Sie sind auf Seite 1von 1079

https://web.facebook.

com/groups/parmenides52 (Facebook group - Romantics of Geometry)

259 Geometry Writings by


Christopher J.Bradley
full pdf articles, separate pages with titles and summaries 09 -02-2017
collected by Chronopoulos Takis (parmenides51)

Obituary *

CHRISTOPHER J. BRADLEY (19382013)

Christopher Bradley was a popular and effective Fellow and Tutor


in Mathematics at Jesus College from 1964 to 1977, after which he
entered schoolteaching, notably at Clifton College, near Bristol. As
Deputy Leader of the British Mathematical Olympiad team and
for many years Secretary of the British Mathematical Olympiad Committee, he
was known for his elegant Olympiad problems. He
was also involved with the UK Maths Trust and the Mathematical Association.
While in Oxford he co-authored a major study
of the mathematical theory of symmetry in solids, and later wrote introductory
texts for the UK Maths Trust and a book on
Challenges in geometry: for mathematical Olympians past and present.

The following problems are taken from one of Christopher Bradleys books:
Prove that one member of a Pythagorean triple is always divisible by 5, and that
the area of any right-angled triangle with integer sides is divisible by 6.

* taken from: https://www.maths.ox.ac.uk/system/files/legacy/3277/newsletter_0414_0.pdf

Contains separate pages with all titles and all summaries.


For your own convenience use the bookmarks on the left of this pdf.

Articles Source: http://people.bath.ac.uk/masgcs/bradley.html

parmenides51 :
facebook : http://parmenides51.blogspot.gr
: http://parmenides52.blogspot.gr
Christopher Bradley's Geometry Writings

37 More Circles in the Cyclic Quadrilateral


Titles Configuration
38 Eight Points on a line and Seven Circles through a
1 The Story of Hagge and Speckman
Point
2 The four Hagge Circles
39 The 60, 75, 45 triangle and its Euler line
3 Generalizations of Hagge's Theorem
40 The 120, 45, 15 triangle and its Euler line
4 Studies in Similarity
41 The 105, 60, 15 Triangle and its Euler Line and
5 Omega circles
Circles
6 On the Ten point Wood-Desargues' Configuration
42 A Cyclic Quadrilateral, 29 Points and 33 Lines
7 Generalization of the Wallace-Simson line.
43 The Altitudes and Radii of a triangle and its
8 A singular Miquel configuration and the Miquel
circumcircle
direct similarity
44 A natural Enlargement resulting in a Collineation
9 Miquel circles and Cevian lines
45 On the Circles defining the Brocard points
10 A new construction to find any circle through a
46 A Mean of two Cevian points and the
given point.
Construction of Triangle centres
11 Eight Circles through the Orthocentre.
47 A Cevian point and its six harmonics
12 Circles concentric with the Circumcircle
48 The Symmedian point and its Harmonics
13 Ex-points and Eight Intersecting Circles
49 Perpendicular Bisectors and Angle Bisectors
14 Intersecting Circles having Chords the sides of a
50 Three triangles in mutual triple perspective
Cyclic Quadrilateral
51 When one Conic produces two more
15 Four concurrent Euler Lines
52 Three triangles in mutual Triple Reverse
16 The Direct Similarity of the Miquel Point
Perspective
Configuration
53 Hagge Circles touching at H
17 Harmonic Ranges in a Coaxal system of Circles
54 Four Triangle Conics
18 Some Special Circles in a Triangle
55 Perpendiculars to a Triangle's sides through its
19 Circular Perspective
Vertices
20 On the Nine Intersections of two Cevian Triangles
56 The Cevian point Conic
21 Significant Points on Circles Centre the
57 Perpendiculars to the Cevians at the Cevian Point
Circumcentre
58 Additional results for the Miquel Configuration
22 Six Points on a Circle
59 7 Points on any Circle not through a Vertex
23 The Symmedian point and the Polar Line
60 Centroid-centred similar ellipses
24 The Thirteen Point Circle
61 Concurrent lines in a triangle with a Circle cutting
25 When Quadrangles are completely in Perspective
the Sides
26 A Circle concentric with the Incircle
62 A Triangle with an arbitrary Conic cutting its
27 Porisms with a circular circumconic
Sides
28 Some circles in a Cyclic Quadrilateral
63 A Cascade of Conics
29 The Miquel Circles for a Quadrilateral
64 Porism constructed by the Circumcircle and
30 More on Circular Perspective
Triangles in Perspective
31 More cases of Circular Perspective
65 When I replaces K and Ge replaces H and Mi
32 The GH Disc and anogther case of Triple Circular
replaces O
Perspective
66 On the mean of two Cevian points
33 29 Circles,
67 A Special Tucker Circle
34 Some circles in the Cyclic Quadrilateral
68 Six Collinear Points in a Special Cyclic
configuration
Quadrilateral
35 Where 7 circles meet Part 1
69 Two Cyclic Quadrilaterals and Two Coaxal
36 Where 7 circles meet Part 2
Systems

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 2


Christopher Bradley's Geometry Writings

70 What happens when a Triangle is Rotated about 104 How any Six Points on a Circle create Two
its Orthocentre Conics
71 More Special Cyclic Quadrilaterals 105 External Squares on the Sides of a Triangle
72 What happens when you reflect a Triangle in any 106 On Perpendiculars at the corners of a Cyclic
Line Quadrilateral
73 Constructing Two Coaxal systems by Incidence 107 Conics through three points with Centres on a
and Reflection Fixed Line
74 The Simson Line Porism 108 A Property of 3 Circles passing through a fixed
75 The Double Simson Line Circle Point
76 The Double Simson Line Conic 109 Four Perspectives given a Triangle and its
77 The Circle centre O and radius OH Circumcircle
78 More on the Seven-Point CircleMore on the 110 Three Concurrent Pascal lines and an Auxiliary
Seven-Point Circle Conic .
79 Conics in the Ex-circle Configuration 111 Perpendicular bisectors of Three Radii of a
80 Incircle and Excircle Conics Circle .
81 A Radical Centre that lies on OI 112 Parallels from the feet of Cevians create a Conic
82 Cevian derived Conics .
83 Conics generated by Points on a Curve of degree 113 Circles passing through the Miquel Point of the
Five feet of Cevians
84 The G Circles and the Conic they determine 114 Properties of a Particular Tucker Circle
85 A Converse of the Pascal Line Property 115 Six Circles and their Centres .
86 The Perpendiculars to Three Segments at a Point 116 The Geometry of the Brocard Axis and
determine a Conic Associated Conics
87 Two connected Rectangular Hyperbolae 117 When the Cevians of triangle ABC meet the
88 A Singular Cyclic Quadrilateral Circumcircle at D, E, F
89 Perpendiculars in a Cyclic Quadrilateral 118 Two Triangles in Perspective and inscribed in a
90 Perpendiculars from the vertices of a Cyclic Conic
Quadrilateral 119 The Nine-point Conic and a Pair of Parallel
91 6 Conics Lines
92 Three special Cyclic Quadrilaterals 120 An Important Line through the Centre of a
93 Circles formed by an Isosceles Trapezium Cevian Inellipse
94 Generalization of the Steiner Point 121 Some Lines through the Centroid G
95 A Nine Point Rectangular Hyperbola 122 Some lines through the Incentre I
96 More on the Nine Point Rectangular Hyperbola 123 Properties of the Intangential Triangle
97 How the Excentres create Points on the 124 Two Cevian Points Collinear with a Vertex and
Circumcircle Thirteen Conics
98 Ex-points and their Sets of Intersecting Circles 125 Properties of the Extangential Triangle
99 A Triangle and its Image under a Half Turn 126 Some Properties of the Ex-Circle configuration
100 Circles though a point in an Equilateral Triangle 127 When a Circle passes through a Vertex and cuts
101 Intersections at the vertices of the Second the Opposite Side
Brocard Triangle 128 The Symmetry of a Scalene Triangle
102 Two Triangles, their Ex-Symmedians and four 129 The Neuberg Cubic
Conics 130 An Extension to the Theory of Hagge Circles
103 The Altitudes Create Four Circles, Four Conics 131 Construction of Circles always having centre the
and a Polar line Nine-Point Centre
132 The Cevian Conic

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 3


Christopher Bradley's Geometry Writings

133 Conics in a semi-regular hexagon 165 A description of properties of Pascal's hexagon


134 When the Cevian Conic is a Circle 166 Analytic treatment of a Romanian problem
135 Constructing Triangles with Coincident 167 Basic Properties of a Quadrangle possessing an
Centroids Incircle
136 Properties of the symmedian point in a cyclic 168 Properties of a pair of Diametrically Opposite
quadrilateral Triangles
137 Affine tranform of the properties of the Eulerian 169 Two In-Perspective Triangles inscribed in a
triangle Conic
138 Explaining some collinearities among triangle 170 The Brocard Conics
centres 171 A Conic through the feet of two Cevians may
139 The harmonic cevian conic lead to a second Conic
140 The centroid of centroids 172 The Inner Circle of a Cyclic Quadrilateral
141 Three centroids created by a cyclic quadrilateral 173 The Auxiliary Circles of a Cyclic Quadrilateral
142 Incircle conjugation 174 A Cevian Circle leads to Perspective Triangles
143 The nine-point circle of the diagonal point 175 When two Triangles intersect in Conics and in
triangle Circles
144 Four nine-point circles 176 Orthologic Triangles
145 Concyclic circumcentres in the Steiner 177 Special Pascal Lines
configuration 178 More Circles centred on the Brocard Axis
146 More conics in a semi-regular hexagon 179 The Second Brocard Triangle
147 The midpoint rectangular hyperbola again 180 Circles through the Brocard points and the
148 A circle through two vertces, three circumcentres Circumcentre
and a Miquel point 181 Circles through the Brocard points and the
149 Circles with a common point in a cyclic Symmedian Point
quadrilateral 182 Problems requiring Proof
150 When a point is a circumcentre, an incentre and 183 Conics and Triangles in Perspective
an orthocentre 184 Cevian Perspectivity
151 The Orthocentre Conics 185 The Four Conic Theorem
152 Generating Circles from the Symmedian point 186 6 Points and 4 perspectives
153 More properties of the Incentre 187 A Perspective in a pair of Cyclic Quadrilaterals
154 On the Incircle and Excircles of a Cevian 188 Are three Triangles ever in Mutual Perspective?
Triangle 189 The Midpoint Conic
155 The Circumconic of a pair of Cevian Triangles 190 The Brocard Circles' Twin Conic
156 Circumcentre Conics 191 Three Circles and their Centres
157 The General Inellipse 192 Generalization of the Triplicate Ratio Circle
158 Six Point Circles and their Associated Conics 193 A Two Circle Problem Involving the GK axis
159 A Twelve Point Configuration and Carnots 194 The KG Nine-Point Conic
Theorem 195 The Orthocentre Circle of a Cyclic Quadrilateral
160 Tangents to a Conic from the vertices of a 196 The Two Central Lines in a Cyclic Quadrilateral
Triangle 197 When Two Pairs of Diagonals are Concurrent
161 When 12 points display eight 6-point Conics and 198 The Two Cevians' Perspective
6 Concurrent Lines 199 Three Circles with collinear Centres
162 When 24 Points form three 8-point Conics and 200 Problem on Quadrilateral with an Incircle
12 Concurrent Lines 201 All about a Cyclic Quadrilateral and its cousin
163 A Theorem on the Complete Quadrilateral 202 The Midpoint Theorem
164 The Transversal of a Quadrilateral 203 Another Circle with Centre on the Brocard axis

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 4


Christopher Bradley's Geometry Writings

204 Between the Triplicate Ratio Circle and the 235 The Truth and the whole Truth about a
Circumcircle Quadrilateral
205 The Brocard Lines K and K' 236 A Nice Conic and a Nasty Circle
206 Dividing OH into Five or Seven Equal Parts 237 How to Construct the Ex-points of a given Point
207 The Median Conic 238 When Two Quadrilaterals are in Complete
208 A Pair of Isogonal Conjugates produce a Conic Perspective
through 6 Points 239 More on Complete perspective
209 Three Concentric Circles 240 Pentagons in Complete Perspective
210 12 Points, 8 six-point Conics, 4 Conics through 241 Quadrilaterals in Perspective inscribed in a
each Point Conic
211 A cyclic quadrilateral and its midpoint circles 242 Perspective Quadrilaterals
212 Quarter size Circles at Triangle Vertices 243 Some properties of a Special Hexagon
213 When an Incircle produces a Circumcircle 244 When two Triangles in Perspective create Conics
214 The Super-Cevian Triangles, their Conic and 245 Some Simple Algebraic Results
Two Perspectives 246 Pascal points
215 Multiplication of Points using Barycentric Co- 247 The G Conic
ordinates 248 The H Conic
216 A Point, two Triangles and Two Conics 249 Twin Conics
217 An interesting Perspective in the 250 Cevian Circles
Anticomplementary Triangle 251 The Four Brocard Triangles
218 Properties of the Triangle of Excentres 252 Odds and Ends
219 The Twelve point Circle 253 Triangles in Perspective the Whole Truth
220 A Conical Hexagon with Main Diagonals 254 Circles through the midpoints of sides of a
Concurrent Cyclic Quadrilateral
221 Four Special Conical Hexagons all with same 255 Five Conics and a Polar line
Polar 256 Two Quadrangles in Perspective in a Conic
222 The Missing Point on the Euler Line 257 Quadrangles in Perspective Part II The three
223 When a non-regular Cyclic Pentagon leads to lines
Another 258 The Miquel Cyclic Quadrilateral
224 The Sixty Pascal Poles 259 Two Triangles and an Eight Point Conic
225 A Typical Pascal line Drawing
226 Special Conical Hexagons
227 The Remarkable Eight-Point Conic
228 How 2 Conics through 4 Points generate 6 more
such Conics
229 When a Conical Quadrilateral produces two 6-
point Conics
230 A Quadrilateral and the Conics and Points
arising
231 A Quadrilateral and resulting Conics Part 2 A
Special Case showing the role of the New Points
X and Y
232 The Coconic Hexagon with Main Diagonals
Concurrent
233 The Conic and the Lines that are Created (Part 1)
234 A Conic generates five 8-point Conics (Part 2)

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 5


Christopher Bradley's Geometry Writings

Titles and summaries


Article 1 is The Story of Hagge and Speckman . This concerns theory first developed by Hagge and
Speckman in the Edwardian era. Speckman investigated triangles which were simultaneously in
perspective, and indirectly similar. On the other hand Hagge studies circles which pass through the
orthocentre of a given triangle. Superficially these subjects look unrelated, but this is not the case.

Article 2 is The four Hagge Circles . This describes properties of the four Hagge circles of triangles BCD,
ACD, ABD, ABC when ABCD is a cyclic quadrilateral.

Article 3 is Generalizations of Hagge's Theorem . Two possible generalizations of Hagge circles are
considered. The first describes those properties preserved when the circle passes through a point other
than the orthocentre. The second when a pair of orthologic triangles are involved.

Article 4 is Studies in Similarity . The six points: the orthocentre, the Brocard points and the points where
the medians intersect the orthocentroidal circle involve nine circles through the vertices of ABC and either
the orthocentre or one of the Brocard points. The triangles formed by the centres of these circles exhibit
many similarities.

Article 5 is Omega circles . It is shown how circles through either of the Brocard points have properties
similar to Hagge circles.

Article 6 is On the Ten point Wood-Desargues' Configuration . The ten pairs of directly similar triangles
in perspective in the Wood-Desargues' configuration have ten common Hagge circle centres. It is shown
in this paper that these centres lie at the vertices of five cyclic quadrangles, which are similar to the five
cyclic quadrangles of the initial Wood-Desargues' configuration.

Article 7 is Generalization of the Wallace-Simson line. In this article it is shown how a direct similarity
transformation of triangle ABC into another triangle enables one to identify the Double Wallace-Simson
line of the new triangle as a generalized Wallace-Simson line of ABC.

Article 8 is A singular Miquel configuration and the Miquel direct similarity. In the Miquel configuration
for a triangle the centres of the three concurrent circles form a triangle similar to ABC. It is shown how
the direct similarity between the two triangles involved produces further results of significance.

Article 9 is Miquel circles and Cevian lines. The Miquel configuration has interesting additional properties
when the points on the sides of ABC are the feet of Cevian lines.

Article 10 is A new construction to find any circle through a given point. This is another generalization of
the Hagge construction when the point involved is not the orthocentre.

Article 11 is Eight Circles through the Orthocentre.

Article 12 is Circles concentric with the Circumcircle. It is shown that such circles contain seven points
with special properties.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 6


Christopher Bradley's Geometry Writings

Article 13 is Ex-points and Eight Intersecting Circles. Points internal to a triangle have three points
external to the triangle associated with them. Properties of these Ex-points are investigated.

Article 14 is Intersecting Circles having Chords the sides of a Cyclic Quadrilateral. The circles pass
through the centre of the cyclic quadrilateral or the intersection of its diagonals.

Article 15 is Four concurrent Euler Lines. In this paper we consider a cyclic quadrilateral PQRS in which
the diagonals PR and QS meet at a point E and we prove a number of results about the triangles PQE,
QRE, RSE, SPE.

Article 16 is The Direct Similarity of the Miquel Point Configuration . This continues work begun in
Article 8 and is best read in conjunction with it.

Article 17 is Harmonic Ranges in a Coaxal system of Circles. A circle cuts a triangle in six points and the
condition that the harmonic conjugates of those points also lie on a circle is investigated. The coaxal
system involving the polar circle, the circumcircle and the orthocentroidal circle feature.

Article 18 is Some Special Circles in a Triangle. There is one special circle through each point in the plane
of ABC, not on the sides or the circumcircle. Eight points on each circle have similar properties to those
in Hagge circles.

Article 19 is Circular Perspective. Two triangles ABC and UVW are in circular perspective when circles
AVW, BWU, CUV all pass through the same point. It is shown that this concept is symmetric and two
triangles in double circular perspective are automatically in triple circular perspective.

Article 20 is On the Nine Intersections of two Cevian Triangles. A Cevian triangle is one whose sides join
the three feet of a Cevian point. Two Cevian triangles intersect in nine points which exhibit a vast number
of properties.

Article 21 is Significant Points on Circles Centre the Circumcentre. Given a triangle ABC with
circumcentre O and a point P not on its sides or their extensions and not on the circumcircle, it is shown
that one may construct on the circle centre O and radius OP six significant points.

Article 22 is Six Points on a Circle. A construction with a triangle ABC and the point X76 is described,
yielding a circle bearing a distinct similarity to the Triplicate Ratio Circle.

Article 23 is The Symmedian point and the Polar Line. A construction based on the Symmedian point is
described yielding six lines through each of three points on the polar line. Two related porisms are
constructed.

Article 24 is The Thirteen Point Circle. In this article we give an account of the properties of the coaxal
system of circles passing through the two Brocard points and having the Brocard axis as line of centres.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 7


Christopher Bradley's Geometry Writings

Article 25 is When Quadrangles are completely in Perspective . In this article we establish a sufficient
condition for when a pair of quadrangles have what may be appropriately called a Desargues' axis of
perspective.

Article 26 is A Circle concentric with the Incircle. A construction is given in a triangle using Gergonne's
point yielding a circle concentric with the incircle.

Article 27 is Porisms with a circular circumconic. The result we prove in this article is if we take the
orthocentre of all the triangles in the porism in a case when the circumconic is the circumcircle, then the
path traced out by the orthocentre is circular (or linear).

Article 28 is Some circles in a Cyclic Quadrilateral. The properties of some circles in a cyclic quadrilateral
passing through its vertices, midpoints of sides and centre are investigated.

Article 29 is The Miquel Circles for a Quadrilateral. It is shown how these may be constructed starting
from the Miquel points of triangles involving the diagonal points.

Article 30 is More on Circular Perspective. It is shown how a triangle may be in circular perspective with
three points on a line (rather than a second triangle). An example involving the intersections of the
tangents at the vertices of ABC with the opposite sides is given.

Article 31 is More cases of Circular Perspective . This article involves triangles ABC and PQR, where P,
Q, R are the intersections of the medians with the orthocentroidal circle.

Article 32 is The GH Disc and anogther case of Triple Circular Perspective, and continues the work on
triple circular perspective, this time in connection with the orthocentroidal circle.

Article 33 is 29 Circles, and continues the work on exploring triple circular perspective.

Article 34 is Some circles in the Cyclic Quadrilateral configuration. When ABCD is a cyclic quadrilateral,
centre O, and BA^CD = E, BC^AD = F and AC^BD = G, then circles ABF, CDF, BCE, ADE meet at a
point T on EF. The perpendicular from T to EF passes through G and O. The centres of the four circles, O
and T lie on a circle.

Article 35 is Where 7 circles meet Part 1. The orthocentroidal circle S of triangle ABC on GH as diameter
possesses an unusual property. If you draw circles BHC, CHA, AHB to meet S again at points X, Y, Z and
circles BGC, CGA, AGB to meet S again at U, V, W then the following property holds: circles AYX,
BZX, CXY, AVW, BWU, CUV all pass through a point Q on the circumcircle of ABC.

Article 36 is Where 7 circles meet Part 1. The result of Article 35 is generalized to involve an arbitrary
circle.

Article 37 is More Circles in the Cyclic Quadrilateral Configuration. Quadrilateral ABCD, circle centre O
has external diagonal points E and F. Circles on AO, CO as diameters meet at Y and circles on BO, DO as
diameters meet at Z. It is shown that T, Y, Z are collinear, where T is the midpoint of EF.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 8


Christopher Bradley's Geometry Writings

Article 38 is Eight Points on a line and Seven Circles through a Point. Points D, X, Y, Z and E, U, V, W
lie on a transversal of triangle ABC. It is shown what the condition is connecting D and E for circles ABC,
AYZ, BZX, CXY, AVW, BWU, CUV to meet at a point.

Article 39 is The 60, 75, 45 triangle and its Euler line. Key results concerning the intersection of the
Euler line and the sides of the triangle are established.

Article 40 is The 120, 45, 15 triangle and its Euler line. Many results involving the triangle, its Euler
line and various circles are established.

Article 41 is The 105, 60, 15 Triangle and its Euler Line and Circles. Some remarkable properties of
the 105, 60, 15 triangle and its intersections with its Euler line are investigated.

Article 42 is A Cyclic Quadrilateral, 29 Points and 33 Lines. It is shown that the quadrilateral formed by
the symmedian points of triangle BDC, ACD, ABD, ABC of a cyclic quadrilateral ABCD has the same
diagonal point triangle as the cyclic quadrilateral ABCD.

Article 43 is The Altitudes and Radii of a triangle and its circumcircle. In triangle ABC, circumcentre O
and orthocentre H if ab is the intersection of AH and OB, then the midpoints of pairs ab, ac and ba, bc and
ca, cb lie on the circle on OH as diameter and circles ab bc ca and ba ac cb both pass through H.

Article 44 is A natural Enlargement resulting in a Collineation . This is a short survey of the properties of
the Exsimilicentre X56 and shows that it is on the line joining Feuerbach point and the Orthocentre. This
is a known collineation.

Article 45 is On the Circles defining the Brocard points . A construction is given in which points X, Y, Z
lie on a circle containing the first Brocard point and points X', Y', Z' lie on a circle containing the second
Brocard point. The six circles such as BXC are the Brocard circles and the six circles such as AYZ always
pass through Tarry's oint.

Article 46 is A Mean of two Cevian points and the Construction of Triangle centres . A construction is
given which gives rise to about 9 million more triangle centres. It is part serious and part jocular, but its
aim is to question the activity of hunting triangle centres.

Article 47 is A Cevian point and its six harmonics If L is the foot of a Cevian point on BC then B1 is the
harmonic conjugate of L and C and C1 is the harmonic conjugate of L and B. When similar points on CA
and AB are defined a configuration is produced with many interesting features.

Article 48 is The Symmedian point and its Harmonics Results of Article 47 are applied when the Cevian
point is the Symmedian point.

Article 49 is Perpendicular Bisectors and Angle Bisectors When these are drawn for a given triangle the
six non-trivial points of intersection create a figure with many significant results

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 9


Christopher Bradley's Geometry Writings

Article 50 is Three triangles in mutual triple perspective . A configuration with pairs of points that are
isotomic conjugates produces a figure with three triangles in mutual triple perspective. Six of the
perspectrices pass through a given point.

Article 51 is When one Conic produces two more . ABC is a triangle, circumcentre O. The given conic is
ABCLMN, where L, M, N lie on AO, BO, CO. Twelve points are then formed from these points which
are shown to lie 6 by 6 on two conics.

Article 52 is Three triangles in mutual Triple Reverse Perspective . A construction and supporting analysis
is given that ensures the perspectives. When two of the triangles have vertices on the circumcircle the
Brocard porism is established and the third triangle degenerates and becomes the polar of the Symmedian
point with respect to the circumcircle.

Article 53 is Hagge Circles touching at H . A construction of two triangles is described whose


circumcircles turn out to be a pair of touching Hagge circles. One of the triangles has a vertex at H and an
interesting connection with the nine-point centre is established.

Article 54 is Four Triangle Conics . The construction of the Triplicate Ratio Circle and the 7-Point Circle
when the generating point is not the Symmedian point is described. Analysis shows that four (and not just
two) Conics emerge. Those that might be termed the Triplicate Ratio Conic and the 7-Point Conic are
similar and have the same centre.

Article 55 is Perpendiculars to a Triangle's sides through its Vertices Six points on the sides are generated
(other than the feet of the altitudes). If one constructs the three circles through B, C and C, A and A, B that
cut the circumcircle orthogonally, the six points are the intersections of these circles with the sides of
ABC.

Article 56 is The Cevian point Conic In triangle ABC let P be a Cevian point with D, E, F the feet of the
Cevians on BC, CA, AB respectively. Points L, M, N lie on AD, BE, CF respectively and are such that P
is the midpoint of AL, BM, CN. It is found that the points A, B, C, L, M, N lie on a conic which we call
the Cevian point conic of P. Several other properties are established.

Article 57 is Perpendiculars to the Cevians at the Cevian Point The points where these perpendiculars
meet corresponding sides are shown to be collinear.

Article 58 is Additional results for the Miquel Configuration The Miquel point P is the common point of
circles AMN, BNL, CLM when L, M, N lie on BC, CA, AB respectively. If perpendiculars to BC at L,
CA at M, AB at N are drawn, producing six points on the other sides of ABC, then a variety of results
hold. In particular three new circles may be drawn, pairs of which each have a common point with one of
the Miquel circles.

Article 59 is 7 Points on any Circle not through a Vertex Three Miquel circles AMN, BNL, CLM with L,
M, N on the sides meet at a point P. S is any other circle through P (not through a vertex). AMN meets S
at G and X, BNL meets S at G and Y, CLM meets S at G and Z. It is proved that AX, BY, CZ concur at a
point lying on S.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 10


Christopher Bradley's Geometry Writings

Article 60 is Centroid-centred similar ellipses The three ellipses are the outer and inner Steiner ellipses
and an intermediate on passing through points a quarter and three quarter away along the sides. Certain
sets of collinear points are identified.

Article 61 is Concurrent lines in a triangle with a Circle cutting the Sides A circle cuts the sides of a
triangle at the feet of two Cevian points. When this construction is carried out a number of other sets of
three lines are concurrent and sets of three points are collinear.

Article 62 is A Triangle with an arbitrary Conic cutting its Sides A conic is chosen that cuts the sides BC,
CA, AB of a triangle ABC in points L, U; M, V; N, W (in that order anticlockwise) respectively. The
chords LW, MU, NV form a triangle DEF. A large number of concurrencies and collinearities are
identified.

Article 63 is A Cascade of Conics A pair of in-perspective triangles are inscribed in a conic. Their non-
corresponding sides meet at six points, which turn out to lie on a second conic and form a pair of in-
perspective triangles. The process may therefore be repeated indefinitely forming the cascade of conics in
the title.

Article 64 is Porism constructed by the Circumcircle and Triangles in Perspective A hexagon is formed by
the tangents to the circumcircle at the vertices of the two triangles and a conic passes through the vertices
of the hexagon.

Article 65 is When I replaces K and Ge replaces H and Mi replaces O An analogue of the triplicate ratio
circle and the 7-point circle is created under these circumstances (Mi is the Mittelpunkt and Ge is
Gergonne's point).

Article 66 is On the mean of two Cevian points The feet of their Cevians are used in a construction of a
mean that has a geometrical significance.

Article 67 is A Special Tucker Circle The Tucker circle passes through the feet of the perpendiculars to
other two sides from the foot of the altitude through their common vertex. The properties of the resulting
hexagon are reviewed.

Article 68 is Six Collinear Points in a Special Cyclic Quadrilateral The cyclic quadrilateral has its
diagonals at right angles and a review is given of the main properties of such a quadrilateral.

Article 69 is Two Cyclic Quadrilaterals and Two Coaxal Systems It is shown how two cyclic
quadrilaterals inscribed in the same circle centre O, each with their diagonals at right angles, generate four
circles whose centres lie two by two on lines through O. In each case a coaxal system is generated.

Article 70 is What happens when a Triangle is Rotated about its Orthocentre This is concerned with the
properties of two triangles related by a rotation.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 11


Christopher Bradley's Geometry Writings

Article 71 is More Special Cyclic Quadrilaterals This article is concerned with a cyclic quadrilateral
ABCD in which AC is perpendicular to BD and its image A'B'C'D' after a rotation of 90o about E =
AC^BD.

Article 72 is What happens when you reflect a Triangle in any Line A triangle ABC and its circumcircle
are mapped by a reflection line L with image triangle A'B'C' and its circumcircle. What happens is that the
line through A' parallel to BC, the line through B' parallel to CA and the line through C' parallel to AB are
concurrent at a point P that always lies on circle A'B'C'.

Article 73 is Constructing Two Coaxal systems by Incidence and Reflection The cyclic quadrilateral
ABCD, centre O, with diagonals AC and BD at right angles at E is reflected in the line AC to form the
cyclic quadrilateral A' B'C'D', centre O'. From this configuration two systems of coaxal circles are
obtained.

Article 74 is The Simson Line Porism Let ABC be a triangle and let A'B'C' be its image under a 180
degree rotation about the circumcentre O. Now let D be any point on the circumcircle S. The locus of the
point of intersection of the Wallace-Simson lines of D with respect to the two triangles is an inconic of
both triangles, creating a porism.

Article 75 is The Double Simson Line Circle ABC is a triangle and A'B'C' is its image under a rotation by
180 degrees about the circumcentre O. With P a point on the circumcircle the Double Simson lines of P
with respect to the triangles are drawn, meeting at a point Q. The locus of Q as P moves round the
circumcircle turns out to be the circle centre O.

Article 76 is The Double Simson Line Conic A transversal LMN of a triangle is drawn through the
orthocentre, which is the Double Simson line of a point P on the circumcircle. The corresponding Cevian
point of the harmonic conjugates of L, M, N is a point Q, whose locus as P moves is a circumconic of
ABC.

Article 77 is The Circle centre O and radius OH A number of related results are established about the
circle centre the circumcentre O of a triangle ABC and radius OH, where H is the orthocentre of triangle
ABC.

Article 78 is More on the Seven-Point CircleMore on the Seven-Point Circle Various properties are
established about the Brocard circle and the circles BKC. CKA, AKB through the symmedian point K of a
triangle ABC.

Article 79 is Conics in the Ex-circle Configuration From the points of contact of the three ex-circles of a
triangle ABC four conics are drawn and their properties are investigated.

Article 80 is Incircle and Excircle Conics A configuration of four conics passing through the points of
contact of the incircle and excircles of a triangle ABC is investigated. The intersections and centres of
these conics define sets of points whose joins provide sets of concurrent lines.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 12


Christopher Bradley's Geometry Writings

Article 81 is A Radical Centre that lies on OI The configuration of a triangle ABC with incentre I and the
three circles BCI, CAI, ABI is investigated and three new circles are defined whose radical centre is a
point J on OI where O is the circumcentre. J is shown to be the isogonal conjugate of Gergonnes point.

Article 82 is Cevian derived Conics Lines through the feet of Cevians of a point are drawn parallel to the
sides of a triangle and meet the sides in six points lying on a conic. Properties of this conic are
investigated.

Article 83 is Conics generated by Points on a Curve of degree Five From the feet of Cevians of a point P
lines are drawn perpendicular to the sides meeting them in six point. It turns out that these six points lie on
a conic if, and only if, P lies on a curve of degree five whose equation is determined. Well known points
lying on this curve are O, H, I, the circumcentre, orthocentre and incentre of ABC, and also the
excentres.

Article 84 is The G Circles and the Conic they determine From the feet of Cevians of a point P lines are
drawn perpendicular to the sides meeting them in six point. It turns out that these six points lie on a conic
if, and only if, P lies on a curve of degree five whose equation is determined. Well known points lying on
this curve are O, H, I, the circumcentre, orthocentre and incentre of ABC, and also the excentres.

Article 85 is A Converse of the Pascal Line Property It is shown how any transversal of a triangle and any
three additional points lying one on each side of the triangle (but not at the vertices) may be used to find
six points on the sides of the triangle through which a conic always lies. The proof involves a particular
case of the converse of the Pascal line theorem.

Article 86 is The Perpendiculars to Three Segments at a Point determine a Conic It is shown how any
transversal of a triangle and any three additional points lying one on each side of the triangle (but not at
the vertices) may be used to find six points on the sides of the triangle through which a conic always lies.
The proof involves a particular case of the converse of the Pascal line theorem.

Article 87 is Two connected Rectangular Hyperbolae Given a Cyclic Quadrilateral and the Rectangular
Hyperbola through the midpoints P, Q, R, S of its sides and its centre O, it is shown that this conic
automatically passes through the diagonal points. A slight extension of this famous result is documented.

Article 88 is A Singular Cyclic Quadrilateral When a Cyclic Quadrilateral ABCD is such that the tangents
at A and C and the line BD are concurrent, it follows that the tangents at B and D and the line AC are also
concurrent. Additional properties of this configuration are obtained.

Article 89 is Perpendiculars in a Cyclic Quadrilateral. Perpendiculars from A and C on to opposite sides of


a cyclic quadrilateral ABCD produce four more cyclic quadrilaterals and two sets of parallel lines one set
containing five lines and the other set containing six lines.

Article 90 is Perpendiculars from the vertices of a Cyclic Quadrilateral . The perpendiculars from A to AB
and DA and the six similar perpendiculars from B, C, D to the sides of a cyclic quadrilateral create two
more cyclic quadrilaterals that are coaxal with ABCD. Their sides and their diagonal point lines exhibit
some surprising properties.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 13


Christopher Bradley's Geometry Writings

Article 91 is 6 Conics . When a circumconic is drawn to triangle ABC the tangents at A, B, C form a
triangle DEF which is in perspective with ABC. If the lines AD, BE, CF meet at U and meet the
circumconic at R, S, T then the tangents at R, S, T form a triangle LMN, It is shown that D, E, F, L, M, N
lie on a conic, Then, by drawing tangents and chords four other conics may be constructed with centres all
lying on a line through U.

Article 92 is Three special Cyclic Quadrilaterals . In any cyclic quadrilateral ABCD if AB and CD meet at
F and AD and BC meet at G then FG is one side of the diagonal point triangle. It is then always the case
that the tangents at B and D and the tangents at A and C meet at points U and V respectively lying on FG.
The first special cyclic quadrilateral is when AC passes through U. If the tangents at A and B meet at P
and Q, R, S are similarly defined then the second special cyclic quadrilateral is when AC is perpendicular
to BD and then P, Q, R, S are concyclic. The third special cyclic quadrilateral is when AC is parallel to
BD and then again P, Q, R, S are concyclic.

Article 93 is Circles formed by an Isosceles Trapezium . Given an isosceles trapezium ABCD with AD
parallel to BC, if tangents to the cyclic quadrilateral ABCD are drawn to produce six points of
intersection, then six more circles may be drawn, all passing through the centre O of ABCD. Five of these
circles are obvious but the fact that the sixth circle passes through O is an interesting result and in this
paper a proof is given using Cartesian co-ordinates.

Article 94 is Generalization of the Steiner Point . The outer Steiner ellipse passes through A, B, C and the
images L, M, N of those vertices in a rotation of 180 degrees about the centroid G. The Steiner point is the
fourth point of intersection of the outer Steiner ellipse and the circumcircle of ABC. A generalization is
obtained by replacing G by another point P internal to the triangle ABC. But more occurs as circles AMN,
BMN, CNL intersect at a point U that lies on both the circumcircle and the ellipse ABCLMN and U is the
generalized Steiner point.

Article 95 is A Nine Point Rectangular Hyperbola . The configuration consisting of the rectangular
hyperbola passing through the incentre, the excentres, the centroid and deLongchamps point in a triangle
and three other significant points exhibits some interesting properties which are investigated.

Article 96 is More on the Nine Point Rectangular Hyperbola . The circumcircle of a triangle ABC is the
Nine-Point circle of the triangle IJK of its excentres. The line JK passes through A and the midpoint D of
JK, which lies therefore on the circumcircle. With E, F similarly defined there are thus two triangles ABC
and DEF and it is shown that their incentres and excentres lie on the nine point rectangular hyperbola
discussed in CJB/2010/95. Other interesting properties emerge such as the deLongchamps point of DEF is
the same point as the incentre of ABC.

Article 97 is How the Excentres create Points on the Circumcircle . The circles through pairs of vertices of
a triangle and the excentres opposite the third vertex have centres lying on the circumcircle and pass
through the incentre of the triangle. The triangles with these centres as vertices exhibit properties that are
described.

Article 98 is Ex-points and their Sets of Intersecting Circles . The ex-symmedian points and the ex-points
of an arbitrary point internal to the triangle are shown to lie on a hyperbola. The configuration involving

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 14


Christopher Bradley's Geometry Writings

ex-points of a given point and vertices of a given triangle is shown to produce a set of three circles having
a point in common (similar to the Fermat point). The relationship between these points is obtained.

