Given a circle C, a line L that doesn't intersect C, and a point P that is not on L and not inside C,...
geometry/construction/tangent.circle New
New entry, edited from material collected for the archive and not yet in a released version.
Problem
Given a circle C, a line L that doesn't intersect C, and a point P that is not on L and not inside C, construct with compass and straightedge a circle passing through P and tangent to both L and C.
Show solution
I first saw that one back in high school, many (too many) years ago. The solution was a complicated ruler-and-compass construction. However, it is easy to understand such a solution if you use the idea of inversion, which I didn't know about way back then. First, P and C must both be on the same side of L, for otherwise there is obviously no solution. Let X denote the unknown circle we have to find. Invert with centre P and any convenient radius. Let L', C' and X' denote the inverses of L, C and X respectively. Then L' is a circle passing through P, and it is easy to construct it. C' is another circle, which is also easily constructed. The unknown X' will be a straight line that is tangent to both L' and C'. After you have drawn L' and C', it is easy to find all their common tangent lines, and X' can be any one of them. Then invert X' to get X. This shows that there should be four possible answers. (Parenthetical remark -- we are given that L does not intersect C, so L' does not meet C', and these two circles will have four common tangent lines. However, this approach will also work even if L meets C, as long as P is outside C. In this more general case, there will be fewer than four answers.) There are many steps involved, so the full detailed construction will be lengthy. However, this approach makes it easy to understand that lengthy step-by-step construction. One simplification results from choosing a good value for "any convenient radius". Let c be the centre of C, and let D be a point on C such that the line PD is perpendicular to the line Dc. (You can construct D by drawing a circle with diameter Pc.) If you use PD as the radius of inversion, then you get the simplification that C' = C, so the steps used to construct C' can be omitted. -- Barry Wolk <wolk@ccm.umanitoba.ca> [ Editor: the problem was posted by Jonathan Haas <positron@Starbase.NeoSoft.COM> from a Microsoft internal puzzles group. ]