Four Points on a Circle

Tom Davis
tomrdavis@earthlink.net http://www.geometer.org/mathcircles October 9, 2002

1 What is an Interesting Theorem?

What makes a Euclidean geometry theorem interesting? There are probably a lot of answers to this, but quite often it is when something occurs that is too lucky to be a coincidence. For example, if you draw some diagram and two of the points of the diagram lie on a line, that is not interesting at all, since every pair of points lie on some line. But if three points lie on a line when they were not dened to be on that line in the rst place, the situation is usually interesting. In fact, if you randomly place three points on a nite portion of a Euclidean plane, selected in a uniform way, the probability that all three will lie on the same line is exactly zero. Similarly, if two non-parallel lines meet at a point, that is not at all interesting, but if three or more do, it may be very interesting. The same sort of idea applies to other gures. If two circles in the plane intersect in two points, or do not intersect at all, this is usually not interesting. But if they intersect at exactly one point (they are tangent, in other words), the situation is usually very interesting. We begin this document with a short discussion of some tools that are useful concerning four points lying on a circle, and follow that with four problems that can be solved using those techniques. The solutions to those problems are presented at the end of the document.

2 Points on Circles
If two points lie on a circle, that is not interesting at all. In fact, for any two Euclidean points, there are an innite number of circles that pass through both of them. Three points on a circle is also not interesting. Unless the three points happen to lie on the same straight line (which has probability zero of occurring by chance), they lie on a circle. To see this, construct the perpendicular bisectors of the segments joining any two pairs of the points. Since the points do not lie in a line, the perpendicular bisectors are not parallel, and must meet somewhere. Any point on the perpendicular bisector of two points is equidistant from those points, so the point at the intersection of the two perpendicular bisectors is equidistant from all three points. Thus a circle centered

at this intersection passing through any one of the three points must pass through the other two. Since the rst three points dene a circle, it is almost impossible for the fourth point to lie on that circle by chance. Thus if, in a diagram, four points that are not in some sense dened to lie on a circle do lie on a circle, that is an interesting occurrence and it may indicate that an interesting theorem can be found. Obviously, if ve, six, or more points lie on a circle, that may be even more interesting, and one theorem we will examine here concerns the famous nine point circle where nine points related to any triangle all lie on the same circle. See Section 4.1.

3 Circle Preliminaries

B O A C /2

B O A

Figure 1: Central Angle and Inscribed Angle Theres basically one theorem we need to know to show a large number of interesting things about sets of points on a circle. Arcs of circles are measured by the central angle. In the left part of Figure 1, is the center of the circle, and the measure of the arc is the same as the measure of the central angle. Both are called in the gure. The right part of Figure 1 illustrates the theorem. If an angle is inscribed in the circle that intersects the circle in an arc of size , then the measure of the angle is . This is not obvious, and requires proof, but it is proved in every high school geometry course. We must be a little careful here. In the gure, the two points and really determine two arcsin this case a short one thats perhaps    in the gure, the rest of the circle (which is about     in the gure). If there is any question, we will name the two points that bound the arc in such a way that the arc is the part of the circle from the rst

to the second in a counter-clockwise direction. Thus in the gure, arc  is about    and arc   is about ! "   . If, in Figure 1, the point # were on the arc  (the small one), the angle $ #% would measure half of about ! "   . This basic fact about angles inscribed in a circle and the fact that a complete circle contains ! &  degrees, allows us to prove a wide variety of interesting theorems.

A D d a B c b C
Figure 2: Cyclic Quadrilateral Since we will be considering the case where four (or more) points lie on a circle, consider some easy (and not so easy) results displayed in Figure 2. In that gure, the four vertices  ,  , # , and ' lie on a circle in order. We will let ( , ) , 0 , and 1 represent the lengths of the segments  ,  # , #' , and '2 , respectively.  #%' is a quadrilateral that is inscribed in a circle, and is called a cyclic quadrilateral (or sometimes a concyclic quadrilateral). The most interesting and useful result is this: $  #434$ #%'2657" 8   (and also, of course, $ '9@3A$ #'B56" 8   ). This is easy to see, since $  # is half the measure of arc # (measured counterclockwise) and $ #%'2 is half the measure of arc # . But arcs # and #% together make the entire circle, or ! &   , so the two opposite angles in any cyclic quadrilateral are supplementary, or in other words, add to make the straight angle. It is not difcult to see the converse; namely, that if two opposite angles in a quadrilateral add to " 8   , then the quadrilateral is cyclic. A very important special case of this is when the two opposite angles are right angles, or C   . This happens often, and when it does, the resulting quadrilateral is cyclic.