Article 99 is A Triangle and its Image under a Half Turn . A triangle ABC and its image DEF under a half
turn results in circles BCD, CAE, ABF intersecting at a point P on circle DEF. A point Q on circle ABC is
similarly defined. Also it is established that a conic passes through A, B, C, D, E, F, P, Q.

Article 100 is Circles though a point in an Equilateral Triangle . Given three points D, E, F lying on the
medians of an equilateral triangle ABC the conditions are determined for the circles BCD, CAE, ABF to
have a common point.

Article 101 is Intersections at the vertices of the Second Brocard Triangle . If D, E, F are the ex-
symmedian points it is proved that circles BCD, CAE, ABF intersect the Brocard circle at the
circumcentre of ABC and at the vertices of the second Brocard triangle. A case of circular perspective is
thereby determined.

Article 102 is Two Triangles, their Ex-Symmedians and four Conics . The two triangles are in perspective,
vertex the centroid of one of them. The six ex-symmedian points lie on a conic, the tangents at the vertex
circumscribe the circumcircle and also the intersections of various lines produce six points lying three by
three on two parallel lines. Two other conics of interest are also created.

Article 103 is The Altitudes Create Four Circles, Four Conics and a Polar line . If the altitudes of a
triangle ABC meet the circumcircle at D, E, F, then the six interior intersections of the two triangles and
the tangents at the above six points produce a configuration in which four circles and four conics feature.
An analysis is given in which the polar of the orthocentre plays a major role.

Article 104 is How any Six Points on a Circle create Two Conics . Given three chords of a circle,
perpendiculars to those chords from their end points create three pairs of parallel lines. Of their fifteen
points of intersection three lie on the line at infinity and the other twelve lie six by six on two conics. Note
that the six points must lie on a circle and not a general conic.

Article 105 is External Squares on the Sides of a Triangle . If squares are drawn externally on the sides of
a triangle the configuration exhibits two points similar to the Fermat points. Also the six intersections of
various lines define a circle.

Article 106 is On Perpendiculars at the corners of a Cyclic Quadrilateral . If perpendiculars to the sides are
drawn at the vertices of a cyclic quadrilateral then a configuration is created that consists of two circles
coaxal with the first circle, each containing quadrilaterals similar to each other. Their external diagonal
points also lie three by three on two collinear lines. Two other circles also emerge.

Article 107 is Conics through three points with Centres on a Fixed Line . Conics through three non-
collinear points A, B, C having their centres on a fixed line have three radical axes, BC, CA, AB. If one
takes a fixed point P on one of these axes, say BC, then the polars of P with respect to all the conics pass
through another fixed point P', which as one would expect, is the harmonic conjugate of P with respect to
B and C.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 15


Christopher Bradley's Geometry Writings

Article 108 is A Property of 3 Circles passing through a fixed Point . Let ABC be a triangle and P a fixed
point not on the sides. Circle BPC meets AB at W and AC at M. Circle CPA meet BC at U and BA at N.
Circle APB meets CA at V and CB at L. The perpendicular bisectors of LU, MV, NW meet at a point Q.

Article 109 is Four Perspectives given a Triangle and its Circumcircle . Given a triangle, its circumcircle
and a point not on the circumcircle, it is shown how by varying the position of the point along a line four
perspectives may be created.

Article 110 is Three Concurrent Pascal lines and an Auxiliary Conic .

Article 111 is Perpendicular bisectors of Three Radii of a Circle .

Article 112 is Parallels from the feet of Cevians create a Conic . From the feet of Cevians through a point
parallels to the other two sides are drawn. The six points created on the sides of the triangle lie on a conic.

Article 113 is Circles passing through the Miquel Point of the feet of Cevians . If LMN are the feet of a set
of Cevians, then circles AMN, BNL, CLM meet at a Miquel point Q. If now circles BQC, CQA, AQB are
drawn and circles AMN and BQC meet at R with S, T similarly defined, then circle RST passes through
the Miquel point Q. The centres of the seven circles also exhibit some remarkable properties showing Q to
be the Miquel point of a second triangle, a cascade process that can be carried on indefinitely.

Article 114 is Properties of a Particular Tucker Circle . Three circles determine a particular Tucker circle
and their centres exhibit some interesting properties.

Article 115 is Six Circles and their Centres . In triangle ABC centroid G, circles BGC, CGA, AGB are
drawn meeting the sides in six points that lie on a conic. Three more circles are drawn and the six circle
centres exhibit some interesting properties.

Article 116 is The Geometry of the Brocard Axis and Associated Conics . From a triangle and its
symmedian point K and its circumcentre O, various triangles, the Brocard inellipse and two further conics
are constructed having their centres on the Brocard axis KO. Two porisms are defined and a special line is
identified. Most, if not all, of the content of this article is known and indeed should be known by all
geometers.

Article 117 is When the Cevians of triangle ABC meet the Circumcircle at D, E, F . In a triangle ABC
when the Cevians through the centroid G meet the circumcircle at points D, E, F and the tangents are
drawn at the six points, two triangles are created. The intersections of the sides of the four triangles in the
configuration have some interesting properties.

Article 118 is Two Triangles in Perspective and inscribed in a Conic . Given two triangles ABC, DEF in
perspective at P and inscribed in a conic , a configuration is produced consisting of the 6 Pascal lines and
2 Steiner points arising. One of the Pascal lines has further properties, as it is not only a Pascal line, but
also a Desargues axis of perspective and also the polar of P with respect to . Further properties
involving the nine lines AD, AE, AF, BD, BE, BF, CD, CE, CF and the tangents at the six vertices are

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 16


Christopher Bradley's Geometry Writings

deduced leading to a conic, further properties of the key Pascal line and numerous sets of collinear points
and concurrent lines.

Article 119 is The Nine-point Conic and a Pair of Parallel Lines . An affine transformation of the nine-
point circle leads to the nine-point conic. The orthocentre transforms into a general point P which is
collinear with the centre of the nine-point conic and the centroid. We prove that the polar of P with respect
to the nine-point conic is parallel to the Desargues axis of perspective of the triangle and the image of the
orthic triangle.

Article 120 is An Important Line through the Centre of a Cevian Inellipse . A Cevian inellipse is one that
touches the sides of a triangle internally at points that are the feet of Cevian lines. It is shown in this short
article that the line through the Cevian point and the centre of a Cevian inellipse contains other significant
points.

Article 121 is Some Lines through the Centroid G . Using areal co-ordinates we catalogue some lines
through the centroid G of a triangle ABC. Co-ordinates of points and equations of lines are given.

Article 122 is Some lines through the Incentre I . Using areal co-ordinates we catalogue some lines
through the incentre I of a triangle ABC. Co-ordinates of points and equations of lines are given.

Article 123 is Properties of the Intangential Triangle . Triangle ABC and its incircle and ex-circles are
given. Common tangents to the incircle and an ex-circle are four in number; three being the sides of ABC
and the fourth is called an intangent. There are three intangents, one for each ex-circle and these form a
triangle A'B'C' called the intangential triangle. Properties of the intangential triangle are studied using
areal co-ordinates with ABC as triangle of reference. By construction its incircle is the same as that of
triangle ABC. The triangle of its excentres has sides parallel to those of ABC.

Article 124 is Two Cevian Points Collinear with a Vertex and Thirteen Conics . Given triangle ABC, a
point P on BC and lines LMP and UVP with L, U on AB and M, V on CA, then two Cevian points may be
constructed collinear with vertex A. Two further Cevian points occur and the resulting configuration of
points and lines results in the appearance of thirteen conics and two harmonic ranges.

Article 125 is Properties of the Extangential Triangle . The external common tangents to the three ex-
circles form the extangential triangle A'B'C'. Triangles ABC and A'B'C' are in perspective with perspector
the orthocentre of the intouch triangle. The axis of perspective plays a prominent part in the configuration.
Triangle A'B'C' and the orthic triangle are homothetic through the Clawson point.

Article 126 is Some Properties of the Ex-Circle configuration . The ex-circle configuration is studied and
it is shown in particular how the midpoints of the sides of the ex-central triangle lead to several circles and
perspectives, as well as a new (to me) triangle with vertices on the sides of the original triangle and two
new (to me) points on the circumcircle.

Article 127 is When a Circle passes through a Vertex and cuts the Opposite Side . A surprising
configuration emerges when a circle passes through a vertex and cuts the opposite side. Two pairs of
touching circles and two circles each passing through seven key points emerge and additional points on

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 17


Christopher Bradley's Geometry Writings

the original circle are also created. It seems unlikely that a synthetic solution will exist, as some of the co-
ordinate working is technically very difficult.

Article 128 is The Symmetry of a Scalene Triangle . Points dividing each side of a scalene triangle in a
fixed ratio leads to a configuration with two sets of three equal circles and a conic. Many results are self-
evident but the equations of the circles and conic are recorded.

Article 129 is The Neuberg Cubic . Analysis is made of the Neuberg triangle cubic. This is the locus of
points P with the property that the reflection of P in the sides of a triangle ABC leads to a triangle which is
in perspective with ABC.

Article 130 is An Extension to the Theory of Hagge Circles . An extension of the theory of Hagge circles
is presented.

Article 131 is Construction of Circles always having centre the Nine-Point Centre . From any general
point P reflections in BC. CA, AB give D, E, F. The midpoints of AD, BE, CF are U, V, W respectively.
U, V, W are shown to lie on a circle with centre the Nine-Point centre.

Article 132 is The Cevian Conic . If P is a Cevian point in a triangle ABC and D, E, F are the feet of the
Cevians, then three circles PEF, PFD, PDE meet the sides again in six other points. It is shown that these
six points lie on a conic, which we call the Cevian Conic.

Article 133 is Conics in a semi-regular hexagon. Hexagon AFBDCE with ABC an equilateral triangle and
AD, BE, CF concurrent at the centroid P of ABC is inscribed in a conic, and as such is defined as a semi-
regular hexagon. It is proved that the six circumcentres of triangles AFP, FBP, BDP, DCP, CPE, EPA are
co-conic as are their six orthocentres. It is also proved that the intersections of contiguous Euler lines are
co-conic. Cabri indicates that the six in-centres and six nine-point centres are also co-conic.

Article 134 is When the Cevian Conic is a Circle . When the Cevian point is Gergonne's point, the Cevian
Conic is a Circle with centre at the incentre.

Article 135 is Constructing Triangles with Coincident Centroids . Given two triangles with coincident
centroids it is shown how to construct an unlimited number of triangles with the same centroid.

Article 136 is Properties of the symmedian point in a cyclic quadrilateral. If ABCD is a cyclic
quadrilateral then the symmedian points of triangles ABC and DBC are shown to have some remarkable
properties. Many pairs of lines may be constructed whose intersections lie on the sides of the diagonal
point triangle.

Article 137 is Affine tranform of the properties of the Eulerian triangle. Given a triangle ABC and a point
K a configuration is constructed containing a triplicate ratio conic, a circumconic, a nine-point conic and a
7 point conic. When K is actually the symmedian point K(a2 b2, c2), then these conics are the familiar
circles associated with a triangle in the Euclidean plane. In this article we take K, a point not on the sides
of ABC, to be an arbitrary point K(f, g, h) and what emerges is an affine map of the Euclidean plane.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 18


Christopher Bradley's Geometry Writings

Article 138 is Explaining some collinearities among triangle centres. Collinearities are preserved by an
Affine Transformation, so that for every collinearity of triangle centres, there is another when an affine
transformation has taken place and vice-versa. We illustrate this by considering what happens to three
collinearities in the Euclidean plane when an affine transformation takes the symmedian point into the
incentre.

Article 139 is The harmonic cevian conic. If D, E, F are the feet of Cevians with D on BC etc. and if L is
the harmonic conjugate of C with respect to B and D and U is the harmonic conjugate of B with respect to
C and D and if M, N, V, W are similarly defined then a conic passes through the six points L, U, M, V, N,
W. There is one internal Cevian point for which the conic is a circle.

Article 140 is The centroid of centroids. Let P be a point not on the sides of a triangle and suppose the
centroids of triangle PBC, PCA, PAB are L, M, N respectively. Now let J be the centroid of triangle LMN.
Then J lies on line PG, where G is the centroid of triangle ABC. The point of concurrence Q of AL, BM,
CN also lies on PG.

Article 141 is Three centroids created by a cyclic quadrilateral . The centroid of the quadrilateral
considered to be an area of constant density is G, the centroid of the quadrilateral considered as having
unit masses at its vertices is N, the centroid of the quadrilateral considered as having unit masses at its
vertices and mass of two units at E (the intersection of its diagonals) is F. It is shown that E, F, N, G are
collinear.

Article 142 is Incircle conjugation . A construction is presented in which points (f, g, h) in barycentric co-
ordinates are transformed into points (a/f, b/g, c/h). This is clearly a conjugation and we call it incircle
conjugation.

Article 143 is The nine-point circle of the diagonal point triangle . An analysis is conducted of the ten
point rectangular hyperbola through the mid-points of a cyclic quadrilateral and of the nine-point circle of
the diagonal point triangle.

Article 144 is Four nine-point circles . In any quadrilateral ABCD the four nine-point circles of triangles
BCD, ACD, ABD, ABC have a point in common that lies on the hyperbola through the mid-points and on
the circumcircle of the diagonal point triangle.

Article 145 is Concyclic circumcentres in the Steiner configuration . In triangle ABC with centroid G, the
points D, E, F lie on AG, BG, CG respectively and are such that AG = GD, BG = GE, CG = GF. It is
proved that the circumcentres of triangles AFG, FBG, BGD, DGC, CGE, EGA are concyclic.

Article 146 is More conics in a semi-regular hexagon . ABC is an equilateral triangle and AFBDCE is a
hexagon inscribed in a conic. Triangles ABC and DEF are in perspective with vertex P, the centroid of
ABC. G1-G6 are the centroids of triangles EAF, FBD, DCE, BDC, CEA, AFB. It is shown that these
points lie on a conic. Also triangles G1G2G3 and G4G5G6 are congruent and in perspective. The six
interior points of intersection of triangles ABC and DEF also lie on a conic.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 19


Christopher Bradley's Geometry Writings

Article 147 is The midpoint rectangular hyperbola again . If A, B, C, D are four points on a circle and E is
any other point, then the locus of the centre of conic ABCDE as E varies is the rectangular hyperbola
through the midpoints of the quadrangle ABCD.

Article 148 is A circle through two vertces, three circumcentres and a Miquel point . Starting with a
triangle ABC and a point D on BC it is shown how the circumcentres of triangles ABC, ADB, ADC lie on
a circle through A. If this circle intersects circle ABC at A and P, then it is shown that triangle PBC
exhibits a Miquel point Q with points D, M, N on its sides.

Article 149 is Circles with a common point in a cyclic quadrilateral If ABCD is a cyclic quadrilateral it is
possible to find points P, Q, R, S, T, U on sides AB, BC, CD, DA, AC, BD respectively so that circles
BPQU, APST, CQRT, DRSU have a common point M. These circles have centres that are concyclic, as
are points M, U, I, T where I = AC^BD.

Article 150 is When a point is a circumcentre, an incentre and an orthocentre . Given a triangle ABC,
circumcentre O, and the centres D, E, F of circles BOC, COA, AOB respectively, then O is the incentre of
triangle DEF, as well as being the orthocentre of the triangle of excentres.

Article 151 is The Orthocentre Conics . When circles BHC, CHA, AHB in a triangle ABC with
orthocentre H are drawn their centres D, E, F have interesting properties.

Article 152 is Generating Circles from the Symmedian point . Given a triangle ABC and its symmedian
point K circles BKC, CKA, AKB are drawn. Their other points of intersection with the sides of ABC are
shown to lie on a Tucker circle. These points also lie in pairs on the sides of another triangle and the other
intersections of these sides with the circles BKC, CKA, AKB are also concyclic. Centres of these circles
are shown to lie on the Brocard axis.

Article 153 is More properties of the Incentre . In a triangle ABC, with incentre I, the midpoints of AI, BI,
CI are located. Perpendicular bisectors of AD, BE, CF are drawn. These meet the interiors of the sides of
ABC at six points that are shown to lie on an ellipse. As the figure develops to include the excentres
several other circles and conics are located.

Article 154 is On the Incircle and Excircles of a Cevian Triangle . Given triangle ABC, if D, E, F are the
feet of the Cevians AD, BE, CF and L, M, N are the points of contact of the incircle of triangle DEF with
the sides of triangle DEF then AL, BM, CN are con ncurrent, as are AL', BM', CN', where L', M', N' are
the points of contact of the ex-circles with EF, FD, DE respectively.

Article 155 is The Circumconic of a pair of Cevian Triangles . In triangle ABC two Cevian triangles are
drawn and their points of intersection and the vertices of ABC lead to six sets of collinear points. Several
conics are identified.

Article 156 is Circumcentre Conics . Given triangle ABC, let triangle UVW be a reduction of ABC by an
enlargement factor t (0 < t < 1) about the circumcentre O. The sides of triangle UVW cut the sides of ABC
in six points that lie on a conic. It is shown that the centres of all such conics lie on the Brocard axis.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 20


Christopher Bradley's Geometry Writings

Article 157 is The General Inellipse . The general inellipse in a triangle touches its sides at the feet of
Cevians. An analysis is given that relates the Cevian point to the centre of the inellipse and to two related
circumconics.

Article 158 is Six Point Circles and their Associated Conics . If P is a point internal to a triangle ABC and
AP, BP, CP meet the circumcircle again at A', B', C' respectively and U, V, W are the midpoints of AP,
BP, CP and D, E, F are the midpoints of PA', PB', PC' respectively then U, V, W, D, E, F are obviously
concyclic. But the configuration has other properties which are pointed out.

Article 159 is A Twelve Point Configuration and Carnots Theorem . A twelve point configuration occurs
when six of the points lie on a conic and the remaining six points are the vertices of two triangles in
perspective. In this article we consider the configuration in which a conic cuts the three sides of a triangle
in real points. A number of theorems create a variety of conditions on the points of intersection equivalent
to Carnots theorem.

Article 160 is Tangents to a Conic from the vertices of a Triangle . Given a conic which cuts the sides of a
triangle in six real points, tangents A1, A2, B3, B4, C5, C6 are drawn to the conic. The tangents meet each
other at 12 points (15 if you count A, B, C). The se form a 12 point configuration (see Article 159) so that
six of the twelve points lie on a conic.

Article 161 is When 12 points display eight 6-point Conics and 6 Concurrent Lines . Six lines being
tangents to a circle at vertices of triangles in perspective create twelve points of intersection. Six conics
with six points each can be drawn through the twelve points, each point lying on three of the conics. Also
six lines can be drawn through the twelve points, lines which are concurrent at the vertex of perspective.

Article 162 is When 24 Points form three 8-point Conics and 12 Concurrent Lines . A construction is
described involving two chords of a conic that results in 24 points that form three 8-point Conics and 12
concurrent Lines.

Article 163 is A Theorem on the Complete Quadrilateral . Given a complete quadrilateral ABCDEF and
two points P and Q then it is well-known that conics ABCPQ, AFEPQ, BFDPQ, and CDEPQ have a
common point R. It is proved that when P, Q are the midpoints of the diagonals BE, CF respectively, then
R lies on the third diagonal AD.

Article 164 is The Transversal of a Quadrilateral . A transversal LMNP intersects a convex quadrilateral
ABCD with L on AB etc. The harmonic conjugates of L, M, N, P are L', M', N', P'. It is proved that L'P'
and M'N' intersect on the diagonal BD and that L'M' intersects N'P' on the diagonal AC.

Article 165 is A description of properties of Pascal's hexagon . A description is given of properties of


Pascals hexagon, illustrating how 6 Pascal lines fall into two sets of 3 concurrent lines, defining 2 Steiner
points.

Article 166 is Analytic treatment of a Romanian problem. . A triangle ABC is given and a conic that cuts
each of its sides in two real points. The tangents at these points are drawn providing a hexagon of tangents
circumscribing the conic, whose vertices are labelled DF'ED'FE as in Fig.1. It is shown that AD, BE, CF

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 21


Christopher Bradley's Geometry Writings

are concurrent as are AD', BE', CF' and DD', EE', FF'. The positions of the resulting perspectives P, P' and
Q are located. In the Romanian question the conic was the nine-point circle and then P lies on the
hyperbola ABCKOi, where K is the symmedian point and Oi is the isotomic conjugate of the circumcentre
O. The isotomic conjugate of P then lies on the diameter of the circumcircle joining the Steiner point and
the Tarry point. It is also shown that in the general case AD', EF', FE' are concurrent as are BE', DF', FD'
and CF', DE', ED'. Areal co-ordinates are used throughout with ABC the triangle of reference.

Article 167 is Basic Properties of a Quadrangle possessing an Incircle . Properties of a quadrangle


possessing an incircle.

Article 168 is Properties of a pair of Diametrically Opposite Triangles . Starting from a triangle ABC and
its circumcircle centre O, triangle DEF is such that AD, BE, CF pass through O. Tangents at the six points
and the nine lines joining their vertices create a configuration that is explored in this article.

Article 169 is Two In-Perspective Triangles inscribed in a Conic . Triangles ABC and DEF are inscribed
in a conic and are in perspective through a point O. Polar lines LMN and PQR are drawn and six points of
concurrency are shown to define a number of important collinearities. Two conics also emerge.

Article 170 is The Brocard Conics . Triangles ABC and DEF are inscribed in a circle and AD, BE, CF are
concurrent at the symmedian point K. Sides of the two triangles are extended and the tangents at the
vertices are drawn. The polar lines of the two triangles coincide and points of intersection of the sides and
tangents are shown to produce in addition four six-point conics all with their centres on the Brocard axis
OK, where O is the circumcentre of ABC.

Article 171 is A Conic through the feet of two Cevians may lead to a second Conic. Points P and Q are
two Cevian points so that the feet U, V, W, U, V', W' carry a conic then the chords UV, WV', W'U'
intersect as shown if and only if ABCDEF is a conic. Other properties (not proved in this article) hold as
may be inferred from Fig. 1.

Article 172 is The Inner Circle of a Cyclic Quadrilateral. Given a cyclic quadrilateral ABCD with
midpoints of sides P, Q, R, S and midpoints of diagonals U, V, then it is well-known that a rectangular
hyperbola passes through P, Q, R, S, U, V as well as O, the centre and H, J the other two diagonal points.
We find the equation of this hyperbola and that of the inner circle OUVT, centre G. The midpoint W of
UV is also the midpoint of PR and of QS.

Article 173 is The Auxiliary Circles of a Cyclic Quadrilateral. First the cyclic quadrilateral ABCD is
drawn. The midpoints of AB, BC, CD, DA, AC, BD are labelled P, Q, R, S, U, V, the centre of ABCD is
the point O and the diagonal points are T = ACCD, H = ABDC and K = ADBC. (These are all finite
points.) The 10 point rectangular hyperbola P, Q, R, S, O, U, V, T, K, H is shown. We establish the
existence of the three midpoint circles OUVT, OPRH, OQSK and also the Semi-Diagonal point Circle
LMNXYZ, where L, M, N are the midpoints of OT, OH, OK and X = OKHT, Y = OHKT and Z =
OTH.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 22


Christopher Bradley's Geometry Writings

Article 174 is A Cevian Circle leads to Perspective Triangles. A circle is drawn through U, V, W the feet
of a Cevian point T. Three other points are created on the sides of triangle ABC enabling a second triangle
DEF to be drawn. It is shown that triangles ABC and DEF are in perspective.

Article 175 is When two Triangles intersect in Conics and in Circles. If P is internal to a triangle and
points D, E, F are defined on AP, BP, CP respectively so that AD/AP = BE/BP = CF/CP then the
intersections of triangles ABC and DEF lie on a conic. If P is the symmedian point K then the conic is a
circle.

Article 176 is Orthologic Triangles. Triangles ABC and DEF are orthologic if the perpendiculars from the
vertex of one of them to the corresponding sides of the other are concurrent. The property is symmetric
but not transitive.

Article 177 is Special Pascal Lines. By considering the hexagon as two triangles in perspective and
labelling their vertices as well as those of the defining conic, it is shown that two of the Pascal lines pass
through points that are points of concurrence of three lines.

Article 178 is More Circles centred on the Brocard Axis. Circles through B and C and the Brocard points
lead to two circles centred on the Brocard axis and several concurrences.

Article 179 is The Second Brocard Triangle. Given the Brocard points , ' in a triangle the point of
concurrence of the intersection of the line AK with the 7-point circle and circles AC, BOC, A'B (D in
the above diagram) and two similar points E, F found by cyclic change of A, B, C forms what is known as
the Second Brocard Triangle (David Monk, private communication). An analysis is the content of this
document.

Article 180 is Circles through the Brocard points and the Circumcentre. With Brocard points and ',
circles are drawn through A', BOW', COW and they meet at a point D on the circumcircle. Points E, F
are similarly defined. It is proved that the lines AD, BE, CD are all parallel to '. Further, if circle A'
meets AB at X and AC at Y then XY is also parallel to ', as are two further lines defined in a similar
fashion to XY by cyclic change.

Article 181 is Circles through the Brocard points and the Symmedian Point. Circles BK, CK both pass
through the same point D on BC. E and F are defined by cyclic change and lie on CA and AB. DEF turn
out to lie on a line parallel to the tangent at K to the 7-point circle (and hence perpendicular to the Brocard
axis).

Article 182 is Problems requiring Proof. Here are four problems. In fact the solution to problem 4 is
known, but the other problems may well be open.

Article 183 is Conics and Triangles in Perspective. It is shown how a pair of triangles in perspective lead
to a conic and conversely.

Article 184 is Cevian Perspectivity. Triangle ABC and PQR with Cevian points D and S respectively are
said to be in Cevian perspective if the points 11 = ADPS, 22 = BDQS, 33 = CDRS are collinear. If

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 23


Christopher Bradley's Geometry Writings

points are labelled as follows: 12 = ADQS,13 = ADRS, 21 = BDPS, 23 = BDRS, 31= CDPS, 32 =
CDQS, then the following results hold: (i) triangles 13 23 21 and 31 32 12 are in (ordinary) perspective
through a vertex X, (ii) the points 13 23 21 31 32 12 are co-conic and (iii) X lies on the line SD.

Article 185 is The Four Conic Theorem. Given two sets of four lines which intersect one on one in four
points in a straight line, then the twelve remaining points carry four conics in which through point two
conics pass.

Article 186 is 6 Points and 4 perspectives If 6 points provide 1 pair of triangles in perspective then there
are 3 more pairs of perspective triangles, with the same vertex, in the same diagram. Their Desargues' axes
create the sides of a triangle and a related transversal. (It is assumed the original pair of triangles are not in
triple perspective.)

Article 187 is A Perspective in a pair of Cyclic Quadrilaterals. When two cyclic quadrilaterals in the same
circle are in perspective then an axis of perspective joining the intersections of all six pairs of
corresponding sides exists.

Article 188 is Are three Triangles ever in Mutual Perspective? The question in the title has the answer
`Yes', for example three triangles in the Brocard porism are in fact in triple reverse perspective with each
other. However, we study in this very short article the condition that three triangles must obey if they are
to be in mutual perspective. (Perspectives, of course, are symmetric, but not generally associative.)

Article 189 is The Midpoint Conic. It is familiar that the feet of a pair of Cevians are co-conic. We prove
here that their midpoints are also co-conic.

Article 190 is The Brocard Circles' Twin Conic The construction of the Triplicate Ratio Circle and the
Brocard Circle is replicated but starting with the circumcentre O rather than the symmedian point K. It is
found that the Brocard Circle is replaced by a conic (the twin conic) passing through O and K and having
centre the midpoint of OK. This result leads to the concept of a conjugation whose properties we briefly
outline.

Article 191 is Three Circles and their Centres Given a cyclic quadrilateral and a point P not on a side, lines
may be drawn through P parallel to its sides each line meeting the adjacent sides in a point. The eight
points created in this way determine two circles and the three circle centres and P form a parallelogram.

Article 192 is Generalization of the Triplicate Ratio Circle Given a triangle ABC a point P not on its sides
is selected and lines parallel to AB and AC are drawn through P to meet the sides in four distinct points.
The position of P is determined when these four points are concyclic.

Article 193 is A Two Circle Problem Involving the GK axis Through a point X lines are drawn parallel to
the sides of ABC. From the resulting six points two circles are drawn and the condition on X for their
common chord to pass through X is determined, the result being that X must lie on the GK axis.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 24


Christopher Bradley's Geometry Writings

Article 194 is The KG Nine-Point Conic The properties of a Nine-Point Conic are exhibited using Cevian
point K (as well as the centroid G). The symmedian point Km of the medial triangle (X141) is shown to
lie on the KG axis, the midpoint of KKm being the centre X of the nine-point conic.

Article 195 is The Orthocentre Circle of a Cyclic Quadrilateral With ABCD a cyclic quadrilateral
perpendiculars are drawn from A to BC and to CD meeting those sides at L and N respectively and
perpendiculars are drawn from B to DA and to CD meeting those sides at Z and X respectively. Further
the perpendiculars from C to AB and DA are drawn meeting those sides at M and P respectively and
finally perpendiculars are drawn from D to AB and BC meeting those sides at Y and W respectively. Point
A' = BZ^DY, point B' = CM^AL, point C' = BX^DW and point D' = AN^CP. It is found that A', B', C', D'
are concyclic and The circle we term the orthocentre circle as it mirrors the construction of the
orthocentre in a triangle. Nine other circles arise out of the figure and their properties are investigated.

Article 196 is The Two Central Lines in a Cyclic Quadrilateral There are five main central points in a
cyclic quadrilateral: the circumcentre O, the centroid F, the centre of mass G, the intersection of the
diagonals E and the intersection of the lines joining the midpoints of opposite sides T. It is proved in this
article that GTE is a straight line and that OTF is a straight line. Furthermore ET = 3TG and OT = 3TF
and so GF is parallel to OE. (The anticentre also lies on OTF.)

Article 197 is When Two Pairs of Diagonals are Concurrent If ABCD is a cyclic quadrilateral and the
tangents at A, B, C, D form a quadrilateral PQRS, the diagonals AC, BD and PR, QS all pass through a
given point. Here we identify this point.

Article 198 is The Two Cevians' Perspective In a triangle Cevians are drawn through P and Q. The points
L, M, N, R, S, T are defined as follows: L = BQ^CP, M = BQ^AP, N = AQ^CP, R = CQ^BP, S = AQ^BP,
T = CQ^AP. It is proved that triangles LMN and RST are in perspective.

Article 199 is Three Circles with collinear Centres With ABCD a cyclic quadrilateral centre O and
midpoints of sides PQRS, lines PO, QO, RO, SO meet the opposite sides at T, U, V, W respectively.
Circles SUQW and TVPR have centres F and E respectively. It is proved that FOE is a straight line. Some
other obvious results are mentioned.

Article 200 is Problem on Quadrilateral with an Incircle Given a quadrilateral with an incircle there are
vertices A, B, C, D and points of contact P, Q, R, S. Lines AQ, AR, CP, CS are drawn intersecting in K
and M. Lines BS, BR, DP, DQ are drawn intersecting in N and L. It is proved that K, M lie on BD and N,
L lie on AC.

Article 201 is All about a Cyclic Quadrilateral and its cousin Given a Cyclic Quadrilateral ABCD with
circumcentre O and diagonals AC, BD, EF then the midpoints of the diagonals P, Q, R respectively are
well known to be collinear. The points U, V, W are the intersections of the diagonals. It is first shown that
circle UPQ passes through O and that if the line RU meets this circle at S, then the following circles all
pass through S. These circles are VRP, WRQ, UVW, EFO, ACR, BDR. If the circles WVA, WVB, WVC,
WVD meet the circle ABCD at points A', B', C', D' then RAA', RBB', RCC', RDD' are all straight lines. It
is not unreasonable to say that S is the most important point in the figure.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 25


Christopher Bradley's Geometry Writings

Article 202 is The Midpoint Theorem Let ABCD be a quadrangle, with L, M, N the midpoints of AB, BC,
CA and U, V, W the midpoints of AD, BD, CD, then the conics AULMN, BVLMN, CWLMN have a
fourth point in common. (D must not lie on the medians of ABC.)

Article 203 is Another Circle with Centre on the Brocard axis In triangle ABC with symmedian point K,
circumcentre O, circles BKC, CKA, AKB meet the sides at six further points, two on each side. It is
shown these points lie on a circle with centre at a point O' on the Brocard axis such that OK = 2KO'.

Article 204 is Between the Triplicate Ratio Circle and the Circumcircle When UVW is a triangle
homothetic (and similar) to triangle ABC through K then the six points where UV, VW, WU intersect the
sides of ABC lie on a circle. When UVW becomes a point at K one gets the Triplicate Ratio Circle and
when U, V, W reach A, B, C one gets the Circumcircle. The centres O' of these circles move along the
Brocard axis.

Article 205 is The Brocard Lines K and K' Circles A', B', C' meet the sides of ABC also in
six further points which lie on two lines K, K' (which we call the Brocard lines). These six points also
lie on four circles, each of the six points lying on two of the circles.

Article 206 is Dividing OH into Five or Seven Equal Parts By dividing each side of a triangle into quarters
and then drawing the resulting nine Cevian lines, six points are created internally through which two
circles are drawn. Their centres lie on the Euler line and divide OH in the ratios 1:4 and 4:3.

Article 207 is The Median Conic Given a triangle ABC and its centroid G, the midpoints of AG, BG, CG
are denoted by U, V, W. Conics BCUVW, CAUVW, ABUVW are drawn and meet the sides again at
point D, D', E, E', F, F'. It is proved that D, D', E, E', F, F' lie on a conic.

Article 208 is A Pair of Isogonal Conjugates produce a Conic through 6 Points If P and Q are an isogonal
conjugate pair then circles APQ, BPQ, CPQ intersect the sides of triangle ABC in six points and it is
found that a conic passes through them. The construction only produces a conic when P and Q are
isogonal conjugates. When P and Q are the Brocard points then the conic degenerates into a pair of lines.

Article 209 is Three Concentric Circles Starting with a cyclic quadrilateral ABCD with centre O, the
midpoints A', B', C', D' of OA, OB, OC, OD respectively form a second cyclic quadrilateral with half the
size. It is shown that points B, B', C', C lie on a circle and so do C, C', D', D etc. The centre of BB'C'C is
labelled bc, and that of CC'D'D is labelled cd etc. Five pairs of these circles are immediately seen to be
coaxal and analysis is given for one such pair.

Article 210 is 12 Points, 8 six-point Conics, 4 Conics through each Point Given a triangle ABC from the
midpoints of each side perpendiculars are drawn to the other two sides. These lines are labelled Lines 1 to
6 as in the Figure. Their intersections create 12 finite points, so that, for example 35 is the intersection of
Lines 3 and 5. It is found that 8 six-point conics can be drawn through these 12 points, with 4 of the conics
through each point. The nine-point circle is, of course, one of the conics. In a later article some properties
of these conics are established.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 26


Christopher Bradley's Geometry Writings

Article 211 is A cyclic quadrilateral and its midpoint circles A cyclic quadrilateral its midpoints and The
Six Midpoint Circle

Article 212 is Quarter size Circles at Triangle Vertices A cyclic quadrilateral ABCD with AC
perpendicular to BD is such that the tangents at A, B, C and D form a cyclic quadrilateral. This
configuration is described analytically. If AC^BD = X and the centres of the two circles are O and Y, then
it is proved that YOX is a line.

Article 213 is When an Incircle produces a Circumcircle A cyclic quadrilateral ABCD with AC
perpendicular to BD is such that the tangents at A, B, C and D form a cyclic quadrilateral. This
configuration is described analytically. If AC^BD = X and the centres of the two circles are O and Y, then
it is proved that YOX is a line.

Article 214 is The Super-Cevian Triangles, their Conic and Two Perspectives In the above Figure, ABC is
a triangle, P is a Cevian point and D, E, F the feet of the Cevians on BC, CA, AB respectively. The
triangle DEF is the Cevian triangle. A triangle XYZ is constructed by drawing through A a line parallel to
EF and through B and C lines parallel to FD and DE respectively. Also a triangle UVW is constructed by
drawing lines parallel to EF, FD, DE through D, E, F respectively. These two triangles we term as Super-
Cevian triangles. We then prove that ABCUVW is a conic. Q is the perspector of triangles UVW and
XYZ and R is the perspector of triangles ABC and XYZ. These results are established in this article, using
areal co-ordinates with ABC as triangle of reference.

Article 215 is Multiplication of Points using Barycentric Co-ordinates If you have points with barycentric
co-ordinates (f, g, h), (u, v, w) then the rule of multiplication is that the result has co-ordinates (fu, gv,
hw). It is shown in this article how to perform this multiplication using a geometric construction . The
method is as follows: First draw the circumconic fyz + gzx + hxy = 0. Next take the isotomic conjugation
of (u, v, w) to get the point (1/u, 1/v, 1/w). To conjugate this point finally use the circumconic to perform
the second conjugation taking (1/u, 1/v, 1/w) to the product point (f/(1/u), g/(1/v), h/(1/v)) = (fu, gv, hw).
This is evidently a commutative product and so the product may also be obtained using the conic ux + vy
+ wz = 0 and operating the two conjugations on the point (f, g, h).

Article 216 is A Point, two Triangles and Two Conics A triangle ABC is given along with its circumcircle
and any internal point P. Lines AP, BP, CP are extended to meet at U, V, W respectively. Triangle
UVW is drawn and their sides intersect internally in six points. BC meets AU at a, with b and c similarly
defined. AU meets side VW at u with v and w similarly defined. It is shown in this article that a conic
passes through the first six points and that a conic also passes through a, b, c, u, v, w.

Article 217 is An interesting Perspective in the Anticomplementary Triangle A triangle ABC and its
anticomplementary triangle A' B'C' are drawn and a point P is selected through which Cevians A'P, B'P,
C'P are drawn meeting the sides of ABC in points L', M', N' and the sides of A', B', C' in points L, M, N.
Circles LMN and L'M'N' are drawn meeting the sides of A'B'C' and ABC respectively in points U', V', W'
and U, V, W. Several perspectives are created, but the most interesting is that of triangles A'B'C' and
UVW with perspector Q.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 27


Christopher Bradley's Geometry Writings

Article 218 is Properties of the Triangle of Excentres In the triangle of excentres the orthocentre is the
incentre I of ABC, the Symmedian point K+ is the Mittenpunkt of ABCand the nine-point centre is the
circumcentre O of ABC. The circumcentre O+ lies on the line OI and is such that O+O = OI. These
properties are proved as also the fact that KI passes through K+.

