Let ABC be an acute-angled triangle with AB<AC, incenter I, and let M be the midpoint of major arc BAC. Suppose the perpendicular line from A to segment BC meets lines BI, CI, and MI at points P, Q, and K respectively. Prove that the A-median line in △AIK passes through the circumcentre of △PIQ.
Solution
Solution A Observe that ∠PIQ=90−2∠A, ∠IQP=90−2∠C and ∠IPQ=90−2∠B. Thus, if △DEF is the orthic triangle of △IPQ, then it is similar to △ABC. Let ℓ be the I midline in △IPQ and O be the circumcenter. Now, let X=ℓ∩AO and K′ be the reflection of I across X. Then, we just want that K′,K coincide or equivalently that I,M,X are collinear. Now, with respect to triangle IPQ, AI is tangent to the circumcircle of IPQ as ∠AIQ=90−2∠B as required. Let H be the orthocenter of △IPQ and N be the midpoint of PQ. Then, we have that ∠AIM=∠NHD so we just want ∠NHD+∠AIX=180∘. But ∠AIX=90∘+∠OIX and ∠NHD=90∘−∠HND. Thus, we just want that ∠HND=∠OIX Taking homothety with dilation factor +2 from I, we have X going to K′ which is now the intersection of PQ and the line through the antipode of I in IPQ and the point R on (IPQ) such that AR is tangent to (IPQ). Now let HN∩(IPQ)=S1,S2 where S2 is the antipode of I and let H′ be the reflection of H in PQ. Now, we just want ∠HND=∠HS2H′=∠S1IH. Thus, we want ∠S1IH=∠OIK′ or that IK′ and IS1 are isogonal in ∠QIP. Now, performing bc and reflection in △IPQ, we get that S1 and K′ interchange. Thus, IK′ and IS1 are isogonal in ∠QIP as required. Thus, we are done. □
Let O be the circumcenter of △PIQ, S be the point on circumcircle of △PIQ so that IS⊥AJ, R be the antipode of I in the circumcircle of △PIQ. I will use the following: Lemma 1.1 Let T be a point on the line through I perpendicular to AI. Then (IB,IC;IT,IM)=−1. Proof. Let IB,IC be the B,C-excenters. Then we can project onto IBIC and reduce to showing that M is the midpoint of IBIC which is known. □ Now, we begin investigating the problem at hand: Lemma 1.2IO⊥AI. Proof. This is direct: ∠AIP=90∘−B/2 and ∠PIO=90∘−∠PQO=90∘−(90∘−B/2)=B/2. □ Proposition 1.3S,R,K are collinear. Proof. Let the circumcircle of △PIQ be denoted by ω. From Lemma 1.2, AI is tangent to ω. Therefore, IS is the I-symmedian of △PIQ. Hence, (P,Q;S,I)=−1. Projecting from R, we get (RP,RQ;RS,RI)=−1, hence if RI∩PQ=X and RS∩PQ=K′, then (P,Q;X,K′)=−1. On the other hand, we know (IB,IC;IR,IM)=−1 from Lemma 1.1, so projecting onto PQ we get (P,Q;X,K)=−1. Thus K=K′ which finishes the proof. □ Proposition 1.4AO bisects IK. Proof.RS⊥SI since RI is the diameter, hence RK∥AO. But O is the midpoint of RI, hence AO bisects IK by midpoint theorem. □
Solution C Let J be the A-excentre of △ABC and L the point on BC with line LJ is tangent to the A-excircle. Let N be the midpoint of arc BC not containing A of the circumcircle of ABC. Let S be the point on BC with ∠ANS=90∘. Let T be the point on BC with TN⊥MI. Claim.S is the midpoint of LT.
