Solution

A Solution to Crux Mathematicorum Problem 4941

Crux Mathematicorum Problem 4941, followed by an incircle-inversion proof using the medial triangle of the contact triangle.

Methods Circle inversion · Contact triangle · Homothety

Problem 4941

Proposed by Michel Bataille

Let a non-equilateral triangle ABCABC be inscribed in a circle Γ\Gamma with center OO. Its incircle γ\gamma, with center II, touches BCBC, CACA, and ABAB at DD, EE, and FF, respectively. The circle ΓA\Gamma_A passes through AA and is tangent to IDID at II; the circles ΓB\Gamma_B and ΓC\Gamma_C are defined cyclically. Prove that ΓA\Gamma_A, ΓB\Gamma_B, ΓC\Gamma_C, and the line OIOI have a common point other than II.

A scalene triangle ABC with circumcircle Gamma and incircle gamma. The three circles Gamma A, Gamma B, and Gamma C pass through I and meet again at J star on the line OI.
Figure 1. The three circles and their second common point on the line OI.

Main idea

Invert with respect to the incircle. The three circles become three lines, while AA, BB, and CC become the midpoints of the sides of the contact triangle. A negative homothety then reveals the point at which those lines meet.

Solution

Proof

We first treat the scalene case, in which all three circles ΓA\Gamma_A, ΓB\Gamma_B, and ΓC\Gamma_C are nondegenerate.

Let

P=IA∩EF,Q=IB∩FD,R=IC∩DE,P=IA\cap EF, \qquad Q=IB\cap FD, \qquad R=IC\cap DE,

and let JJ be the circumcenter of △PQR\triangle PQR.

Since AE=AFAE=AF and IE=IFIE=IF, both AA and II lie on the perpendicular bisector of EFEF. Thus PP is the midpoint of EFEF. Moreover, FPFP is the altitude to the hypotenuse AIAI of the right triangle AIFAIF, and hence

IA⋅IP=IF2.IA\cdot IP=IF^2.

Cyclically,

IB⋅IQ=ID2,IC⋅IR=IE2.IB\cdot IQ=ID^2, \qquad IC\cdot IR=IE^2.

Therefore, inversion in γ\gamma maps PP, QQ, and RR to AA, BB, and CC, respectively.

The contact triangle DEF and its medial triangle PQR. The circumcenter J of PQR lies on OI; the lines JP, JQ, and JR are respectively parallel to ID, IE, and IF.
Figure 2. The contact triangle, its medial triangle, and the concurrent lines used in the inversion.

The points PP, QQ, and RR are the midpoints of EFEF, FDFD, and DEDE. The homothety centered at the common centroid of △DEF\triangle DEF and △PQR\triangle PQR, with ratio −12-\tfrac12, maps

(D,E,F)⟼(P,Q,R).(D,E,F)\longmapsto(P,Q,R).

It also maps the circumcenter II of △DEF\triangle DEF to the circumcenter JJ of △PQR\triangle PQR. Consequently,

JP∥ID,JQ∥IE,JR∥IF.JP\parallel ID, \qquad JQ\parallel IE, \qquad JR\parallel IF.

The circumcircle of △PQR\triangle PQR is mapped by the inversion to Γ\Gamma. The centers of two inverse circles are collinear with the center of inversion; hence II, JJ, and OO are collinear.

The points II and OO are distinct. Indeed, if the incenter and circumcenter coincided, the three sides of △ABC\triangle ABC would be chords at the same distance from the center of Γ\Gamma, so they would have equal lengths. The triangle would then be equilateral. We also have J≠IJ\ne I: otherwise the inverse of the circumcircle of △PQR\triangle PQR would be centered at II, forcing O=IO=I.

None of the lines JPJP, JQJQ, and JRJR passes through II. For example, suppose that I∈JPI\in JP. Since P∈IAP\in IA and P≠IP\ne I, the lines JPJP and IAIA pass through the distinct points II and PP, so they coincide. Since JP∥IDJP\parallel ID, the line IAIA has the same direction as IDID; as both pass through II, they coincide. The angle bisector from AA would then also be perpendicular to BCBC, forcing AB=ACAB=AC, contrary to the scalene assumption. The other two cases are cyclic.

Let J∗J^\ast be the inverse of JJ with respect to γ\gamma. A line not passing through the center of an inversion is mapped to a circle through that center, and the tangent to the image circle there is parallel to the original line. Thus JPJP, JQJQ, and JRJR are mapped to ΓA\Gamma_A, ΓB\Gamma_B, and ΓC\Gamma_C, respectively. The line OIOI is mapped to itself.

All four lines JPJP, JQJQ, JRJR, and OIOI pass through JJ. Their images therefore pass through J∗J^\ast. Since J≠IJ\ne I, the point J∗J^\ast is distinct from II, which proves the result.

□