Article 219 is The Twelve point Circle In the triangle of excentres the orthocentre is the incentre I of
ABC, the Symmedian point K+ is the Mittenpunkt of ABCand the nine-point centre is the circumcentre O
of ABC. The circumcentre O+ lies on the line OI and is such that O+ O = OI. These properties are proved
as also the fact that KI passes through K+. Wow!

Article 220 is A Conical Hexagon with Main Diagonals Concurrent A hexagon ABCDEF is inscribed in a
conic with diagonals AD, BE, CF concurrent at a point P. It is shown that this property is replicated and
that both conics have the same polar with respect to P. This polar is one of the Pascal lines.

Article 221 is Four Special Conical Hexagons all with same Polar Given a hexagon ABCDEF inscribed in
a conic with AD, BE, CF concurrent at P, the tangents tA, tB, tC, tD, tE, tF are drawn. Points jk = tj^tk are
determined and it is shown that 12, 23, 34, 45, 56, 61 lie on a conic. Lines 61 12, 34 45 meet at point 1,
lines 12 23, 45 56 meet at 2 and lines 23 34, 56 61 meet at 3. Points 1, 2, 3 are collinear and it is shown
that 123 is the polar of P with respect to both conics.

Article 222 is The Missing Point on the Euler Line Given a triangle ABC, circumcentre O, orthocentre H,
let the midpoints of sides be L, M, N and suppose AO, BO, CO meet the sides BC, CA, AB respectively at
U, V, W then the centre of the ellipse LMNUVW is a point S on the Euler line such that OH = 4OS. There
appears to be no reference to this point in the literature, and so we call it The Missing Point.

Article 223 is When a non-regular Cyclic Pentagon leads to Another It can be arranged that a non-regular
cyclic pentagon ABCDE, by special adjustment of D and E, produces a second cyclic pentagon A'B
'C'D'E', where A' = BD^CE, B' = AC^DE, C'= AB^DE, D'= AB^CE and E' = AC^BD.

Article 224 is The Sixty Pascal Poles When six points lie on a conic then sixty Pascal lines may be
constructed. In this paper it is shown how the poles of sixty these lines may be constructed without
drawing tangents. Their properties are, of course, the duals of those of the Pascal lines.

Article 225 is A Typical Pascal line Drawing ABCDEF is a cyclic hexagon (though the results remain true
if the circle is replaced by a conic). A'B'C' is a Pascal line, where A' = BC^AF, B' = CD^AF and C' =
DE^AB. We denote this by (AEC, DBF), where care must be taken with the order of letters. Points X, Y,
Z are given by X = AC^DF, Y = AE^DB, Z = EC^BF. Points X', Y', Z' are given by X' = AD^CF, Y' =
AD^EB, Z' = CF^EB. In Article 225 it is shown that XX', YY', ZZ' are concurrent at a point P which is the
pole of A'B'C' with respect to the circle.

Article 226 is Special Conical Hexagons Hexagon ABCDEF inscribed in a Conic is said to be special if
AD, BE, CF are concurrent at a point P. When this happens hexagons formed by taking the intersections
of neighbouring tangents are proved also to be special and conical with the same point P.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 28


Christopher Bradley's Geometry Writings

Article 227 is The Remarkable Eight-Point Conic Two conics intersect at A, B, C, D and AC meets BD at
P. A line through P is drawn meeting one of the conics at F, G and the other at E, H. Tangents are drawn
at F, G, E, H. Tangents at F and G meet at Q and tangents at E and H meet at R. Tangents at H and G meet
at V, tangents at H and F meet at S, tangents at E and F meet at T and tangents at E and G meet at U. A
conic now passes through the 8 points A, B, C, D, S, T, U, V and the line QR is the polar of P with respect
to all three conics. The general case is technically too difficult to establish, so here we provide a numerical
case.

Article 228 is How 2 Conics through 4 Points generate 6 more such Conics Two Conics through 4 point
A, B, C, D are drawn (in the figure the light blue and the orange). Their tangents to the first (the orange) at
A, B, C, D are labelled 1, 2, 3, 4 and the tangents to the second (the light blue) are labelled 5, 6, 7, 8.
Intersections of these tangents are labelled 12, 13, 14, 23, 24, 34 and 56, 57, 58, 67, 68, 78. It is now found
that six 6 point conics may be drawn.

Article 229 is When a Conical Quadrilateral produces two 6-point Conics If a quadrilateral ABCD is
drawn in a Conic and the tangents are drawn at A, B. C. D then the sides of ABCD and the intersections of
the tangents produce two six point conics and a polar line.

Article 230 is A Quadrilateral and the Conics and Points arising ABCD is an arbitrary quadrilateral
inscribed in a general conic. The tangents at its vertices form a second quadrilateral PQRS. There emerges
the polar line FGTU and 8 points of intersection of ABCD and the tangential quadrilateral PQRS. Many
conics result, but two of them in particular meet on the diagonal PR at points X, Y in the figure. X and Y
appear to have properties as important as any of the 8 vertices of the quadrilaterals and the four points on
the polar line. Not by any means are all the conics that can be drawn featured in this article and a second
article is envisaged. What determines why the diagonals PR and QS have such different properties are (as
far as we are concerned) open and challenging questions.

Article 231 is A Quadrilateral and resulting Conics Part 2 A Special Case showing the role of the New
Points X and Y In Article 230 we investigated properties of a configuration of a quadrilateral ABCD
inscribed in an ellipse and the common points of tangents to the ellipse at the vertices A, B, C, D. It turns
out that there are two remarkable points X and Y through which many lines may be drawn and also
through which many conics pass. The analysis of Part 1 was too complicated to illustrate the results, so in
this Article we give a numerical presentation that illustrates the properties of the point X and Y, for which
the results all hold in the general case. It seems unlikely the general case can be treated analytically, but
possibly a pure approach would succeed as the points are members of a Desargues' involution. Figure 1
above illustrates the line properties of X and Y and Figure 2 below illustrates their Conical properties.

Article 232 is The Coconic Hexagon with Main Diagonals Concurrent ABCDEF is a hexagon inscribed in
a conic and is such that its main diagonals AD, BE, CF are concurrent at a point G. Sides AB, BC, are
labelled 1, 2, . Points such as 14 are then the intersection of AB and DE. Tangents are drawn at A, B,
, so that the tangents at A and B meet at P. Point Q is the intersectionof the tangents at B and C and so
on. The tangents at A, B, are also labelled tA, tB, and points such as ac are thus the intersections of
tA and tC. It is proved that PQRSTU is a conic and that the six points ac, bf, ae, df, ce, bd lie on a conic. It
is also proved that 13, 26, 15, 46, 35, 24 lie on a conic.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 29


Christopher Bradley's Geometry Writings

Article 233 is The Conic and the Lines that are Created (Part 1) A Conic is defined by 5 points, no three of
which are collinear, Pairs of lines joining these points and the tangents to the conic at its defining points
create 10 lines and these lines generate 15 other key lines and many subsidiary ones.

Article 234 is A Conic generates five 8-point Conics (Part 2) A conic ABCDE with sides 1 = AB, 2 = BC
etc. and its tangents a, b, at the vertices AB, produce 20 intersections through which five 8-point
Conics may be drawn.

Article 235 is The Truth and the whole Truth about a Quadrilateral A quadrilateral is inscribed in a conic
and its sides and diagonals and the tangents at its vertices are drawn. A study is made of the lines and
conics that pass through the points of intersection. In all a total of eighteen conics emerge three of which
pass through eight points and fifteen through six points.

Article 236 is A Nice Conic and a Nasty Circle A triangle ABC and its circumcircle S are given. The line
through B parallel to CA and the line through C parallel to AB meet at L. Points M, N are similarly
defined by cyclic change. Tangents to S at B and C meet at U, with V, W defined by cyclic change. It is
proved that LMNUVW is a conic and the centre of circle UVW is shown to lie on the Euler line of ABC.

Article 237 is How to Construct the Ex-points of a given Point It is shown how knowing the Ex-points of
one point it is possible to construct the Ex-points of other points.

Article 238 is When Two Quadrilaterals are in Complete Perspective If two quadrilaterals ABCD and
PQRS are such that AP, BQ, CR, DS are concurrent at X then S may be moved into one and only one new
position on DX so that the four Desargues' axes coincide.

Article 239 is More on Complete perspective It is shown that if two quadrilaterals ABCD and PQRS are
such that AP, BQ, CR, DS are concurrent at a point X (and with the notation that AB^PQ = ab etc.) then if
ab, bc, cd are collinear the quadrilaterals are in complete perspective (see Article 238).

Article 240 is Pentagons in Complete Perspective If ABCDE and PQRST are pentagons such that AP, BQ,
CR, DS, ET are concurrent at a point X. Using the notation AB^PQ = ab etc. it is proved that if ab, bc, cd,
de are collinear then the pentagons are in complete perspective, that is the ten Desargues axes merge into
the one line ab bc.

Article 241 is Quadrilaterals in Perspective inscribed in a Conic If ABCD is a quadrilateral inscribed in a


conic S and P is any point not on S and AP, BP, CP, DP meet S again in points A', B', C', D' then the
quadrilaterals ABCD and A'B'C'D' are in complete perspective. In particular if AC^BD = E and A'C'^B'D'
= E' then E, E', P are collinear..

Article 242 is Perspective Quadrilaterals Two quadrilaterals ABCD and A'B'C'D' are such that AA', BB',
CC', DD' are concurrent at a point X. It is found that a single condition is such that the four Desargues
axes of perspective coincide. It is also the case that when this happens the diagonal point triangles EFG
and E'F'G' are also in perspective with vertex of perspective X and the same Desargues axis of
perspective. See also Article 238.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 30


Christopher Bradley's Geometry Writings

Article 243 is Some properties of a Special Hexagon The hexagon ABCDEF is such that AD, BE, CF are
concurrent at a point U. Given the quadrilaterals ABDE, ACDF, BCEF the diagonal points (other than U)
are labelled V, W, V', W', V'', W''. It is proved that these six points lie on four lines with three points on
each line and such that each of the six points lies on two of the lines.

Article 244 is When two Triangles in Perspective create Conics When triangles ABC and DEF are in
perspective through a point I a conic does not normally pass through A, B, C, D, E, and F. It is proved
here if the position of F is moved along CI a position may always be found for F so that a conic passes
through all six points. When this is so other conics may be drawn through suitably created points.

Article 245 is Some Simple Algebraic Results


Part 1 The conic passing through (1, 0, 0), (0, 1, 0), (0, 0, 1), (u, v, w), (p, q, r).
Part 2 Some results about the conic y2 = zx.
Part 3 The conic yz + zx + xy = 0
Part 4 A conic containing 3 pairs of points in perspective

Article 246 is Pascal points An analysis of a Pascal point is given, using the conic y2 = zx and general
points.

Article 247 is The G Conic Through the centroid G lines parallel to the sides are drawn intersecting the
sides in six points. A conic passes through these six points, which we call the G Conic. Its centre is G and
we obtain its equation.

Article 248 is The H Conic Lines are drawn parallel to the sides through the orthocentre H of a triangle
ABC meeting the sides in six points. The conic through these six points we term the H conic. We obtain
its equation and the co-ordinates of its centre X. It is shown that X, H and K, the symmedian point, are
collinear.

Article 249 is Twin Conics In a triangle ABC if lines through the symmedian point K are drawn parallel to
the sides to meet opposite sides in 6 points, then it is well known that the conic through these six points is
a circle called The Triplicate Ratio Circle. If K is exchanged for another point Q and the same
construction is carried out then a conic passes through the 6 points which we call the Q conic. If X is the
centre of this conic then we show there is a second point Q' also having X as centre and arising from the
same construction. If Q is at the centroid G then Q' is also at G. We show that the Triplicate Ratio Circle is
twinned to the O conic and we determine the twin of the H conic. The general point Q is first treated and
the equation of the Q conic is obtained as also its centre X. The twin conic, the Q' conic, is also obtained.

Article 250 is Cevian Circles The feet of a set of Cevians through a point S are three points through which
a circle may be drawn. The other three points where the circle cuts the sides are also the feet of a set of
Cevians through a point T. The co-ordinates of T are obtained in terms of those of S. When S is the
centroid G then T is the orthocentre. There do not appear to be any other important circles deserving the
title Cevian Circles.

Article 251 is The Four Brocard Triangles The six circles used to construct the Brocard points are
analysed to produce the First and Second Brocard triangles situated on the circumference of the Brocard

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 31


Christopher Bradley's Geometry Writings

circle and similar constructions produce the Third and Fourth Brocard triangles which turn out to be the
Isogonal Conjugates of the first two.

Article 252 is Odds and Ends


Part 1: Equation of Circle from radius and centre
nd th
Part 2: Change of Brocard points and the construction of the 2 and 4 B-Triangles

Article 253 is Triangles in Perspective the Whole Truth Triangles ABC and 123 are in perspective if A1,
B2, C3 are concurrent at a point X commonly called the perspector. It then follows that the points AB^12,
BC^23, CA^31 are collinear. The line is called the Desargues axis of perspective and nowadays is often
called the perspectrix. That is all that is taught to elementary classes. However, it is in fact far from the
end of the matter. If we define the cross-over triangle to have vertices B1^A2, C2^B3, A3^C1 then we
show that B1^A2, C2^B3 and CA^31 are collinear as are B1^A2, A3^C1and BC^23 and A3^C1, C2^B3
and AB^12. In other words there is not just an axis of perspective but a complete quadrilateral of
perspective. But even that is not all. We prove also that the remaining six points A2^B3, B1^C2, B3^C1,
C2^A3, A3^B1, C1^A2 lie on a conic. This last result I have not yet located in past literature.

Article 254 is Circles through the midpoints of sides of a Cyclic Quadrilateral Let ABCD be a cyclic
quadrilateral with diameter AC, with centre O and with external diagonal points Q and R and let S, T be
the midpoints of AQ, AR and U, V the midpoints of BC, CD. Then circles QUVR and BOSTD exist and
their line of centres passes through the midpoint of AO.

Article 255 is Five Conics and a Polar line Two triangles inscribed in a triangle are also in perspective.
Five Conics are now created and a polar line which is also two Pascal lines.

Article 256 is Two Quadrangles in Perspective in a Conic Two Quadrangles in Perspective in a Conic are
drawn. We prove their six diagonal points lie on a conic.

Article 257 is Quadrangles in Perspective Part II The three lines Two quadrangles are drawn in a conic
and are in perspective. Three lines emerge with six points of intersection on each line. They are like Pascal
lines but appropriate to Quadrilaterals in perspective and not to pairs of triangles. It is difficult to be sure
but they may be a new discovery.

Article 258 is The Miquel Cyclic Quadrilateral Let ABCD be a cyclic quadrilateral with points L, M, N on
the line segments AB, CD, DA respectively. Circles ANL and DMN meet at N and a point P (which we
assume does not lie on AC or BD otherwise adjust L, M, N). Circle BPL now meets BC at K. We now
prove that circle CPM passes through K. We thus have a Miquel point P with four circles through it. The
Miquel point is not automatic since the point K has to arise naturally. Circles ANL and CMK meet at P
and another point R that we show lies on AC. Circles DMN and BLK meet at P and another point S that
we prove lies on BD. We then show circle PRS passes through Q the intersection of AC and BD.

Article 259 is Two Triangles and an Eight Point Conic Let ABC and TUV be two triangles with L, M, N
the points BC^UV, CA^VT, AB^TU respectively Suppose adjustments are made so that L, M, N are the
feet of Cevians of both triangles. Then we prove a conic passes through A, B, C, T, U, V and the two
Cevian points.

Romantics of Geometry (https://web.facebook.com/groups/parmenides52/) 32


Article 1

The Story of Hagge and Speckman


Christopher J Bradley

1. Introduction

In 1905 Speckman [1] wrote a paper on indirectly similar triangles in perspective and in
1907 Hagge [2] wrote another paper on the construction of circles that always pass
through the orthocentre of a given triangle, which have become known as Hagge circles.
It is evident that Hagge was unaware of Speckmans work, and indeed it appears that the
contents of Speckmans paper, possibly because it is written in Dutch, seem to be little
known. Hagges work, on the other hand has been developed, notably by Peiser [3], and
more lately by Bradley and Smith [4], [5]. Both the 1905 and 1907 papers are ground
breaking papers in classical geometry, and the work of Speckman, in particular, deserves
to be better known. The interesting thing is that the two pieces of work are related in a
manner that at first sight is not at all obvious. However, if Hagge had read Speckmans
paper it is almost certain that he would have understood its relevance and we would not
have had to wait so long to see the developments of Hagges work that are a
consequence. For reasons that will become clear it is best to describe Hagges work first
and then to summarise Speckmans work, showing how it may be applied, step by step, to
Hagges configuration.

2. Hagge circles

For the construction of a Hagge circle we refer to Fig.1.

Given a triangle ABC and a point P not on a side or the extension of a side and not on the
circumcircle, the Hagge circle of P with respect to ABC is defined as follows: draw AP,
BP, CP to meet at D, E, F respectively; now reflect D, E, F in the sides BC, CA, AB to
obtain the points U, V, W. The Hagge circle (P) is now defined to be the circle UVW. In
his paper [2] Hagge proved
(i) that (P) always passes through H, the orthocentre of triangle ABC;
(ii) that, if (P) intersects the altitudes AH, BH, CH at X, Y, Z respectively, then UPX,
VPY, WPZ are straight lines;
(iii) that the Hagge circle of G, the centroid of ABC, with respect to the medial triangle
(the triangle joining the midpoints of the sides) is the Brocard circle on OK as
diameter, where K is the symmedian point of ABC, O being the circumcentre of
ABC and thus the orthocentre of the medial triangle. At the time, Brocard
geometry was a subject of particular interest, which may account for why roughly
half of Hagges paper is devoted to this application.

1
Peisers [3] main contribution was concerned with the centre Q of the circle (P). Using
the algebra of complex numbers he was able to show that the image of Q under a half-
turn about the nine-point centre T is the same point as the isogonal conjugate Pg of P
with respect to triangle ABC. Since T is the midpoint of OH it follows that the figure
PgHQO is a parallelogram. Another property not mentioned by earlier geometers is that
the midpoints of AU, BV, CW all lie on the nine-point circle.


Pg
Y
F H
T G
O U
W
P E
Q
Z
B C
V
X

Fig. 1

If P lies on the circumcircle the lines APg, BPg, CPg are parallel, so Pg lies on the line at
infinity. The consequence is that the Hagge circle degenerates into a line through H,
which being the line through the reflections of P in the sides BC, CA, AB is parallel to the
Wallace-Simson line of P and is sometimes called the double Wallace-Simson line of P.

When P is the incentre I the Hagge circle is the Fuhrmann circle and this circle was
known about many years before Hagges publication. The Fuhrmann circle not only
passes through H but also passes through Nagels point Na. An account of the Fuhrmann
circle is given by Honsberger [5], who shows that HNa is one of its diameters. When P is
the symmedian point K the Hagge circle is the orthocentroidal circle on GH as diameter.

2
In general there is a straightforward result that PgG meets the Hagge circle at the point at
the other end of the diameter through H. See Fig. 1.1.

In their first paper Bradley and Smith [4] give a synthetic proof of Peisers result and also
show how a consistent definition can be given for a Hagge circle if P lies on a side of the
triangle or at a vertex. In their second paper Bradley and Smith [5] provide another proof
of Hagges result (ii).

Now if you consider the quadrilateral HXYZ, since CZH and BHY are perpendicular to the
sides, we have ZXY = ZHY = BAC. Similar angle relations show that triangles
ABC and XYZ are similar, and since the ordering of the letters in the two triangles is
opposite, these triangles are indirectly similar. The same applies to triangles UVW and
DEF. Bradley and Smith [4] also give a synthetic proof of the important result that there
is an indirect similarity carrying ABCDEFP to XYZUVWP in which P is the unique fixed
point. In Fig. 2 we provide an illustration of the dilative reflection through P and its axis
that provides this indirect similarity.

3
F
A

V'

X'

V E
Z' X
Z
P
Q W
U

H
Y
U' W'
C
B

Y'

Fig. 2

On reflection in the axis through P the figure ABCDEF is mapped to XYZUVW and
then by dilation through the centre P by a factor XY/AB the figure XYZUVW is
mapped to XYZUVW. Note that if the perpendicular axis is used instead, then a rotation of
180o is required in addition to the dilation and reflection.

It is also significant, as we shall see in due course, that triangles ABC and XYZ are in
perspective with vertex of perspective H, and that triangles XYZ and UVW are in
perspective with vertex of perspective P.

4
Further results proved by Bradley and Smith [5] are
(i) that, if VW meets AH at U, WU meets BH at V and UV meets CH at W, then U,
V, W, P are collinear, and
(ii) that, if U1, V1, W1, X1, Y1, Z1 are the midpoints of DU, EV, FW, AX, BY, CZ, then
U1, V1, W1, X1, Y1, Z1 lie on a conic, which they call the midpoint conic.

As regards result (ii), it is in fact the case that if A, B, C, D, E, F are six points on a conic
and if X, Y, Z, U, V, W are their images on a conic under an indirect similarity then
the midpoints of AX, BY, CZ, DU, EV, FW lie on a conic. In fact one can go further and
say that the points that divide these six segments in any fixed ratio lie on a conic.

3. Speckmans theorems on indirectly similar triangles in perspective

Speckman announced proofs of his theorems in April 1903 at a Congress in


-Gravenhage and they were published [1] in 1905. Their relevance to the work of Hagge
lies in the fact that in the Hagge configuration, described in Section 2, triangles ABC and
XYZ are indirectly similar and are in perspective with vertex of perspective H.

If Hagge had read Speckmans work, then it seems highly likely that he would have
applied the theorems to obtain a sequence of additional results. And if Speckman had
known about Hagges work it seems almost certain he would have referred to it
subsequently. As far as we are aware the application of Speckmans results to the Hagge
configuration has never been pointed out and is a story waiting to be told. We tell this
story by relating the more important of Speckmans results paragraph by paragraph,
pointing out in each case its significance in relation to Hagges configuration. Whilst it is
true in an obvious sense that Speckmans results are more general than Hagges (he deals
with the case when the vertex of perspective is a general point and not the orthocentre
and consequently derives more general results), it is possible to hold an alternative point
of view that Hagges configuration is illustrative of all the properties one might need to
know about pairs of indirectly similar triangles in perspective. For if you have an
arbitrary pair of indirectly similar triangles in perspective, one of them is always related
to the other by being directly similar to one of its partners Hagge triangles (defined to be
the triangle XYZ inscribed in a Hagge circle).

This is illustrated in Fig. 3. It shows two arbitrary triangles ABC and ABC that are
indirectly similar by means of a reflection in the line l through P followed by dilation
through P. They are chosen to be in perspective with vertex a point Q. Triangle XYZ is
the Hagge triangle of P and triangle XYZ shows a dilation of XYZ through P with an
appropriate scale factor. It can now be observed that triangles ABC and XYZ are
images of one another under a rotation about P. If the point P is not known it may be
obtained as the other point of intersection besides Q of the two rectangular hyperbolae
ABCHQ and ABCHQ. See the first paragraph of Section 4.

5
Q

To Q
l

C'
Y'

X' A'
X P
H
H'
Z

B C
B'

Z'

Fig. 3

4. Paragraphs 1 to 9 of Speckmans paper

We refer to Fig. 3 in which H and H are the orthocentres of triangles ABC and ABC, Q
is the perspector and P is what Speckman calls the double-point of inverse similarity. He
proves in the first paragraph that rectangular hyperbolas may be drawn passing through
ABCHPQ and ABCHPQ and possessing asymptotes that are parallel.

From now on we describe Speckmans results primarily as they apply to the Hagge
configuration. The more general results, if not stated, should be clear enough.

In Fig. 4 the result of the first paragraph is illustrated when the second triangle is the
Hagge triangle XYZ. Note that Q now coincides with H and the asymptotes are parallel to
the axes through P, which Speckman calls the double lines of inverse similarity. Also he

6
shows in the second paragraph that these lines are parallel to the angle bisectors of
corresponding lines of triangles ABC and XYZ. The reflection of XYZ in one of these axes
is the triangle XYZ, which is a dilation of triangle ABC through P. Triangles ABC and
XYZ are orthologic, as indeed are all pairs of indirectly similar triangles (a fact we prove
later in this article) and the orthologic centres of the triangles are D and H, where the
point D lies on both the circumcircle of ABC and the rectangular hyperbola ABCHP.
Also, as shown in Bradley and Smith [4], the orthocentre h of triangle XYZ, P and D are
collinear. The centres of the two rectangular hyperbolas are denoted by M and m. Also
illustrated in Fig. 4 is the fact, proved in Speckmans third paragraph, that J, j and P are
collinear where j is the 180o rotation of j about m, and J, j are any pair of corresponding
points on the two hyperbolae under the similarity.

J
Co Bo

D
j'
A
Z

X'

h
M
m X
Ao
P
H
Y' Z'
Y
j
B C
J'

Fig. 4

In paragraph four there is a construction, whereby triangle ABC is rotated by 180 about
the centre M of the hyperbola ABCHP to produce a triangle A0B0C0 and it is proved that
triangles XYZ and A0B0C0 are also inversely similar and in perspective, but now the

7
vertex of perspective is P, the double point of inverse similarity is H and the orthocentre
of triangle A0B0C0 is D. In the more general case the interchange of the roles of the
double point of inverse similarity P and the vertex of perspective Q arising from this
construction is an intriguing result.

In paragraph five there is proof of the result that corresponding points on the two
hyperbolae together with the points obtained by half-turns about their respective centres
M and m are concyclic. An arc of the circle JjjJ can be seen in Fig. 4.

In paragraph seven there is a proof that in the general case (when the perspector is not at
one of the orthocentres) the two orthocentres are collinear with the perspector. This
result is trivial in the Hagge configuration, since Q then coincides with H. In paragraph
nine it is proved that the lines connecting the orthological centre of one triangle to the
orthocentre of the other passes through the double point of inverse similarity. When
applied to the Hagge configuration this implies that the line hD passes through P, a result
which is also proved in Bradley and Smith [5].

8
M
To M'

Z
H
U
N'
N V' U' V
P h D
Y
B L C
W
X

L'

Fig. 5

5. Axes of perspective in the Hagge configuration

In Fig. 5 we show the axis UVW defined at the end of Section 2 and also the Desargues
axes LMN, LMN of perspective arising respectively from the perspectives of triangles
ABC, XYZ and of triangles XYZ, UVW. Paragraph ten of contains the proof that the axis
LMN is perpendicular to the line HD joining the orthologic centres.

9
6. Paragraphs 11 to 18 of Speckmans paper

Fig. 6 illustrates the results of paragraphs 11-16. Most of the notation has previously been
defined. Pg is the isogonal conjugate of P with respect to triangle ABC and pg is the
isogonal conjugate of P with respect to triangle XYZ.

In paragraph eleven it is noted that the line through A parallel to YZ, the line through B
parallel to ZX and the line through C parallel to XY are known to meet and it is proved
they do so at a point S on the circumcircle of ABC. Speckman calls this point the
metapole of triangle ABC with respect to triangle XYZ. Such a point is now called a
paralogic centre, so it turns out that we are dealing with a case in which the triangles are
not only orthologic but also paralogic. (It is not always the case that orthologic triangles
are also paralogic.)The point s is defined similarly. In the next two paragraphs there are
proofs about the properties of the lines DS and Hs. They are diameters, respectively, of
circles ABC and XYZ. It is proved that the line Hs is parallel to the isogonal conjugate of
the rectangular hyperbola ABCHP, which is a line through Pg, and that DS is parallel to
the isogonal conjugate of the rectangular hyperbola XYZHPh, which is a line through pg.
Speckman calls the lines DS and Hs orthological diameters of the circles. A curious fact
is that in the Hagge configuration the intersection R of the two orthological diameters lies
on the rectangular hyperbola ABCHP.

In paragraph fourteen the midpoints of the sides of ABC are denoted by Ma, Mb, Mc so that
triangle MaMbMc is indirectly similar to triangle XYZ. It is shown that the perpendiculars
from Ma on to YZ, Mb on to ZX and from Mc on to XY meet at a point W, which is the
midpoint of HS. The point w is similarly defined, and is the midpoint of hs. Also HS/hs =
AB/XY.

In paragraph fifteen it is proved that the paralogic centre of triangle MaMbMc with respect
to triangle XYZ is the midpoint of DH. This is the centre of the hyperbola ABCHPD. In
paragraph seventeen it is shown that the Desargues axis of perspective LMN, see Fig. 5,
bisects the line segment Ss joining the paralogic centres.

In paragraph sixteen it is shown that if you reflect triangle XYZ in the Desargues axis
LMN to get triangle XYZ, then triangle XYZ is in perspective with triangle ABC with
vertex of perspective the point where DH meets the circumcircle ABC. In paragraph
eighteen it is then noted that if the axis LMN meets the common chord of circles ABC and
XYZ at E, then E is the radical centre of circles ABC, XYZ and XYZ. Hence the line
through E perpendicular to OQ is the radical axis of the circumcircle and the Hagge
circle. The construction of the radical axis of the circumcircle and the Hagge circle is a
splendid flourish with which to end the paper.

10
To E

A
Isogonal
conjugate of
D Z common chord
ABCHP my of ABC and
s pg M
X'Y'Z'
w h
X X'

g P
Q Mb
mx Mc mz

O
G

H
Z'
Y Pg

B L Ma
C
Y'

Isogonal
conjugate of
XYZHP
To N
Fig. 6

11
7. Which of all these properties hold if there are two indirectly similar triangles
that are not necessarily in perspective?

Y X
C
B

Fig. 7

We first provide a construction. Take a triangle ABC, draw any circle (other than the
circumcircle) and choose any point T lying on it. Draw lines through T parallel to the
altitudes of triangle ABC to meet the circle again at points X, Y, Z. See Fig. 7. Simple
angle arguments show that triangle XYZ is indirectly similar to triangle ABC. As T moves
around this circle the triangle XYZ also rotates around the circle and CABRI shows there
may be up to three positions of T for which triangles ABC and XYZ are also in
perspective, but that is not a feature that interests us in this paragraph. The construction is
such that the lines through X, Y, Z perpendicular to BC, CA, AB respectively are
concurrent at T and hence triangles XYZ and ABC are not only indirectly similar, they are
orthologic.

12
Theorem 1.1
Every pair of indirectly similar triangles are orthologic and the orthologic centres lie on
the circumcircles of the triangles.

Proof
Given a triangle XYZ that is indirectly similar to a triangle ABC, we can construct the
circumcircle of XYZ and choose any point T on this circumcircle. We can then generate a
triangle X'Y'Z' by drawing lines through T parallel to the altitudes of triangle ABC. Now
all indirectly similar triangles inscribed in a given circle are congruent, so we can rotate T
until triangles X'Y'Z' and XYZ coincide. The construction given at the outset of Section 7
now ensures that triangle XYZ is orthologic to ABC with orthologic centre at the final
position of T.

Triangles DEF and UVW are now defined as in the Hagge construction and they are
similar as one is the image of the other under the indirect similarity. Paralogic centres S
and s may be defined, even when there is no perspective. However, when there are no
perspectives, there are no rectangular hyperbolas. But it is true in the more general case
that the double lines of inverse symmetry are parallel to the angle bisectors of
corresponding sides of triangle ABC and XYZ.

When there is no perspective there is no Desargues axis of perspective LMN. However


Cabri confirms that the axis U'V'W'P exists, as defined at the end of Section 2. The lines
KS and Ts pass through the centres of their respective circles and so may justifiably be
called orthologic diameters

8. What happens when the centre of inverse symmetry lies at the orthocentre

When Speckman wrote his paper [1] on indirectly similar triangles in perspective he kept
his account general and did not take account of particular cases. In other words his
account assumes that the centre of inverse symmetry P, the perspector Q and the
orthocentres of triangles ABC and XYZ are distinct points. The result is that he missed
two special cases that are highly interesting. When the perspector Q coincides with H, the
orthocentre of ABC, one gets triangles inscribed in circles through H and this gap was
filled three years later by the ground breaking work of Hagge [2]. What has never been
considered in the period from 1905 to the present day is what happens when the centre of
inverse symmetry P lies at H. In fact it produces cases that are just as interesting as the
Hagge circles that arise when Q lies at H.

When P lies at H there are an infinite number of rectangular hyperbolas passing through
A, B, C and H depending on the direction of the lines of inverse symmetry. And sure
enough it turns out that if you choose any axis through H and then use it to form a
triangle XYZ that is inversely similar to ABC, XYZ always turns out to be in perspective
with ABC. The circumcircle of triangle XYZ obviously cannot be reduced by dilation
through H to become a non-degenerate Hagge circle. In other words there are a whole set
of pairs of indirectly similar and in perspective triangles and a whole set of circles, whose

13
properties have not been studied before. Of course, since Speckman just starts with a
given set of points A, B, C, P and Q, his results are applicable when P lies at H, but he
never considered what additional properties are true when either P or Q lies at H. And
these additional properties are, as we shall see, quite substantial. Hagge circles are the
outcome of what happens when Q lies at H. We now describe what happens when P lies
at H.

First, if you reflect ABC in any line through H, and use any enlargement factor k (k 0,1)
for an enlargement (or reduction) centre H, then the resulting triangle XYZ is always
similar to and in perspective with triangle ABC, with vertex of perspective at some point
Q. Now let AH, BH, CH meet the circumcircle of ABC at D, E, F respectively. Through
D draw a line parallel to AX to meet XP at U and define V, W similarly. Then it transpires
that points U, V, W lie on the circumcircle of triangle XYZ and that triangle UVW is the
image of triangle DEF under the indirect similarity. We know from work in Section 7,
since XYZ is indirectly similar to ABC, there must exist points T' and T that are orthologic
centres of the two triangles and that these points lie on the circles XYZ and ABC
respectively and have the properties that the perpendiculars from T' on to BC, CA, AB
pass through X, Y, Z respectively and the perpendiculars from T on to YZ, ZX, XY pass
through A, B, C respectively. It also follows that T' lies on the rectangular hyperbola
XYZHQ and T lies on the rectangular hyperbola ABCHQ. See Fig. 8 for illustration of all
these properties. The earlier properties involving triangles XYZ, DEF and UVW require
analytic proof and we now turn our attention to setting up this analysis.

14
Q

Z
T'
E
U
J'
V
F
T
J
Y H
C
B

W D
X

Fig. 8

Choosing co-ordinates

In dealing with a triangle it is possible to choose three co-ordinates, such as the angular
dispositions of the vertices and thereby make use of symmetry, or it is possible, by
choosing the axes in some preferred directions, to choose just two co-ordinates. In the
latter case symmetry is lost, but the expressions for the equations of lines and co-
ordinates of points may be less complicated. In the present calculation which features an
axis through H and dilation through H it is desirable for this point to be taken as origin.
We therefore present a calculation in which the vertices have co-ordinates as follows:

15
A( 2 2vw, 0), B( 2vw, 2v), C( 2vw, 2w). It may now be checked that H is the origin
and the circumcentre O has co-ordinates (1 3vw, v + w). The radius of the circumcircle
R is given by R2 = (1 + v2)(1 + w2).

The equations of BC, CA, AB are respectively


x + 2vw = 0, (8.1)
y wx = 2w(1 + vw), (8.2)
y vx = 2v(1 + vw), (8.3)

and the equations of AH, BH, CH are respectively


y = 0, (8.4)
wy + x = 0, (8.5)
vy + x = 0. (8.6)

The equation of the circumcircle is


x2 + y2 + 2(1 + 3vw)x 2(v + w)y + 8vw(1 + vw) = 0. (8.7)

The indirect similarity

First perform the dilation through H by a factor of k (k 0, 1) and then the images of A,
B, C are respectively X', Y', Z' with co-ordinates X'( 2k(1 + vw), 0), Y'( 2kvw, 2kv), Z'(
2kvw, 2kw). Since H is the origin we may suppose the axis of reflection through H has
equation y = mx. We quote the result that the image of the point with co-ordinates (c, d)
after reflection in this line has co-ordinates {1/(1 + m2)}(2dm + c(1 m2), 2cm d(1
m2)). Using the co-ordinates of X', Y', Z' as c, d we obtain the co-ordinates of the images
X, Y, Z as

X: {1/(1 + m2)}( 2k(1 m2)(1 + vw), 4km(1 + vw)),


Y: {2kv/(1 + m2)}(2m w(1 m2), 2mw (1 m2)),
Z: {2kw/(1 + m2)}(2m v(1 m2), 2mv (1 m2)).

The perspective

We now prove that triangles ABC and XYZ are in perspective. Since this is true for all
values of k 0, 1 and for all values of m this is a very general result that implies that any
axis through H may be used for the reflection and any enlargement factor (k = 1would
mean that AX, BY, CZ are parallel). An interesting case that we do not provide separate
analysis for is when the axis through H is the Euler line and the enlargement factor is 0.5.
The image of the circumcircle is then the nine-point circle and the points X, Y, Z are
additional natural points on the nine-point circle.

The equations of AX, BY, CZ turn out to be

2kmx + (1 + m2 k(1 m2))y + 4km(1 + vw) = 0, (8.8)

16
(1 + m2 + k(1 + 2wm m2))x + (w(1 + m2) k(w 2m wm2)y = 4kv(m(1 w2) w(1
m2), (8.9)
(1 + m + k(1 + 2vm m ))x + (v(1 + m ) k(v 2m vm )y = 4kw(m(1 v ) v(1 m2),
2 2 2 2 2

(8.10)
respectively.

