, BEB E, and CFC F are concurrent.

Solution

In all three solutions, we denote θ=∠A−∠D=∠C−∠F=∠E−∠B\theta = \angle A - \angle D = \angle C - \angle F = \angle E - \angle B and assume without loss of generality that θ⩾0\theta \geqslant 0.

Solution 1. Let x=AB=DEx = A B = D E, y=CD=FAy = C D = F A, z=EF=BCz = E F = B C. Consider the points P,QP, Q, and RR such that the quadrilaterals CDEPC D E P, EFAQE F A Q, and ABCRA B C R are parallelograms. We compute
∠PEQ=∠FEQ+∠DEP−∠E=(180∘−∠F)+(180∘−∠D)−∠E=360∘−∠D−∠E−∠F=12(∠A+∠B+∠C−∠D−∠E−∠F)=θ/2 \begin{aligned} \angle P E Q & = \angle F E Q + \angle D E P - \angle E = (180^\circ - \angle F) + (180^\circ - \angle D) - \angle E \\ & = 360^\circ - \angle D - \angle E - \angle F = \frac{1}{2}(\angle A + \angle B + \angle C - \angle D - \angle E - \angle F) = \theta / 2 \end{aligned}
Similarly, ∠QAR=∠RCP=θ/2\angle Q A R = \angle R C P = \theta / 2.

Figure 1

If θ=0\theta = 0, since △RCP\triangle R C P is isosceles, R=PR = P. Therefore AB∥RC=PC∥EDA B \parallel R C = P C \parallel E D, so ABDEA B D E is a parallelogram. Similarly, BCEFB C E F and CDFAC D F A are parallelograms. It follows that ADA D, BEB E and CFC F meet at their common midpoint.

Now assume θ>0\theta > 0. Since △PEQ\triangle P E Q, △QAR\triangle Q A R, and △RCP\triangle R C P are isosceles and have the same angle at the apex, we have △PEQ∼△QAR∼△RCP\triangle P E Q \sim \triangle Q A R \sim \triangle R C P with ratios of similarity y:z:xy : z : x. Thus
△PQR is similar to the triangle with sidelengths y,z, and x. \begin{equation*} \triangle P Q R \text{ is similar to the triangle with sidelengths } y, z, \text{ and } x. \tag{1} \end{equation*}
Next, notice that
RQQP=zy=RAAF \frac{R Q}{Q P} = \frac{z}{y} = \frac{R A}{A F}
and, using directed angles between rays,
(R Q, Q P) = (R Q, Q E) + (Q E, Q P) = (R Q, Q E) + (R A, R Q) = (R A, Q E) = (R A, A F) .\text{(R Q, Q P) = (R Q, Q E) + (Q E, Q P) = (R Q, Q E) + (R A, R Q) = (R A, Q E) = (R A, A F) .}
Thus △PQR∼△FAR\triangle P Q R \sim \triangle F A R. Since FA=yF A = y and AR=zA R = z, (1) then implies that FR=xF R = x. Similarly FP=xF P = x. Therefore CRFPC R F P is a rhombus.

We conclude that CFC F is the perpendicular bisector of PRP R. Similarly, BEB E is the perpendicular bisector of PQP Q and ADA D is the perpendicular bisector of QRQ R. It follows that ADA D, BEB E, and CFC F are concurrent at the circumcenter of PQRP Q R.