3.1 Ptolemys Theorem

There is a very easy way to determine whether four points lie on a circle if you know the distances between them. Conversely, if you know that four points lie on any circle, the formula below, known as Ptolemys Theorem, holds.

In Figure 2, if:

then the four points , , , and all lie on a circle. We will prove one direction of this result in Section 10.

DEAF % G H4IPE GQF H9DQRSDGQF E H D E G H

3.2 Brahmaguptas Formula

Again, we will not prove this result here since it is not used in the problem set, but it also relates to cyclic quadrilaterals. The usual proof is straightforward, but involves a lot of patience with trigonometric manipulations that begin with the observation that the laws of sines and cosines apply to various triangles in theD gure. EGH9V DE G% H , the area of the Brahmaguptas formula provides a method to calculate T9U cyclic quadrilateral . Here it is: DEG H I a I cyclic V b c quadrilateral, whose sides have lengths W , X , Y , and ` , then if If R a U W I X is Y ` is the semiperimeter, its area is given by:

T9U

DE G%H2VdRAe

V V V Vg a f X U af Y U af ` U af W U

Notice that from Brahmaguptas formula it is trivial to deduce Herons formula for the area of any triangle (remember that DE every G triangle is concyclic). Here is Herons R I V b c formula for the area of a I triangle h having sides of lengths W , X , and Y and UW X Y : semiperimeter a

DE G%VdRAe a af V af V aif V g U W U X U Y H G DE G The derivation of Herons Hformula is obvious: G pick a point V near of h , but on af ` in Brahmaguptas formula its circumcircle. Then let approach .D The E G term H U DEG a , and the quadrilateral will tend toward will become more and more like the triangle h . T9U h
4 Problems
4.1 The Nine Point Circle

This is perhaps one of the most amazing theorems in Euclidean geometry. It states Dp E with p any G%ptriangle, the nine points consisting of the midpoints of the that if you begin sides (points , , and in Figure 3), the feet of the altitudes (points q , r , and s ), and the midpoints of the segments from the orthocenter to the vertices of the triangle (points t , u , and v ), all lie on the same circle. (The orthocenter w is the intersection of the three altitudes.) For the proof, see Section 5.

B R F C Q A G B H A S C E

Figure 3: Nine Point Circle

4.2 Miquels Theorem

Figure 4: Miquels Theorem Figure 4 shows two examples of Miquels theorem. Choose any triangle and then choose a point on each of its sides. Construct circles passing through each vertex of the triangle and through the points on the two adjacent sides. All three of those circles meet at a point. The gure shows that the result seems to hold in two cases, but note that the point of intersection of the three circles may not lie within the original triangle. In fact, the theorem even holds if the points on the sides of the triangle do not lie between the vertices of the triangle. For the proof, see Section 6

4.3 Napoleons Theorem

Napoleons Theorem (see Figure 5) says that given any triangle, if you erect equilateral triangles on the edges of that triangle, all pointing outward, that the centers of those three triangles will form an equilateral triangle. In the gure three circles are drawn to give a hint as to how the proof should proceed. Notice that those three circles all seem to meet at a point. 5

S A O R E C Q D B

Figure 5: Napoleons Theorem For the proof, see Section 7

4.4 The Simson Line

C B M A A

B C

Figure 6: The Simson Line Let x be any point on the circumcircle of y2 . From x , drop perpendiculars to each of the sides of the triangle. Show that the points of intersection of the perpendiculars with the sides all lie on a line called the Simson Line. See Figure 6 For the proof, see Section 8