These lines are concurrent at the point Q with co-ordinates

Q: (4k/{(1 k2)(1 + m2)2})((k(m4vw + m3(v + w) + 2m2 m(v + w) + vw) + m4vw + m3(v


+ w) + m(v + w) vw), m(k(m2(vw 1) + 2m(v + w) vw + 1) + (1 + m2)(1 + vw))).
(8.11)

Of significance is the equation of circle XYZ and this is


(1 + m2)(x2 + y2) 2k(m2(3vw + 1) + 2m(v + w) (1 + 3vw))x 2k(m2(v + w) 2m(3vw
+ 1) (v + w))y + 8k2vw(1 + vw)(1 + m2) = 0. (8.12)

Triangle UVW and its properties

We define points U, V, W as the points of intersection of the lines XH, YH, ZH with circle
XYZ. This means that UVW is inscribed in the same circle as triangle XYZ and is in
perspective with it. Since X, Y, Z are the images of A, B, C in the indirect similarity and H
is the fixed point of the similarity, it follows that if AH, BH, CH meet the circumcircle at
points D, E, F respectively, then U, V, W are the images of D, E, F in the similarity and
consequently triangles DEF and UVW are similar.

The equation of XH is
(1 m2)y = 2mx. (8.13)

This meets circle XYZ again at U, with co-ordinates {4k/(1 + m2)}( ( 1 m2), 2m). The
equation of YH is
(m2w + 2m w)y = (m2 2mw 1)x. (8.14)

This meets circle XYZ again at V, with co-ordinates


{4kw(1 + vw)/(1 + m2)(1 + w2)}(m2w + 2m w, m2 2mw 1).

The equation of ZH is
(m2v + 2m v)y = (m2 2mv 1)x. (8.15)

This meets circle XYZ again at W, with co-ordinates


{4kv(1 + vw)/(1 + m2)(1 + v2)}(m2v + 2m v, m2 2mv 1).

The equation of AH is
y = 0. (8.16)
This meets the circumcircle at the point D, with co-ordinates ( 4vw, 0).

17
The equation of BH is
wy + x = 0. (8.17)

This meets the circumcircle at the point E, with co-ordinates


{4w(1 + vw)/(1 + w2)}( w, 1).

The equation of CH is
vy + x = 0. (8.18)

This meets the circumcircle at the point F, with co-ordinates


{4v(1 + vw)/(1 + w2)}( v, 1).

Having defined U, V, W in the above manner their main property is that the line through
D parallel to AX passes through U, the line through E parallel to BY passes through V and
the line though F parallel to CZ passes through W.

The line AX has equation given by (8.8), so the line parallel to AX through D has equation

2kmx + (1 + m2 k(1 m2))y + 8kmvw = 0. (8.19)

It is now straightforward to show this meets the line XH, with equation (8.13), at the
point U with co-ordinates {4k/(1 + m2)}( ( 1 m2), 2m). Similarly the line parallel to
BY through E meets YH at V and the line parallel to CZ through F meets ZH at W.

References
1. H.A.W. Speckman, Perspectief Gelegen, Nieuw Archief, (2) 6 (1905) 179 188.
2. K. Hagge, Zeitschrift fr Math. Unterricht, 38 (1907) 257-269.
3. A.M. Peiser, The Hagge circle of a triangle, Amer. Math. Monthly, 49 (1942) 524-
527.
4. C. J. Bradley and G. C. Smith, Hagge circles and isogonal conjugation, Math.
Gaz., 91 (2007) p202.
5. C. J. Bradley and G.C. Smith, On a construction of Hagge, Forum. Geom.
7(2007) 231-247.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

18
19
Article 2

The Four Hagge circles

Christopher J Bradley

A H2

H3

G2 N2
N3 D
H
G3 GN
H4
G4
N4 M
O
G1 N1
C

B H1

Fig. 1

1. Preliminary analysis

1
In later sections we state and prove a number of theorems about the four Hagge circles, with
respect to an appropriately chosen centre of inverse similarity P, of the four triangles BCD, ACD,
ABD, ABC that comprise a cyclic quadrilateral ABCD. It is desirable therefore to give a
description of the basic configuration involved and to prove some straightforward facts that are
required later.

We use vectors with origin O, the centre of the circumscribing circle, and with A, B, C, D having
vector positions a, b, c, d each of which has magnitude R, the radius of the circle. The triangles
BCD, ACD, ABD, ABC are denoted by 1, 2, 3, 4 respectively. Suffices are used for key
points in these triangles. Thus Hk is the orthocentre of k, Nk is the nine-point centre of k and
Gk is the centroid of k, k = 1, 2, 3, 4. The vector positions corresponding to the points are h1 = b
+ c + d, h2 = a + c + d, h3 = a + b + d, h4= a + b + c and, as in any triangle, nk = hk and gk =
hk. The following facts are now easy to establish:
(i) AH1, BH2, CH3, DH4 are concurrent at the point N with vector position (a + b + c +
d);
(ii) AN1, BN2, CN3, DN4 are concurrent at the point G with vector position (a + b + c +
d);
(iii) AG1, BG2, CG3, DG4 are concurrent at a point M with vector position (a + b + c +
d);
(iv) H1, H2, H3, H4 lie on a circle of radius R with centre H, which has position vector (a +
b + c + d) and moreover H1H2H3H4 is the image of ABCD under a180 rotation about
N;
(v) N1, N2, N3, N4 lie on a circle of radius R with centre N and N1N2N3N4 is homothetic
with H1H2H3H4, with centre O and enlargement factor ;
(vi) The nine-point circles of triangles 1, 2, 3, 4 , each having radius R therefore all
pass through N;
(vii) A known result is that a rectangular hyperbola through the vertices of a triangle
passes through it orthocentre. It follows that the rectangular hyperbola through A,
B, C, D passes through all of H1, H2, H3, H4;
(viii) Another known result is that the centre of a rectangular hyperbola passing through the
vertices of a triangle lies on its nine-point circle. In consequence of (vii) the centre of
is the point N;
(ix) G1, G2, G3, G4 lie on a circle of radius R with centre G, which has position vector
(a + b + c + d);
(x) The points O, M, G, N, H are collinear and if O and H are given co-ordinates 0, 1 on
this line, then M, G, N have co-ordinates , , respectively.

See Fig. 1 for an illustration of these results.

2
2. Setting the scene

It is clear from Speckmans work on indirect similar perspective triangles that if we now choose
any point P on the rectangular hyperbola defined in Section 1 Result (vii), it is possible to find
the Hagge circle of P with respect to each of the four triangles 1, 2, 3, 4 and to draw them
on the same diagram. These are the four Hagge circles 1, 2, 3, 4 of the article heading and
they are shown in Fig. 2. Note that as the axes of inverse similarity through P are parallel to the
asymptotes of the rectangular hyperbola , these axes coincide for each of the Hagge circles.

A consistent notation is essential and this we now describe. See Fig. 2. The centres of the circles
are denoted by Qk, k = 1, 2, 3, 4. For reasons that become clear shortly we re-label the four
orthocentres A1, B2, C3, D4, so that, for example D4 is the orthocentre of triangle ABC. The lines
AP, BP, CP, DP meet the circumcircle at points A', B', C', D'. We now consider the labelling
of points on the Hagge circle 4, which is the Hagge circle of P with respect to triangle ABC.
They all carry the subscript 4. The reflection of A' in BC we denote by A4', the reflection of B' in
CA we denote by B4' and the reflection of C' in AB we denote by C4'. These replace the labels U,
V, W used in Article 1. The points where A4'P, B4'P, C4'P, D4P meet 4 are denoted by A4, B4, C4,
D4' respectively. The first three of these replace the labels X, Y, Z in Article 1. Points with
suffices 1, 2, 3 are similarly defined on circles 1, 2, 3. The labels Pg1, Pg2, Pg3, Pg4 are given
to the isogonal conjugates of P with respect to triangles 1, 2, 3, 4 respectively. The double
lines of inverse symmetry through P are denoted by L and L'. The centre of the rectangular
hyperbola is denoted by M.

In Section 3 we use Cartesian co-ordinates, origin M and the asymptote parallel to L is chosen as
the x-axis. In this way has equation xy = 1 (by choice of scale), and we use a parameter t on the
hyperbola so that its points have co-ordinates (t, 1/t). The points A, B, C, D, P are given
parameters a, b, c, d, p, where it is known, since ABCD is cyclic, that abcd = 1 and the
parameters of A1, B2, C3, D4, since they are the orthocentres, have parameters a, b, c, d
respectively.

The following theorems now hold:

Theorem 1
The centres, Q1, Q2, Q3, Q4, of the four Hagge circles are collinear.

Theorem 2
The points A1, A2, A3, A4, A1', A2', A3', A4' are collinear and similarly for the points Bk, Bk', k = 1,
2, 3, 4 and Ck, Ck', k = 1, 2, 3, 4 and Dk, Dk', k = 1, 2, 3, 4.

3
Theorem 3
The quadrilateral Pg1Pg2Pg3Pg4 is similar to the quadrilateral A'B'C'D'.

L' C3
B2
C2'
L
Pg3 Pg2 A3'
C4
D3'' D
To D3'
C' A B3'
D4' D2
B' To A3'
B4'
A3
C1' B1 A4

D1 A1'
M
Q3 Q4 Q1 Q2
A1 P

B1'
D1' D4
D3
O C1
To A3'
A4' A2 Pg1 ToD2'
C4''
B4
B Pg4
A3' B2' C
C3'
To C2 D2'
B3 A'

D'
C2

Fig. 2

3. The analysis

The co-ordinates of O

The chord AB has equation


aby + x = a + b. (3.1)

4
The midpoint of AB has co-ordinates ((a + b), (a + b)/(ab)). It follows that the perpendicular
bisector of AB has equation
2ab(y abx) = (a + b)(1 a2b2). (3.2)

The perpendicular bisector of BC has equation


2bc(y bcx) = (b + c)(1 b2c2).

These meet at O, the centre of , at the point with co-ordinates ((a + b + c + d), (1/a + 1/b +
1/c + 1/d). Here we have used abcd = 1.

It can be shown that the radius of the circumcircle is


(a2 + b2 + c2 + d2 + 1/a2 + 1/b2 + 1/c2 + 1/d2). (3.3)

The equation of the circumcircle

From the above information this is easily proved to be


2x2 2x(a + b + c + d) + 2y2 2y(1/a + 1/b + 1/c + 1/d) +
(ab + ac + ad + bc + bd + cd + 1/ab + 1/ac + 1/ad + 1/bc + 1/bd + 1/cd) = 0. (3.4)

Alternatively, in terms of a, b, c only it has the form

abcx2 (abc(a + b + c) + 1)x + abcy2 (a2b2c2 + ab + bc + ca)y +


(abc)(ab + bc + ca) + a + b + c = 0. (3.5)

A digression

The reflection of the point with co-ordinates (f, g) in the line with equation lx + my = n is the
point with co-ordinates (h, k) where
h = {f(m2 l2) + 2l(n gm)}/(l2 + m2),
k = {2mn + g(l2 m2) 2flm}/(l2 + m2).

This is straightforward and is left to the reader. This is used later to reflect ABCD in the line L
and then dilate through P to obtain the Hagge circle A4B4C4D4.

The indirect similarity

We now determine the equations governing the indirect similarity which maps the circumcircle
into the Hagge circle of P with respect to 4. This is done by following what happens to the point
D(d, 1/d). We know the image of this point is the orthocentre D4( 1/abc, abc) and we know
the mapping is effected by a reflection in the line L, with equation y = 1/p followed by an

5
enlargement (reduction) by a factor PD4/PD. As far as the reflection is concerned we may use
the analysis of the last paragraph with l = 0, m = 1, n = 1/p. The result of this reflection on D is to
produce the point with co-ordinates (d, 2/p 1/d).

The enlargement (reduction) factor (comparing x-co-ordinates of the points concerned) is equal
to (p + d)/(p d) = (abcp + 1)/(abcp 1). It may now be checked that the reflection in L
followed by the dilation with this enlargement (reduction) factor takes the point with co-
ordinates (h, k) to the point with co-ordinates (x, y) where
x = (abchp 2p + h)/(abcp 1), (3.6)
and
y = (abc(kp 2) + k)/(1 abcp). (3.7)

Using h = a, k = 1/a we deduce the co-ordinates of A4 to be


((a2bcp + a 2p)/(abcp 1), (2a2bc abcp 1)/(a(abcp 1))),
with similar expressions for the co-ordinates of B4 and C4 by using b and c, in x and y above,
instead of a.

The equations of the Hagge circles

From here it is quite an elaborate calculation to obtain the equation of the Hagge circle A4B4C4
and check the fundamental theorem that D4 lies on this circle. The computer algebra package
DERIVE was used to perform the calculations, and the result is that 4 has equation
abc(abcp 1)(x2 + y2) (a2b2c2p(a + b + c) + abc(a + b + c) 3abcp + 1)x
+ (a3b3c3p 3a2b2c2 + (abcp + 1)(bc + ca + ab))y
2a3b3c3 + a2b2c2p(a + b + c) + abc(bc + ca + ab) (abcp + 1)(a + b + c) 2p = 0. (3.8)

The equations of the other Hagge circles follow immediately by using other triplets of
parameters instead of a, b, c. The co-ordinates of the centres of these circles may now be written
down and it has been checked, using DERIVE, that any three of the four centres are collinear.
Theorem 1 is now proved, but it also follows from a geometrical argument. The centre O of the
circumcircle is the common circumcentre of all four triangles 1, 2, 3, 4. If we now reflect O
in the two lines of inverse similarity, it maps into a pair of points collinear with P. Since the
centres of the four Hagge circles are now the images of these points by means of dilations
through P with enlargements (reductions) using different factors, the resulting images all lie on
this line, and this line also passes through P. Theorem 2 follows by a similar argument, bearing
in mind that we already know primed and unprimed pairs of points such as A4, A4' are collinear
with P.
The quadrilateral A'B'C'D'

The equation of the line AP is

6
x + apy = a + p. (3.9)

This meets the circumcircle again at the point A' with co-ordinates (x, y) where
x = {ap2(abc(b + c) + 1) p(a2b2c2 + a(b + c) bc) + abc}/{bc(a2p2 + 1)}, (3.10)
y = {abcp2 + p(a2bc abc(b + c) 1) + ab2c2 + b + c}/{bc(a2p2 + 1)}. (3.11)

The point D' has co-ordinates similar to these, but with d replacing a.
From these co-ordinates we can work out (A'D')2 and the result is
{(a d)2(b p)2(c p)2(b2c2 + 1)}/{b2c2(a2p2 + 1)(d2p2 + 1)}. (3.12)

The quadrilateral Pg1Pg2Pg3Pg4

The co-ordinates (h, k) of Pg4, the isogonal conjugate of P with respect to triangle ABC are given
by
h = (a + b + c p)/(1 abcp), (3.13)
k = (p(bc + ca + ab) abc)/(abcp 1). (3.14)

This may be checked as follows: Let the line from A to (h, k) meet at A'', then it is easy to show
that A'A'' is parallel to BC. The symmetry of h, k with respect to a, b, c now proves that (h, k) are
the co-ordinates of Pg4, as this calculation reflects one of the standard constructions for an
isogonal conjugate. Indeed the parallel A'A'' to BC indicates that AP and APg4 are reflections of
each other in the internal bisector of angle A. The co-ordinates of Pg1 now follow from (h, k) by
exchanging a and d.

From these co-ordinates we can work out (Pg1Pg4)2 and the result is
{(a d)2(b2c2 + 1)(b2p2 + 1)(c2p2 + 1)}/{(abcp 1)2(bcdp 1)2}. (3.15)

The similarity of the quadrangles


The ratio of the squares of the side lengths and diagonals of the quadrangles Pg1Pg2Pg3Pg4 and
A'B'C'D' may now be calculated and is the totally symmetric expression
{(a2p2 + 1)(b2p2 + 1)(c2p2 + 1)(d2p2 + 1)}/{(a p)2(b p)2(c p)2(d p)2}. (3.16)

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

7
Article 3

Generalizations of Hagges Theorems

Christopher J Bradley

1. Introduction

There are two unrelated generalizations of the Hagge configuration.

The first generalization involves moving from the orthocentre H to another point K at the
expense of losing the indirect similarity. It also appears that the material of Article 2 cannot be
completely generalized, as there is a restriction on the facility for creating four circles from a
cyclic quadrilateral. Also it does not appear there is a prescription that relates the point of
perspective P between the triangles XYZ and UVW and the centre Q of the circle through K on
which they lie. In the case of the Hagge construction Peiser [1] showed that the isogonal
conjugate Pg of P is located in such a way that the nine-point centre N is the midpoint of QPg. In
Sections 3 6 we give proofs of the main results that form the extensions of Hagges theorems
and the four Hagge circle property. In the Hagge construction AP, BP, CP are drawn to meet the
circumcircle at points D, E, F respectively. These points are then reflected in the sides BC, CA,
AB to create three new points U, V, W. The first Hagge theorem is that the circle UVW passes
through the orthocentre H. Next one draws AH, BH, CH to meet this circle at X, Y, Z respectively
and the second Hagge theorem is that UX, VY, WZ are concurrent at P. It is interesting that in this
generalization one does not start with a point P, but with the centre Q of the proposed circle
UVWK.

The second generalization is designed to maintain the property that triangles ABC and XYZ have
orthologic points with respect to each other and this leads to these triangles and their
circumcircles being related by an indirect similarity and also to triangles DEF and UVW being
related by the same indirect similarity. What is lost is the reflection property mentioned in the
last paragraph. We describe these generalizations in turn and the following results hold for the
first:

2. Results for the first generalization

Theorem 1
Let ABC be a triangle and K any point not lying on its sides or extensions. With centre any point
Q draw a circle to pass through K. The intersections with of circles BKC, CKA, AKB are
denoted by U, V, W respectively. Then

1
(BU/CU)(CV/AV)(AW/BW) = 1. (2.1)

In this expression (BU/CU), for example, is taken as positive if KBUC is convex and negative
otherwise. The converse is also true, that if U, V, W lie on a circle and this relation holds, then
the circle passes through K.

This result is due to Boreico [2].

Theorem 2
Given the configuration of Theorem 1, suppose now AK, BK, CK meet at X, Y, Z respectively,
then UX, VY, WZ are concurrent at a point P.

In other words the pairs (U, X), (V, Y), (W, Z) are in involution on , by means of a perspective
point. These results are shown in Fig. 1.

3. Proof of Theorem 1

Invert with respect to K and denote inverse points by primes. We require U', V', W' collinear,
which is the case by Menelaus Theorem if, and only if, (B'U'/C'U')(C'V'/A'V')(A'W'/B'W') = 1.
However, B'U'/C'U' = (BU/CU)(KB'/KC'). Multiplying three such relationships we get the
required result.

4. Proof of Theorems 2, 3 and 4

Theorem 2
We use Cartesian co-ordinates with K as origin and let have centre Q with co-ordinates ( ,
). This merely imposes a scale and a direction on the configuration, so there is no loss of
generality. Then has equation
x2 + y2 + x + y = 0. (4.1)

We parameterize by taking lines through K to be of the form y = mx and then m serves as the
parameter of the point where this line meets again.

Circle BKC and meet on the line KU with equation (p 1)x + (q 1)y = 0, so the parameter for
the point U is a = (p 1)/(q 1). Similarly the parameters for V and W are respectively
c = (s 1)/(t 1) and e = (u 1)/(v 1). Now circles CKA and AKB meet at A and K, so
their common chord is the line AK, which therefore has equation (s u)x + (t v)y = 0. It follows
that the parameter of X is b = (s u)/(t v). Similarly Y and Z have parameters that are
respectively d = (p u)/(q v) and e = (p s)/(q t).

2

A

V
Z

X
P
U K

Y W

B C

Fig. 1

To prove that UX, VY, WZ are concurrent it is sufficient to show that the pairs (a, b), (c, d), (e, f)
are in involution. This is because pairs in involution on a conic must arise from a vertex of
perspective. Since an involution of pairs (h, k) is specified by an equation of the form lhk + m(h
+ k) + n = 0 for some real numbers l, m, n (m2 nl) it follows that the pairs are in involution if

3
the determinant with rows(ab, a + b, 1)(cd , c + d, 1), (ef, e + f, 1) vanishes. DERIVE verifies
this is indeed the case.

Theorem 3
Let VW meet AKX at L, with M, N similarly defined, then L, M, N, P are collinear.

Proof
Consider the hexagon VWUXKY on the circle . VW^XK = L, WU^KY = M and UX^YV = P. It
follows by Pascals theorem that L, M, P are collinear. Similarly M, N, P are collinear.

The above theorems may be generalized even further, as the following theorem shows:

Theorem 4
Let ABC be a triangle and D, E, F any three generally situated points. Draw any conic through
D, E, F. Let conic BCDEF meet at U, conic CADEF meet at V and conic ABDEF meet at
W. Further let AD, BD, CD meet at X, Y, Z respectively, then UX, VY, WZ are concurrent.

Proof
Project E, F to the circular points at infinity.

Theorem 4 is illustrated in Fig. 2

4
A

V
D
Y E
P
W
C
F
B U
Z

Fig. 2

5. Restricted generalization of the four Hagge circle configuration

We refer back to Article 2 for a detailed description of this configuration, but in brief what
happens is as follows: Take a triangle ABC inscribed in a circle . Let its orthocentre be D4. Take
a point P anywhere (not on or the sides of ABC), that will serve as a centre of inverse
similarity. Draw the rectangular hyperbola through A, B, C, D4 and P. Its centre, as is well
known, is a point M on the nine-point circle of ABC. Now perform a 180 rotation of A, B, C, D4
about M to get the points A1, B2, C3, D. Then, wherever the initial selection of P is made, it is
always the case that D lies on and the points A1, B2, C3 are the orthocentres of triangles BCD,
ACD, ABD respectively. The four Hagge circles of P with respect to the triangles BCD, ACD,
ABD, ABC can now be drawn using the same inverse spiral symmetry about P in each case.

If we now choose D4 to be a general point of triangle ABC other than the orthocentre, the same
construction does not work. If P is chosen anywhere (not on or the sides of ABC), then it turns
5
out that the point D does not lie on . In fact there is only one conic through A, B, C, D4 for
which this turns out to be true. The point M that is the centre of this conic lies somewhere on the
circular locus of points for which D4M produced to meet at D is such that D4M = MD. P is
restricted to lie anywhere on this conic, which may be a hyperbola or an ellipse.

The next question to ask is how to draw the four circles that have the point P as their involution
point. As we no longer have either Peisers [1] prescription or the inverse spiral symmetry to rely
upon, there has to be some other method of obtaining the four circle centres from P. We draw on
the property of the four Hagge circle configuration to provide the answer. In that configuration it
turned out that points Ak (k = 1, 2, 3, 4) lie on a line through P. We call this the A-line. Similarly
there is a B-line, a C-line and a D-line. We insist for the generalization that these lines must
exist. We now describe how to obtain the points A4, B4, C4. A4 is the intersection of A1P with
AD4, B4 is the intersection of B2P with BD4 and C4 is the intersection of C3P with CD4. The circle
4 is now defined to be the circle A4B4C4.

In order to justify the validity of the construction certain theorems have to be proved.

Theorem 5
The circle 4 passes through D4 and similarly 1, 2, 3 pass through A1, B2, C3

Theorem 6
Let circle BCD4 meet 4 at A4' then A4A4' passes through P. Similarly if circle CAD4 meets 4 at
B4' then B4B4' passes through P etc.

Altogether there must be twelve such lines passing through P, three for each of the circles k, k =
1, 2, 3, 4. However because of the symmetry of the configuration only one such case needs to be
established.

Theorem 7
The centres Qk of the circles k, k = 1, 2, 3, 4 are collinear. Their radii are in proportion to their
distances from P.

In Fig. 3 we illustrate all the above properties.

6
C3 B2 D
line
D2'
C2'
D B3'
A
C1 D3

B1'

Q3 M D1
A3' C4' B4
A2 D4'
A1' Q1
P
A4
Q4
A4' A3
D4 A1
O A
B4'
D1' C4 Q2 line
C1' A2'
B1
D3' C3' C
B
D2
B3

B Q
line line
C
B2' line

C2

Fig. 3

6. Analysis and proofs

Consider the circle with equation


4m2(x2 + y2) + m(m2 + 1)(bdc + acd + abd + abc a b c d)x (m2 + 1) (bdc + acd + abd
+ abc + a + b + c + d)y + (m2 + 1)(ab + ac + ad + bc + bd + cd) 2(m2 1) = 0. (6.1)

It may be verified that it contains the points (x, y) where

7
x = {1/(2m)}(t 1/t), y = {(1/2)}(t + 1/t), (6.2)

for t = a, b, c, d, provided abcd = 1 These define parametrically the hyperbola with equation y2
m2x2 = 1 and the points with parameters a, b, c, d may be taken to define the points A, B, C, D,
the cyclic quadrilateral inscribed in . The centre M of the hyperbola is the origin and the image
of ABCD under the 180 rotation about M defines the congruent quadrilateral A1B2C3D4 and
moreover the points A1, B2, C3, D4 lie on the hyperbola and have parameters a, b, c, d
respectively. The condition that four points on the hyperbola are concyclic is that the product of
their parameters is 1. It follows that the sets of points (ABC3D4), (AB2CD4), (AB2C3D),
(A1BCD4), (A1BC3D), (A1B2CD) are concyclic and that their equations may be obtained by
altering signs in the equation of as appropriate. The equation of the chord of the hyperbola
joining points with parameters s and t is

m(1 st)x + (1 + st)y = s + t. (6.3)

The point P is now chosen on the hyperbola with parameter p. We can now put (s, t) successively
equal to (a, d), ( a, p) and get the equations of the lines D4A and A1P. Their intersection is by
definition the point A4, which has co-ordinates (x, y), where

x = {(a2 + 1)(d + p) 2a(dp + 1)}/{2am(p d)}, (6.4)


y = {(a2 1)(d + p) + 2a(1 dp)}/{2a(p d)}. (6.5)

The co-ordinates of B4, C4 may now be obtained by using parameters b, c rather than a. Points
B1, C1, D1 follow by using parameter a instead of d and b, c, d respectively instead of a. Points
A2, C2, D2 follow by using parameter b instead of d and a, c, d respectively instead of a. Points
A3, B3, D3 follow by using parameter c instead of d and a, b, d respectively instead of a. The
equation of the line A1P is worth recording as it is what we have termed the A-line. It has
equation

m(1 + ap)x + y(1 ap) = p a. (6.6)

Proof of Theorem 5
Take the co-ordinates of A4, B4, C4, D4 and construct a 4 x 4 matrix consisting of rows with
entries (x2 + y2, x, y, 1) for each of the four points. Then take its determinant and factorize and
DERIVE provides the answer

. (6.7)

Since abcd = 1 it follows that the four points are concyclic on a circle we denote by 4. A similar
proof establishes the existence of circles 1, 2, 3.

8
We now determine the equation of circle 4. This is done using derive by the same method as in
the proof of Theorem 5, but by using current co-ordinates instead of those of D4. The result is

(6.8)

The line AP meets 4 at the point with co-ordinates (x, y) where

x = (1/(2bcm(a2p2(m2 + 1) + 2ap(m2 1) + m2 + 1))) x (a2bcp(m2 + 1)(b(c p) cp + 1)


a(b2c(c p)(m2 +1) b(c2p(m2 + 1) c(3m2(p2 1) p2 + 1) p(m2 + 1)) + p(c p)(m2 + 1)) +
b(m2 + 1)(1 cp) + c(m2 +1) m2p p), (6.9)

y = (1/(2bc(a2p2(m2 + 1) + 2ap(m2 1) + m2 + 1))) x (a2bcp(m2 + 1)(b(c p) cp 1) + a(b2c(c


p)(m2 + 1) b(c2p(m2 + 1) + c(m2(p2 + 1) 3p2 3) + p(m2 + 1)) p(c p)(m2 + 1)) b(m2 +
1)(cp + 1) c(m2 + 1) + m2p + p). (6.10)

When these co-ordinates are substituted into the equation of circle BD4A1C they are found to
satisfy it, and hence they are the co-ordinates of the point A4' lying on the A-line and the two
circles. Similar analysis provides all twelve pairs of points on circles 1, 2, 3, 4 that are in
involution through the involution point P. This completes the proof of Theorem 6.

The equations of the circles k, k = 1, 2, 3, 4 having been obtained it is now possible to find the
co-ordinates of their centres. The co-ordinates of Q1, Q4, P may then be used to show these
points are collinear. This is done by forming the determinant whose three rows are (x, y, 1) where
(x, y) are successively the co-ordinates of the three points. The value of this determinant turns out
to be

(1/(32ab2c2d(abcp 1)(bcdp 1))) x ((a d)(m2 + 1)(abcd 1)(bcdp + 1)(b3c2p(m2 +1) +


b2c2(cp(m2 + 1) 2(m2(p2 1) + p2 + 1)) b(2c(m2(p2 1) p2 1) + p(m2 + 1)) cp(m2 + 1))).

(6.11)

9
This vanishes, on account of the factor (abcd 1) and so the three points are indeed collinear.
Similarly any other two circle centres are collinear with P, so we have identified the existence of
a Q-line containing all the circle centres and the point P. This establishes the first part of
Theorem 7.

The second part of Theorem 7 is not difficult to check. The co-ordinates of all points are now
known and it soon follows that triangles such as A4B4C4 and A1B1C1 are similar and that the
enlargement factor is the same as PQ1/PQ4 (distances along the Q-line are signed, depending on
which side of P the centres are, and this corresponds to whether the triangles are directly
homothetic through P or whether a 180 twist is involved as well). See Fig. 3 again.

7. Results for the second generalization

Choose any point T, not on the sides of triangle ABC or their extensions, which will act as a
pseudo-orthocentre. In the original Hagge configuration H acts as a centre of perspective and as
an orthologic centre of triangles ABC and XYZ. The generalization that is appropriate is to ensure
that T is an orthologic centre. An arbitrary circle is now drawn through T, which will serve as the
generalized Hagge circle. CABRI indicates that there are generally two positions of T on the
circle for which it will also act as a centre of perspective, but the configuration shown in Fig. 4 is
not one of these cases, as that is not necessary for the generalization.

To force T to be an orthologic centre drop the perpendiculars from T onto BC, CA, AB and
suppose these lines meet the circle through T at points X, Y, Z respectively. This forces XYZ to
have an orthologic centre with respect to ABC. Also the perpendiculars mean that, as may be
proved by simple angle chasing, that triangle XYZ is similar to ABC and ordering of labels show
that it is indirectly similar to ABC. It follows that the orthologic centre of ABC with respect to
XYZ exists and is the point J in the figure. J and T will be corresponding points in the indirect
similarity that necessarily arises as a result of the orthologic property, so J will lie on the
circumcircle of ABC.

Standard methods may be used to find the centre P of inverse similarity and the axes of
reflection (lines of inverse similarity). Only one is shown in Fig. 3.4.

Having located P draw AP, BP, CP to meet the circumcircle at D, E, F respectively. Find the
images U, V, W under the indirect similarity. These are bound to lie on the circle through T. Also
triangles DEF and UVW are bound to be similar, as they are related by the indirect similarity.

10
J
A

T F

Q
O
T'
X

W C
B D
V
Z U
E P

J'

Fig. 4

The following theorem now holds:

Theorem 8

XPU, YPV, ZPW are straight lines. In other words triangles XYZ, UVW are in perspective with
centre P, or since they all lie on a circle X, U and Y, V and Z, W are pairs in an involution on the
circle through T by projection through P.

11
Despite there not being a perspective, some of Speckmans results still hold or generalize. In
particular the conics ABCPJ and XYZPT will be images of each other in the indirect similarity.
These conics are not necessarily hyperbolae; and even then CABRI confirms that their
asymptotes are parallel only when a perspective exists between triangles ABC and XYZ.
However, triangles ABC and XYZ are paralogic, with paralogic centres at the opposite ends of the
diameter to T and J respectively. These are labeled T' and J' in the figure.

Theorem 9

Let VW meet XT at L, with M, N similarly defined, then LMN is a straight line.

Theorem 10

The midpoint conic (midpoints of AX, BY, CZ, DU, EV, FW) exists and is illustrated in Fig. 5
along with the axis LMN.

Theorem 10 is, of course, a general theorem concerning six points on a pair of conics connected
by an indirect similarity and needs no separate proof. Theorem 9 is unproved, but CABRI
indicated. See Fig. 5.

12
J
A

T F
M
Q
O
T'
X
L
W C
B D
Z V
N U
E P

J'

Fig. 5

Proof of Theorem 8

This is trivial since it is a direct consequence of the indirect similarity. The points A, P and D are
mapped by the indirect similarity on to points X, P and U. Since APD is a straight line, it follows
that XPU is a straight line, P being the only invariant point and lines being mapped into lines.
Similarly YPV and ZPW are straight lines.

13
References
1. A.M. Peiser, The Hagge circle of a triangle, Amer. Math. Monthly, 49 (1942) 524-527.
2. I. Boreico, Result attributed to him (Private communication).

Flat 4
Terrill Court
12-14 Apsley Road,
BRISTOL BS8 2SP.

14
ARTICLE 4

STUDIES IN SIMILARITY

Christopher J Bradley

1. Introduction

Given a scalene triangle ABC there are precisely six points J such that the angles BJC, CJA, AJB
have values = 180o A, = 180o B, = 180o C in some order. Three of these points are
well known to be the orthocentre H and the two Brocard points and, which in this article,
for reasons that will almost immediately become clear, we denote by H+ and H-. The other three
points are less familiar. We denote them by aH, bH, cH and they lie on the circumference of the
orthocentroidal circle (diameter GH), where the medians meet this circle (other than at the
centroid G).

The angles BJC, CJA, AJB at these points are shown in the Table 1 below.

J BJC CJA AJB


H
H+
H-
aH
bH
cH

The notation used for the six points is prompted by Table 1, which reflects the group of
symmetries of an equilateral triangle.

Each Brocard point may be constructed by drawing three circles and these circles together with
the circles BHC, CHA, AHB produce nine circles. We first show how these nine circles may be
used to locate the points aH, bH, cH. We determine the equations of the nine circles and the co-
ordinates of aH, bH, cH showing that they have the locations stated above. We also find the co-
ordinates of the centres of these nine circles. The nine centres have the following remarkable
properties. They form six triangles with each centre appearing as a vertex of two of the triangles.
Each of these triangles is similar to triangle ABC, three of them being directly similar and three
being indirectly similar (with vertices in the opposite order). Also the three that are indirectly
similar are mutually in perspective with O, the circumcentre of ABC, as perspector.

Given a triangle ABC with orthocentre H then, if we draw a general circle through H, its points
of intersection X, Y, Z with the altitudes AH, BH, CH respectively, determine a triangle XYZ that
is indirectly similar to triangle ABC. The circles through H are known as Hagge circles [1]. Since

1
the similarity is indirect there is a centre of inverse symmetry P and lines through P, which are
the axes of inverse symmetry. If AP, BP, CP meet the circumcircle again at D, E, F and XP, YP,
ZP meet the Hagge circle again at U, V, W then triangles DEF and UVW are also similar. There
is an obvious connection between the vertices of the two triangles, which is that U is the
reflection of D in BC, V the reflection of E in CA and W the reflection of F in AB.

The reason a Hagge circle carries a triangle that is indirectly similar to ABC by means of the
construction mentioned above is that the angles subtended at H by the sides BC, CA, AB are , ,
respectively. The proof of this is a straightforward argument using angle chasing.

One might therefore hope, that if we use either of the Brocard points H+ or H- and draw a
general circle through that Brocard point and find its intersections with the line segments from
that Brocard point to the vertices, then these intersections might determine a triangle ZXY or
YZX, which is indirectly similar to triangle ABC. This indeed proves to be the case and again
there is a centre and axis of inverse symmetry, see Bradley [2].

If one follows the same procedure with any of the three points aH, bH, cH we get triangles that
turn out to be directly similar to triangle ABC. Direct similarities have rather different properties
from indirect similarities and these properties are the subject of a comprehensive survey by
Wood [3]. We point out how our new circles and triangles fit into Woods scheme.

2. The nine circles and the six triangles

We first consider three circles that determine H. If you reflect O in the side BC to obtain the
point aA and then draw a circle centre aA of radius R (the circumradius of ABC), then this circle
is the circle BHC. It is also the case that, if bB and cC are the reflections of O in CA and AB
respectively, then circles centres bB and cC of radius R are respectively the circles CHA and
AHB. The common point of these three circles is therefore the orthocentre H. In terms of areal
co-ordinates, since the co-ordinates of H are given by (1/(b2 + c2 a2), 1/(c2 + a2 b2), 1/(a2 + b2
c2)), the equations of these circles are

BHC: a2yz + b2zx + c2xy (b2 + c2 a2)x(x + y + z) = 0, (2.1)


CHA a2yz + b2zx + c2xy (c2 + a2 b2)y(x + y + z) = 0, (2.2)
AHB a2yz + b2zx + c2xy (a2 + b2 c2)z(x + y + z) = 0. (2.3)

Using the fact that the centre of a circle is the pole of the line at infinity we find the
unnormalised co-ordinates of the centres are
aA( a2(b2 + c2 a2), a4 c4 + a2b2 + 2c2a2 + b2c2, a4 b4 + 2a2b2 + c2a2 + b2c2), (2.4)
4 4 2 2 2 2 2 2 2 2 2 2 4 4 2 2 2 2 2 2
bB( b c + 2b c + a b + c a , b (c + a b ), b a + b c + 2a b + c a ), (2.5)
cC( c4 b4 + c2a2 + 2b2c2 + a2b2, c4 a4 + 2c2a2 + b2c2 + a2b2, c2(a2 + b2 c2)). (2.6)

Triangle aAbBcC is congruent to triangle ABC being the image of ABC under an 180o rotation
about the nine-point centre N.

