3

I have read a lot about the nine point circle. Finding the nine point circle of a triangle is straightforward. But what about going the other way?

Let $w$ be a circle. Given nine arbitrary concyclic points on $w$, how can one find the triangle (if it exists) for which $w$ is the nine-point circle?


In other words, let's say there exists a triangle, $t$, where

  • $P$, $Q$, and $R$ are the midpoints of each side of $t$
  • $P'$, $Q'$, and $R'$ are the foot of each altitude of $t$
  • $P''$, $Q''$, and $R''$ are the midpoints of the line segments from each vertex of $t$ to the orthocenter
  • $P$, $Q$, $R$, $P'$, $Q'$, $R'$, $P''$, $Q''$, and $R''$ all lie on circle, $c$

Given $P$, $Q$, $R$, $P'$, $Q'$, $R'$, $P''$, $Q''$, $R''$, and $c$, find $t$

Note: there may exist multiple triangles which fit the criteria for $t$, or there may exist no triangles which fit the criteria for $t$. Thus, there are two halves to the problem:

a) Determine of a triangle $t$ exists whose significant points are $P$, $Q$, $R$, $P'$, $Q'$, $R'$, $P''$, $Q''$, and $R''$

b) If $t$ exists, find the vertices of $t$

retrovius
  • 223
  • Are you sure that every set of nine concyclic points are the nine-point circle for some triangle? Suppose the points are all very close together one one side of a large circle? – Paul Sinclair Feb 29 '20 at 16:03
  • @retrovius You imply that if nine arbitrary points are given on a circle, then the originating triangle can be determined (upto rotation maybe) right? – Narasimham Feb 29 '20 at 19:36

1 Answers1

6

I've significantly re-worked this answer. For a previous version, see the Edit History.


"The" nine points of a triangle fall into three types:

  1. the three midpoints of its sides ($P$, $Q$, $R$ in the figure),
  2. the three points ($P'$, $Q'$, $R'$) half-way between its orthocenter and its vertices (we'll call these "ortho-midpoints"), and
  3. the three feet ($P''$, $Q''$, $R''$) of its altitudes.

These points naturally determine three concyclic right triangles:

enter image description here

By Thales' Theorem, each hypotenuse is necessarily a diameter of the nine-point circle. Each diameter has a triangle midpoint and an ortho-midpoint as its endpoints. The left-over points are the altitude feet, which determine the original triangle's orthic triangle. These elements happen to be related in an interesting way:

Fun Fact. The midpoint/ortho-midpoint diameters must be the perpendicular bisectors of the sides of the orthic triangle.

The figure shows the situation for diameter $\overline{PP'}$ and orthic triangle side $\overline{Q''R''}$:

enter image description here

For proof, we note that, since $\angle BQ''C$ and $\angle BR''C$ are right angles, Thales tells us that $Q''$ and $R''$ live on a circle with diameter $\overline{BC}$ (and, therefore, with center $P$). Similarly, with $G$ the orthocenter of $\triangle ABC$, we have right angles $\angle AQ''G$ and $\angle AR''G$, so that $Q''$ and $R''$ live on a circle with center $P'$. Thus, $\overline{Q''R''}$ is a chord common to the two circles, so that it must be perpendicular to, and bisected by, the line $\overline{PP'}$ connecting the centers. $\square$

The Fun Fact provides the "only if" portion of this characterization of viable nine-point sets:

Theorem. A set of nine distinct concyclic points is "the" nine-point set of some triangle if and only if it can partitioned into three diametric pairs $\{P,P'\}$, $\{Q,Q'\}$, $\{R,R'\}$ and a triad $\{P'', Q'', R''\}$ such that $$\overline{PP'}\perp\overline{Q''R''} \qquad \overline{QQ'}\perp\overline{R''P''} \qquad \overline{RR'}\perp\overline{P''Q''}$$

For the "if" part, we provide a construction a quartet of solution triangles. To begin, one can show (as @mathlove did in this answer)