Proof. Note that by the Incenter-Excenter Lemma, N is the midpoint of IJ so JL⊥AI and by standard mixtilinear facts, TI⊥AI. Thus, L,S,T are obtained by drawing perpendicular lines to line AI from the points J,N,I respectively and intersecting with line BC. Since N is the midpoint of JI, then S is the midpoint of LT as desired. □ Now observe that △IPQ∼△JBC so the spiral similarity mapping IPQ to JBC consists of a 90° rotation as PQ⊥BC. Since AI is tangent to (PIQ) and A lies on PQ, the spiral similarity sends A to L and the line passing through A and the circumcentre of PIQ to the line LN. So it suffices to show that the line through A perpendicular to LN bisects IK. Thus we want cross ratio of lines AP,AI, line through A parallel to IK and line through A perpendicular to LN form a harmonic bundle. Taking directions perpendicular to this pencil from N, we get the four lines N∞BC (as BC⊥AP), NS (as NS⊥AI), NL and the line NT as NT⊥KI. Projecting this pencil at N to BC we get that this cross ratio equals since (∞BCS;LT)=−1 as S is the midpoint of LT. This concludes the proof. □
Solution D Let D,E,F be the foot of perpendiculars from I to BC,CA,AB respectively. Let N be the midpoint of BC. Suppose the foot of perpendicular from N onto EF is L. We use that I,L,M are collinear. This can be proven in many ways: one sketch of the argument goes as follows: If X (resp. Y) is the foot of perpendicular of B (resp. C) onto CI (resp. BI), then X,Y lie on EF, and N is the midpoint of XY. On the other hand, if IB,IC are the B-, C-excenters, then M is the midpoint of IBIC. The result then follows on noting that XY∥IBIC, and considering a homothety at I. Now toss the diagram on the complex plane. Let the incircle of △ABC be the unit circle, so I is the origin, and let D=1, E=e, F=f without loss of generality. Now, it is well known that the intersection of tangents to the unit circle at points x,y is x+y2xy. Therefore, A=a=e+f2ef, B=b=f+12f, C=c=e+12e. Thus, the midpoint of BC is N=e+1e+f+1f. Hence, the foot of perpendicular from N to EF is L=l=21(e+f+e+1e+f+1f−ef(e+11+f+11))=2(e+1)(f+1)(e+f)(e+f+2). We know D=1, therefore the line perpendicular to BC from A is the set of points z with Im(z)=Im(A), i.e. z−zˉ=e+f2(ef−1). Now, P lies on BI, therefore p=λb for some λ∈R. Note that bˉ=f1⋅b, therefore, pˉ=f1⋅p. Thus, p−pˉ=f(f−1)⋅p⟹p=(e+f)(f−1)2f(ef−1)⋅Similarly, q=(e+f)(e−1)2e(ef−1). With the same idea, one checks that lˉ=2ef(e+1)(f+1)(e+f)(2ef+e+f)=l⋅ef(e+f+2)(2ef+e+f), therefore, k−kˉ=ef(e+f+2)(ef−1)(e+f)⋅k⟹k=(e+f)22ef(e+f+2). We know that the circumcenter of a circle passing through origin, p,q is given by pˉq−pqˉpq(pˉ−qˉ). Note that p=(e+f)2(ef−1)⋅p1, and q=(e+f)2(ef−1)⋅q1 where p1,q1 are f−1f,e−1e respectively. Thus the circumcenter of △IPQ is (e+f)2(ef−1)⋅t where t is the circumcenter of △IP1Q1. This simplifies calculation, and we get The circumcenter of △IPQ=(e+f)2(ef−1)⋅(1−f)(e−1)e−(1−e)(f−1)f(e−1)(f−1)ef(1−f1−1−e1)=(e+f)(e−1)(f−1)2ef(ef−1). Thus, we need to show that (e+f)2ef, (e+f)(e−1)(f−1)2ef(ef−1), (e+f)2ef(e+f+2) are collinear. Multiplying by 2ef(e+f), we reduce to proving 1, (e−1)(f−1)ef−1, 2(e+f)e+f+2 are collinear for which we consider: 2(e+f)e+f+2−1(e−1)(f−1)ef−1−1=2(e+f)2−e−f(e−1)(f−1)e+f−2=−(e−1)(f−1)2(e+f) The above quantity may be easily checked to equal the conjugate, so we are done.
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.