We next consider the three circles defining the Brocard point H+, which has co-ordinates (1/b2,
1/c2, 1/a2). These are well known to be (i) the circle through A and B touching BC at B, (ii) the

2
circle through B and C and touching CA at C and (iii) the circle through C and A touching AB at
A. We denote their centres by aB, bC, cA respectively. The equations of these circles are:
AH+B: a2z2 + (a2 b2)zx c2xy = 0, (2.7)
2 2 2 2 2
BH+C: b x + (b c )xy a yz = 0, (2.8)
CH+A: c2y2 + (c2 a2)yz b2zx = 0. (2.9)

The unnormalised co-ordinates of their centres are


aB(2c2a2, a4 b4 + 2a2b2 + c2a2 + b2c2, c2(c2 + a2 b2)), (2.10)
bC( a2(a2 + b2 c2), 2a2b2, b4 c4 + 2b2c2 + a2b2 + c2a2), (2.11)
cA( c4 a4 + 2c2a2 + b2c2 + a2b2, b2(b2 + c2 a2), 2b2c2). (2.12)

We next consider the three circles defining the Brocard point H-, which has co-ordinates (1/c2,
1/a2, 1/b2). These are well known to be (i) the circle through A and C touching BC at C, (ii) the
circle through B and A and touching CA at A and (iii) the circle through C and B touching AB at
B. We denote their centres by aC, bA, cB respectively. The equations of these circles are:

CH-A: a2y2 b2zx + (a2 c2)xy = 0, (2.13)


AH-B: b2z2 c2xy + (b2 a2)yz = 0, (2.14)
BH-C: c2x2 a2yz + (c2 b2)zx = 0. (2.15)

The unnormalised co-ordinates of their centres are


aC(2a2b2, b2(a2 + b2 c2) , c4 a4 + a2b2 + 2c2a2 + b2c2), (2.16)
bA( a4 b4 + b2c2 + 2a2b2 + c2a2, 2b2c2, c2(b2 + c2 a2)), (2.17)
cB( a2(c2 + a2 b2), b4 c4 + c2a2 + 2b2c2 + a2b2, 2c2a2). (2.18)

The co-ordinates of the nine centres are unnormalised but in each case they sum to the same
amount (a + b + c)(b + c a)(c + a b)(a + b c). The nine circles are illustrated in Fig. 1.
Also shown in Fig. 1 is the orthocentroidal circle on GH as diameter, with equation shown in
Bradley and Smith [4] to be
(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2 a2yz b2zx c2 xy = 0. (2.19)

The median AG meets this circle again at the point aH with co-ordinates
aH(a2, b2 + c2 a2, b2 + c2 a2). To prove that this is indeed the point aH featured in Section 1,
one may verify that it lies on each of the circles with centres aA, aB, aC and consequently the
sides BC, CA, AB subtend at aH the angles , , these being the angles in the segments of
those circles. Similarly bH and cH lie on the orthocentroidal circle and on the medians BG and
CG respectively and have co-ordinates bH(c2 + a2 b2, b2, c2 + a2 b2) and cH(a2 + b2 c2, a2 +
b2 c2, c2). The point bH lies on the circles with centres bA, bB, bC and cH lies on the circles
with centres cA, cB, cC.

3. Lengths of segments and the similarities

If ABC is an acute-angled triangle the distances from the vertices to the orthocentre and to the
Brocard points are known, see Shail [5], to be
AH = 2R cos A, BH = 2R cos B, CH = 2R cos C;
AH+ = 2Ro sin B/sin A, BH+ = 2Ro sin C/sin B, CH+ = 2Ro sin A/ sin C;

3
AH- = 2Ro sin C/sin A, BH- = 2Ro sin A/sin B, CH- = 2Rosin B/sin C,
where Ro = abc/(b2c2 + c2a2 + a2b2)1/2. The reader may verify that the corresponding formulae
for aH, bH, cH are

A aH = 4Ro cos A, B aH = 2Ro sin A/sin B, C aH = 2Ro sin A/ sin C,


A bH = 2Ro sin B/ sin A, B bH = 4Ro cos B, C bH = 2Ro sin B/ sin C,
A cH = 2Rosin C/sin A, B cH = 2Ro sin C/sin B, C cH = 4Ro cos C,

where Ro is, in each case, the radius of the pedal triangle of the corresponding point. In an
obtuse-angled triangle these become signed lengths. For the point aH the value of
Ro = bc/(2b2 + 2c2 a2)1/2 with a similar formula for the points bH, cH.

aB
A

bA
cC
bB
cA
H+ bH aC
G
O
aH
H
H-
cH

C
B
bC

aA

cB

Fig. 1(a)

We next consider the triangle cA aB bC and show that it is similar to triangle ABC.

4
The displacement bC aB = (a2(a2 + b2 + c2), a4 b4 + c2a2 + b2c2, b4 b2 c2 a2b2 + 2c2a2
Using the areal metric , see Bradley [6], we find that (bC aB)2 = ka2, where
k = (3a2b2c2(a2 + b2 + c2) + 2(b4c4 + c4a4 + a4b4) a6(b2 + c2) b6(c2 + a2) c6(a2 + b2)) (3.1)
all divided by {(a + b + c)(b + c a)(c + a b)(a + b c)}2. Similarly (cA bC)2 = kb2 and (aB
cA)2 = kc2, from which the result follows. Another short calculation shows that triangle bA cB aC
is congruent to triangle cA aB bC. These two triangles together with ABC and aA bB cC are all
directly similar to one another and fall into two congruent pairs.

We now consider the three triangles aA aB aC, bA bB bC and cA cB cC. Note first that the
centres aA, bC, cB all lie on the perpendicular bisector of the side BC. Similarly the centres bB,
cA, aC all lie on the perpendicular bisector of CA and the centres cC, aB, bA all lie on the
perpendicular bisector of AB. It follows that triangles aA aB aC, bC bA bB, cB cC cA are
mutually in perspective with the common perspector O.

5
aB
A

bA

cC
bB
cA
H+ bH aC
G
O
aH
H
H-
cH

C
B
bC

aA

cB

Fig. 1(b)

The displacement aB aC = (2a2(b2 c2), a2(a2 3b2 c2), a2(b2 + 3c2 a2)). Using the areal
metric we find (aB aC)2 = la2, where
l = {a2(2b2 + 2c2 a2)}/{(a + b + c)(b + c a)(c + a b)(a + b c)}. (3.2)
2 2 2 2
Similar calculations show that (aC aA) = lb and (aA aB) = lc . It follows that triangle aA aB
aC is similar to triangle ABC.
In similar manner we find (bB bC)2 = ma2, (bC bA)2 = mb2 and (bA bB)2 = mc2, where
m = {b2(2c2 + 2a2 b2)}/{(a + b + c)(b + c a)(c + a b)(a + b c)}, (3.3)
showing that triangle bA bB bC is also similar to triangle ABC.
And again in similar manner we find (cB cC)2 = na2, (bC bA)2 = nb2 and (bA bB)2 = nc2, where
n = {c2(2a2 + 2b2 c2)}/{(a + b + c)(b + c a)(c + a b)(a + b c)}, (3.4)

6
showing that triangle bA bB bC is also similar to triangle ABC.

All three triangles are directly similar to one another and inversely similar to triangle ABC. The
areas of the three triangles are in the ratio l : m : n. The six triangles are shown in Fig.1(b).

References

1. K. Hagge, Zeitschrift fr Math. Unterricht, 38 (1907) 257-269.


2. C.J. Bradley, Omega Circles (Article 5 in this series).
3. F.E. Wood, Amer. Math. Monthly 36:2 (1929) 67-73.
4. C.J.Bradley & G.C.Smith, The locations of Triangle Centres, Forum Geom., 6(2006) 57-
70.
5. R.Shail, Some Properties of Brocard Points, Math. Gaz., 80 (1996) 485-491.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

7
ARTICLE 5

Omega Circles

Christopher J Bradley

S

D
O
U
X
T
Q Y
W P
R V
B C
Z

Figure

1. Introduction

Every circle passing through H, the orthocentre of a triangle ABC is called a Hagge circle,
named after the first person [1] to detail their most important properties. These are (i) If AH, BH,
CH meet a Hagge circle at points X, Y, Z respectively, then triangle XYZ is indirectly similar
1
to triangle ABC, (ii) If circles BHC, CHA, AHB meet at U, V, W respectively, then UX, VY,
WZ meet at a point P and (iii) P is also the centre of inverse similarity, meaning there is an axis
through P such that if ABC is reflected in this axis followed by an appropriate enlargement
(reduction) centre P, it is mapped on to triangle XYZ. It may be added that if triangle UVW is
mapped back (by the inverse transformation) from to the circumcircle to form a triangle DEF
then DEF is inversely similar to UVW and because of property (ii) APD, BPE, CPF are straight
lines.

If you take a general point J rather than H and repeat the construction then the triangle XYZ is no
longer similar to ABC. That might have been the end of the story, but it now appears that circles
through the Brocard points and ' have properties nearly the same as Hagge circles, but with
subtle differences. To be particular we restrict our discussion for the time being to Omega
circles, those passing through the Brocard point , which if AB > AC is further from BC than
the Brocard point ' and which has areal co-ordinates (1/b2, 1/c2, 1/a2). The properties of an
Omega circle are (i) If A, B, C meet again at Z, X, Y respectively, then triangle XYZ is
indirectly similar to triangle ABC, (ii) If circles BC, CA, AB meet at U, V, W then UZ,
VX, WY meet at a point P. However, the centre of inverse similarity R is in general distinct from
P, so that when the points D, E, F are obtained by finding the inverse images of U, V, W
respectively, the lines AE, BF, CD meet at a point T which is the inverse image of P under the
indirect similarity centre R. The exception is the seven-point circle, when R and P coincide.
These properties are all exhibited in the figure above, which is drawn using CABRI II plus. There
is good reason to believe that H, , ' are the only three points that exhibit the indirect
similarities, but there are three other points which provide direct similarities rather than indirect
similarities between ABC and XYZ. There are numerous other properties that emerge from both
Hagge circles and Omega circles such as the fact that pairs of triangles are orthologic and that
the perspectives also create axes of importance. In this article, we do not attempt to catalogue all
these properties. The interested reader is referred to the work of Speckman [2] and to previous
articles in this series.

2. The equation of an Omega circle

We use areal co-ordinates with A, B, C the triangle of reference. The point has co-ordinates
(1/b2, 1/c2, 1/a2). From the equations of A, B, C we may write down possible co-ordinates
of Z, X, Y which are Z(1, l/c2, l/a2), X(m/b2, 1, m/a2), Y(n/b2, n/c2, 1). However, l, m, n are
related to one another because the circle XYZ passes through . The condition for this is
l = a2b2c2(a2(b2c2 + n(c2 m)) + c2m(b2 + n)) all divided by
a4b2(b2c2+ n(c2 m)) + a2c2b4(m n) b2m(c2 + n) c2mn b4c2mn. (2.1)

What Equation (2.1) implies is that there are a doubly infinite number of Omega circles given by
varying the values of m and n. By eliminating l in favour of m and n means that symmetry is lost.

2
Despite that further working becomes easier to check using a programme such as DERIVE,
which is the one we use. The equation of the Omega circle, in terms of m and n, is
b4x2(a2(b2c2 + n(c2 m)) + c2m(b2 + n)) + b2x(y(a2(b4c2 b2(c4 + mn) c4n) + c2m(b4 b2c2
c2n)) + z(a4n(b2 + m) a2(b4n + b2m(c2 + n) + c2mn) b4mn)) + a2(c2my2(b2(c2 + n) + c2n)
c2yz(a2(b4 + b2(m + n) + mn) + m(b4 b2c2 c2n)) + b2nz2(a2(b2 + m) + b2m)) = 0. (2.2)

3. The indirect similarity

The fact that triangles ABC and XYZ are indirectly similar follows by angle chasing as a result
of the facts that angle BC = 1800 C, angle CA = 1800 A, and angle AB = 180o B. (If
is replaced by H then C, A, B are replaced by A, B, C in these equations, involving a cyclic
rearrangement of A, B, C.)

4. The circles BC, CA, AB

The circle BC has equation


b2x2 xy(c2 b2) + a2yz = 0. (4.1)

The equations of circles CA, AB may be written down by cyclic change of x, y, z and a, b, c.

5. The points U, V, W and P

If a triangle ABC is given and a point J and any circle through J, it is a general result that if AJ,
BJ, CJ meet again in points Z, X, Y and circles BJC, CJA, AJB meet again in points U, V, W
then XV, YW, ZU are concurrent at a point P, so we do not repeat the proof, but simply give the
co-ordinates of the points in the present context. (The result can then be checked in this case.)
Note that the perspective is quite independent of the similarity in this case between ABC and
XYZ.

U has co-ordinates (x, y, z) where

x = a2m(b2c2 + n(b2 + c2)),


y = b2n(a2b2 + m(a2 + b2)), (5.1)
2 2 2 2 2 2 2 2 2 2 2 2
z = (b m(a (b n + c (m n) + 2mn) +mn(b c ))(b (c + n) + c n))/(n(a (b + m) + b m)).

V has co-ordinates (x, y, z) where

x = (n(a2(b2 + m) + b2m)(a4(b2c2 + n(c2 m)) a2c2(b2(c2 m + n) + n(c2 m)) c2m(b2(c2 +


n) + c2n)))/(b2(a2(b2c2 + n(c2 m)) + c2m(b2 + n))),
y = n(a2(b2 + m) + b2m), (5.2)

3
z = a2(b2c2 + n(c2 m)) + c2m(b2 + n).

W has co-ordinates (x, y, z) where

x = a2m(b2(c2 + n) + c2n),
y = (b2(a2c2(b4 + b2(m + n) +mn) b4mn)(a2(b2c2 + n(c2 m)) +c2m(b2 + n)))/(c2m(b2(c2+ n) +
c2n)), (5.3)
2 2 2 2 2 2 2
z = b (a (b c + n(c m)) + c m(b + n)).

P has co-ordinates (x, y, z) where

x = (c2mn(a2(b2 + m) + b2m))(b2(c2 + n) + c2n),


y = (b2n(a2(b2 + m) + b2m))(a2(b2c2 + n(c2 m)) + c2m(b2 + n)), (5.4)
z = (b2(a2(b2c2 + n(c2 m)) + c2m(b2 + n)))(m(b2(c2 + n) + c2n)).

It may be proved that VW^A, WU^B, UV^C, P are collinear, but that we leave to the
reader. As with all indirect similarities the midpoints of AX, BY, CZ, DU, EV, FW lie on a
conic.

6. The points R, T, S

The points R, T, S are the additional points in the figure.

In view of the indirect similarity between ABC and XYZ there are bound to be a point R and two
axes through R that are the centre and axes of inverse similarity. Nothing additional is proved by
working out the co-ordinates of R, as it is known to exist. Its position is constructed as follows.
The directions of the axes are parallel to the bisectors of any pair of corresponding sides such as
BC and YZ. Take any such line and reflect triangle ABC in it. It will now be found that AX, BY,
CZ are concurrent at a point that does not lie on the line. Move the line parallel to itself and
when the point lies on the line that will be R, the centre of inverse similarity. This construction,
of course, cannot be done with complete precision. But the construction can be made precise by
noting that R must lie on certain lines. For example the point lies on the circle XYZ. Its
angular position relative to these points can be mirrored precisely on the circumcircle of ABC.
Call that point o. Call the same Brocard point of triangle XYZ the point 1. The passage from
o to 1 is now the product of two indirect similarities and is thus an enlargement, so that o1
must pass through R. A second such point, leading to a second line through R, might be the
second point of intersection of the circle XYZ with the seven-point circle. The two lines fix the
position of R exactly.

4
The point T is the inverse image of the in direct similarity of the point T and if DEF is the
inverse image of triangle UVW then AE, BF, CD are concurrent at T.

Since circles BUC, CVA, AWB are concurrent at , it follows that circle AVW, BWU, CUV are
concurrent at a point S. As indicated by CABRI II plus, S lies on the circumcircle of ABC. This
curious fact we have been unable to prove, as DERIVE found the equations too technically
complicated to solve.

7. Omega prime circles

Y S
E
Q D
R
X W
P
O

U T Z '

B C

Figure 2

5
Circles through ' (1/c2, 1/a2, 1/b2) have very similar properties to those passing through , the
difference arising from the labelling of the points X, Y, Z. Now A', B', C' meet the Omega
prime circle at points Y, Z, X respectively. Then there is an indirect similarity between triangles
ABC and XYZ. This similarity follows again by simple angle chasing bearing in mind that angle
B'C = 180o B, angle C'A = 180o C and angle A'B = 180o A.

8. aH bH cH circles

A'
J X

V E
F' bH
O

Q G E'

B'
F H
W
S D' Z
B C

Y
D C'
U
K
T

P
F IGURE 3

aH, bH cH are the intersections of the lines AG, BG, CG with the orthocentroidal circle on GH
as diameter, where G is the centroid of ABC and H is its orthocentre. We illustrate in Figure 3
what happens when a circle is drawn through bH. The lines AbH, BbH, CbH meet again at

6
Z, Y, X respectively. Since angle BbHC = 180o C, angle CbHA = 180o A, angle AbHB =
180o B it follows by simple angle chasing that triangle XYZ is directly similar to triangle ABC.
The intersections of with the circumcircle are denoted by J and K. According to the theory of
Wood [3] either J or K may be used as a perspector to project triangle XYZ into a triangle
A'B'C', where A' is the intersection of JX with the circumcircle and B' and C' are similarly
defined. Triangle A'B'C' is directly similar to XYZ and so directly congruent to ABC. It may be
seen in the figure that they are related by a rotation about O. If is not large enough to intersect
the circumcircle it may be enlarged about its centre Q until it does. All that happens is that an
extra step is required to illustrate the direct similarity. Once this has been done the rest follows as
with Omega circles. Points U, V, W are defined as intersections with of the circles BbHC,
CbHA, AbHC respectively and UZ, VY, WX intersect at P. Points D, E, F are the inverse images
of U, V, W and DC, EB, FA intersect at T. The point S is the intersection of circles AYZ, BZX,
CXY and it also lies on the circumcircle.

References
1. K. Hagge, Zeitschrift fr Math. Unterricht, 38 (1907) 257-269.
2. H.A.W. Speckman, Perspectief Gelegen, Nieuw Archief, (2) 6 (1905) 179 188.
3. F.E. Wood, Amer. Math. Monthly 36:2 (1929) 67-73.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

7
Article 6

On The Ten Point Wood-Desargues Configuration


Christopher J Bradley and Geoff C Smith

Abstract

The ten pairs of directly similar triangles in perspective in the Wood-Desargues


configuration have ten common Hagge circle centres. These centres are known to lie one
each of the ten perspectrices. It is shown in this paper that these centres also lie at the
vertices of five cyclic quadrangles, which are similar to the five cyclic quadrangles of the
initial Wood-Desargues configuration and that these five new cyclic quadrangles have a
common circumradius.

1. Introduction

The Wood-Desargues configuration is built up as follows: Start with two circles


intersecting at points J and K. In the first circle inscribe a triangle ABC. Draw AK, BK,
CK to meet the second circle at points a, b, c. It follows that triangles ABC and abc are
directly similar triangles in perspective with J the centre of direct similarity. By this we
mean that triangle abc may be obtained by from triangle ABC by a rotation about J
followed by a dilation centre J. Suppose that BC meets bc at the point 1, CA meets ca at
the point 2 and AB meets ab at the point 3. Then 123 is the Desargues axis of
perspective, commonly referred to nowadays as the perspectrix. There are in all ten
points A, B, C, K, a, b, c, 1, 2, 3 which form what may be called the Wood-Desargues
configuration. The ten points turn out to be equivalent in the following sense. Each of
them is a vertex of perspective of two triangles in the configuration, the remaining three
points being collinear and serving as the perspectrix. This situation is illustrated in Figure
1. In his landmark paper Wood [1] proved that the quadrangles Aa23, Bb31, Cc12 are
cyclic in addition to the starting quadrangles ABCK and abcK. These circles are shown in
Figure 1 and for reasons, which will emerge shortly, their centres are also labelled ABCK,
abcK etc. In his paper Wood showed that the five circle centres are themselves concyclic
and lie on a circle passing through J. The ten pairs of triangles in perspective, together
with their vertices of perspective and their perspectrices are shown in Table 1. The orders
of the vertices in pairs of triangles correspond and the order of letters in the perspectrix is
standard, so, for example 1, is the intersection of BC and bc, BC and 32 and bc and 32.

Triangle 1 Triangle 2 Vertex Perspectrix


ABC abc K 123
KBC a32 A 1cb
AKC 3b1 B c2a
ABK 21c C ba3
Cc2 Bb3 1 aAK
1
Aa3 Cc1 2 bBK
Bb1 Aa2 3 cCK
Kbc A32 a 1CB
Kca B13 b 2AC
Kab C21 c 3BA

Table 1: The ten perspectives

In Section 2 we consider the orthocentres of the twenty triangles and show that they lie
four at a time on five quadrangles similar to the initial five quadrangles and also that the
same twenty orthocentres lie four at a time on five lines. This follows from well-known
results about cyclic quadrangles and quadrilaterals, but is presented, as the result is
required in what follows.

In Section 3 we establish our main results. Each pair of triangles in the list of ten
perspective pairs have orthocentres, say H and F, so that the circle JHF is a common

2
Hagge circle of the two triangles. Its centre we denote by h, and it is the intersection of
the perpendicular bisectors of JH and JF. It is known that these ten points h lie one each
on the ten perspectrices, see [1]. We show that they also form the vertices of five cyclic
quadrangles similar to the initial five cyclic quadrangles, and that each of the five circles
involved has the same circumradius. The method of proof involves the pentagon formed
by the five circle centres of the initial five circles. The result follows from lemmas that
are proved and from straightforward extensions of the work of Wood [1].

2. The twenty orthocentres

In Figure 2 we show the circle ABCK and its four component triangles ABC, KBC, AKC,
ABK with orthocentres H(K), H(A), H(B), H(C) respectively. In the same figure we also
show the four lines 123, b1c, ac2, 3ba, which are the four perspectrices of the four
perspectives associated with the triangles of the cyclic quadrangle ABCK. The four
orthocentres of the triangles formed by these four lines, namely triangles abc, a32, 3b1
and 21c (see Table 1) are denoted by F(K), F(A), F(B), F(C) respectively.

Proposition 1
The quadrangle H(A)H(B)H(K)H(C) is congruent to the quadrangle ABKC.

Proof
The result is true for any cyclic quadrangle and does not depend upon A, B, C, K being
part of the Wood-Desargues configuration. Let the centre of the circle ABCK be the
origin of vectors and suppose A, B, C, K have vector positions a, b, c, k respectively.
Then H(K) has vector position a + b + c and H(A) has vector position b + c + k. It follows
that H(A)H(K) = a k = KA. It is evident that H(A)H(B)H(K)H(C) is the image of ABKC
under a 180o rotation about the point whose vector position is (a + b + c + k).

Proposition 2
The points F(A), F(B), F(C), F(K) are collinear.

Proof
The result is true for the orthocentres of the four triangles formed by any set of mutually
non-parallel lines and does not depend on the lines 123, b1c, ac2, 3ba being part of the
Wood-Desargues configuration. The result is immediate from the work of Steiner on the
complete quadrilateral. See, for example, Durrell [2].

Now there are five circles and associated quadrangles in the initial Wood-Desargues
configuration, so there are similar results to Propositions 1 and 2 for each of the other
four. This means that each of the twenty orthocentres of the twenty triangles of the ten
perspectives appears twice. Each will appear once as an H point of one of the five
quadrangles and once as an F point of one of the five lines. The five resulting cyclic
quadrangles are congruent in each case to one of the initial five quadrangles.

3
c

1
F(C)
J

A F(B)
h(A)
h(B)
F(K) b
C

H(B) 2
h(K)
H(K)

B h(C)

H(C)
a

H(A)

3
F(A)

Figure 2

The perpendicular bisectors of JH(K) and JF(K) meet at a point we denote by h(K). This
is the joint Hagge centre of triangles ABC and abc. It is known to lie on the perspectrix
123. The points h(A) and h(B) and h(C) are similarly defined for the pairs of triangles
KBC, a32 and AKC, 3b1 and ABK, 21c respectively and lie on the lines b1c, ac2, 3ba
respectively. In Figure 2 we also show these points and the quadrangle h(A)h(B)h(K)h(C).
There are five such quadrangles and each of the h points appears on two of them. For
example h(K) lies on the quadrangle h(a)h(b)h(c)h(K) as well. In Section 3 we show that
these five quadrangles are directly similar to the five initial quadrangles in the Wood-
Desargues configuration.

4
3. The ten common Hagge centres

First we establish a very simple Lemma that at first sight does not appear to be related the
Wood-Desargues configuration at all, but is in fact a useful step in establishing two of
the theorems we want to prove. The Lemma is illustrated in Figure 3.

Lemma 1
Suppose that P, Q, R are three non-collinear points and lines p, q, r through P, Q, R
respectively meet at a point S on circle PQR. Then lines p , q , r through P, Q, R
respectively, perpendicular to p, q, r respectively, are concurrent at a point T
diametrically opposite to S.
Proof
Let the lines p and q meet at T. Then, since SPT = SQT = 90o, T lies at the other
end of the diameter to S of circle SPQ. But R lies on this circle, and r and r are at right
angles. It follows that r passes through T.

It is not needed here, but it is easy to verify that the condition that S lies on circle PQR is
necessary as well sufficient for the concurrence of p , q , r .

5
We now establish the crucial Lemma in helping to establish what we wish to prove. One
possible configuration is shown in Figure 4, but the proof given using directed angles is
diagram independent.

Lemma 2
Let J, O, L be three points on the circumference of a circle S1 and let S2 and S3 be the
circles through J, centres L and O respectively. Suppose that S2 and S3 meet again at A, S3
and S1 meet again at Z and S1 and S2 meet again at W, then A, L and Z are collinear as are
A, O and W.

Proof
Figure 5 is a copy of Figure 1, but with some additional points and lines added. First,
however, note that the centres of the five circles have been relabelled, so that circle
ABCK has centre U, circle abcK has centre V, circle Aa23 has centre L, circle Bb13 has
centre M and circle cC12 has centre N. Consider the point L, the centre of circle Aa23.

6
This is clearly the point of intersection of the perpendicular bisectors of JA, Ja, J2, J3.
Now A, a; A, 2; A, 3; a, 2; a, 3, 2, 3 are six corresponding pairs of vertices amongst the
thirty such pairs in the ten perspectives in Table 1. The same applies to each of the five
circle centres and this fact accounts for all thirty such points of intersection.

Additional points in the diagram are X, Y, Z, W. These points occur in Propositions we


establish next. Lines ALZ, AUW, BMZ, CNZ are also shown, and these collinearities will
soon be established.

Proposition 3
Let Z be the point of intersection of circles JUVLMN and JABCK, and let W be the point
of intersection of circles JUVLMN and JAa23, then ALZ, AUW, BMZ, CNZ are straight
lines.

Proof
In any ten point Wood-Desargues configuration, the conditions of Lemma 2 apply,
which accounts for the straight lines ALZ and AUW. Application of the same Lemma to
other pairs of circles shows that BMZ and CNZ are also straight lines.

7
Proposition 4
Triangles ABC and LMN are directly similar triangles in perspective with centre of direct
similarity J and vertex of perspective Z.

Proof
This follows immediately as a result of Proposition 3.

Since circle JUVLMN cuts the five initial circles in five different points, there are thus
five such vertices of perspective and all twenty triangles of the initial Wood-Desargues
configuration get mapped by perspectives on to the triangles of the pentagon UVLMN.
We have thus established

Proposition 5
Consider the pentagon UVLMN. Omit one vertex, say M. Then the quadrangle UVLN is
directly similar to the quadrangle Bb31 (a little care has to be taken over the order of the
vertices). The same applies when any other vertex of the pentagon is omitted.

In this way the pentagon inscribed in the circle UVLMN carries all the angular
information of the five quadrangles of the initial configuration.

If we now use Lemma 1 with X the point diametrically opposite Z on the circle JABCK
we recover Theorem 12 of Wood [1], namely

Proposition 6
Tangents at A, B, C to circles Aa23, Bb13, Cc12 are concurrent at a point X on circle
JABCK.

Proof
These tangents are perpendicular to the normals AL, BM, CN so Lemma 1 applies.

There are, of course other cases in which sets of tangents are concurrent.

Lemma 1 also provides proof of the following proposition.

Proposition 7
Lines through L, M, N parallel to the tangents at A, B, C in Proposition 6 are concurrent at
a point Y on circle JUVLMN, which is diametrically opposite Z in that circle.

This allows us to identify the centre of the circle JUVLMN. It is the midpoint of YZ.

Recall now that h is the intersection of the perpendicular bisectors of JH and JF and this
is how all ten common Hagge centres are constructed. We have shown above that if we
take the perpendicular bisectors of JA and Ja we obtain L, if we take the perpendicular
bisectors of JB and Jb we get M and if we take the perpendicular bisectors of JC and Jc
8
we get N. But we have shown in Proposition 5 that triangles ABC and LMN are in
perspective and similar with vertex. It follows that h(K) is the orthocentre of triangle
LMN. Similar considerations hold for all the other common Hagge centres, which
together account for all ten orthocentres of the triangle formed by the pentagon UVLMN
and its diagonals.

Proposition 8
The quadrangle h(A)h(B)h(C)h(K) is similar to the quadrangle ABCK.

Proof
We know from Proposition 1 that the quadrangles ABCK and H(A)H(B)H(C)H(K) are
congruent. For the same reason quadrangles LMNV and h(A)h(B)h(C)h(K) are congruent.
But from Proposition 5 the quadrangles ABCK and LMNV are similar.

There are five such cyclic quadrangles formed by the common Hagge centres and since
the orthocentres of the triangles formed by UVLMN and its diagonals all derive from the
same circle we have proved

Proposition 9

The ten common Hagge centres of the Wood-Desargues configuration are inscribed in
five circles of equal radius and each Hagge centre lies on two such cyclic quadrangles.

The configuration is illustrated in Figure 6, in which the initial configuration and its five
circles and quadrangles are shown, together with the perspectrices and the common
Hagge centres lying on the perspectrices and forming five cyclic quadrangles of equal
radius.

9
References
1. F.E. Wood, Amer. Math. Monthly 36:2 (1929) 67-73.
2. C.V. Durell, Modern Geometry, Macmillan (1946).

10
ARTICLE 7
Generalization of the Wallace-Simson line
Christopher J Bradley

1. Introduction

If ABC is a triangle and J is a point on its circumcircle , then the Wallace-Simson line property
is that the feet of the perpendiculars L, M, N from J on to the sides of ABC are collinear and LMN
is called the Wallace-Simson line. Furthermore, if H is the orthocentre of ABC then the midpoint
Q of JH lies on LMN. If instead of taking the feet of the perpendiculars on to the sides, we take
the reflections L', M', N' of J in the sides, then L'M'N' is also a straight line which we refer to as
the Double Wallace-Simson line of J. As this line is a degenerate Hagge circle it is known that
L'M'N' passes through the orthocentre H.

In this chapter we provide a generalization of this construction, so that the Wallace-Simson line
is just a particular case, being one of an infinite number of lines that we refer to as generalized
Wallace-Simson lines. Later we identify these lines as ones that are already known, providing an
explanation of why these lines are the same as those we describe.

Figure 1 illustrates the general construction. The point Q is now varied from its fixed position at
the midpoint of JH from which the well known Wallace-Simson line results and becomes a
variable point on the perpendicular bisector of JH. The circle S is drawn, centre Q, passing
through J and H. The points where the altitudes of ABC meet S are denoted by X, Y, Z, where AX,
BY, CZ are perpendicular to BC, CA, AB respectively. Circles AJX, BJY, CJZ are then drawn and
L is the point of intersection of BJY and CJZ with M, N similarly defined. As we prove in due
course, the following results hold:

1. L, M, N lie on BC, CA, AB respectively;


2. L, M, N are collinear.
3. The line LMN contains the point Q.

The line LMN is, of course, what we mean by a generalized Wallace-Simson line and as Q varies
we get an infinite number of such lines all defined with respect to the one fixed point J on the
circumcircle of ABC.

Various subsidiary results emerge in the course of establishing these propositions, and we now
briefly indicate the plan we adopt to prove them.

1
M
Z
X

K
Ao

A
J

Q
Co

N
Bo

Y
L C
B

Fig. 1

The method we use is to take J to be the centre of the direct similarity that maps H to Q. We
then apply this direct similarity to ABC to get a triangle AoBoCo and it transpires that o, the

2
circumcircle of AoBoCo, shares the point J with and a second point K such that AAo, BBo,
CCo concur at K. This is typical of directly similar triangles in perspective, see Wood [1].
When Q is the midpoint of JH triangle AoBoCo has its sides parallel to ABC and is half the
size. In all cases it turns out that the line LMN we want is the Double Wallace- Simson line of
J with respect to triangle AoBoCo and, being a Double Wallace-Simson, line it automatically
passes through the orthocentre of AoBoCo, which by construction is Q. Synthetic
considerations may be used to show that L, M, N lie on BC, CA, AB respectively and indeed
that LMN is a straight line. The bonus of the analysis is that it highlights the role played by
the direct similarity.

2. Defining the direct similarity

We use Cartesian co-ordinates and take J to be the origin (0, 0) and the centre O of triangle
ABC to the point (1, 0). Throughout the calculation the computer algebra package DERIVE
was used. If, for the moment, we suppose H has co-ordinates (h, k), then it is straightforward
to show that a variable point Q on the perpendicular bisector of JH has co-ordinates (h kt,
k + ht), where t is a variable parameter.

The direct similarity through J that maps H to Q is therefore given by the matrix
(2.1)

If considered as a dilation of magnitude d with centre of enlargement J and a rotation by an


angle about J we see that d and are related by the equation d = 1/(2 cos) and also that t
= tan . (In all that follows we suppose without loss of generality that cos is positive and
that < < .)

The equation of the circumcircle is


x2 + y2 2x = 0. (2.2)
We suppose that the point A has co-ordinates ( , where a is a parameter, and that
B, C have similar co-ordinates with b, c respectively replacing a. Applying the direct
similarity to A we obtain a point Ao with co-ordinates . The points Bo, Co, the
images of B, C respectively under the direct similarity have similar co-ordinates with b, c
respectively replacing a.

We now prove that triangle AoBoCo is in perspective with triangle ABC. The equation of AAo
is
(a 2t)x (1 + 2at)y + 4t = 0. (2.3)

3
This meets the circumcircle with Equation (2.2) at A and at the point K with co-ordinates
( ). Since the co-ordinates of K are independent of a, b, c it follows that AAo, BBo,
CCo are concurrent at K and the triangles are not only similar, but in perspective. From the
theory of Wood [1] on directly similar triangles in perspective it follows that the circle o, the
circumcircle of triangle AoBoCo intersects at the points J and K. See Fig. 1.

We now consider the circle, centre Ao passing through J. Its equation is soon found to be
(1 + a2)x2 + (1 + a2)y2 2(1 2at)x 2(a + 2t)y = 0. (2.4)
It may now be checked that this circle passes through A. So this circle is now identified as
circle JAX, though we have yet to find the co-ordinates of X and to show it lies on the Hagge
circle, centre Q passing through H and J. If the parameter a in Equation (2.4) is replaced by
b, c we have the equations of the circles that may be identified as circles JBY, JCZ
respectively. Note that in view of the direct similarity triangles JAoA, JBoB, JCoC are similar.

The co-ordinates of H may be obtained from those of O, A, B, C and are

(2.5)
From this we may obtain the equation of the altitude AH, which is
(1 + a2)(b + c)x (1 + a2)(1 bc)y + 2(a + b + c abc) = 0. (2.6)

This meets the circle centre Ao through J at the point A and the point X with co-ordinates
). (4.2.7)

The co-ordinates of Y and Z may be found from those of X by cyclic change of the symbols a,
b, c. The equation of the circle through X, Y, Z may now be calculated and is
(1 + a2)(1 + b2)(1 + c2)(x2 + y2) + 2(a2b2c2 + 2abct(bc + ca + ab) + 2t(a2b + a2c + b2c +
b2a + c2a + c2b) + 2t(a + b + c) a2 b2 c2 2)x + 2(2a2b2c2t abc(bc + ca + ab) 2t(a2 +
b2 + c2) (a2b + a2c + b2c + b2a + c2a + c2b) (a + b + c + 4t))y = 0. (2.8)

This circle clearly passes through J and it may be verified by substitution that it also passes
through H. Equation (2.8) therefore represents the Hagge circle, centre Q and passing through
J and H. Fig. 1 has now been completely analysed and from here the results stated in the
introduction follow by synthetic arguments and other known results.

3. Results

4
The angle arguments given below are appropriate to the figure at the start of the article and as
such are diagram dependent. However, they are easily modified for other possible figures, or
may be written in a diagram independent way by working mod .

Theorem 4.1
Angle JLB = Angle JYB = Angle JYH = 180o Angle JZH = 180o Angle JZC = Angle JLC.
This shows that L lies on BC. Similarly, M and N lie on CA and AB respectively.

Theorem 4.2
Now angle JNL = Angle JBL = = 180o Angle JBC = Angle JAC = Angle JAM = 180o
Angle JNM and so LMN is a straight line.

Theorem 4.3
Circles JBY and JCZ have centres Bo and Co respectively. The common chord JL is
perpendicular to the line of centres BoCo and bisected by it. L is therefore the reflection of J in
the side BoCo of triangle AoBoCo. The same holds for M and N with respect to the sides CoAo
and AoBo respectively. LMN is therefore the Double Wallace-Simson line of J with respect to
triangle AoBoCo. But, by construction, Q is the orthocentre of this triangle and therefore lies on
LMN.

Since LMN is drawn making right angles with the sides of a triangle which is similar and at a
fixed angle to the sides of ABC, it is clear that these lines make equal angles with the sides of
ABC. It is interesting to observe that these generalized Wallace-Simson lines (sometimes
called oblique Wallace-Simson lines) exist only because Double Wallace-Simson lines exist
for a smaller similar triangle. The original proof seems to have been by Carnot, the best
reference I could find being in Exercises by Frre Gabriel-Marie [].

