Two circles and have common external tangents and meeting at . Suppose touches at and touches at . A circle through and intersects again at and again at , such that quadrilateral is convex.
Suppose lines and meet at point , while lines and meet at point . Show that , , are collinear.
Solution
First solution, elementary (original) We have , from which it follows
Moreover, if we denote by and the radii of and , then
the last step by the law of sines.
This solves the problem up to configuration issues: we claim that and both lie inside . WLOG .
* The former is since lies outside segments and , since we assumed was convex.
* For the latter, we note that lies inside both and in fact on the radical axis of the two circles (since was an interior point of both chords and ). In particular, is contained inside , and moreover , and this is enough to imply the result.
Second solution, invasive This is based on the solution posted by kapilpavase on AoPS. Consider the inversion at swapping and ; we let it send to , to , to , to , as shown. Draw circles and .
Claim — Points and lie on the radical axis of and .
Proof. Because and .
Claim — Point has equal power to and .
Proof. Since , quadrilateral is cyclic too, so by radical axis with and we find lies on . Similarly, lies on . Thus, is the center of negative inversion between and .
But since and moreover
we conclude that and are congruent. As was the center of negative inversion between them, we're done.
Third solution, projective (Nikolai Beluhov) We start with some definitions. Let touch at , touch at , , , line meet again at , line meet again at , and lines , , and meet line at , , and . Thus the desired statement is equivalent to .
Claim — .
Proof. Note that , so is cyclic. Likewise is cyclic.
Consider the inversion with center which swaps and ; it also swaps the pairs and . Since is cyclic, is on , and is on , it also swaps ; similarly it swaps .
Thus as desired.
With this claim, the remainder of the proof is chasing cross-ratios:
implies as desired.
Fourth solution by untethered moving points Fix , and , and let and meet at and . By the radical axis theorem, lies on .
Thus we instead as a variable point on line and let , . By definition, has degree 1 and has degree 0.
We apply Zack's lemma to untethered point . Note that and move projectively on conics, and therefore have degree 2. Then, lines and each have degree at most , and so their intersection has degree at most . But when , the lines and are the same, so Zack's lemma implies that
Thus the assertion that are collinear (which for example can be seen as a certain vanishing determinant) is a statement of degree at most . Thus it suffices to find 5 values of (other than , which we used already). This is remarkably easy:
1.
When or , then and the statement is obvious
2.
When , say, then and so lies on as well. The case is symmetric.
3.
Finally, take at infinity along . Then and are the other tangency points of the circles with , and so , so .
This finishes the problem.