Solution 2. Let X=CD∩EFX = C D \cap E F, Y=EF∩ABY = E F \cap A B, Z=AB∩CDZ = A B \cap C D, X′=FA∩BCX' = F A \cap B C, Y′=BC∩DEY' = B C \cap D E, and Z′=DE∩FAZ' = D E \cap F A. From ∠A+∠B+∠C=360∘+θ/2\angle A + \angle B + \angle C = 360^\circ + \theta / 2 we get ∠A+∠B>180∘\angle A + \angle B > 180^\circ and ∠B+∠C>180∘\angle B + \angle C > 180^\circ, so ZZ and X′X' are respectively on the opposite sides of BCB C and ABA B from the hexagon. Similar conclusions hold for X,Y,Y′X, Y, Y', and Z′Z'. Then
∠YZX=∠B+∠C−180∘=∠E+∠F−180∘=∠Y′Z′X′ \angle Y Z X = \angle B + \angle C - 180^\circ = \angle E + \angle F - 180^\circ = \angle Y' Z' X'
and similarly ∠ZXY=∠Z′X′Y′\angle Z X Y = \angle Z' X' Y' and ∠XYZ=∠X′Y′Z′\angle X Y Z = \angle X' Y' Z', so △XYZ∼△X′Y′Z′\triangle X Y Z \sim \triangle X' Y' Z'. Thus there is a rotation RR which sends △XYZ\triangle X Y Z to a triangle with sides parallel to △X′Y′Z′\triangle X' Y' Z'. Since AB=DEA B = D E we have R(AB→)=DE→R(\overrightarrow{A B}) = \overrightarrow{D E}. Similarly, R(CD→)=FA→R(\overrightarrow{C D}) = \overrightarrow{F A} and R(EF→)=BC→R(\overrightarrow{E F}) = \overrightarrow{B C}. Therefore
0→=AB→+BC→+CD→+DE→+EF→+FA→=(AB→+CD→+EF→)+R(AB→+CD→+EF→) \overrightarrow{0} = \overrightarrow{A B} + \overrightarrow{B C} + \overrightarrow{C D} + \overrightarrow{D E} + \overrightarrow{E F} + \overrightarrow{F A} = (\overrightarrow{A B} + \overrightarrow{C D} + \overrightarrow{E F}) + R(\overrightarrow{A B} + \overrightarrow{C D} + \overrightarrow{E F})
If RR is a rotation by 180∘180^\circ, then any two opposite sides of our hexagon are equal and parallel, so the three diagonals meet at their common midpoint. Otherwise, we must have
AB→+CD→+EF→=0→ \overrightarrow{A B} + \overrightarrow{C D} + \overrightarrow{E F} = \overrightarrow{0}
or else we would have two vectors with different directions whose sum is 0→\overrightarrow{0}.

Figure 2

This allows us to consider a triangle LMNL M N with LM→=EF→\overrightarrow{L M} = \overrightarrow{E F}, MN→=AB→\overrightarrow{M N} = \overrightarrow{A B}, and NL→=CD→\overrightarrow{N L} = \overrightarrow{C D}. Let OO be the circumcenter of △LMN\triangle L M N and consider the points O1,O2,O3O_1, O_2, O_3 such that △AO1B\triangle A O_1 B, △CO2D\triangle C O_2 D, and △EO3F\triangle E O_3 F are translations of △MON\triangle M O N, △NOL\triangle N O L, and △LOM\triangle L O M, respectively. Since FO3F O_3 and AO1A O_1 are translations of MOM O, quadrilateral AFO3O1A F O_3 O_1 is a parallelogram and O3O1=FA=CD=NLO_3 O_1 = F A = C D = N L. Similarly, O1O2=LMO_1 O_2 = L M and O2O3=MNO_2 O_3 = M N. Therefore △O1O2O3≅△LMN\triangle O_1 O_2 O_3 \cong \triangle L M N. Moreover, by means of the rotation RR one may check that these triangles have the same orientation.

Let TT be the circumcenter of △O1O2O3\triangle O_1 O_2 O_3. We claim that ADA D, BEB E, and CFC F meet at TT. Let us show that CC, TT, and FF are collinear. Notice that CO2=O2T=TO3=O3FC O_2 = O_2 T = T O_3 = O_3 F since they are all equal to the circumradius of △LMN\triangle L M N. Therefore △TO3F\triangle T O_3 F and △CO2T\triangle C O_2 T are isosceles. Using directed angles between rays again, we get
(T F, T O 3 ) = (F O 3, F T ) and (T O 2, T C ) = (C T, C O 2 ) . 2\text{(T F, T O 3 ) = (F O 3, F T ) and (T O 2, T C ) = (C T, C O 2 ) . 2}
Also, TT and OO are the circumcenters of the congruent triangles △O1O2O3\triangle O_1 O_2 O_3 and △LMN\triangle L M N so we have (T O 3, T O 2 ) = (O N, O M)\text{(T O 3, T O 2 ) = (O N, O M)}. Since CO2C O_2 and FO3F O_3 are translations of NON O and MOM O respectively, this implies
(T O 3, T O 2 ) = (C O 2, F O 3 ) . 3\text{(T O 3, T O 2 ) = (C O 2, F O 3 ) . 3}
Adding the three equations in (2) and (3) gives
(T F, T C) = (C T, F T) = - (T F, T C)\text{(T F, T C) = (C T, F T) = - (T F, T C)}
which implies that TT is on CFC F. Analogous arguments show that it is on ADA D and BEB E also. The desired result follows.