References
1. F.E. Wood, Amer. Math. Monthly 36:2 (1929) 67-73.
2. Frre Gabriel-Marie (listed as simply F. G.-M.) Exercices de gomtrie, comprenant
l'expos des mthodes gomtriques et 2000 questions rsolues, 6me dition. Paris: J.
Gabay, 1991. 1302.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

5
ARTICLE 8

A Singular Miquel Configuration and the Miquel Direct Similarity

Christopher J Bradley

P
D

S R

E
F
B
Q C
U

Fig. 1

1
1. Introduction

A Miquel configuration is one in which points P, Q, R are chosen on the sides AB, BC, CA
respectively (but not at the vertices A, B, C) and the circles ARP, BPQ and CQR are drawn,
illustrating that these three circles always pass through a common point S. The configuration we
have called a singular Miquel configuration is that once P has been chosen arbitrarily (but not at
B or C), then Q and R are chosen so that Q lies on circle CAP and R lies on the circle BCP. We
prove in this paper that in these circumstances ASQ and BSR are straight lines and that the
centres U, V, D, E, F of the five circles BCP, CAP, ARP, BPQ, CQR respectively do themselves
lie on a circle passing through P. The Figure shows a case in which P lies between A and B, but
the results still holds when P is external and the algebra we present is independent of the position
of P on AB. We also show that triangle DEF is directly similar to ABC and as this is true in any
Miquel configuration we give a proof when P, Q and R are arbitrary. In the analysis of the
singular Miquel configuration we use areal co-ordinates in which ABC is the reference triangle.
For the study of the direct similarity we use Cartesian co-ordinates.

2. Finding Q and R from P

Suppose P is chosen to have co-ordinates (n, (1 n), 0), n 0, 1. The general equation of a circle
in areal co-ordinates is
(2.1)
Here u, v, w are determined by substituting the co-ordinates of three points lying on the circle.
Using this method repeatedly we find the equations of CAP and BCP are respectively are
(2.2)
and
(2.3)

The co-ordinates of Q are therefore (0, a2 c2n, c2n) and the co-ordinates of R are (b2 + c2(n 1),
0, c2(1 n)). The point S we define to be AQ^BR and has co-ordinates (n(b2 + c2(n 1)),
(1 n)(a2 c2n), c2n(1 n)).

3. The three Miquel circles

The circles ARP. BPQ have equations

(3.1)
(3.2)
and it may be checked that both these circles pass through S.

Finally, the circle CQR has equation

2
(3.3)
and, of course, it too passes through S.

4. The centres of the 5 circles

If a circle is expressed in the form of equation (2.1), then, see Bradley [1], its centre has co-
ordinates x, y, z), where

(4.1)
.

Using Equation (4.1) and values of u, v, w from the working in Section 2 we find the co-
ordinates of the centre of circle CAP are

(4.2)

and the co-ordinates of the centre of circle BCP are

(4.3)

The co-ordinates of the centre of circle ARP are

(4.4)

The co-ordinates of the centre of circle BPQ are

(4.5)

Finally the co-ordinates of the centre of the circle CQR are

3
(4.6)

The last three of these points was used to compute the circumcircle of triangle DEF and it was
then checked that it does actually contain U, V and P. The circumcircle has a cumbersome
equation which is

(4.7)

4
A

A'
R
D

F C'
S

E
B Q C

B'

Figure 2

5. The direct similarity when P, Q, R are arbitrary

For the more general Miquel configuration, when P, Q, R are chosen as arbitrary points on AB,
BC, CA respectively, it is easier to use Cartesian co-ordinates. An economically chosen set of
parameters involve co-ordinates (0, 0) for A, (2, 2v) for B and (2, 2w) for C, where without loss
of generality we take w > v. Points P, Q, R on AB, BC, CA respectively are now assigned co-
ordinates (k, kv), (2, 2u), (h, hw) respectively, where u v, w; h, k 0,2. See Figure 2.

The equations of the circles BPQ, CQR, ARP may be obtained using standard methods involving
4 x 4 determinants and are respectively
x2 + y2 + (2(uv 1) k(v2 + 1))x 2(u + v)y + 2k(v2 + 1) = 0, (5.1)
2 2 2 2
x + y + (2(uw 1) h(w + 1))x 2(u + w)y + 2h(w + 1) = 0, (5.2)
2 2 2 2 2 2
(v w)(x + y ) + (kw(v + 1) hv(w + 1))x + (h(w + 1) k(v + 1))y = 0. (5.3)

5
It may be checked that these three circles have a common point S. We do not record its co-
ordinates as they are complicated and are not needed in what follows. The co-ordinates of their
centres can now be determined and are
D (centre of ARP): {1/(2(v w))}( hv(w2 + 1) kw(v2 + 1), k(v2 + 1) h(w2 + 1)),
E (centre of BPQ): ((k(v2 + 1) 2(uv 1)), u + v),
F (centre of CQR) : ((hw2 + 1) 2(uw 1)), u + w).
Calculations may now be carried out to show that DE/EF = AB/BC = {(v2 + 1)}/(w v), and
DF/DE = AC/AB = {(w2 + 1)/(v2 + 1)} and this establishes the similarity between triangle XYZ
and ABC. It may be shown that triangle ABC may be moved by appropriate rotation to A'B'C'
and enlargement (reduction) about S as centre of direct similarity to coincide with triangle DEF.

Flat 4,
Terrill Court,
12-14 Apsley Road
BRISTOL BS8 2SP

6
Article 9

Miquel circles and Cevian lines

Christopher J Bradley

Q
W
U
N
P
V

C
B L

Figure 1

1
1. Introduction

Given a triangle ABC it is well-known that if you choose L, M, N on the sides BC, CA, AB
respectively (but not at the vertices), then circles AMN, BNL, CLM have a common point Q
commonly known as the Miquel point [1, 2]. An interesting question is whether there exist any
further properties when L, M, N are the feet of Cevian lines through a point P. We answer the
question in the affirmative and establish the theorem that both P and Q lie on a circle UVW,
where U is the intersection of AP with circle AMN, V is the intersection of BP with circle BNL
and W is the intersection of CP with circle CLM. See Figure 1. We also have a second result
which also only holds when L, M, N are the feet of Cevians and this is that circles AQL, BQM,
CQN have a common point R. See Figure 2. The proofs given below involve areal co-ordinates.
The definition of areal co-ordinates and how to use them may be found in Bradley [3].

M
N
R
Q
P

B C
L

Figure 2

2
2. The Miquel circles

Let P have co-ordinates (l, m, n) so that the co-ordinates of L, M, N are respectively L(0, m, n),
M(l, 0, n), N(l, m, 0). The equation of any circle is
(2.1)

For circle AMN, we insert the co-ordinates of A, M, N in Equation (2.1) and find three equations
for u, v, w, which give We now substitute back in Equation (2.1)
to get the equation of AMN, which is

(2.2)
The equations of BMN, CNL may now be written down by cyclic change of x, y, z and a, b, c
and l, m, n.

3. The Miquel Point Q

It may now be checked that these three circles have a common point Q with co-ordinates (x, y,
z), where

(3.1)

4. The points U, V, W

The point U is the intersection of AP and circle AMN. AP has equation ny = mz and AMN has
Equation (2.2). It follows that U has co-ordinates (x, y, z), where

(4.1)

The co-ordinates of V, W now follow by cyclic change of x, y, z and a, b, c and l, m, n.

5. The circle UVWPQ

The co-ordinates of U, V, W are now put into an equation of the form (2.1). The result is three
equations to determine u, v, w. These are now substituted back and the result is the equation of
circle UVW, which is

3
(5.1)
It may now be verified that both P and Q lie on this circle.

6. The circles AQL, BQM, CQN

Starting from Equation (2.1) we may insert the co-ordinates of A, Q, and L to obtain three
equations for u, v, w. Substituting these values back into the equation and simplifying we obtain
the equation of circle AQL, which is

(6.1)

The equations of the circles BQM, CQN may now be written down by cyclic change of x, y, z
and a, b, c and l, m, n. Unfortunately the algebra computer package DERIVE was not powerful
enough to deduce the co-ordinates of the point of intersection of these three circles, other than Q.

7. The centres of the three circles

As the three circles are known to have the common point Q, then they will have a second
common point R if, and only if, their centres lie on a line. This is because they are then bound to
form part of a coaxal system of circles. We refer again to Bradley [3] for how to obtain the centre
of a circle. What has to happen is that the polar of the centre is the line at infinity x + y + z = 0.
This means that the centre of the conic with equation
(7.1)
must have co-ordinates X : Y : Z, where

(7.2)
Using Equation (6.1) we find the centre of circle AQL ha co-ordinates (X, Y, Z), where

(7.3)

4
The centres of circles BQM and CQN may now be written down by cyclic change of X, Y, Z and
a, b, c and l, m, n. A 3 x 3 determinantal test with the co-ordinates of the centres of these three
circles results in a zero value for the determinant, implying that the three centres are collinear,
and hence that there must be a second point of intersection R.

References

1. Miquel, A. "Mmoire de Gomtrie." Journal de mathmatiques pures et appliques de


Liouville 1, 485-487, 1838.
2. Honsberger, R. Episodes in Nineteenth and Twentieth Century Euclidean Geometry.
Washington, DC: Math. Assoc. Amer., p. 81, 1995.
3. Bradley, C. J., The Algebra of Geometry, Highperception, Bath (2007).

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

5
Article 10

A New Construction to find any Circle through a Given Point


Christopher J Bradley

G
P
O U
W U' K
W' F
Q D
V' D'
V
B C

Figure

1. Introduction

In 1907 Hagge wrote an article [1] in which he provided a construction that always
provides a circle, now called a Hagge circle, through the orthocentre H of a triangle ABC.
In the construction there is a key point D such that lines ADE, BDF, CDG meet the
circumcircle at E, F, G respectively. And then E, F, G are reflected in the sides BC, CA,
AB respectively to provide three points U, V, W that are concyclic with H. When you vary
1
D you get a different circle through H. Later Peiser [2] showed that the isogonal conjugate
of D, when rotated by 180o about the nine-point centre coincides with the centre of the
Hagge circle. Furthermore triangles EFG and UVW are indirectly similar by a spiral
similarity through the point D, as shown by the work of Speckman [3]. In the similarity
triangle ABC is similar to a triangle XYZ inscribed in the Hagge circle, where triangle
XYZ is in perspective to triangle UVW with vertex of perspective D. The disadvantage of
the construction is the curious connection between the centre of the Hagge circle and the
point D that generates it. The advantage is that because of the indirect similarity many
interesting results, applications of the work by Speckman [3], exist for the configuration.

It has long been an objective that we should be able to provide a similar construction to
provide special circles through any point P (not lying on the sides of ABC) but that has
remained elusive until now. At the expense of the similarity we now have a method, which
has the advantage that one can predetermine immediately the radius and centre of the
resulting circle through P.

The method is as follows: If a special circle through P is required, then draw a segment
POQ, where O is the circumcentre of triangle ABC and O is the midpoint of PQ. Then
choose a point D so that OPKD is a parallelogram, where K is the desired centre of the
circle and PK = OD is its desired radius. Then draw lines ADE, BDF, CDG where E, F, G
lie on the circumcircle. Next construct the points U, V, W such that they are the fourth
points of the parallelograms AQEU, BQFV, CQGW respectively. The key result is that
circle UVW passes through P and has centre K and radius OD.

A second result is that if the diagonals of these three parallelograms meet at U', V', W'
respectively, then circle U'V'W' is a circle passing through O and D, with OD as diameter,
and with half the radius of the special circle through P. In fact QUU', QVV', QWW' are
straight lines, so that the circles and the disposition of the points on them are directly
similar by an enlargement factor two through Q (showing incidentally that one can
construct similar circles through any point on the line QOP by choosing points on the
diagonals QU, QV, QW suitably). All these result are illustrated in the figure above.

In the following sections we prove these results using Cartesian co-ordinates, taking the
circumcircle to have radius 1 and centre at O. The working is technically quite elaborate,
but the final results are beguilingly simple.

2. Position of the special circle and the points U, V, W

Take P to have co-ordinates ( k, 0), so that Q being the 180o rotation of P about O has co-
ordinates (k, 0). Let D have co-ordinates (m, n). What we shall establish is that the radius
of the special circle UVWP is (m2 + n2) and that its centre K has co-ordinates (m k, n).
Let A have co-ordinates (2a/(1 + a2), (1 a2)/(1 + a2)), and let B, C have corresponding
parameters b, c.

The equation of the line AD is


{(n + 1)a2 + (n 1)}x {(1 + a2)m 2a}y + (1 a2) m 2an = 0. (2.1)

2
This line meets the circumcircle with equation x2 + y2 = 1 at the point E with co-ordinates
(x, y), where
x = 2{m(n + 1)a2 (1 + m2 n2)a + m(1 n)}/{(m2 + (n + 1)2)a2 4ma + m2 + (1 n)2},
y = {(n + 1)2 m2)a2 4mna + m2 (1 n)2}/{(m2 + (n + 1)2)a2 4ma + m2 + (1 n)2}.
(2.2)
Having obtained E the co-ordinates of the displacement QE may be obtained. Bearing in
mind that AU = QE, the co-ordinates of U may now be found and they are (x, y), where
x = (1/s){ (k(m2 + (n + 1)2) 2m(n + 1))a4 + 4(km + n(n + 1))a3 2(k(m2 + n2 + 1) +
2m)a2 + 4(km + n(n 1))a k(m2 + n2 2n + 1) 2m(n 1)},
y = (1/s) { 2(m2a4 + 2m(n 1)a3 4na2 + 2m(n + 1)a m2)}, (2.3)
and
s = (1 + a2) {(m2 + (n + 1)2)a2 4ma + m2 + (n 1)2}. (2.4)
The co-ordinates of V and W may now be obtained by replacing the letter a in Equations
(2.2) (2.4) by letters b and c respectively.

3. The circle UVWP

The equation of the circle with centre ( g, f) and radius (g2 + f2 t) is


x2 + y2 + 2gx + 2fy + t = 0. (3.1)
Inserting the co-ordinates of U, V, W we obtain three equations for f, g, t. When these are
substituted back in Equation (2.5), we obtain the equation of circle UVW, which is
x2 + y2 + 2x(k m) 2mny + k(k 2m) = 0. (3.2)
It can now be seen that circle UVW passes through P( k, 0), has centre K(m k, n) and
radius (m2 + n2), as required to be proved.

4. The points U'V'W' and the circle U'V'W'DO

Having obtained the co-ordinates of U, it is straightforward to obtain the co-ordinates of


U', the midpoint of QU. Its co-ordinates (x, y) are independent of K, since U' is also the
midpoint of AE. They are given by
x = (1/s){(n(1 + a2) (1 a2))(2an m(1 a2))},
y = (1/s){(2a m(1 + a2))(2an m(1 a2))}, (4.1)
where s is given by Equation (2.4). The co-ordinates of V' and W' may be written down
from Equation (4.1) by replacing the letter a by letters b and c respectively.

The equation of circle U'V'W' may now be obtained and is


x2 + y2 mx ny = 0. (4.2)
This circle clearly passes through both O and D and has centre (m, n). In fact OD is a
diameter and its radius is (m2 + n2).

5. Another circle through O

Finally by letting P, O, Q coincide at O and forming the parallelograms AOEU'', BOFV'',


COGW'' the circle U''V''W'' passes through O and has centre D.

3
References
1. K. Hagge, Zeitschrift fr Math. Unterricht, 38 (1907) 257-269. (German)
2. A.M Peiser, Amer. Math. Monthly, 49 (1942) 524-527.
3. H. A. W. Speckman, Perspectief Gelegen, Nieuw Archief, (2) 6 (1905) 179 188.
(Dutch)

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

4
Article 11

P1
Eight circles through the Orthocentre

Christopher J Bradley

W2 V3

W1
N'
A

N
P2
W' M

V
U G H M'
Q
V' U'

E B W C V1

L'
P3

L U2

U3

Figure

1. Introduction

An appreciation of the Figure (drawn with CABRI II plus) requires a description of how it was
drawn and the order in which the various constructions were made. Start with a triangle ABC
and its centroid G. AG, BG, CG meet the circumcircle of ABC at points L, M, N respectively.

1
Tangents at L, M, N to are then drawn with those at M and N meeting at D. E and F are
similarly defined. then becomes an incircle of triangle DEF. Next DA, EB, FC are drawn and
these turn out to meet at a point Q and when extended meet again at L', M', N' respectively.
Lines AL, BN', CM' now prove to be concurrent at P1; BM, CL', AN' prove to be concurrent at
P2; and CN, AM', BL' prove to be concurrent at P3. Furthermore it transpires that P1, P2, P3 are
collinear lying on the polar of Q with respect to . Twelve more points are now obtained as
reflections of L, M, N, L', M', N' in the sides of ABC. These are U, V, W the reflections of L, M,
N respectively in the sides BC, CA, AB; U', V', W' are the reflections of L', M', N' respectively in
the sides BC, CA, AB; U2, U3 are the reflections of M', N' respectively in the side BC; V3, V1 are
the reflections of N', L' respectively in the side CA; and finally W1, W2 are the reflections of L',
M' respectively in the side AB. Now five circles can be drawn, all of which pass through the
orthocentre H, UVW is the Hagge circle of G, U'V'W' is the Hagge circle of Q, UV3W2 is the
Hagge circle of P1, U3VW1 is the Hagge circle of P2 and U2V1W is the Hagge circle of P3. For
the original account of Hagge circles see Hagge [1], for a modern account see Bradley and Smith
[2]. Three more circles may now be drawn, circles BHC, CHA, AHB, and these circles
remarkably contain the points U, U', U2, U3 and V, V', V3, V1 and W, W', W1, W2 respectively. It is
also the case that lines V1W1, W2U2 and U3V3 all pass through H.

The analysis that follows in later sections proves all the above results using areal co-ordinates
with ABC as triangle of reference. For an account of these co-ordinates and how to use them, see
Bradley [3]. It must now be confessed that the point G was used in the construction and in the
analysis that follows in order to provide an analysis that is as easy as possible, the areal co-
ordinates of G being simply (1, 1, 1). The truly remarkable thing is that all the concurrences,
collinearities, and dispositions of points on circles holds not just for the starting point G, but for a
any internal point of ABC taken as starting point (except H itself and except the symmedian
point K when in both cases the figure degenerates). The problem, of course, is that if one starts
with P(l, m, n) the analysis becomes technically very difficult (because of the three extra
variables) even with a good algebra computer package, but not only that, it becomes virtually
unprintable. No excuses are therefore made for having taken the easiest option.

2. The points L, M, N, D, E, F

The equation of the line AG is y = z and this meets the circumcircle with equation
(2.1)

at the point L with co-ordinates L( a2, b2 + c2, b2 + c2). M and N follow by cyclic change and
have co-ordinates M(c2 + a2, b2, c2 + a2) and N(a2 + b2, a2 + b2, c2).

The equation of the tangent to at any point (d, e, f) is


(2.2)

2
It follows that the equation of the tangent at L is
(2.3)

The tangents at M and N are respectively


(2.4)
and
(2.5)

The tangents at M and N meet where both equations (2.4) and (2.5) hold. This is at the point D
with co-ordinates D( (a4 + a2(b2 + c2) + 2b2c2), b2(a2 + b2 c2), c2(c2 + a2 b2)). The co-
ordinates of E and F follow by cyclic change and are E(a2(a2 + b2 c2), (b4 + b2(c2 + a2) +
2c2a2, c2(b2 + c2 a2)) and F(a2(c2 + a2 b2), b2(b2 + c2 a2), (c4 + c2(a2 + b2) + 2a2b2)).

3. The points Q, L', M', N', P1, P2, P3

The critical result that makes the configuration possess so many interesting details is the
concurrence of the lines DA, EB, FC at the point Q. This is the point that links L, M, N with the
points L', M', N'. Without that link the figure would not contain enough circles to be of
significance.

The equations of the lines DA, EB, FC are


(3.1)
(3.2)
(3.3)

respectively. These lines are concurrent at the point Q with co-ordinates (x, y, z), where
Note that the isogonal of Q is the isotomic of H.
Now the line DA with Equation (3.1), when extended, meets , with Equation (2.1), at the point
L', which therefore has co-ordinates L'( 1, 2b2/(c2 + a2 b2), 2c2/(a2 + b2 c2)). The co-ordinates
of M' and N' follow by cyclic change and are M'(2a2/(b2 + c2 a2), 1, 2c2/(a2 + b2 c2)) and
N'(2a2/(b2 + c2 a2), 2b2/(c2 + a2 b2), 1).

The polar of a point (d, e, f) with respect to has the same form as Equation (2.2) and
accordingly the equation of the polar of Q with respect to is
(3.4)

Next it emerges that the three lines AL, BN', CM' are concurrent at the point P1 with co-ordinates
P1(2a2, a2 b2 c2, a2 b2 c2), and the three lines BM, CL', AN' are concurrent at P2 with co-
ordinates P2(b2 c2 a2, 2b2, b2 c2 a2), and the three lines CN, AM', BL' are concurrent at P3

3
with co-ordinates P3(c2 a2 b2, c2 a2 b2, 2c2). Finally P1, P2, P3 are collinear and lie on the
line with Equation (3.4), which is the polar of Q with respect to .

Note that we have now shown that triangle ABC is in perspective with each of the five triangles
LMN, L'M'N', LN'M', N'ML', M'L'N, the five vertices of perspective being G, Q, P1, P2, P3
respectively.

4. Definition of the points U, V, W, U', V', W', U2, U3, V3, V1, W1, W2

The points U, V, W are the reflections of L, M, N respectively in the sides BC, CA, AB. The
points U', V', W' are the reflections of the points L', M', N' respectively in the sides BC, CA, AB.
The points U2, U3 are the reflections of M', N' respectively in the side BC. The points V3, V1 are
the reflections of the points N', L' respectively in the side CA and the points W1, W2 are the
reflections of the points L', M' in the side AB. To go through all the details of obtaining the co-
ordinates of these points would be both dull and unnecessary. Instead we give expressions for the
reflection of a general point with co-ordinates (d, e, f) in the sides BC, CA, AB and then quote
the co-ordinates of those points that are necessary to establish the features illustrated in the figure
and explained in section 1.

The treatment of perpendicularity when using areal co-ordinates is tiresome. Details of how to
proceed are given in Bradley [3, 4]. Following the plan explained above we find the reflection of
(d, e, f) in the side BC has co-ordinates ( d, e + d(a2 + b2 c2)/a2, f + d(c2 + a2 b2)/a2), the
reflection of (d, e, f) in the side CA has co-ordinates (d + e(a2 + b2 c2)/b2, e, f + e(b2 + c2
a2)/b2) and the reflection of (d, e, f) in the side AB has co-ordinates (d + f(c2 + a2 b2)/c2, e +
f(b2 + c2 a2)/c2, f).

5. The five Hagge circles

Using the results of Section 4 we find the co-ordinates of U to be (a2, 2c2 a2, 2b2 a2), the co-
ordinates of V to be (2c2 b2, b2, 2a2 b2) and the co-ordinates of W to be (2b2 c2, 2a2 c2, c2).
The general equation of a circle in areal co-ordinates is
(5.1)
We put the co-ordinates of U, V, W into Equation (5.1) and obtain three equations for u, v, w.
These are then reinstated in Equation (5.1) and simplified to obtain the Hagge circle of G, whose
equation is

(5.2)
It may be checked that this circle passes through the orthocentre H with co-ordinates (1/(b + c2
2

a2), 1/(c2 + a2 b2), 1/(a2 + b2 c2)).

4
We now repeat this exercise to find the co-ordinates of the points U', V', W' and hence determine
the equation of the Hagge circle of Q. The co-ordinates of U' turn out to be
(1, (a4 2a2b2 (b2 c2)2)/{a2(c2+ a2 b2)}, (a4 2c2a2 (b2 c2)2)/{a2(a2 + b2 c2)}). The
co-ordinates of V' and W' may be written down by cyclic change. Using the method described
above we find the Hagge circle of Q to be

(5.3)
Again this circle contains H. The three other Hagge circles are UV3W2, VW1U3 and WU2V1 and
derive from the points P1, P2, P3 respectively. We do not give the equations of these circles.

6. Three lines and three more circles

The reflections of the point L', in the sides BC, CA, AB, are the points U', V1, W1 respectively.
These are known to lie on a line through the orthocentre H, sometimes known as the Double
Simson line of L'. Similarly the points V', W2, U2, H are collinear as are W', U3, V3, H.

The other three circles in the figure (besides ) are the circles BHC, CHA and AHB. These are
well-known circles. The equation of CHA, for example, is
(6.1)
The interesting result is that this circle passes through all the points V, V', V1 and V3. The co-
ordinates of V and V' have already been given, those of V1 are (x, y, z), where

(6.2)
and the co-ordinates of V3 are (x, y, z), where

, (6.3)
Similarly circle BHC contains U, U', U2, U3 and circle AHB contains the points W, W', W1, W2.

7. When the starting point is H or K

When the starting point is the orthocentre H, then the points U, V, W coincide with H and one of
the Hagge circles becomes a point. Circles such as UV2W3 are ill-defined, though, of course, the
circle HV2W3 can be drawn.

When the starting point is the symmedian point K, then, as D, A, L are collinear, L, L' coincide,
as do M, M' and N, N'. There are only four Hagge circles as Q coincides with K. It is interesting
however, that triangle ABC is in perspective with the four triangles LMN, LNM, NML and

5
MLN. The latter three provide what is called a triple reverse perspective and triangles ABC and
LMN become triangles that define the Brocard porism. The line P1P2P3 is now the polar of K and
it contains three further interesting points AB^DE, BC^EF and CA^FD.

References
1. K.Hagge, Zeitschrift Fr Math. Unterricht, 38, pp257-269 (1907).
2. C. J Bradley & G.C.Smith, Math. Gaz. pp202 -207 July 2007.
3. C. J. Bradley, Challenges in Geometry, Oxford (2005).
4. C. J .Bradley, The Algebra of Geometry, Highperception, Bath (2007).

Flat 4,
Terrill Court,
12/14 Apsley Road
Clifton
BRISTOL BS8 2SP

6
Article 12
Circles concentric with the Circumcircle
Christopher J Bradley

C'
A

E
F
V
U O B'
Q P
X
Y
B C
W

D
A'

Figure

1. Introduction

Let ABC be a triangle with circumcircle S, centre O, and suppose P is a point not on the sides of
ABC nor on S. In this short article we show that the circle, centre O, through P contains seven

1
points U, V, W, X, Y, Z, Q with special properties. The figure above illustrates these properties
which are a result of the following three theorems.

Theorem 1
Let AO, BO, CO meet S at D, E, F respectively. Let AP, BP, CP meet S at A', B', C' respectively.
Draw lines l, m, n through P parallel to A'D, B'E, C'F respectively. Draw the line through D
perpendicular to l to meet l at U. Define V, W similarly using E, m and F, n respectively. Then
U, V, W lie on the circle through P, centre O.

Theorem 2
Draw the diameters UOX, VOY, WOZ of then X, Y, Z lie on AP, BP, CP respectively.

Theorem 3
AU, BV, CW are concurrent at a point Q lying on .

In the proofs that follow we use Cartesian co-ordinates.

2. Proof of Theorem 1

Take O to be the origin, S to be the unit circle, with equation x2 + y2 = 1, and A to have co-
ordinates ((1 a2)/(1 + a2), 2a/(1 + a2)), with B and C having parameters b, c respectively. Let P
have co-ordinates (c, 0), c 0, 1.

The equation of AP is therefore


2ax + y(a2(c + 1) + c 1) = 2ac. (2.1)

This line meets S again at the point A' with co-ordinates (x, y), where
x = {a2(c + 1)2 (c 1)2}/{a2(c + 1)2 + (c 1)2}, (2.2)
y = {2a(c2 1)}/ {a2(c + 1)2 + (c 1)2}. (2.3)

Since AOD is a diameter, the co-ordinates of D are the negatives of those of A and thus the
equation of A'D is
x(a2(c + 1) + (c 1)) 2ay = a2( c + 1) (c 1). (2.4)

The line l is parallel to this through P and therefore has equation


x(a2(c + 1) + (c 1)) 2ay = c(a2(c + 1) + (c 1)). (2.5)

The equation of the line through D perpendicular to A'D is


2ax + y(a2( c + 1) + (c 1)) = 2ac. (2.6)

2
The lines with equations (2.5) and (2.6) meet at the point U with co-ordinates (x, y), where
x = {c(a4(c + 1)2 +2a2(c2 3) + (c 1)2)}/{a4(c + 1)2 + 2a2(c2 + 1) + (c 1)2}, (2.7)
2 4 2 2 2 2
y = {4ac(a ( c + 1) + (c 1)}/ {a (c + 1) + 2a (c + 1) + (c 1) }. (2.8)

The sum of the squares of these co-ordinates is c2, so U lies on the circle , centre O, radius c,
which passes through P. Similarly V, W lie on .

3. Proof of Theorems 2 and 3

Since UOX is a diameter of the co-ordiantes of X are the negatives of those of U and it is soon
checked that X lies on AP with Equation (2.1).

The line DU with equation (2.6) meets again at the point Q with co-ordiantes ( c, 0) and since
its co-ordinates are independent of Q also lies on EV and FW. It also lies on the diameter PO.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

3
Article 13

Ex-points and eight intersecting circles

Christopher J Bradley

H'

P3
W' L'
A
V'

N E P2

M
Q V
U P
O J H
W
M'
C
B R N'

To
T S

U'

P1
D

Figure

1. Introduction

1
In this article we use areal co-ordinates throughout. See Bradley [1, 2 ] for a review of how to
use them. Incentres and ex-centres are well-known and if (a, b, c) are the co-ordinates of the
incentre, then the three ex-centres I1, I2, I3 have co-ordinates ( a, b, c), (a, b, c), (a, b, c)
respectively. Less well-known, perhaps, are the ex-symmedian points D, E, F, which are the
intersections of the tangents at the vertices of the circumcircle, D being the intersection of the
tangents at B and C, with E, F similarly defined. These points can be seen in the Figure. Their
co-ordinates are ( a2, b2, c2), (a2, b2, c2), (a2, b2 c2) respectively. It is also true that if P (l, m,
n) is any point internal to a triangle ABC, then there are three associated ex-points P1( l, m, n),
P2(l, m, n), P3(l, m, n). We show how these points may be constructed using compass and
straight edge only. It turns out that the construction is a development of the construction of the
Hagge circle [3] of the point P, so we are able to draw four Hagge circles, one for each of P, P1,
P2, P3, all of which, of course, pass through the orthocentre H of triangle ABC. As a bonus there
are in addition four other circles and the CABRI II geometry computer software package
indicates that these other four circles also have a common point, labelled H' in the Figure. The
additional four circles have equations that are considerably more complicated than those of the
four Hagge circles, so we have been unable to find the co-ordinates of H'. However, it may be
proved by inversion that these four circles do indeed have a common point. This proof has been
supplied by David Monk [4] and I am extremely grateful to him for providing the proof.

We now describe the construction of the ex-points, starting from ABC and P. First draw the
tangents at A, B, C to the circumcircle to provide D, E, F. Next draw the lines AP, BP, CP to
meet the circumcircle at L, M, N respectively. Now draw DL, EM, FN to meet the circumcircle
again at L', M', N' respectively. Note that these three lines are concurrent at a point Q, which we
call the generating point. Now AL, BM', CN' are found to be concurrent at the point P 1, AL',
BM, CN' are concurrent at P2, and AL', BM', CN are concurrent at P3. Section 2 provides the
algebraic proof of these remarks. Q is called the generating point, because rather than start with
P, one can start with Q, which must be internal to triangle DEF, and then reverse the
construction, thus generating all 4 points P, P1, P2, P3. We defer the description of the
construction of the 8 circles until later in the text.

2. The ex-points P1, P2, P3 and the generating point Q

For those unfamiliar with areal co-ordinates for points on the circumcircle we give a brief
account of what is required. The co-ordinates of a general point on are ( a2t(1 t), b2(1 t),
c2t), where t is a parameter that can take any value (including infinity). Thus the points A(1, 0,
0), B(0, 1, 0), C(0, 0,1) have parameters , 0, 1 respectively and the equation of the circumcircle
is
. (2.1)
For checking purposes it is useful to have the equation of the chord of joining points with
parameters t and s. This is

2
(2.2)

Starting with P(l, m, n) we find the equation of AP to be ny = mz. This meets , with Equation
(2.1), at the point L with co-ordinates L( a2mn/(b2n + c2m), m, n). Similarly BP, CP meet at
M, N respectively with co-ordinates M(l, b2nl/(c2l + a2n), n) and N(l, m, c2lm/(a2m + b2l).

The tangents to with Equation (2.1) at A, B, C are respectively b2z + c2y = 0, c2x + a2y = 0, a2y
+ b2x = 0.The point D, that is the intersection of the tangents at B and C, therefore has co-
ordinates D( a2, b2, c2) and E, F respectively have co-ordinates E(a2, b2, c2), F(a2, b2, c2). It
can now be seen that D, E, F are the ex-symmedians, as the symmedian point K has co-ordinates
(a2, b2, c2). The line DL has equation
(2.3)

This meets again at the point L' with co-ordinates L'(a2mn/(b2n c2m), m, n) Note that this
differs from L simply by a change of sign of m (or n). We can now write down, by cyclic
change, the co-ordinates of M' and N', which are M'(l, b2nl/(c2l a2n), n) and N'( l, m,
c2lm/(a2m b2l)).

Now EM has equation that may be derived from that of DL by cyclic change and then the
generating point Q is determined as DL^EM. The result is that Q has co-ordinates

It may now be checked that Q also lies on the line FN.

The equations of the lines AL', BM', CN' can now be determined and they are respectively
(2.4)

We now have enough to determine the co-ordinates of the points P1 = AL^ BM'^CN', P2 =
AL'^BM^CN' and P3 = AL'^BM'^CN, and, as you will have guessed, these are the ex-points of P
with co-ordinates P1( l, m, n), P2(l, m, n), P3(l, m, n). It may be checked that EP2 and FP3
intersect at a point R on BC, with S, T on CA, AB respectively defined similarly.

In the Figure we have chosen P to be the isogonal conjugate of the nine-point centre. This point
generates the Hagge circle with OH as diameter.

3. A 9 point hyperbola

Using the standard 6 x 6 determinantal method to obtain the conic through P, its 3 ex-points and
K the symmedian point we find the equation of a hyperbola
(3.1)

3
Its form immediately indicates that not only does this hyperbola pass through the points just
mentioned, but it also passes through D, E, F the ex-symmedians. It may be checked that it also
passes through the generating point Q. Thus, it passes through 9 key points of the figure. As P
may be any point, we have the result, for example, that a 9 point hyperbola passes through the
incentre, the ex-centres, the symmedian point, the ex-symmedians and the point Q, which in this
case is the circumcentre O. It is a rectangular hyperbola, since the incentre and ex-centres form
an orthocentroidal quartet. If P lies at O, then the hyperbola is also rectangular, since the
orthocentre of triangle O1O2O3 lies on the hyperbola.

4. The reflected points

The points in the Figure labelled U, V, W, U', V', W' are the reflections of the points L, M, N, L',
M', N' in the sides BC, CA, AB, BC, CA, AB respectively. This is the same as in the
construction of Hagge circles. The algebra for obtaining the co-ordinates of the reflected points
is covered in detail in Bradley and Smith [3] and is not repeated here. We therefore state the co-
ordinates of these points without proof and they are as follows:
(4.1)
(4.2)
(4.3)
(4.4)
(4.5)
(4.6)

5. The 8 circles

There is, of course a circle through any 3 of these points, but there are 8 special circles found by
choosing one of U, U', one of V, V' and one of W, W'. First we consider the circle UVW. This is
the Hagge circle generated by P. The method is to write down the general equation of a circle,
which is
(5.1)
Then we insert the co-ordinates of U, V, W in turn, to provide three equations for the real
numbers u, v, w. Their values are then substituted back in Equation (5.1). The result is

(5.2)

It may be checked that this circle passes through H, the orthocentre of ABC. The circle UV'W'
has the equation derived from Equation (5.2) by changing the sign of l. It is therefore the Hagge
circle generated by P1. Similarly changes of sign of m, n in Equation (5.2) produce the Hagge
circles of P2, P3 respectively.

4
The other four circles U'V'W', U'VW, UV'W, UVW' have equations that are extremely lengthy
and complicated, and the algebra computer package DERIVE was unable to solve the
simultaneous quadratics of any pair of them to show that they have the common point, labelled
H' in the figure. However, David Monk [4] has kindly provided the following proof: Invert the
figure in H so that circles UVW, UV'W', U'VW', U'V'W become the lines uvw, uv'w', u'vw',
u'v'w. The resulting figure is a complete quadrilateral, and thus the four triangles u'v'w', u'vw,
uv'w, uvw' have circumcircles that possess a common point h'. Inverting back gives the result
required. This proof does not tell us anything about the dependence of H' on the point P. All that
we can say from drawings is that the locus of H' as P moves on a line is a curve that is not a
conic.

References
1. C.J.Bradley, Challenges in Geometry, Oxford (2005);
2. C.J.Bradley, The Algebra of Geometry, Highperception, Bath (2007);
3. Christopher Bradley and Geoff C. Smith, On a construction of Hagge, Forum
Geometricorum p231247 (2006).
4. D. Monk (private communication).

Flat 4,
Terrill Court,
12-14 Apsley Road,
Clifton,
BRISTOL BS8 2SP

5
Article 14

Intersecting Circles having chords the sides of a Cyclic Quadrilateral

Christopher J Bradley

F
A

B
L

S K
O
P
N X
E
T
H

G
C
M

Figure 1

1. Introduction

1
We consider a cyclic quadrilateral in which the centre O of the circle ABCD is distinct
from the intersection E of the diagonals AC and BD.

In this configuration a number of results hold.


(i) Circles AOB, COD, AED, BEC are concurrent at a point S;
(ii) Circles AOD, BOC, AEB, CED are concurrent at a point T;
(iii)O, S, E, T are concyclic;
(iv) The centre of circle OSET is X, the midpoint of OE, and it follows that
angle OSE and OTE are each 90o;
(v) FXG, NXP, HXK, LXM are straight lines, where F, G, H, K, L, M, N, P are the
centres of circles AOB, COD, AOD, BOC, AEB, CED, AED, BEC
respectively;
(vi) The following angles are each 90o: angles HXL, NXF, KXL, PXF, HXM,
NXG, KXM and PXG and consequently angles HXN, FXL, GXM, KXP are
equal.

