Maths Olympiad Prep

Library / /6 of 25

Geometry Difficulty 7.7 National Olympiad, round 2 Prove it European Girls' Mathematical Olympiad (EGMO)

Problem:

Let Ω\Omega be the circumcircle of the triangle ABCA B C. The circle ω\omega is tangent to the sides ACA C and BCB C, and it is internally tangent to Ω\Omega at the point PP. A line parallel to ABA B and intersecting the interior of triangle ABCA B C is tangent to ω\omega at QQ.
Prove that ACP=QCB\angle A C P=\angle Q C B.

Solutions — 6

Solution 1

Solution:

Assume that ω\omega is tangent to ACA C and BCB C at EE and FF, respectively and let PE,PF,PQP E, P F, P Q meet Ω\Omega at K,L,MK, L, M, respectively. Let II and OO denote the respective centres of ω\omega and Ω\Omega, and consider the homothety H\mathscr{H} that maps ω\omega onto Ω\Omega. Now KK is the image of EE under H\mathscr{H}, and EIACE I \perp A C. Hence OKACO K \perp A C, and thus KK is the midpoint of the arc CAC A. Similarly, LL is the midpoint of the arc BCB C and MM is the midpoint of the arc BAB A. It follows that arcs LML M and CKC K are equal, because
B M = M A B L + L M = M K + K A L C + L M = M K + C K 2 L M + M C = M C +2 C K\text{B M = M A B L + L M = M K + K A L C + L M = M K + C K 2 L M + M C = M C +2 C K}
Thus arcs FQF Q and DED E of ω\omega are equal, too, where DD is the intersection of CPC P with ω\omega. Since CEC E and CFC F are tangents to ω\omega, this implies that DEC=CFQ\angle D E C=\angle C F Q. Further, CE=CFC E=C F, and thus triangles CEDC E D and CFQC F Q are congruent. In particular, ECD=QCF\angle E C D=\angle Q C F, as required.

Figure 1

A Variant. As above, we show that arcsFQ\operatorname{arcs} F Q and DED E of ω\omega are equal, which implies that DEFQD E F Q is an isoceles trapezoid, and so we have FED=QFE\angle F E D=\angle Q F E. Together with FQ=DE|F Q|=|D E|, this implies that, since EE and FF are images of each other under reflection in the angle bisector CIC I of C\angle C, so are the segments [EQ][E Q] and [FD][F D], and, in particular, DD and QQ. In turn, this yields ECD=QCF\angle E C D=\angle Q C F, as required.

Remark. Let JJ denote the incentre of ABCA B C. By Sawayama's theorem, JJ is the midpoint of [EF][E F], i.e. PJP J is a median of PFEP F E. Since CC is the intersection of the tangents ACA C and BCB C to the circumcircle of PFEP F E at EE and FF, respectively, PCP C is a symmedian of PFEP F E. Thus CPE=FPJ\angle C P E=\angle F P J. But, since the arcs FQF Q and DED E of ω\omega are equal, CPE=FPQ\angle C P E=\angle F P Q. This shows that JJ lies on the line PQP Q.

Another Variant. We show that arcsQE\operatorname{arcs} Q E and FDF D are equal, and then finish as in the main solution. Let BPB P meet ω\omega again at ZZ. Consider the homothety H\mathscr{H} that maps ω\omega onto Ω\Omega. Under H,DC\mathscr{H}, D \mapsto C and ZBZ \mapsto B, so DZCBD Z \parallel C B. (This also follows by considering the common tangent to ω\omega and Ω\Omega, and tangential angles.) Now, by power of a point,
BF2=BZBP,CF2=CDCP B F^{2}=B Z \cdot B P, \quad C F^{2}=C D \cdot C P
Now DZCBD Z \parallel C B implies BZ/BP=CD/CPB Z / B P=C D / C P, and so, dividing the two previous equations by each other, and taking square roots, BF/CF=BP/CPB F / C F=B P / C P. Hence PFP F bissects angle BPC\angle B P C. Now let BPF=FPC=β\angle B P F=\angle F P C=\beta. By tangential angles, it follows that CFD=β\angle C F D=\beta. Further, BAC=BPC=2β\angle B A C=\angle B P C=2 \beta. Let the tangent to ω\omega through QQ and parallel to ABA B meet ACA C at XX. Then QXC=2β\angle Q X C=2 \beta, so, since XQ=XEX Q=X E by tangency, QEX=β\angle Q E X=\beta. By tangential angles, it follows that arcsFD\operatorname{arcs} F D and QEQ E are equal, as claimed.