Solution 3. Place the hexagon on the complex plane, with AA at the origin and vertices labelled clockwise. Now A,B,C,D,E,FA, B, C, D, E, F represent the corresponding complex numbers. Also consider the complex numbers a,b,c,a′,b′,c′a, b, c, a', b', c' given by B−A=aB - A = a, D−C=bD - C = b, F−E=cF - E = c, E−D=a′E - D = a', A−F=b′A - F = b', and C−B=c′C - B = c'. Let k=∣a∣/∣b∣k = |a| / |b|. From a/b′=−kei∠Aa / b' = -k e^{i \angle A} and a′/b=−kei∠Da' / b = -k e^{i \angle D} we get that (a′/a)(b′/b)=e−iθ\left(a' / a\right)\left(b' / b\right) = e^{-i \theta} and similarly (b′/b)(c′/c)=e−iθ\left(b' / b\right)\left(c' / c\right) = e^{-i \theta} and (c′/c)(a′/a)=e−iθ\left(c' / c\right)\left(a' / a\right) = e^{-i \theta}. It follows that a′=ara' = a r, b′=brb' = b r, and c′=crc' = c r for a complex number rr with ∣r∣=1|r| = 1, as shown below.

Figure 3

We have
0=a+cr+b+ar+c+br=(a+b+c)(1+r). 0 = a + c r + b + a r + c + b r = (a + b + c)(1 + r) .
If r=−1r = -1, then the hexagon is centrally symmetric and its diagonals intersect at its center of symmetry. Otherwise
a+b+c=0. a + b + c = 0 .
Therefore
A=0,B=a,C=a+cr,D=c(r−1),E=−br−c,F=−br. A = 0, \quad B = a, \quad C = a + c r, \quad D = c(r - 1), \quad E = -b r - c, \quad F = -b r .
Now consider a point WW on ADA D given by the complex number c(r−1)λc(r - 1) \lambda, where λ\lambda is a real number with 0<λ<10 < \lambda < 1. Since D≠AD \neq A, we have r≠1r \neq 1, so we can define s=1/(r−1)s = 1 / (r - 1). From rrˉ=∣r∣2=1r \bar{r} = |r|^2 = 1 we get
1+s=rr−1=rr−rrˉ=11−rˉ=−sˉ. 1 + s = \frac{r}{r - 1} = \frac{r}{r - r \bar{r}} = \frac{1}{1 - \bar{r}} = -\bar{s} .
Now,
W is on BE⟺c(r−1)λ−a∥a−(−br−c)=b(r−1)⟺cλ−as∥b⟺−aλ−bλ−as∥b⟺a(λ+s)∥b. \begin{aligned} W \text{ is on } B E & \Longleftrightarrow c(r - 1) \lambda - a \| a - (-b r - c) = b(r - 1) \Longleftrightarrow c \lambda - a s \| b \\ & \Longleftrightarrow -a \lambda - b \lambda - a s \| b \Longleftrightarrow a(\lambda + s) \| b . \end{aligned}
One easily checks that r≠±1r \neq \pm 1 implies that λ+s≠0\lambda + s \neq 0 since ss is not real. On the other hand,
W on CF⟺c(r−1)λ+br∥−br−(a+cr)=a(r−1)⟺cλ+b(1+s)∥a⟺−aλ−bλ−bsˉ∥a⟺b(λ+sˉ)∥a⟺b∥a(λ+s), \begin{aligned} W \text{ on } C F & \Longleftrightarrow c(r - 1) \lambda + b r \| -b r - (a + c r) = a(r - 1) \Longleftrightarrow c \lambda + b(1 + s) \| a \\ & \Longleftrightarrow -a \lambda - b \lambda - b \bar{s} \| a \Longleftrightarrow b(\lambda + \bar{s}) \| a \Longleftrightarrow b \| a(\lambda + s), \end{aligned}
where in the last step we use that (λ+s)(λ+sˉ)=∣λ+s∣2∈R>0(\lambda + s)(\lambda + \bar{s}) = |\lambda + s|^2 \in \mathbb{R}_{>0}. We conclude that AD∩BE=CF∩BEA D \cap B E = C F \cap B E, and the desired result follows.

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.