See Figure 1 for an illustration of these properties.

Now let AB and CD meet at F, and let AD and BC meet at G. Then another set of
properties hold. In properties (vii) to (xi) points S, T, E are as before but points F, G, H, K
are different points.
(vii) Circles FBC, FAD, GCD, GAB meet at a point H lying on FG;
(viii) Circles FBD, GAC pass through S and circles FAC, GBD pass through T;
(ix) Points O, E, H are collinear;
(x) OS^ET = F and OT^SE = G;
(xi) Points F, G, S, T are concyclic and the centre K of the circle FGST is the midpoint
of FG.

See Figure 2 for an illustration of these properties.

In the following sections we prove these results, using Cartesian co-ordinates.

2. Circles AOB and COD and the point S

We take the equation of ABCD to be x2 + y2 = 1 and the co-ordinates of A to be (2a/(1 +


a2), (1 a2)/(1 + a2)), with B, C, D having similar co-ordinates with parameters b, c, d
respectively. The co-ordinates of O are, of course, (0, 0).

Since it passes through O the equation of circle OAB is of the form


(2.1)

Inserting the co-ordinates of A and B in Equation (2.1) we find f and g and substituting
back we obtain the equation of circle OAB to be
(2.2)

2
The equation of OCD may be obtained from Equation (2.2) by writing c, d instead of a, b.
The intersections of these two circles are the origin O and the point labelled S, whose co-
ordinates are (x, y), where
(2.3)

(2.4)
and

. (2.5)

3. Circles AOD and BOC and the point T

The equations of circles AOD and BOC may be written down by altering the parameters in
Equation (2.2) to a, d and to b, c respectively. The co-ordinates of T may be written down
from Equations (2.3) to (2.5) by exchanging a and c (or b and d, but not both).

4. The point of intersection E of the diagonals AC and BD

The equation of AC is
(4.1)

and the equation of BD is


(4.2)

These lines meet at E with co-ordinates (x, y), where


(4.3)
(4.4)
and

5. The circles AED, BEC, AEB, CED

The general equation of a circle is of the form


(5.1)

If we now insert the co-ordinates of A, E, D into Equation (5.1) we obtain three equations
for f, g, k. When these are substituted back into Equation (2.1) we find the equation of
AED, which is

3
. (5.2)
The equations of circles BEC, AEB, CED may now be written down from Equation (5.2).
Bearing in mind that the co-ordinates of E are invariant under the interchange of a and c
and of b and d, it is now straightforward to write down the equations of circles BEC, AEB,
CED.

6. (i) and (ii) and the circle OSET

It may now be verified that circles AED and BEC pass through S and circles AEB and
CED pass through T.

The circle OES has an equation of the form (2.1) and inserting the co-ordinates of E, given
in Equations (4.3) to (4.5) and the co-ordinates of S, given in Equations (2.3) to (2.5) we
get two equations for f and g. When these values are reinserted in (2.1) we obtain the
equation of OES, which is

(6.1)

It may now be verified that T lies on this circle.

7. Centres of circles and properties (iv), (v), (vi)

The centre of the circle with Equation (5.1) is, of course ( g, f). It follows that the centre
X of circle OSET has co-ordinates
(1/k)(ac bd, (7.1)
where
k= . (7.2)

Noting Equations (4.3) to (4.5) we see that X is the midpoint of OE. This establishes
property (iv).

The perpendicular bisector of AB passes through the centres of circles ABCD, AOB, AEB.
It follows that points F, L, O are collinear. Similarly H, N, O and P, K, O and M, G, O are
collinear, all four lines meeting at O.

From Equation (2.2) the centre F of circle AOB has co-ordinates ((a + b)/(1 + ab), (1
ab)/((1 + ab)). Similarly the centre G of circle COD has co-ordinates ((c + d)/(1 + cd),
(1 cd)/((1 + cd)). A 3 x3 determinantal test using the co-ordinates of X, F, G in the first
two columns and 1 as the entry in each row of the third column gives a zero result. Hence
F, X, G are collinear. Similarly H, X, K are collinear.

4
From Equation (5.2) the centre N of circle AED has co-ordinates (1/k)((a + d)(c b), (c
b)(1 ad)), where k = a(b(c d) + cd + 1) b(1 + cd) + c d. The co-ordinates of the
centre P of circle BEC may now be written down by exchanging b and d and also a and c.
It may now be verified that N, X, P are collinear and similarly L, X, M are collinear. This
now establishes property (v).

We now have the co-ordinates of N, X and F and can therefore find the equations of the
lines FX and NX. The equation of FX is

(7.3)

and the equation of NX is

(7.4)

From these equations it is evident that the product of their gradients is 1 and hence angle
FXN = 90o. Similarly the angles LXH, PXF, KXL, GXN, MXH, MXK and GXP are all
90o. This completes the proof of property (vi).

8. The points F and G and the line FG

The intersection of lines AB and CD is the diagonal point F and its co-ordinates are
therefore (x, y), where
(8.1)

(8.2)

5
A

O B
S

T E
F

HK

Figure 2

where
(8.3)

The diagonal point G is the intersection of lines AD and BC and therefore has co-ordinates
(x, y), where
(8.4)
(8.5)
and
(8.6)

The equation of the line FG may now be obtained and is

(8.7)

9. OT and SE pass through G and OS and TE pass through F

6
The equation of OT is
(9.1)

It may now be verified, using Equations (8.4) to (8 .6) that G lies on this line.

The equation of SE is

(9.2)

It may be verified that G also lies on this line. Similarly F lies on OS and TE. This
establishes property (x).

10. Circles FBC and FAD, the point H, the line OEH and the circle FSTG

Following the usual procedure we obtain the equation of circle FBC, which is

(10.1)
And the equation of circle FAD is

(10.2)
The intersection of these circles, other than the point F, is the point we define to be the
point H and its co-ordinates are (x, y), where

(10.3)

(10.4)
where

. (10.5)

It may now be verified that H lies on FG with Equation (8.7). The equations of circles
GCD and GAB follow by carefully rearranging parameters in Equations (10.1) and (10.2)
and then it may be checked that H lies also on these circles. This establishes property (vii).

The line OE passes through H and as O is the orthocentre of triangle EFG it follows that
OE cuts FG at right angles.

We have already shown that angles OSE and OTE and that OS passes through F and OT
passes through G. It follows that angles FSG and FTG are both right angles. Thus S, T, F,
7
G are concyclic and the centre K of circle FSTG is the midpoint of FG. We have now
established properties (ix) and (xi).

11. Circles FBD and FAC pass through T and circles GBD and GAC pass
through S

From the co-ordinates of F, B, D we may now obtain the equation of circle FBD, which is

. (11.1)

It may now be checked that T lies on this circle. Similarly T lies on circle FAC, S lies on
circle GBD and also on circle GAC. This establishes property (viii).

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

8
Article 15
Four Concurrent Euler Lines

Christopher J Bradley

1. Introduction

In this paper we consider a cyclic quadrilateral PQRS in which the diagonals PR and QS meet at
a point E and we prove a number of results about the triangles PQE, QRE, RSE, SPE.

These are:
1. The Euler lines of the four triangles are concurrent at a point X;

1
2. If the circumcentres of the four triangles are denoted by O1, O2, O3, O4 respectively then
O1O2O3O4 is a parallelogram;
3. If the orthocentres of the four triangles are denoted by H1, H2, H3, H4 respectively then
H1H2H3H4 is a parallelogram;
4. The parallelograms O1O2O3O4 and H1H2H3H4 are inversely similar, with O1O2 parallel to
H1H4 etc.;
5. O1H3, O3H1, O2 H4, O4H2 are concurrent at E.

Results 2 and 3 are known results, see Bradley [1], but they are inevitably given explicit proof in
the course of the calculations required to prove the other results. The results are illustrated in the
figure. It is also the case that the analysis takes no account of the order of the vertices P, Q, R, S
and is equally valid when P and R are adjacent vertices. This means that Results 1-5 are also true
for the cyclic quadrilateral PQRS when E is replaced by F = PQ^RS or G =QR^SP, with the four
triangles replaced in an obvious way. Analysis is carried out using Cartesian co-ordinates.
Singular cases can occur when one or other of the key points recede to infinity. We assume in
our analysis that all points arising are finite and distinct. Calculations were carried out using the
algebra package DERIVE and have been subject to various consistency tests.

2. Preliminary results

We consider a triangle ACE, with E(0, 0), A(a, b), C(c, d), (ad bc), and work out the co-
ordinates of its circumcentre O, its orthocentre H and the equation of its Euler line OH. In
subsequent sections we identify the vertices A, C successively as the pairs P, Q; Q, R; R, S; S, P.
The equations of EA, EC, AC are immediate and are respectively
ay = bx, (2.1)
cy = dx (2.2)
and
x(b d) + y(c a) + ad bc = 0 (2.3)

The perpendicular bisectors of EA and EC are respectively


2(ax + by) = a2 + b2 (2.4)
and
2(cx + dy) = c2 + d2. (2.5)

These meet at the circumcentre O with co-ordinates (x, y), where


x= (2.6)

y= . (2.7)
It may be checked that this also lies on the perpendicular bisector of the line AC. The three
altitudes have equations

2
y(b d) = x(c a), (2.8)
ax + by = ac + bd, (2.9)
cx + dy = ac + bd. (2.10)

These meet at the orthocentre H with co-ordinates (x, y), where



x= (2.11)
y= . (2.12)

We may now calculate the equation of the line OH, the Euler line of triangle ACE, and the result
is
(3ac(a c) + ad(2b d) + bc(b 2d))x + (3bd(b d) + ad(a 2c) + bc(2a c))y
(a2 + b2 c2 d2)(ac + bd) = 0. (2.13)

It may be checked that the centroid (a + c, b + d) lies on this line.

3. Co-ordinates of key points in the configuration

The ease with which a calculation may be carried out depends critically on the choice of the co-
ordinates of E, P, Q, R, S. As in Section 2 the point E is taken as origin. For the points P, Q, R, S
we take QS to be the x-axis and PR to be inclined to it in a direction determined by the unit
vector ( ) and use the converse of the intersecting chord theorem to specify their co-
ordinates as P(km(1 t2), 2kmt), Q(mn(1 + t2), 0), R( nl(1 t2), 2nlt), S(kl(1 + t2), 0). With
these co-ordinates the equation of the circle PQRS is
2t(x2 + y2) + 2t(1 + t2)(mn kl)x + (1 + t2){(nl km)(1 + t2) + (kl mn)(1 t2)}y
2klmnt(1 + t2)2 = 0. (3.1)

With E as origin then the analysis of Section 2 may be applied to each of the triangles PQE,
QRE, RSE, SPE in turn by varying the values of a, b, c, d.

Result 1
The four Euler lines of these triangles are concurrent at the point X with co-ordinates (x, y),
where
x= , (3.2)

y= . (3.3)

3
We now catalogue the co-ordinates of the points O1, O2, O4 the circumcentres of triangles of
triangles PQE, QRE, SPE and their corresponding orthocentres H1, H2, H4. The co-ordinates of
O3, H3 for triangle RSE are omitted as results involving these points follow by similar reasoning.

O1 ( mn(1 + t2), (1/4t)(m(1 + t2)){k(1 + t2) + n(1 t2)});


O2 ( mn(1 + t2), (1/4t)(n(1 + t2)){m(1 t2) l(1 + t2)});
O4 ( kl(1 + t2), (1/4t)(k(1 + t2)){m(1 + t2) l(1 t2)});
H1 (km(1 t2), (1/2t)(m(1 t2)){k(1 t2) + n(1 + t2)});
H2 ( ln(1 t2), (1/2t)(n(1 t2)){l(1 t2) m(1 + t2)});
H4 (km(1 t2), (1/2t)(k(1 t2)){l(1 + t2) m(1 t2)}).

4. The remaining results

Since Results 2 and 3 are known we do not provide further details, except to say that these
results are a consequence of the fact that O1O2, O3O4 are perpendicular to QS, as are H1H4 and
H3H2 all having infinite gradient, and the fact that the other four sides are perpendicular to PR
having gradient .

Result 4
We are now able to compute the squares of various lengths and find that
O1O22 = ; (4.1)

H1H22 = ; (4.2)

O1O42 = ; (4.3)

H1H42 = ; (4.4)

From which we deduce


O1O22/H1H22 = O1O42/H1H42 = , (4.5)

which, together with Results 2 and 3, establishes Result 4 that the two parallelograms are
inversely similar. It is interesting to note that the enlargement factor of the similarity depends
only on the angle between PR and QS.

Result 5

Inspection of the co-ordinates given at the end of Section 3 shows that O1O2O3O4 is also in
perspective with H3H4H1H2 with vertex of perspective the origin E.

4
Reference
1. C. J. Bradley, The Algebra of Geometry, Bath: Highperception (2007).

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

5
Article 16
The Direct Similarity of the Miquel Point Configuration
Christopher J Bradley

1. Introduction

Given a triangle ABC and points L, M, N, other than the vertices, lying on BC, CA, AB
respectively, then circles AMN, BNL, CLM share a common point P called the Miquel point. For
a given triangle the position of P obviously depends on the points L, M, N.

In this article we show that the centres X, Y, Z of the three circles form a triangle XYZ that is
directly similar to ABC and that the centre of the similarity is the point P itself, by which we
mean that XYZ is obtained from ABC by a rotation about P followed by dilation with centre P.

We also give an analysis of the converse result. What this amounts to is that given a triangle
ABC and a point P not on the sides, then it is always possible to find circles centres X, Y, Z to
provide a configuration having P as a Miquel point, but the options are more limited than one
might suppose. A direct similarity centre P is involved and any angle of rotation (other than a
right angle) is possible, but then the scale factor of the dilation is fixed and depends on the angle
of rotation. When the angle of rotation is 0 the scale factor is and triangle LMN is the pedal
triangle of P.

We also consider the problem of where L, M, N have to be placed in order that the circles AMN,
BNL, CLM should have equal radius. This problem results in equations that have to be solved
numerically in order to find the relationship between the points L, M, N.

Finally L, M, N are collinear if, and only if, P lies on the circumcircle of ABC and then AX, BY,
CZ are concurrent, so that triangles XYZ and ABC are not only similar but in perspective. It
follows from the work of Wood [1] on similar in perspective triangles that the perspector Q lies
on both the circumcircles of ABC and XYZ. The circle XYZ may be identified as the circle of
centres in the Wood configuration, see Bradley and Smith [2]. It therefore contains O, the
circumcentre of ABC.

In Fig. 5.1 we show the Miquel configuration for an arbitrarily chosen point P, when the angle of
rotation of the similarity is 30.

1
2. Figure arising from arbitrary L, M, N on the sides and the consequential similarity

We start with a general, but economically parameterised triangle ABC, with A(0, 0), B(2, 2v),
C(2, 2w), (w > v). The points L, M, N are now chosen on the sides BC, CA, AB respectively and
to these are assigned co-ordinates L(2, 2u), M(h, hw), N(k, kv) (u v, w; h, k 0, 1).

The equations of the circles BNL, CLM, AMN may be obtained using standard methods involving
4 x 4 determinants and are respectively
x2 + y2 + (2(uv 1) k(v2 + 1))x 2(u + v)y + 2k(v2 + 1) = 0, (2.1)
2 2 2 2
x + y + (2(uw 1) h(w + 1))x 2(u + w)y + 2h(w + 1) = 0, (2.2)
2 2 2 2 2 2
(v w)(x + y ) + (kw(v + 1) hv(w + 1))x + (h(w + 1) k(v + 1))y = 0. (2.3)

It may be checked that these three circles share a common point P. We do not record its co-
ordinates as they are complicated and are not needed in what follows. The co-ordinates of the
centres of the three circles can now be determined and are

2
X : {1/(2(v w))}( hv(w2 + 1) kw(v2 + 1), k(v2 + 1) h(w2 + 1)),
Y : ((k(v2 + 1) 2(uv 1)), u + v),
Z : ((hw2 + 1) 2(uw 1)), u + w).

Calculations may now be carried out to show that XY/YZ = AB/BC = {(v2 + 1)}/(w v), and
XZ/XY = AC/AB = {(w2 + 1)/(v2 + 1)} and this establishes the similarity between triangle XYZ
and ABC.

We now consider the problem of when the three Miquel circles have a common area, which is
when AL2 = BM2 = CN2. We find
AL2 = (v2 + 1)(w2 + 1)(h2(w2 + 1) 2hk(vw + 1) + k2(v2 + 1))/{4(v w)2}, (2.4)
2 2 2 2 2 2 2
BM = {k (v + 1) 4k(v + 1)(uv + 1) + 4(u + 1)(v + 1)}, (2.5)
2 2 2 2 2 2 2
CN = {h (w + 1) 4h(w + 1)(uw + 1) + 4(u + 1)(w + 1)}. (2.6)

Solving for h, k in terms of u we find the condition for equal areas is


h = (2uv v2 + 1)/(vw + 1), (2.7)
and
k = (2uw w2 + 1)/(vw + 1). (2.8)
l
These formulas give the unique positions for M, N when L is given. A particularly obvious case
is when h = k = 1 and u = (v + w), which is when P is the circumcentre and L, M, N the
midpoints of the sides. No other case exists in which h = k.

3. Figure arising from choosing an arbitrary position for P and constructing a


direct similarity with centre P

In order to investigate the above direct similarity more fully we now investigate how a Miquel
configuration can be constructed by choosing a point P to serve as a Miquel point and creating a
direct similarity centre P for which the image of triangle ABC is a triangle XYZ so that X, Y, Z
have the property that they are the centres of circles APMN, BPNL, CPLM where L, M, N lie on
BC, CA, AC respectively.

It is convenient, since rotation and dilation about P is to be effected, to choose P to be the origin,
which means altering the co-ordinates of points. Now it is A, B, C and P that are fixed and the
similarity that is specified. What happens is this. Triangle ABC is rotated about P by an angle
to give a triangle A'B'C'. Now dilation is carried out by a scale factor d that has to be adjusted so
that X lies on A'P and is also on the perpendicular bisector of AP. This is necessary for X to be
the centre of a circle through A and P. Y and Z are then fixed by the direct similarity and they are
then centres of circles through B and P and through C and P respectively. It turns out that
although any angle ( ) can be chosen, X, Y, Z do not lie on the required perpendicular

3
bisectors unless d = 1/(2cos). Rotation through a right angle is forbidden in order that X, Y, Z
should be finite points.

We choose the circumcircle of ABC to have radius 1 and equation


(x k)2 + y2 = 1. (3.1)

Then the circumcentre has co-ordinates (k, 0) and P has co-ordinates (0, 0). Points on the circle
can be assigned a real parameter t so that they have co-ordinates
{1/(1 + t2)}(t2(k 1) + (k + 1), 2t) and to keep matters sufficiently general we choose A, B, C to
have parameters a, b, c respectively.

As indicated we now perform the direct similarity with rotation and scale factor d = 1/(2cos),
so that the matrix that represents it has rows ( , s/(1 s2)) and (s/(1 s2), ), where s =
tan.

The resulting co-ordinates of X are {(1/(2(1 s2)(1 + a2)}(a2(k 1)(1 s2) 4as + (k + 1)(1 s2),
2(a2s(k 1) + a(1 s2) + s(k + 1))). Those of Y, Z may be obtained by replacing a by b, c
respectively. It may now be checked that XA = XP, YB = YP, ZC = ZP. Circles centres X, Y, Z
through P now form a Miquel configuration in which pairs of circles meet at points (other than
P) on the sides of the triangle.

4. What happens when P lies on the circumcircle of ABC

What happens when P lies on the circumcircle is that the previous analysis may be used with k =
1. A now has co-ordinates {2/(a2 + 1)}(1, a) with similar expressions for B and C. The
circumcircle ABC has equation
x2 + y2 2x = 0. (4.1)

X has co-ordinates {1/((a2 + 1)(s2 1))}(2as + s2 1, a(s2 1) 2s) with similar expressions for
Y and Z. The circle centre X passing through A has equation
(a2 + 1)(s2 1)(x2 + y2) 2(2as + s2 1)x 2(a(s2 1) 2s)y = 0. (4.2)

The circle centre Y passing through B has a similar equation with b replacing a. These two circles
meet at P and at the point N with co-ordinates
{1/((s2 1)(a2 + 1)(b2 + 1))}(2(ab 1)(1 s2) + 2s(a + b), 2(2s(ab 1) + (a + b)(s2 1)), with
similar expressions for L and M. It is straightforward to show these points lie on the sides of
ABC. Forming the 3x 3 determinant whose rows are the co-ordinates of L, M, N (with third
element 1) we find that it vanishes, showing that LMN is a straight line.

The equation of AX is

4
(a(s2 1) + 2s)x + (2as (s2 1))y 4s = 0. (4.3)

This line meets the circumcircle with Equation (3.1) at the point Q with co-ordinates
{(4s/(s2 + 1)2}(2s, 1 s2). As this is independent of a, lines BY, CZ also pass through Q, which
means that triangles ABC and XYZ are also in perspective.

Circle XYZ has equation


(s2 1)(x2 + y2) + (1 s2)x + 2sy = 0. (4.4)

This circle passes through P and also through O, the centre of the circumcircle of ABC, showing
that circle XYZ is the circle of centres in the Wood [1] configuration based on ABCPQ. It may
also be shown that LMN is the double Wallace-Simson line of P with respect to triangle XYZ
triangle. Since the construction is reversible this demonstrates that every transversal of ABC is
the Double Simson line of a point P on its circumcircle with respect to some triangle XYZ that is
directly similar to ABC with centre of similarity P. See Fig. 2 for a record of these results.

5
A P

Z
X
N
H

O J
M

B C L
Y
Q

Fig. 2

6
Z

A K

N
Y
J

C
B L

Figure 3
5. Paralogic Wood configuration

This is a very natural construction. Start with a triangle ABC and a transversal LMN, with L on
BC etc. Now erect perpendiculars at L, M, N to BC, CA, AB respectively to create by their
intersections in pairs the triangle XYZ.
The ABC and XYZ are directly similar and in perspective with LMN the Desargues axis and the
circles ABC and XYZ meet at the perspector K and the centre of inverse symmetry J. Moreover
the circles are orthogonal, so the triangles ABC and XYZ are paralogic. The paralogic centres
are the orthocentres of the two triangles. See Figure 3.

7
Reference
1. F.E. Wood, Amer. Math. Monthly 36:2 (1929) 67-73.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

8
Article 17
Harmonic ranges in a coaxal system of circles
Christopher J Bradley

1. Introduction

Suppose we are given a circle that cuts the sides BC, CA, AB of a triangle ABC at pairs
of points on each side, say P and L on BC, Q and M on CA and R and N on AB, in each
case in the specified order working around the triangle in an anticlockwise direction.
Now let P' and L' be the harmonic conjugates of P and L with respect to B and C, Q' and
M' the harmonic conjugates of Q and M with respect to C and A and R' and N' the
harmonic conjugates of R and N with respect to A and B. First we prove that P', L', Q',
M', R' and N' always lie on a conic. We then investigate the conditions on the given circle
for this conic also to be a circle.

We define such a pair of circles as a harmonic pair of circles and it turns out that each
such pair belongs to the coaxal system of circles of which the polar circle, centre the
orthocentre H of ABC and the circumcircle, centre the circumcentre O of ABC may be
regarded as defining members. Since the polar circle is real only when ABC is obtuse we
remark that other circles in this coaxal system are the nine-point circle, centre T, the
midpoint of OH, and the orthocentroidal circle based on GH as diameter, where G is the
centroid of ABC. If a harmonic pair S and S' have centres at X and X', then X and X'
divide O and H harmonically. See Fig. 1.

Investigation of the equations of the circles forming a harmonic pair enables us to relate
their equations in a canonical form in terms of those of the polar circle and the
circumcircle, with the position of their centres. Analysis that is first applied to the coaxal
system of circles with centres on the Euler line may be transferred to the circles of any
coaxal system.

If one sets up a more general figure in which conics replace circles and one draws the
lines PN, QL, RM to form a triangle DEF and one draws the lines P'N', Q'L', R'M' to
form a triangle UVW then we prove that triangles ABC, DEF, UVW are mutually in
perspective, and moreover the three Desarguess lines of the three perspectives coincide
and the three perspectors are collinear. This is possibly a known result, but as its proof is
straightforward we include it thereby providing an extension of the results about
harmonic pairs of circles.

1
Areal co-ordinates are used throughout the article.

2. -Harmonic conjugate points lie on a conic

Suppose a conic cuts the sides BC, CA, AB of a triangle in points P, L on BC, Q, M on CA and
R, N on AB, then by Carnots theorem
(BP/PC)(BL/LC)(CQ/QA)(CM/MA)(AR/RB)(AN/NB) = 1. (2.1)

If P' is the harmonic conjugate of P with respect to B and C, and L', Q', M', R', N' are similarly
defined, then we have (BP'/P'C) = (BP/PC) etc. And so from Equation 1.1,
(BP'/P'C)(BL'/L'C)(CQ'/Q'A)(CM'/M'A)(AR'/R'B)(AN'/N'B) = 1. (2.2)

2
Now, if L, M, N and P, Q, R are the feet of two Cevians, then L'M'N' and P'Q'R' are two straight
lines, by the converse of Menelauss theorem, and hence form a degenerate conic. Otherwise the
conditions for the converse of Carnots theorem are satisfied and P', L', Q', M', R', N' lie on a
non-degenerate conic.

The degenerate case occurs in at least one familiar case when the conic is the nine-point circle,
and the harmonic conjugates then lie on two straight lines, one being the line at infinity.
This is perhaps a good place to mention that if we work in the complex field the theorems in this
article are true irrespective of whether the points of intersection of the conic cut the triangle sides
in real points or not.

3. The condition for the harmonic conjugate points to lie on a circle

Suppose that P and L have co-ordinates (0, q, r) and (0, m, n) respectively. Then if we put x = 0
in the equation of the given conic it must reduce to (ry qz)(ny mz) = 0. Since P' and L' are the
harmonic conjugates of P and L, their co-ordinates must be (0, q, r) and (0, m, n) and it
follows that if we put x = 0 in the equation of the harmonically related conic it must reduce to (ry
+ qz)(ny + mz) = 0. Thus under harmonic conjugation rny2 (rm + qn)yz + qmz2 becomes rny2 +
(rm + qn)yz + qmz2 and the change is effected simply by altering the sign of the yz term.

If then the original conic has equation


ux2 + vy2 + wz2 + 2fyz + 2gzx + 2hxy = 0, (3.1)

it follows immediately that the harmonically related conic has equation


ux2 + vy2 + wz2 2fyz 2gzx 2hxy = 0. (3.2)

We now investigate the conditions under which both these equations represent circles. From
Section 2.3.10 of Bradley [1] there must exist a constant k such that the following six equations
hold.
v + w 2f = a2, w + u 2g = b2, u + v 2h = c2
v + w + 2f = ka2, w + u + 2g = kb2, u + v + 2h = kc2. (3.3)

The solutions of Equation (2.3) are


f = a2(k + 1), g = b2(k + 1), h = c2(k + 1),
u = (b2 + c2 a2)( k + 1), v = (c2 + a2 b2)( k + 1), w = (a2 + b2 c2)( k + 1). (3.4)

From Section 2.3.6 of Bradley [1] the co-ordinates of the centre of the conic with Equation (3.1)
are proportional to
(vw gv hw f2 + fg + fh, wu hw fu g2 + gh + gf, uv fu gv h2 + hf + hg).

3
Substituting values from Equation (3.4) we find the x-co-ordinate to be proportional to
2a4k + a2(1 + k)(b2 + c2) (k + 1)(b2 c2)2 (3.5)
with y- and z-co-ordinates found from (3.5) by cyclic change of a, b, c.

Now the equation of the Euler line of triangle ABC is


(b2 c2)(b2 + c2 a2)x + (c2 a2)(c2 + a2 b2)y + (a2 b2)(a2 + b2 c2)z = 0. (3.6)

Substituting the co-ordinates of the centre of the circle into Equation (2.6) we find the centre lies
on this line. Similar analysis holds for the conic with Equation (3.2). It follows that whatever the
value of k the centres of the two circles must lie on the Euler line.

4. The coaxal system formed by harmonic pairs of circles

From Section 3 we know that the equations of a harmonic pair of circles are of the form
( k + 1){(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2}
2(k + 1){a2yz + b2zx + c2xy} = 0, (4.1)
and

( k + 1){(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2}


+ 2(k + 1){a2yz + b2zx + c2xy} = 0, (4.2)
Note that one may obtain Equation (4.2) from Equation (4.1) by replacing k by 1/k.

First let us look at some particular cases. When k = 1 one gets the polar circle with equation
(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2 = 0 (4.3)

from both Equations (4.3) and (4.4). It follows that the polar circle is self-conjugate being its
own harmonic partner. Note that the polar circle is real if and only if triangle ABC is obtuse, so
if one wishes to visualise cases for all real k one must draw ABC as obtuse. Similarly if one puts
k = 1 one gets the circumcircle with equation
2(a2yz + b2zx + c2xy) = 0, (4.4)
again from both Equations (4.1) and (4.2). (The factor 2 is introduced in Equation (4.4) rather
than in Equations (4.7) and (4.8) below.) It follows that the circumcircle is also self-conjugate. If
one puts k = 0, Equation (4.1) becomes the nine-point circle with equation
{(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2} 2{a2yz + b2zx + c2xy} = 0. (4.5)

Equation (4.2) factorizes and reduces to the line pair


x + y + z = 0 and (b2 + c2 a2)x + (c2 + a2 b2)y + (a2 + b2 c2)z = 0. (4.6)

The first of these is the line at infinity and the second line we shall identify very soon.

4
Abbreviating equations (4.3) and (4.4) to read SP = 0 and SC = 0 we see that Equations (4.1) and
(4.2) may be written in the form
(1 k)SP (1 + k)SC = 0 (4.7)
and
(1 1/k)SP (1 + 1/k)SC = 0 (4.8)

respectively. It follows that a pair of circles are harmonic if, and only if, their centres are on the
Euler line and they are in the coaxal system of circles defined by the polar circle and the
circumcircle. For an obtuse-angled triangle ABC, these two circles define an intersecting system
of coaxal circles and the second of Equations (4.6) is therefore their common chord. In the case
of an acute-angled triangle, when working over the real field, one may view the coaxal system as
a non-intersecting system of circles, as in Fig.1, defined by the circumcircle and the line
corresponding to the second of Equations (4.6), which is now the radical axis.

5. The location of the centres of a harmonic range of circles

Circle centres may be computed using the formula from Section 2.3.6 of Bradley [1]. Details of
the calculation, carried out by DERIVE, are omitted. Only the x-co-ordinate of each point needs
to be recorded, since the y- and z-co-ordinates follow by cyclic change of a, b, c. Co-ordinates
are normalized, as is necessary for deductions to be made about the disposition of the centres.

For the polar circle the centre is at H with x-co-ordinate


xP = (a2 + b2 c2)(a2 + c2 b2)/(a + b + c)(b + c a)(c + a b)(a + b c). (5.1)

For the circumcircle the centre is at O with x-co-ordinate


xC = a2(b2 + c2 a2)/(a + b + c)(b + c a)(c + a b)(a + b c). (5.2)

For the circle with Equation (4.1) the x-co-ordinate is


xk = (1/2)(1 k)xP + (1/2)(1 + k)xC. (19)

For the circle with Equation (4.2) the x-co-ordinate is


x(1/k) = (1/2)(1 1/k)xP + (1/2)(1 + 1/k)xC. (20)

Since {xP, xC; xk, x(1/k)} = { 1, 1; k, 1/k} = 1 the centres of any harmonic pair of circles
separate the centre of the polar circle and the centre of the circumcircle harmonically.

Another circle in this coaxal system is the orthocentroidal circle, which is the circle with GH as
diameter. It has parameter k = 1/3. Circles centre G, the centroid, and deL, deLongchamps
point, form a harmonic pair with k = 1/3 and k = 3 respectively.

5
6. Perspectives formed by harmonic pairs of circles

The results of this section do not depend on the conics involved being circles, so we establish the
results in the more general case. The notation used, however, is the same as before, with a conic
meeting sides BC, CA, AB respectively at pairs of points P, L; Q, M, R, N. The harmonic
conjugate points are denoted by P', L', Q', M', R', N' and as we have seen in Section 2 these
points also lie on a conic. Suppose now that lines PN, QL, RM are drawn to form triangle DEF
with D = PN^QL etc. and lines P'N', Q'L', R'M' are drawn to form triangle UVW with U =
P'N'^Q'L' etc.

Result 1 Triangles ABC and DEF are in perspective

Proof
BC^EF is the same as RM^PL = I, say. CA^FD is the same as QM^PN = J, say. AB^DE is the
same as RN^QL = K, say. It follows, by Pascals theorem for the hexagon (RMQLPN) inscribed
in the conic S that IJK is a straight line, which is therefore the Desarguess axis of perspective
for triangles ABC and DEF. Using the notation of Fig. 2 the vertex of perspective is the point Z.

Result 2 Triangles ABC and UVW are in perspective

Proof
Since P'L' is the same line as PL, Q'M' is the same line as QM and R'N' is the same line as RN
the proof of Result 2 follows by the same reasoning as for Result 1, but with the hexagon
(R'M'Q'L'P'N') inscribed in the conic S'. Furthermore the intersections are again the points I, J, K
which is the same Desarguess axis as before. The vertex of perspective is shown as the point Y
in Fig. 2.

Result 3 Triangles DEF and UVW are in perspective

Proof
Since VW^EF = RM^R'M' = I etc., Result 3 is immediate and once again the Desargues axis is
the same line. The vertex of perspective is shown as X in Fig. 2.

Result 4 The perspectors X, Y, Z are collinear

Proof
By using a projective transformation we may map the line IJK to the line at infinity and then the
three triangles map into three triangles homothetic in pairs. The question now reduces to whether
the three homothety centres are collinear. Denote the three pairwise homothetic triangles by T1,

6
T2, T3 and let the homothety ij carrying Ti to Tj have centre Cij. Now 12 followed by 23 leaves
the line joining C12C23 invariant. This composition is 13 so C13 is on this line.

I am grateful to Dr Geoff. Smith of the University of Bath for providing the proof of Result 4 and
for having a number of useful conversations with me concerning the contents of this article.

Reference
1. C.J.Bradley, The Algebra of Geometry, Highperception, Bath 2007.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

7
Article 18
Some Special Circles in a Triangle
Christopher J Bradley

F
A

E
Z M
V
P
N
U Q
Y X
B L W C

Figure

1. Introduction

1
We study a construction in a triangle ABC that assigns to any point P, not on the sides,
circumcircle or altitudes of a triangle, a unique circle passing through P. This is done by finding
three points U, V, W that define the circle and once it has been shown that U, V, W, P are
concyclic, then four other points Q, X, Y, Z are determined with the property that UX, VY, WZ
are concurrent at Q.

The construction is as follows. Choose P and draw the circles BPC, CPA, APB, which we denote
by 1, 2, 3 respectively. Now draw AP, BP, CP to meet 1, 2, 3 respectively at D, E, F. Draw
through D the perpendicular to BC to meet 1 again at U, with V, W defined similarly on
perpendiculars to CA, AB respectively. Then it turns out that U, V, W, P are concyclic and lie on
a circle we denote by P. Now draw AP, BP, CP to meet P at X, Y, Z respectively. Then it may
be proved that UX, VY, WZ are concurrent at a point Q. Furthermore AQ, BQ, CQ intersect BC,
CA, AB respectively at points L, M, N that also lie respectively on DU, EV, FW. Those familiar
with the construction of Hagge circles, see Bradley and Smith [1], will recognize the similarity
that exists between the two constructions, except that Hagge circles define all circles passing
through the orthocentre H, whereas our construction gives one circle (for any given triangle)
passing through all points P not lying on the sides or altitudes of ABC. See the figure above for
an illustration of the construction.

In the analysis that follows we take P to be the centroid G, so all that is proved is that the
construction works in this case. However, the computer geometry software package CABRI II
plus is so accurate that we may be sure that the construction holds generally. However, the
algebra involved for a general point is so formidable (even with the aid of a computer algebra
software package) that we offer this as a most potent excuse for not covering the general case.
We use areal co-ordinates throughout, an account of which is given by Bradley [2, 3].

2. The circles BGC, CGA, AGB and the points D, E, F

The equation of a circle using areal co-ordinates is always of the form

(2.1)

where u, v, w are constants to be determined and a, b, c are the side lengths of ABC. To find the
equation of circle BGC we put the co-ordinates of points B, G, C in Equation (2.1) and get three
equations to determine u, v, w. The values obtained are and the
equation of circle BGC is accordingly

(2.2)

2
The equations of circles CGA, AGB may be obtained by cyclic change of x, y, z and a, b, c. The
equation of AG is y = z and this meets the circle BGC at the point D with co-ordinates D( 3a2,
(a2 + b2 + c2), (a2 + b2 + c2)). Points E, F have co-ordinates that may be obtained from those of D
by cyclic change of x, y, z and a, b, c.

3. Lines through D, E, F perpendicular to BC, CA, AB

Finding perpendicular lines when using areal co-ordinates is tiresome, but the results are known,
see [3], and may therefore be quoted.

If we take a point T with co-ordinates (d, e, f) then the foot of the perpendicular from T on the
line BC has co-ordinates and consequently the equation
of the line perpendicular to BC, x = 0, through the point T has equation

(3.1)

If we now put d = 3a2, e = f = (a2 + b2 + c2), we get the equation of DU, which is
(3.2)

The equations of EV and FW may be written down from Equation (3.2) by cyclic change of x, y,
z and a, b, c.

4. Points U, V, W and the circle through G

The line DU, with Equation (3.2), meets 1 with Equation (2.2) at the point U with co-ordinates
(x, y, z), where

(4.1)

The co-ordinates of the points V, W may be found by cyclic change of x, y, z and a, b, c.