Figure 2

Solution 2

Solution:

Let II and OO denote the respective centres of ω\omega and Ω\Omega. Observe that CIC I is the angle bisector of angle C\angle C, because ω\omega is tangent to ACA C and BCB C. Consider the homothety H\mathscr{H} that maps ω\omega onto Ω\Omega. Let MM be the image of QQ under H\mathscr{H}. By construction, IQABI Q \perp A B, so OMABO M \perp A B. Thus the diameter OMO M of Ω\Omega passes through the midpoint of the arcAB\operatorname{arc} A B of Ω\Omega, which also lies on the angle bisector CIC I. This implies that ICM=90\angle I C M=90^{\circ}. We next show that P,I,Q,CP, I, Q, C lie on a circle. Notice that
PQI=9012QIP=9012MOP=90(180PCM)=(PCI+ICM)90=PCI \begin{aligned} \angle P Q I & =90^{\circ}-\frac{1}{2} \angle Q I P=90^{\circ}-\frac{1}{2} \angle M O P=90^{\circ}-\left(180^{\circ}-\angle P C M\right) \\ & =(\angle P C I+\angle I C M)-90^{\circ}=\angle P C I \end{aligned}
Hence P,I,Q,CP, I, Q, C lie on a circle. But PI=IQP I=I Q, so CIC I is the angle bisector of PCQ\angle P C Q. Since CIC I is also the angle bisector of angle C\angle C, it follows that ACP=QCB\angle A C P=\angle Q C B, as required.

Figure 3

A Variant. We show that PIQCP I Q C is cyclic by chasing angles. Define α=BAC\alpha=\angle B A C, β=CBA\beta=\angle C B A and γ=ACP\gamma=\angle A C P. For convenience, we consider the configuration where AA and PP lie one the same side of the angle bisector CIC I of C\angle C. In this configuration,
PCI=12ACBACP=9012α12βγ. \angle P C I=\frac{1}{2} \angle A C B-\angle A C P=90^{\circ}-\frac{1}{2} \alpha-\frac{1}{2} \beta-\gamma .
Now notice that PBA=ACP=γ\angle P B A=\angle A C P=\gamma, and therefore CAP=180βγ\angle C A P=180^{\circ}-\beta-\gamma, whence PAB=180αβγ\angle P A B=180^{\circ}-\alpha-\beta-\gamma. Further, POP O is a diameter of Ω\Omega, and therefore APO=90γ\angle A P O=90^{\circ}-\gamma. Let ABA B and POP O intersect at TT. Then
BTO=180PABAPO=α+β+2γ90 \angle B T O=180^{\circ}-\angle P A B-\angle A P O=\alpha+\beta+2 \gamma-90^{\circ}
But QIABQ I \perp A B by construction, and thus
OIQ=90BTO=180αβ2γQIP=180OIQ=α+β+2γPQI=9012α12βγ. \begin{aligned} & \angle O I Q=90^{\circ}-\angle B T O=180^{\circ}-\alpha-\beta-2 \gamma \\ & \Longrightarrow \quad \angle Q I P=180^{\circ}-\angle O I Q=\alpha+\beta+2 \gamma \\ & \Longrightarrow \quad \angle P Q I=90^{\circ}-\frac{1}{2} \alpha-\frac{1}{2} \beta-\gamma . \end{aligned}
Hence ICQ=PQI\angle I C Q=\angle P Q I, and thus PIQCP I Q C is cyclic. Since PI=QIP I=Q I, it follows that CIC I is the angle bisector of PCQ\angle P C Q, which completes the proof.