4.5 Fagnanos Problem

Fagnanos problem is the following: Given an acute-angled triangle y , nd the triangle with the smallest perimeter whose vertices lie on the edges of y2 . In Figure 7, this triangle with smallest perimeter is y9 . The answer is that it is the triangle whose vertices are the feet of the altitudes of the original triangle. The easiest proof uses some interesting arguments involving not only circles, but reections of the original triangle. For the proof, see Section 9

C F

A D B

Figure 7: Fagnanos Problem

5 Proof: Nine Point Circle

B R F C Q A G B H N S C A E

Figure 8: Nine Point Circle See Figure 8. We will prove that all nine points lie on the circle by rst showing that the six points , , , , and all lie on a circle. The proof will use the line as the base of the triangle. But every triangle has three bases, and if we consider the line to the the base, exactly the same proof would show that the points , , , , , and also lie on the same circle. Since these two circles have three points in common: , , and , they must be the same circle, and hence all nine points lie on the same circle. Since and % are midpoints of the sides of , we know that . Similarly, since and are midpoints of the sides of d , Q% . But since both and are parallel to the same line , they are parallel to each other: ei . Using the same reasoning, since and are midpoints of and d in 2 d , f . If we look at 2 d , repeat the reasoning to show that % B2 . 7

Since g%h i and jh k are both parallel to the same line lm , they are parallel to each other. Thus jh g%h ik is a parallelogram. But it is more than a parallelogram; it is a rectangle. Since lm is part of the altitude of the triangle lmon@jg . Because jh k and ik are parallel to perpendicular lines, jh kPnSik . Therefore jh g%h ik is a rectangle. Let p be the center of the rectangle. Obviously p is equidistant from jh , gh , i , and k , so all four of those points lie on a circle centered at p . In this circle, jh i and gh k are diameters. Since the lines iq and ksr are altitudes, t jh qi4u t gh rkPuSv w x . But these right angles subtend the diameters jh i and g%h k , so q and r lie on the circle centered at p and passing through points jh , g%h , i , and k . Thus we have shown that six of the nine points lie on a circle. As we said above, there is no reason to assume that jg is the only base, and using either jl or lg as base we can show that other sets of six points lie on circles, and since each set has three points in common with the two other sets, all nine points lie on the same circle.

6 Proof: Miquels Theorem

A E F

M D

Figure 9: Miquels Theorem The proof is fairly simple; see Figure 9. Using the labels in Figure 9, begin by considering only two of the circlesjqr and l r%y . They will meet at some point z , and our goal will be to show that the point z also lies on the circle g%qy . Construct lines y9z , qz , and r%z , forming two cyclic quadrilaterals jqzAr and t qjrS{ t qzArAuA| } w x l y9zAr t . Since thet quadrilaterals are cyclic, we know that t y9zArAuA| } w x . We also know that qjr~{ t t rlye{ t t y9gyu and that rlye{ |t } w x since they are the three angles of a triangle, and we know that qzAr{ y2zAr{ qzAy@uS w x . t t qzAyu| } w x . Apply a little algebra to those equations to conclude that y9gqQ{ If two opposite angles in a quadrilateral add to | } w x , we know that it is a cyclic quadri8

lateral, so the points , the point .

and

all lie on a circle. Thus the three circles meet at

7 Proof: Napoleons Theorem

F

S A A O R E C C D Q B B

Figure 10: Napoleons Theorem The proof of Napoleons Theorem is similar in some ways to the previous proof. See Figure 10. We begin by constructing circles around the three equilateral triangles and we will show that they all meet at a point. In the gure, consider the two circles and . In addition to meeting at they will meet at another point . Since quadrilaterals and % are cyclic, we know that opposite angles in them add to . Thus AQ 4 %A4 9ff . But the triangles 2 and % are equilateral so S4 2AS . Thus we can conclude that %@A %@A . Since the angles % , , and % add to make a full , %QA . But then since %SA 4@ , quadrilateral % is cyclic, so the three circles surrounding the exterior equilateral triangles meet in a single point . Not only that, but the lines , , and % all meet at angles. The points , , and are centers both of the equilateral triangles and of the circles, so the lines connecting them are perpendicular to the lines , , and % . Thus s @4 s 4Q and since they add to , quadrilateral s is cyclic and s S 4 . Therefore s Q . But the same argument can be made to show that % A %%S , so 2% is equilateral.