Lemma. The incenter $G$ and excenters $A$, $B$, $C$ of $\triangle P''Q''R''$ form an orthocentric system such that $\triangle ABC$, $\triangle GBC$, $\triangle AGC$, and $\triangle ABG$ have altitude feet $P''$, $Q''$, $R''$. No other triangles have these altitude feet.

enter image description here

The four triangles determined by the orthocentric system share a common nine-point set, namely: the three altitude feet ($P''$, $Q''$, $R''$), as well as the three midpoints of $\overline{AB}$, $\overline{BC}$, $\overline{CA}$ (the midpoints of $\triangle ABC$), and the three midpoints of $\overline{GA}$, $\overline{GB}$, $\overline{GC}$ (the ortho-midpoints of $\triangle ABC$). The roles of some points change for $\triangle GBC$, $\triangle AGB$, $\triangle ABG$, but the point-set itself remains the same.

By the Fun Fact, the midpoints and ortho-midpoints determine diameters perpendicular to the sides of $\triangle P''Q''R''$. Such diameters are unique, so they must coincide the the Theorem's assumed pairs $\{P,P'\}$, etc, so that the midpoints and ortho-midpoints themselves coincide with the points $P$, $P'$, $Q$, $Q'$, $R$, $R'$. Thus, the given nine points are "the" nine points of $\triangle ABC$. $\square$


So, given nine points, the Theorem tells us when the set is viable, and the Lemma tells us how to construct an orthocentric system that yields exactly four solution triangles.

Note that nine points can contain four diametric pairs, raising the possibility of additional solutions. What happens then?

In the case of four diametric pairs, there the altitude-feet triad must contain one of them (so that the other three remain in tact), making the orthic triangle a right triangle (by Thales, yet again). Consequently, we have two midpoint/ortho-midpoint diameters perpendicular to the legs, which makes them perpendicular to each other, while the third midpoint/ortho-midpoint diameter is perpendicular to the hypotenuse diameter. In other words: Those four diameters comprise two mutually-perpendicular pairs.

enter image description here

I'll leave it as an exercise to the reader to show that there is only one choice of altitude-foot triad, unless the diameters make angles of $30^\circ$ and $60^\circ$; in that case, there are two symmetric choice of triad, leading to two orthocentric systems, for a total of eight solution triangles.

enter image description here

Blue
  • 75,673
  • I am reading this and trying to understand XD I really appreciate this solution, I’ll get back to you with questions. If you could add drawings/diagrams to go along with your explanation that would be wonderful, but I know it’s a lot to ask. – retrovius Mar 03 '20 at 19:28
  • @retrovius: Accepting my answer is appreciated, but if you (understandably) still find it unsatisfying, please un-accept so that others aren't discouraged from posting alternative approaches. ... I'm actually still mulling the problem (but have gotten a little distracted). I've run into a few separate cases and am seeking the appropriate perspective by which to unify them, but I'm not quite there yet. – Blue Mar 06 '20 at 15:09
  • I appreciate that, but your solution actually does make sense! I just had to visualize it first. I'll get back if I can implement an algorithm. Let me know if you have other ideas. – retrovius Mar 06 '20 at 15:33
  • When you say "nine points determine at least 4 diameters", do you mean at least 4 distinct simultaneous diameters? – retrovius Mar 09 '20 at 02:06
  • 2
    @retrovius: Diametric points come in pairs. With 9 distinct pts, there can be at most four diametric pairs. As indicated in the answer, the nature of viable 9-pt sets requires at least 3 such pairs. So, if you have 9 pts at random, and there are 4 or more pts whose diametric opposites aren't among those pts, then those pts are not a 9-pt set of any triangle. If there are only 3 un-paired pts, or else just one, the set might be viable. As mentioned, the easiest way to tell may be to pick ostensible midpts from the pairs, construct the triangle, and see if the remaining pts fit the triangle. – Blue Mar 09 '20 at 19:29