Solution 3

Solution:

Let II and OO denote the respective centres of ω\omega and Ω\Omega. Let DD be the second point of intersection of CPC P with ω\omega, and let \ell denote the tangent to ω\omega at DD, which meets ACA C at SS. Hence IDI D \perp \ell. By construction, P,I,OP, I, O lie one a line, and hence the isosceles triangles PIDP I D and POCP O C are similar. In particular, it follows that OCO C \perp \ell, so CC is the midpoint of the arc of Ω\Omega defined by the points of intersection of \ell with Ω\Omega. It is easy to see that this implies that
DSC=ABC \angle D S C=\angle A B C
Under reflection in the angle bisector CIC I of C,\angle C, \ell is thus mapped to a tangent to ω\omega parallel to ABA B and intersecting the interior of ABCA B C, since ω\omega is mapped to itself under this reflection. In particular, DD is mapped to QQ, and thus QCB=ACD\angle Q C B=\angle A C D, as required.

Figure 4

Remark. Conceptually, this solution is similar to Solution 1, but here, we proceed more directly via the reflectional symmetry. Therefore, this solution links Solution 1 to Solution 4, in which we use an inversion.

Solution 4

Solution:

Let the tangent to ω\omega at QQ meet ACA C and BCB C at XX and YY, respectively. Then AC/XC=BC/YCA C / X C=B C / Y C, and thus there is a radius rr such that r2=ACYC=BCXCr^{2}=A C \cdot Y C=B C \cdot X C. Let Γ\Gamma denote the circle with centre CC and radius rr, and consider the inversion I\mathscr{I} in the circle Γ\Gamma. Under I\mathscr{I},
AAA \longmapsto A^{\prime}, the point on the ray CAC A satisfying CA=CYC A^{\prime}=C Y;
BBB \longmapsto B^{\prime}, the point on the ray CBC B satisfying CB=CXC B^{\prime}=C X;
Ω\Omega \longmapsto the line ABA^{\prime} B^{\prime};
ωω\omega \longmapsto \omega^{\prime}, the excircle of CABC A^{\prime} B^{\prime} opposite CC;
PPP \longmapsto P^{\prime}, the point where ω\omega^{\prime} touches ABA^{\prime} B^{\prime};
In particular, ω\omega^{\prime}, the image of ω\omega, is a circle tangent to AC,BCA C, B C and ABA^{\prime} B^{\prime}, so it is either the excircle of CABC A^{\prime} B^{\prime} opposite CC, or the incircle of CABC A^{\prime} B^{\prime}. Let ω\omega be tangent to BCB C at FF, and let FF^{\prime} be the image of FF under I\mathscr{I}. Then CFCF=BCXCC F \cdot C F^{\prime}=B C \cdot X C. Now CF<BCC F<B C, so CF>CX=CBC F^{\prime}>C X=C B^{\prime}. Hence ω\omega^{\prime} cannot be the incircle, so ω\omega^{\prime} is indeed the excircle of CABC A^{\prime} B^{\prime} opposite CC.
Now note that ω\omega is the excircle of CXYC X Y opposite CC. The reflection about the angle bisector of C\angle C maps XX to B,YB^{\prime}, Y to AA^{\prime}. It thus maps the triangle CXYC X Y to CBA,ωC B^{\prime} A^{\prime}, \omega to ω\omega^{\prime} and, finally, QQ to PP^{\prime}. It follows that ACP=ACP=QCB\angle A C P=\angle A C P^{\prime}=\angle Q C B, as required.

Figure 5

Solution 5

Solution:

Let rr be the radius such that r2=ACBCr^{2}=A C \cdot B C. Let J\mathscr{J} denote the composition of the inversion I\mathscr{I} in the circle of centre CC and radius rr, followed by the reflection in the angle bisector of C\angle C. Under J\mathscr{J},
AB,BA;A \longmapsto B, B \mapsto A ;
Ω\Omega \longmapsto the line ABA B;
ωω\omega \longmapsto \omega^{\prime}, the excircle of ABCA B C opposite the vertex CC;
PQP \longmapsto Q^{\prime}, the point where ω\omega^{\prime} touches ABA B;
In particular, note that the image ω\omega^{\prime} of ω\omega under J\mathscr{J} is a circle tangent to AC,BCA C, B C and ABA B, so it is either the incircle of ABCA B C, or the excircle opposite vertex CC. Observe that rmin{AC,BC}r \geqslant \min \{A C, B C\}, so the image of the points of tangency of ω\omega must lie outside ABCA B C, and thus ω\omega^{\prime} cannot be the incircle. Thus ω\omega^{\prime} is the excircle opposite vertex CC as claimed. Further, the point of tangency PP is mapped to QQ^{\prime}.
Now, since the line CPC P is mapped to itself under the inversion I\mathscr{I}, and mapped onto CQC Q^{\prime} under J,CP\mathscr{J}, C P and CQC Q^{\prime} are images of each other under reflection in the angle bisector of C\angle C. But C,Q,QC, Q, Q^{\prime} lie on a line for there is a homothety with centre CC that maps ω\omega onto the excircle ω\omega^{\prime}. This completes the proof.

Figure 6

Solution 6

Solution:

Assume that ω\omega is tangent to ACA C and BCB C at EE and FF, respectively. Assume that CPC P meets ω\omega at DD. Let II and OO denote the respective centres of ω\omega and Ω\Omega. To set up a solution in the complex plane, we take the circle ω\omega as the unit circle centered at the origin of the complex plane, and let POP O be the real axis with o>0o>0, where we use the convention that lowercase letters denote complex coordinates of corresponding points in the plane denoted by uppercase letters.
Now, a point ZZ on the circle Ω\Omega satisfies
zo2=(o+1)2zzo(z+z)2o1=0 |z-o|^{2}=(o+1)^{2} \quad \Longleftrightarrow \quad z z^{*}-o\left(z+z^{*}\right)-2 o-1=0
The triangle ABCA B C is defined by the points EE and FF on ω\omega, the intersection CC of the corresponding tangents lying on Ω\Omega. Thus c=2ef/(e+f)c=2 e f /(e+f), and further
co2=(o+1)2cco(c+c)2o1=0, |c-o|^{2}=(o+1)^{2} \quad \Longleftrightarrow \quad c c^{*}-o\left(c+c^{*}\right)-2 o-1=0,
and this is the equality defining oo. The points AA and BB are the second intersection points of Ω\Omega with the tangents to ω\omega at EE and FF respectively. A point ZZ on the tangent through EE is given by z=2ee2zz=2 e-e^{2} z^{*}, and thus AA and CC satisfy
(2ee2z)zo(2ee2z+z)2o1=0e2z2+(2e+oe2o)z(2eo+2o+1)=0z2(2e+ooe2)z+(2eo+2oe2+e2)=0 \begin{aligned} \left(2 e-e^{2} z^{*}\right) z^{*} & -o\left(2 e-e^{2} z^{*}+z^{*}\right)-2 o-1=0 \\ & \Longleftrightarrow-e^{2} z^{* 2}+\left(2 e+o e^{2}-o\right) z^{*}-(2 e o+2 o+1)=0 \\ & \Longleftrightarrow z^{* 2}-\left(2 e^{*}+o-o e^{* 2}\right) z^{*}+\left(2 e^{*} o+2 o e^{* 2}+e^{* 2}\right)=0 \end{aligned}
since e=1|e|=1. Thus
a+c=2e+ooe2a=2efe+f+o(1e2) a^{*}+c^{*}=2 e^{*}+o-o e^{* 2} \quad \Longrightarrow \quad a^{*}=\frac{2 e^{*} f}{e+f}+o\left(1-e^{* 2}\right)
and similarly
b=2fef+e+o(1f2) b^{*}=\frac{2 f^{*} e}{f+e}+o\left(1-f^{* 2}\right)
Then
ba=2(efef)e+f+o(e2f2)=2ef(f2e2)e+f+o(e2f2)=(f2e2)(2efe+fo)=(f2e2)(co). \begin{aligned} b^{*}-a^{*} & =\frac{2\left(e f^{*}-e^{*} f\right)}{e+f}+o\left(e^{* 2}-f^{* 2}\right) \\ & =\frac{2 e f\left(f^{* 2}-e^{* 2}\right)}{e+f}+o\left(e^{* 2}-f^{* 2}\right) \\ & =\left(f^{* 2}-e^{* 2}\right)\left(\frac{2 e f}{e+f}-o\right)=\left(f^{* 2}-e^{* 2}\right)(c-o) . \end{aligned}
Now let ZZ be a point on the tangent to ω\omega parallel to ABA B passing through QQ. Then
z=2qq2zzq=qq2z=q2(zq) z=2 q-q^{2} z^{*} \quad \Longleftrightarrow \quad z-q=q-q^{2} z^{*}=-q^{2}\left(z^{*}-q^{*}\right)
for q=1|q|=1, and thus
baba=zqzq=q2(zq)zq=q2. \frac{b-a}{b^{*}-a^{*}}=\frac{z-q}{z^{*}-q^{*}}=\frac{-q^{2}\left(z^{*}-q^{*}\right)}{z^{*}-q^{*}}=-q^{2} .
It follows that
q2=baba=(f2e2)(co)(f2e2)(co)=e2f2coco=e2f2(co)2co2=e2f2(co)2(1+o)2 \begin{aligned} q^{2} & =-\frac{b-a}{b^{*}-a^{*}}=-\frac{\left(f^{2}-e^{2}\right)\left(c^{*}-o\right)}{\left(f^{* 2}-e^{* 2}\right)(c-o)}=e^{2} f^{2} \frac{c^{*}-o}{c-o} \\ & =e^{2} f^{2} \frac{\left(c^{*}-o\right)^{2}}{|c-o|^{2}}=e^{2} f^{2} \frac{\left(c^{*}-o\right)^{2}}{(1+o)^{2}} \end{aligned}
where we have used (1). In particular,
q=efco1+o q=e f \frac{c^{*}-o}{1+o}
where the choice of sign is to be justified a posteriori. Further, the point DD satisfies
dp=dpdp=cpcpd=cpcp1=c+1c+1, -d p=\frac{d-p}{d^{*}-p^{*}}=\frac{c-p}{c^{*}-p^{*}} \quad \Longrightarrow \quad d=-\frac{c-p}{c^{*} p-1}=\frac{c+1}{c^{*}+1},
using p=1p=-1 to obtain the final equality.
Now, it suffices to show that (i) DQEFCID Q \parallel E F \perp C I and (ii) the midpoint of [DQ][D Q] is on CIC I. The desired equality then follows by symmetry with respect to the angle bisector of the angle ACB\angle A C B. Notice that (i) is equivalent with
dqdq=efefdq=ef \frac{d-q}{d^{*}-q^{*}}=\frac{e-f}{e^{*}-f^{*}} \quad \Longleftrightarrow \quad d q=e f
for [DQ][D Q] and [EF][E F] are chords of ω\omega. But
dq=efc+1c+1efco1+o=ef(c+1)(co)=(c+1)(1+o)cco(c+c)2o1=0. \begin{aligned} d q=e f & \Longleftrightarrow \frac{c+1}{c^{*}+1} e f \frac{c^{*}-o}{1+o}=e f \quad \Longleftrightarrow \quad(c+1)\left(c^{*}-o\right)=\left(c^{*}+1\right)(1+o) \\ & \Longleftrightarrow c c^{*}-o\left(c+c^{*}\right)-2 o-1=0 . \end{aligned}
The last equality is precisely the defining relation for o,(1)o,(1). This proves (i). Further, the midpoint of [DQ][D Q] is 12(d+q)\frac{1}{2}(d+q), so it remains to check that
dq=d+qd+q=cc=ef d q=\frac{d+q}{d^{*}+q^{*}}=\frac{c}{c^{*}}=e f
where the first equality expresses that [DQ][D Q] is a chord of ω\omega (obviously) containing its midpoint, the second equality expresses the alignment of the midpoint of [DQ],C[D Q], C and II, and the third equality follows from the expression for cc. But we have just shown that dq=efd q=e f. This proves (ii), justifies the choice of sign for qq a posteriori, and thus completes the solution of the problem.

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.