8 Proof: The Simson Line

The following proof works only if lies on the arc between and . If it is on one of the other arcs, rename the points appropriately before beginning.
B A C A B M

Figure 11: The Simson Line Since the lines dropped from are perpendicular to the sides of the triangle, they form right angles, and it is easy to see in Figure 11 that A , A and A% are sets of concyclic points. Obviously A is also concyclic, since is on the circumcircle of 2 . Since A and A are concyclic sets of points, both A and % A are supplementary to . Therefore they are equal. Since A4@ A A and % A 7 A @ A we nd that A7 A . But since each of those angles lies in a circle, A % and Af . The equality of the nal pair of angles proves that the points , and are collinear, since equal vertical angles are guaranteed at .

9 Proof: Fagnanos Problem

C F H E A D B

Figure 12: Equal Angles To prove that the triangle connecting the feet of the altitudes of an acute triangle has the minimum perimeter, we rst need to prove a lemma. We will show that this pedal triangle 9 in Figure 12 makes equal angles with the bases of the triangles. In other words, we need to show that 24A 9 . 10

We begin by considering the two circles passing through and . It is clear that both pass through , the orthocenter of since both quadrilaterals 2d and 9d contain two opposite right angles and are hence concyclic. But %dA d since they are vertical angles, and since the two quadrilaterals 2d and 2d are concyclic, 2 d and 2 %d since they subtend equal arcs of the circles. Thus 24A 9 .

C Y X A B Y X A A B

C
Figure 13: Fagnanos Problem With the lemma above, and with a somewhat miraculous construction, the proof of the main theorem is not difcult. See Figure 13. Reect the original triangle across the line forming a triangle 2 . Reect this one across forming another triangle 2 . Continue in this way for ve reections, as in the gure. Chase angles to show the line is parallel to . It is easy to show the reections of the triangle connecting the feet of the altitudes (the one with solid diamond vertices) forms a straight line after the reections. Any other triangle (like the one in the gure with dotted diamond vertices) will not be reected to form a straight path. The beginning and end points of these two paths ( to ~ and to ) are equally far apart, since they form a parallelogram. It is a parallelogram since is parallel to and the length ~ is the same as the length ~ . But the solid-diamond path is a straight line, so any other path must be longer. Note that the two paths each consist of two copies of the sides of the pedal triangles, so clearly the pedal triangle connecting the bases of the altitudes is the shortest. 11

10 Proof: Ptolemys Theorem

The statement of Ptolemys theorem contains a bunch of products of lengths of segA %4Q 9QSQ d ments: (1) Generally, to prove theorems like this we need to convert those products into ratios, and then use similar triangles or some other technique to establish the ratios. 9 but notice that in equation 1 if we could It turns out not to matter much how we start, get the ratio that might help. But and are not in any pair of similar triangles, so lets just construct a line that creates a pair of similar triangles with the side in one corresponding to side in the other.
B A

C D H

Figure 14: Ptolemys Theorem In Figure 14, construct the line from outside the cyclic quadrilateral that makes the same angle with that does with B and%intersects the line at . Now % we 9 that we want. equal to % have the similarity since both are which were constructed to be equal. Since plus or and the  same circle, are inscribed in they are also equal, so by angle-angle similarity, we . know that %2 %2 But theres another pair of similar triangles. Since and are opposite angles in 2 a cyclic quadrilateral, they are supplementary. f 2 It is also obvious 7 that and are supplementary, so . Since A (by construction), we have, again by angle-angle similarity, that . e4P Finally, notice that (2)

s (3) If we solve for and in equations 3 and substitute them into equation 2, we Q S%4 4 obtain:
From the similarities of the pairs of triangles, we have: When we multiply through by we obtain Ptolemys theorem. 12