We now substitute the co-ordinates of U, V, W into Equation (2.1) to obtain three equations for
u, v, w. These values are then substituted back in Equation (2.1) to obtain the unpromising
equation of the circle G, which is

. (4.2)

3
Terms in y2, z2, and zx, xy may be obtained by cyclic change of a, b, c of those in x2 and yz.
It may now be checked that this circle passes through G(1, 1, 1).

5. The points X, Y, Z and the linking point Q

The equation of AG is y = z and this meets the circle G with Equation (4.2) at the point X with
co-ordinates (x, y, z), where

(5.1)

The co-ordinates of Y, Z, where BG, CG respectively meet G may be obtained from those of X
by cyclic change of x, y, z and a, b, c.

The equations of the lines UX, VY, WZ are lengthy and no good purpose would be served by
writing them down. However, the algebra computer package DERIVE, which we used
throughout, showed that the three lines meet at the point Q, whose co-ordinates (x, y, z) are
given by
(5.2)

The apparent simplicity of this result is very gratifying. We use the term linking point for Q, as it
links the point G with the points lying on circle G. This it does in two ways. The first has just
been explained as Q is the point of concurrency of UX, VY, WZ. The second way is that the
lines AQ, BQ, CQ meet the lines DU, EV, FW respectively at points L, M, N lying on BC, CA,
AB respectively. To see this, note that the equation of AQ is
(5.3)

and from Equation (3.2) it can be seen that DU meets BC at this point also. Note that Q may lie
outside triangle ABC.

References
1. C. J Bradley & G.C.Smith, Math. Gaz. pp202 -207 July 2007.
2. C. J. Bradley, Challenges in Geometry, Oxford (2005).
3. C. J .Bradley, The Algebra of Geometry, Highperception, Bath (2007).

Flat 4,
Terrill Court,
12/14 Apsley Road,
BRISTOL BS8 2SP

4
Article 19
Circular Perspective
Christopher J Bradley

1. Introduction

If two triangles ABC and PQR are such that AP, BQ, CR are concurrent at a point V then the
triangles are said to be in perspective. The vertex V is sometimes called the perspector. This
relation is obviously a symmetric property. When the triangles are in perspective then the dual
property also holds that L = BC^QR, M = CA^RP, and N = AB^PQ are collinear. The line LMN
is called Desargues axis of perspective and more recently is sometimes called the perspectrix.
The main theorem is that if triangles are in double perspective so that ABC is in perspective, say,
with both triangles PQR and QRP, then they are bound to be in triple perspective, meaning that
triangle ABC is in perspective with triangle RPQ as well. There is also the concept of reverse
perspective when one considers ABC in relation to triangles PRQ, RQP or QPR. For example,
pairs of triangles in the Brocard porism are in perspective with perspector the symmedian point
K and are in triple reverse perspective with the three perspectors lying on the polar line of K.

In this paper we introduce a concept which we call circular perspective. This is a relation
between a pair of triangles ABC and PQR in which the vertices of both triangles possess the
property that vertices of one do not lie on the sides or extended sides of the other. The definition
of circular perspective is as follows: ABC and PQR are in circular perspective if circles BCP,
CAQ and ABR have a common point D. It is by no means obvious that this is a symmetric
relation, but we prove that this is the case. In fact we prove two main theorems:
(i) If ABC is in circular perspective with PQR, then PQR is in circular perspective with
ABC.
(ii) If ABC is in circular perspective with both PQR and QRP then it is in circular
perspective with triangle RPQ.

We conclude our study by investigating an interesting case in the geometry of the triangle,
involving the orthocentroidal circle and the Brocard points, in which triple circular perspective
exists. Note that there is no concept of reverse perspective since, for example, circle CPQ is the
same as circle CQP. The theorem stated in (i) is illustrated in Figure 1.

2. Circular perspective is a symmetrical relation

The proof of (i) is straightforward. If D is the point of concurrence of circles BCP, CAQ and
ABR, then invert with respect to D and the configuration becomes one in which (using stars for
inverted points) P* lies on B*C*, Q* lies on C*A* and R* lies on A*B*. It then follows that

1
circles Q*R*A*, R*P*B*, P*Q*C* are concurrent at the point S*, the Miquel point. The inverse
image S of the point S* is now the point of concurrence of circles QRA, RPB and PQC.

P
D
R

B
C

Figure 1

3. Double circular perspective implies triple circular perspective

Suppose now that the circles BCP, CAQ, ABR meet at D and circles BCQ, CAR, ABP meet at
E. We want to prove that circles BCR, CAP, ABQ meet at a point F. As in Section 2 we may
invert with respect to D, and dropping the stars on invoice points it is sufficient to prove that if P,

2
N
Q, R lie on BC, CA, AB respectively and circles BCQ, CAR, ABP meet at a point E, then circles
BCR, CAP, ABQ meet at a point F. This result is illustrated in Figure 2. We prove this result
using areal co-ordinates with ABC as triangle of reference and supposing P has co-ordinates (0,
p, 1 p), Q has co-ordinates (1 q, 0, q) and R has co-ordinates (r, 1 r, 0).

A E

R
F

B P
C

Figure 2

The equation of any circle using areal co-ordinates is

a2yz + b2zx + c2xy (x + y + z)(ux + vy + wz) = 0. (3.1)

To obtain the equation of circle ABP we insert into Equation (3.1) the co-ordinates of A, B, P in
succession and obtain three equations for u, v, w. The solution is then substituted back into
Equation (3.1). The result is u = v = 0 and w = a2p and consequently the circle ABP has equation

3
a2pz2 + a2(1 p)yz + (b2 a2p)zx + c2xy = 0. (3.2)

The equations of the circles BCQ, CAR may now be written down by cyclic change of x, y, z
and a, b, c. Circles ABP and BCQ meet at the point E with co-ordinates (x, y, z), where

x = a2p(c2p + b2q(1 p)),


y = b2pq(a2p b2(1 q)), (3.3)
z = b2q(c2p + b2q(1 p)).

If we now put in the condition that circle CAR also passes through E we get an expression (equal
to zero) that factorizes and discarding one factor that cannot hold because p and q are real, we
find the condition for concurrence to be
r(1 q)/a2 + p(1 r)/b2 + q(1 p)/c2 = 0. (3.4)

We now consider the circles ABQ, BCR, CAP. The equation of circle ABQ is

b2(1 q)z2 + (a2 + b2(1 q))yz + b2qzx + c2xy = 0. (3.5)

The equations of circles BCR and CAP may now be written down by cyclic change of x, y, z and
a, b, c. Circles ABQ and BCR meet at the point F with co-ordinates (x, y, z), where

x = b2(1 q)(a2(1 r) + b2r(1 q)),


y = b2(1 q)(1 r)(c2(1 r) b2q)), (3.6)
z = c2(1 r)(a2(1 r) + b2r(1 q)).

The condition that F also lies on circle CAP, after similar algebra as before, is again found to be
Equation (3.4), which completes the proof.

When properties (i) and (ii) are taken into account it follows that if ABC and PQR are in double
circular perspective then both ABC and PQR, and PQR and ABC are in triple circular
perspective.

It is also the case in projective geometry that if ABC and PQR are two triangles and X, Y are two
distinct points then if conics BCPXY, CAQXY, ABRXY have a common point Z then we may
say that triangles ABC and PQR are in conical perspective by means of triangle XYZ.
Generalizations of properties (i) and (ii) now hold by relating X and Y to the circular points at
infinity.

4. An example of triple circular perspective

4
Consider the orthocentroidal circle on GH as diameter, where H is the orthocentre of ABC and G
is its centroid. Its equation, see Bradley and Smith [1], is

(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2 a2yz b2zx c2xy = 0. (4.1)

The median AG, with equation y = z, meets this circle at a point we call aH and its co-ordinates
are easily calculated to be aH(a2, b2 + c2 a2, b2 + c2 a2). The points bH and cH are similarly

bH
G

'
cH
aH
H
L
B C

Figure 3

defined and have co-ordinates bH(c2 +a2 b2, b2, c2 + a2 b2) and cH(a2 + b2 c2, a2 + b2 c2,
c2). What we show is the following:

5
Triangles ABC and aHbHcH are in triple circular perspective.

The method of proof is as follows:


(i) We show that circles ABaH, BCbH, CAcH all pass through the Brocard point with
co-ordinates (1/b2, 1/c2, 1/a2).
(ii) We then show that circles ABbH, BCcH, CAaH all pass through the Brocard point '
with co-ordinates (1/c2, 1/a2, 1/b2).
(iii) It then follows by property (ii) that the circles ABcH, BCaH, CAbH all pass through a
fixed point, which we show is the orthocentre H.
(iv) It then follows by property (i) that triangles ABC and aHbHcH are in triple circular
perspective.
(v) As a corollary it then follows that circles BaHcH, CbHaH, AcHbH have a common
point L, circles BaHbH, CbHcH, AcHaH have a common point M and circles
BcHbH, CaHcH, AbHaH have a common point N.

These results are illustrated in Figure 3.

Proof of (i)
The equation of the circle BCbH is

b2x2 + (b2 c2)xy a2yz = 0. (4.2)

It is easily checked that this circle passes through (1/b2, 1/c2, 1/a2). The equations of circles
CAcH, ABaH may be written down from Equation (4.2) by cyclic change of x, y, z and a, b, c.
And it may now be checked that both these circles also pass through .

Proof of (ii)
The equation of the circle BCcH is
c2x2 a2yz + (c2 b2)zx = 0. (4.3)

It is easily checked that this circle passes through '(1/c2, 1/a2, 1/b2). The equations of circles
CAaH, ABbH may be written down from Equation (4.3) by cyclic change of x, y, z and a, b, c.
And it may now be checked that both these circles also pass through '.

Proof of (iii)
It now follows from Section 3 that the circles ABcH, BCaH, CAbH all pass through a fixed
point, but it remains to show that fixed point is H. In fact the equation of circle BCaH is
a2yz + b2zx + c2xy (b2 + c2 a2)x(x + y + z) = 0. (4.4)
It may now be checked that this circle passes through H(1/(b + c a ), 1/(c + a b ), 1/(a2 +
2 2 2 2 2 2

b2 c2)).

6
Proof of (iv)
It now follows by Section 2 that triangles ABC and aHbHcH are in triple circular perspective.

Proof of (v)
As a corollary it follows immediately from (iv) that circles BaHcH, CbHaH, AcHbH have a
common point L, circles BaHbH, CbHcH, AcHaH have a common point M and circles BcHbH,
CaHcH, AbHaH have a common point N.

We do not find the co-ordinates of L, M, N, as it is really only their existence which is


significant. In any case experience suggests that the algebra computer package DERIVE, which
we use, would not be able to find them, the solving of algebraic (rather than numerical)
simultaneous quadratics not being possible unless they are very straightforward as in Section 3.

It may just be added that there is a great deal more to this configuration, once the positions of the
centres of the eighteen circles have been identified. Many similarities exist, but that is another
story.

Dedication
This article is dedicated to my great nephew Matthew Joshua Bradley, it having been composed
during the first two weeks of his life.

Reference
1. C.J.Bradley and G.C.Smith, The locations of triangle centres, Forum. Geom., 6 (2006)
57-70.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

7
Article 20
On the Nine Intersections of Two Cevian Triangles
Christopher J Bradley

1. Introduction

Let DEF and LMN be the Cevian triangles of distinct points S and T, with D, L on BC, E, M on
CA and F, N on AB. It is presumed that S and T do not lie on a sideline of triangle ABC nor on a
line parallel to a sideline through an opposite vertex. We introduce the nine intersections of these
two triangles:
P = EF^MN, Q = FD^NL, R = DE^LM;
X = DE^NL, Y = EF^LM, Z = FD^MN;
U = FD^LM, V = DE^MN, W = EF^NL.

1
See the figure above for an illustration in which S and T are internal points, but the following
results are true whether or not S and T are internal. We have four triangles ABC, PQR, XYZ,
UVW and together with the two Cevian triangles they exhibit a number of properties, some of
which are quite striking.

These properties, which we establish in the following sections using areal (barycentric) co-
ordinates - see [1], [2], are as follows:
(i) A, Q, R are collinear as are B, R, P and C, P, Q;
(ii) BC, EN, YZ are concurrent at a point A', CA, FL, ZX are concurrent at a point B' and
AB, DM, XY are concurrent at a point C';
(iii) BC, MF, VW are concurrent at a point A'', CA, ND, WU are concurrent at a point B'',
and AB, LE, UV are concurrent at a point C'';
(iv) A', B', C' are collinear, A'', B'', C'' are collinear and these two lines are concurrent
with ST at a point O;
(v) A, X, U are collinear as are B, Y, V, and C, Z, W.
(vi) (v) implies that triangles ABC, XYZ, UVW are mutually in perspective having a
common perspector J;
(vii) In fact each of these triangles is also in perspective with triangle PQR. Triangles
ABC and PQR with perspector H (off the page of the figure), triangles PQR and XYZ
with perspector I and triangles PQR and UVW with perspector K;
(viii) The perspectors I, J, K are collinear.

2. Setting the scene

Let the Cevian points S and T have co-ordinates (d, e, f) and (l, m, n) respectively. The equations
of their sides are straightforward to obtain and the co-ordinates of the nine points of intersection
introduced in Section 1 are as follows:

3. Result (i): Triangle ABC is inscribed in triangle PQR

2
Since the y- co-ordinates of Q and R are identical, as are also their z- co-ordinates it follows that
the line QR passes through A. Similarly the line RP passes through B and the line PQ passes
through C.

4. Results (ii) and (iii): Pairs of lines concurrent at points on the sides of triangle ABC

The equation of the line YZ is


. (4.1)

The equation of the line EN is


(4.2)

These meet at the point A' with co-ordinates (0, dm, fl) and A' lies on BC. Similarly ZX and FL
meet at a point B'(dm, 0, en) on CA, and XY and DM meet at a point C'( fl, en, 0) on AB.

Likewise the equation of VW is


(4.3)

The equation of MF is
(4.4)

These meet at the point A'' with co-ordinates (0, el, dn) and A'' lies on BC. Similarly WU and
ND meet at a point B''(el, 0, fm) on CA, and UV and LE meet at a point C''( dn, fm, 0) on AB.

5. Result (iv): Three concurrent lines

Points A', B', C' are collinear on the line with equation
(5.1)

Points A'', B'', C'' are collinear on the line with equation
(5.2)

These meet at the point O with co-ordinates

The equation of ST, the line joining the Cevian points, is


(5.3)

It may now be checked that O lies on the line ST.

3
6. Results (v) and (vi): Three perspectives with the same perspector

The lines AUX, BVY, CWZ have equations


(6.1)
(6.2)
(6.3)

respectively. These lines are concurrent at the point J with co-ordinates

It follows that triangles ABC, UVW, XYZ are mutually in perspective with J as the common
perspector.

7. Result (vii): Triangle PQR is in perspective with all of ABC, UVW, XYZ

First consider the pair of triangles ABC and PQR. The equations of AP, BQ, CR are
(7.1)
(7.2)
(7.3)

respectively. These lines are concurrent at the point H with co-ordinates

Thus triangles ABC and PQR are in perspective with H as perspector. Note that as triangle ABC
is inscribed in triangle PQR (Result (i)), it follows that H is a Cevian point of triangle PQR with
ABC as its Cevian triangle. This point has recently featured in a paper by Pohoata and Yiu [3],
who also show that the corresponding sidelines of three Cevian triangles are concurrent if and
only if the three Cevian points are coconic with the vertices of triangle ABC. They call H the
anticevian point of S and T.

Next consider the pair of triangles PQR and XYZ. The equations of PX, QY, RZ are
(7.4)
(7.5)
(7.6)

respectively. These meet at the point I with co-ordinates (x, y, z), where

4
Thus triangles PQR and XYZ are in perspective with perspector I.

Next consider the pair of triangles PQR and UVW. The equations of PU, QV, RW are
(7.7)
(7.8)
(7.9)

respectively. These meet at the point K whose co-ordinates are (x, y, z), where

Thus triangles PQR and UVW are in perspective with perspector K.

8. Result (viii): The perspectors I, J, K are collinear

It may now be checked that the three perspectors I, J, K lie on the line with equation

. (8.1)

9. An alternative construction for the anticevian point H

As is well known the harmonic conjugates of the feet of the Cevians of a point are collinear on a
transversal associated with the point. The equation of the transversal associated with S(d, e, f) is
x/d + y/e + z/f = 0, (21)
and it is easily checked that this line contains the point
H
In fact H lies on the transversal associated with any point on the circumconic ABCST, and as
shown in [3] the triangle PQR is also invariant for all pairs of Cevian triangles involving pairs of
points lying on this circumconic.
References
1. C. J. Bradley, The Algebra of Geometry, Bath: Highperception (2007).
2. C. J. Bradley, Challenges in Geometry, Oxford: University Press (2005).
3. C. Pohoata & P.Yiu, On a product of two points induced by their Cevian triangles,
Forum Geometricorum, 7 (2007) 169-180.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

5
Article 21
Significant Points on Circles Centre the Circumcentre
Christopher J Bradley

P
X

N
V M
O E
F
W
Z Q
Y C
B L
U

Figure 1

1. Introduction

Given a triangle ABC with circumcentre O and a point P not on its sides or their extensions and
not on the circumcircle, it is shown that one may construct on the circle centre O and radius OP
six significant points. The construction of U, V, W are straightforward enough; they are the

1
second intersections of AP, BP, CP with the circle. The other three points X, Y, Z arise in a
somewhat elaborate fashion. The point X, for example, arises as follows: Join P to the midpoint
L of BC and extend it to D, where L is the midpoint of PD. Then reflect D in the side BC to
produce X. Points Y, Z follow similarly using the midpoints M, N of the sides CA, AB
respectively. Alternatively, and more easily, X is the reflection of P in the perpendicular bisector
of BC. Two very interesting properties arise. First, the lines UX, VY, WZ are concurrent at a
point Q. And secondly, triangle XYZ is inversely similar to triangle ABC. We prove the second
result by establishing two lines of inverse similarity passing through O, with the property that
ABC may be mapped into XYZ by using one of these lines and O as a centre of inverse spiral
symmetry. The first property means that the inverse image T of Q under this transformation must
be such that AD', BE', CF' are concurrent at T, where D', E', F' are the inverse images of U, V, W
respectively. As is well known, the point T may now be used to manufacture a circle passing
through the orthocentre H, known as a Hagge circle, see [1, 2]. The Hagge circle is also obtained
by an inverse similarity of triangle ABC with T the centre of inverse similarity, and it follows
from our analysis that the circle through P and the points on it are related to those on the Hagge
circle by a direct congruence, which is just a rotation of 180o about a point midway between the
centres. We do not repeat any of the work on the derivation of Hagge circles in this paper, but it
is our purpose to point out the connections that exist.

The procedure may, of course, be reversed. Start with a Hagge circle and rotate by 180o in a way
that will be explained and you end up with a circle centre O passing through the isogonal
conjugate of the point that is used to construct the Hagge circle.

The article concludes with a conjecture (verified by CABRI, but without algebraic or geometrical
proof) that other direct congruences not only map the Hagge circle and its points on to other
circles with similar properties (which must obviously be the case), but the new circles
manufactured have additional properties that make them interesting and significant. We believe
this removes the mystique of why circles through the orthocentre are somehow thought to be
special. They are no more special than any other circle once the appropriate conjugation and
direct congruence are identified.

In this article we produce an analysis to prove the assertions made about our construction, using
Cartesian co-ordinates. Care is needed in following the text to refer to the appropriate figure, as
changes of notation from one section to the next are inevitable, so many points being involved.

2. The six points X, Y, Z, U, V, W

Let the circumcircle have equation x2 + y2 = 1, with centre O(0, 0) and suppose P has co-
ordinates (p, 0). For the co-ordinates of A we use the familiar expressions (2a/(1 + a2), (1 a2)/(1

2
+ a2)), and similar for B and C with parameters b and c respectively. In this section and the next
the notation is that of Figures 1 and 2.

L, the midpoint of BC, has co-ordinates (k(b + c), k(1 bc)), where
k = (1 + bc)/{(1 + b2)(1 + c2)}. (2.1)

Since L is the midpoint of PD we may obtain the co-ordinates of D as twice those of L minus
those of P, the result being (x, y), where
x = [2{(1 + bc)(b + c)} p(1 + b2)(1 + c2)]/{(1 + b2)(1 + c2)}, (2.2)
2 2
y = [2(1 + bc)(1 bc)]/ {(1 + b )(1 + c )}, (2.3)

The equation of BC is
(b + c)x + (1 bc)y = (1 + bc). (2.4)

After some algebra the equation of the line perpendicular to BC passing through D is
(1 bc)x (b + c)y + p(1 bc) = 0. (2.5)

These lines meet at the midpoint of DX whose co-ordinates are (x, y), where
x = (1/s)( b2c2p + (1 + bc)(b + c) + 2bcp p)), (2.6)
y = (1/s)((1 bc)(bc + (b + c)p + 1)), (2.7)

where s = (1 + b2)(1 + c2). The co-ordinates of X follow and are found to be (x, y), where
x = (p/s)( b2c2 + b2 + c2 + 4bc 1), (2.8)
y = (2p/s)(b + c)(1 bc). (2.9)

The co-ordinates of Y, Z may be written by cyclic change of a, b, c. These co-ordinates may now
be substituted into the general equation of a circle, x2 + y2 + 2gx + 2fy + t = 0, to provide three
equations for f, g, t, which, with help from DERIVE, gives f = g = 0 and t = p2, as required for
the circle XYZ to have centre O and to pass through P.

The equation of AP is
(1 a2)x + ((1 + a2)p 2a)y (1 a2)p = 0. (2.10)

This meets the circle x2 + y2 = p2 again at the point U with co-ordinates (x, y), where
x = (1/q)(p{a4(1 p2) + 4a3p 2a2(p2 + 3) + 4ap p2 + 1}), (2.11)
y = (2/q)(p(1 a2)(a2p 2a + p)), (2.12)
and
q = a4(1 + p2) 4a3p + 2a2(1 + p2) 4ap + p2 + 1. (2.13)

The co-ordinates of V and W may be written by replacing a by b and c respectively.

3
3. The point Q and the indirect similarity

We now show that the lines UX, VY, WZ are concurrent at a point Q. The equation of the line
UX turns out to be
[a2{b(cp 1) c p} + 2a(1 bc) + b(1 + cp) + c p]x + [a2{b(c + p) + cp 1} 2a(b + c) +
b(p c) + cp + 1]y + [a2{b(cp + 1) + c p} + 2a(1 bc) + b(cp 1) c p]p = 0. (3.1)

The equations of VY, WZ may be written down by cyclic change of a, b, c.

These lines are concurrent at the point Q with co-ordinates (x, y), where
x = (1/r){p(a2(b2(c2(p2 3) + p2 + 1) + 4b(c p) + c2(p2 + 1) 4cp + p2 + 1) + 4a(b2(c p) +
b(c2 2cp + 1) c(cp 1)) + b2(c2(p2 + 1) 4cp + p2 + 1) + 4bc(1 cp) + c2(p2 + 1) + p2 3)},
(3.2)
2 2 2 2 2 2
y = (2/r){p(a (b (c p 2c + p) + 2bc(p c) + p(c 1)) 2a(b c(c p) + bp(1 c ) + cp 1) +
b p(c2 1) + 2b(1 cp) c2p + 2c p)},
2
(3.3)
and
r = (1 p2)(1 + a2)(1 + b2)(1 + c2). (3.4)

Two triangles ABC and XYZ are similar when their angles are equal and this is so if their
corresponding sides are in fixed ratio, that ratio being an enlargement (or reduction) about a
given point. When they are indirectly similar it is always the case that there are two axes
perpendicular to one another and passing through a fixed point such that when ABC is reflected
in one of these axes through the fixed point and enlarged (or reduced) by a fixed amount through
the fixed point, it is mapped on to triangle XYZ. The axes are called the double lines of inverse
similarity and the fixed point is called the centre of inverse similarity. Either line of symmetry
may be used, but in one case the enlargement factor is positive, and in the other case it is
negative, implying that a rotation of 180o is also involved.

What we prove now is that the triangles ABC and XYZ are indirectly similar, and we do this by
establishing the double lines of inverse similarity and the centre of inverse similarity. The latter
is, in fact, the circumcentre O. The axes are shown in Figure 2 as the lines m and n passing
through O. In the figure the axis labelled m is used. The reflection through m takes triangle ABC
into triangle A'B'C' and then there is a reduction (since P is inside ABC in the case we have
drawn) by a factor OX/OA' taking A' to X. Since the line m turns out to be symmetric in a, b, c
and OX/OA' = p, this shows that B' is taken to Y and C' to Z, thereby establishing the similarity.

4
m
A A'

P
X
N M
V H
F E'
O E
Q
F' n
C' Z W
T Y C
L
B
U B'
S

D'
D

Figure 2

Suppose the equation of one of the axes of inverse symmetry passes through O and has equation
y = mx. The equation of the line through A perpendicular to this line is
(1 + a2)x + m(1 + a2)y + a2m 2a m = 0. (3.6)

These lines meet at the point with co-ordinates (x, y) where


x = (2a + m a2m)/{(1 + a2)(1 + m2)}, (3.7)
y = m(2a + m a2m)/{(1 + a2)(1 + m2)}. (3.8)

The co-ordinates of A', the reflection of A in y = mx, can now be obtained and are (x, y), where
x = 2(m + a)(1 am)/{(1 + a2)(1 + m2)}, (3.9)
2 2
y = ((a 1) + m(a + 1))((a + 1) m(a 1))/{(1 + a )(1 + m )}. (3.10)

We now determine the possible values of m if A', X, O are collinear. The condition for this is
x1y2 = x2y1, where (x1, y1) are the co-ordinates of A' and (x2, y2) are the co-ordinates of X. This
results in a quadratic equation for m whose solutions are, say, m and n, where mn = 1 and

5
m = (abc + bc + ca + ab a b c 1)/(abc bc ca ab a b c + 1). (3.11)

The enlargement (reduction factor) is obviously p (supposed positive without loss of generality).
As the values of m and the enlargement factor are symmetric in terms of a, b, c and independent
of them respectively, the proof of the indirect similarity is complete.

4. Consequences and the connection with a circle through H of radius p

Now it is well known that a Hagge circle and its key points arise from an indirect similarity of
the circumcircle, A, B, C and three other key points on the circumcircle, so it is a challenge to
discover a direct similarity between the circle centre O through P, that we have so far been
involved with, and a Hagge circle. This proves to be possible and depends on the remarkable
properties of the point T defined to be the inverse image of Q in the indirect similarity described
in Section 3. See Figure 3 for an illustration of what follows.

Since T is the inverse image of Q, which we know is the point of concurrence of UX, VY, WZ, it
follows that if we draw AT, BT, CT to meet the circumcircle at D', E', F' respectively, then not
only is ABC indirectly similar to XYZ, but triangle D'E'F' is similar to triangle UVW. If
therefore we create the Hagge circle generated by T, by reflecting D', E', F' in BC, CA, AB
respectively to get points U', V', W', then triangle U'V'W' is both similar to triangle D'E'F', but
then also to triangle UVW. Furthermore from the theory of Hagge circles it follows that if AH,
BH, CH meet circle U'V'W' at X', Y', Z' respectively, then triangle X'Y'Z' is similar to triangle
ABC and hence to triangle XYZ. Also from the theory of Hagge circles U'X', V'Y', W'Z' meet at
T. This is part of what is needed to show that circle U'V'W'X'Y'Z' is directly similar to circle
UVWXYZ.

The second property of T is that it is the isogonal conjugate of P. This is best proved by drawing
AT and AP to meet the circumcircle at a pair of points whose displacement vector is parallel to
BC and similarly for the other vertices. We are now very much in business, because (again
quoting the theory of Hagge circles) the figure POO'H must be a parallelogram, where O' is the
centre of the Hagge circle. All is now clear, because the direct similarity between U'V'W'X'Y'Z'
and UVWXYZ must be a 180o rotation about the midpoint R of OO'. The conclusive third
property of T is that it is not only the inverse image of Q, but that R is the midpoint of QT.

6
m
m'

A A'

P X
M
N U'
H E'
V O E
F Y' R Q W n
F' Z'
n'
C' W' Z T Y C
O'
B L
V' B'
U
S
X'

D'
D

Figure 3

The whole argument may be reversed. Create a Hagge circle, centre O', from a point T, perform
an 180o rotation about the midpoint of OO' and you get a circle centre O of the same radius, and
the remarkable thing is that this circle passes through P = Tg, the isogonal conjugate of T and, of
course, TgOO'H is a parallelogram. It is also the case that R is the midpoint of UU', VV', WW',
XX', YY', and ZZ'.

Further features in Figure 3 are S, the point on the circumcircle where DD', EE', FF' appear to
meet, the double lines of inverse symmetry m', n' of the indirect similarity between ABC and
X'Y'Z', which are parallel to the axes m, n respectively. Indeed if one reflects ABC in n' to obtain
triangle A''B''C'' it will be found that A''X', B''Y', C''Z' all pass through T. And finally the figure
illustrates the fact that OQO'T is also a parallelogram.

7
What is needed in conclusion is a proof, which we now provide that the three properties leading
to the position of T result in the same values for its co-ordinates. On the way we also find the co-
ordinates of O'. The dilemma facing author and reader is that the algebra involved becomes
technically very involved, even more so than what has preceded; so anyone checking the details
will need an algebra computer package, such as DERIVE, which is the one we used.

In logical order we first find the co-ordinates of the point T, when defined as the pre-image of
the point Q in the indirect similarity. The co-ordinates of Q are given in Equations (3.3) and
(3.4), but as these are very lengthy expressions, for the time being we call them (e, f). We want
the reflection of (e, f) in the line y = mx, where m is given by Equation (3.11). The line
perpendicular to y = mx through (e, f) has equation x + my = e + mf. This meets y = mx at the
point ((e + mf)/(1 + m2), m(e + mf)/(1 + m2)). The reflection of Q has co-ordinates twice these
minus those of Q. After dividing by p and inserting the values of e, f and m, this gives the
following expressions for the co-ordinates of T, which are (x, y) where

x = (2/r)(a2(b2(c2p + c(1 p2) p) b(c2(p2 1) + 2cp p2 1) c2p + c(p2 + 1) p)


a(b2(c2(p2 1) + 2cp p2 1) + 2bp(c2 2cp + 1) c2(p2 + 1) + 2cp + p2 1) b2(c2p c(p2 + 1)
+ p) + b(c2(p2 + 1) 2cp p2 + 1) c2p + c(1 p2) + p), (4.1)

y = (1/r)(a2(b2(c2(p2 3) + 4cp p2 1) + 4bcp(c p) (c2 1)(p2 + 1)) + 4ap(b2c(c p)


+ bp(1 c2) + cp 1) + b2(1 c2)(p2 + 1) + 4bp(cp 1) + c2(p2 +1) 4cp p2 + 3), (4.2)

and where r is given by Equation (3.4).

Next we work out the isogonal conjugate of P with respect to triangle ABC and show that it
coincides with T. The line AP has equation

(a2 1)x (a2p 2a + p)y p(a2 1) = 0. (4.3)

This meets the circumcircle at a point D'' with co-ordinates


(2(a2p a(p2 + 1) + p), (1 p2)(a2 1)) / (a2 (p2 + 1) 4ap + p2 + 1). (4.4)

The line through D'' parallel to BC meets the circumcircle again at the point D'. When the co-
ordinates of A, T and D' are entered into a determinant with 1 in the last column of each row, the
value of this determinant is zero. Therefore the three points are collinear. Cyclic change shows
that B, T, E' and C, T, F' are also collinear and hence T is the isogonal conjugate of P.

8
A

F'
X P

F U'
O E
H
V Q Y E'
Z'
Z Y' R
Pg V'
O'
U
B C
W'
X'

D' S

Figure 4

It follows that a Hagge circle generated by T can be drawn, using the points D', E', F' and if its
centre is denoted by O', we know from the theory of Hagge circles that HTgOO' is a
parallelogram. It follows, since Tg = P has co-ordinates (p, 0) that O' has co-ordinates
(2a/(1 + a2) + 2b/(1 + b2) + 2c/(1 + c2) p, (1 a2)/(1 + a2) + (1 b2) /(1 + b2) + (1 c2)/(1 + c2)).
(4.5)

Our proposition that the direct similarity between the circle, centre O through P, and the Hagge
circle, centre O', is correct, is confirmed if the 180o rotation of Q about R, the mid-point of OO',
takes Q to T. And indeed the co-ordinates of O' minus the co-ordinates of Q do coincide with
those of T, given in Equations (4.1) and (4.2).

9
5. Further Observations

It is also the case that O' is the centre of circle DEF and that the isogonal conjugate of Q with
respect to triangle DEF lies on the Hagge circle centre O'. The circle DEF is therefore directly
similar to circle ABC and the circle centre O through P is the Hagge circle of Q with respect to
triangle DEF. See Figure 4, where the full symmetry of the construction is finally revealed.

We observe that DD', EE', FF' do concur at a point S on the circumcircle. Its co-ordinates are too
complicated to record.

Another rather curious result, as David Monk [3] pointed out, is that the centroid of triangle TPQ
coincides with that of ABC. Using P to mean the vector OP etc., the proof of this is that, since T
+ Q = 2R = O', it follows that T + P + Q = O' + P = H = A + B + C.

6. Conjecture supported by Cabri

We conclude with a conjecture. As we have seen circles centre O carry triangles that are directly
similar to Hagge circles by 180o rotation about the midpoint of the line OO', where O' is the
centre of the Hagge circle.

We now conjecture that any circle and the triangles produced on it by a direct congruence of a
Hagge circle and its triangles have additional properties that make the matter interesting and
significant. We now refer to Figure 5 and the points have meanings as attached to this diagram.
As can be seen the point of rotation of the Hagge circle R is chosen arbitrarily and the angle of
rotation is also arbitrary. Cabri is so accurate that we have no doubt the conjectures we describe
are true. We have not proved them as it seems that algebra is not the medium for doing so, and it
seems unlikely that we could put through a proof algebraically anyway.

10
Angle of
rotation =
84.6405888197

F'

F
C'
X' E'
V' U
O Z
Z'
K
V E
P' P
O' H
U' Q
Y S
R W' X C
B
B'
Y' S'
W
A'
D'
D

N
Figure 5

In the Figure the point P generates the Hagge circle in the usual way with U, V, W the reflections
of D, E, F in the lines BC, CA, AB respectively, where D, E, F are the intersections of AP, BP,
CP with the circumcircle. X, Y, Z are the points on AH. BH, CH lying on circle UVWH. Q is the
centre of the Hagge circle and S is the opposite end of the diameter HQ. The direct congruence
of the Hagge circle by rotation though R is the circle, centre O', with images U', V', W', X', Y',
Z'. Also S' is the image of S. In the case of the circle centre O, the fourth vertex K of the
parallelogram O'QHK is the isogonal conjugate of P. The conjecture is that this point K is now a
conjugate of P in the sense that if AK, BK, CK meet the circumcircle of ABC at D', E', F'
respectively then if L, M, N are the intersections of DD', EE', FF' with BC, CA, AB respectively,
L, M, N are collinear. (In the case of the circles centre O the conjugate is the isogonal conjugate

11
and LMN is the line at infinity.) Finally if S' is the image of S under the direct congruence, then
X', Y', Z' are now reflections of the point S' in axes (dotted in Figure 5), that make angles
86.405... o with lines parallel to the altitudes of triangle ABC. (That part of the conjecture is
obvious.) In Figure 5 the axis through P of indirect similarity relating ABC and XYZ is shown.
The conjugation and the role of the point S at the other end of the diameter to H in the Hagge
circle are the matters that require further investigation, but which we do not intend to pursue
further.

References
1. K. Hagge, Zeitschrift fr Math. Unterricht, 38 (1907) 257-269. (German)
2. Christopher Bradley and Geoff C. Smith, On a construction of Hagge, Forum
Geometricorum (7) 231247(2007).
3. D. Monk, private communication.

Flat 4
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

12
ARTICLE 22

SIX POINTS ON A CIRCLE

Christopher J Bradley

1. Introduction

A construction arising from a triangle ABC and a variable point P is described. In general it leads
to a conic passing through pairs of points on the sides of ABC, but for one, and only one,
particular choice of P the conic is a circle, whatever the shape of the triangle. It is shown that the
special choice of P is the image of a 180o rotation of the isotomic conjugate of the symmedian

1
point (also known as the third Brocard point or X76) about the midpoint of the two Brocard
points. The circle bears distinct resemblance to the triplicate ratio circle.

The general construction is to draw AP, BP, CP to meet BC, CA, AB at points R, S, T
respectively. From R on BC draw lines parallel to BA and CA to meet CA and BA respectively
at E' and F respectively. From S on CA draw lines parallel to CB and AB to meet AB and CB
respectively at F' and D respectively. From T on AB draw lines parallel to AC and BC to meet
BC and AC respectively at D' and E respectively. It then transpires that D', E', F' are the feet of
Cevians through a point Q' and that D, E, F are the feet of Cevians through a point Q. A corollary
of Carnots theorem then implies that D, E, F, D', E', F' lie on a conic. As stated above this conic
is a circle if, and only if, P is chosen to be the image of a 180o rotation of the isotomic conjugate
of the symmedian point about the midpoint of the two Brocard points.

Subsidiary results are (i) that triangle DEF and D'E'F' have the same area, (ii) that if the triangle
A'B'C' is constructed by extending the sides DF', FE and E'D' as shown in the Figure, then
triangles ABC and A'B'C' are in perspective and (iii) if P is chosen at K, the symmedian point,
then the points Q, Q' are the Brocard points and the points where AK, BK, CK meet BC, CA, AB
respectively are the points where Brocards ellipse touches the sides of ABC. As far as (ii) is
concerned there are in all eight such perspectives arising from interchange of primed and
unprimed letters. A similar property holds for the triplicate ratio circle.

Areal co-ordinates are used throughout and in dealing with circles results established in Bradley
[1] are used without proof.

2. The construction

Suppose that P has co-ordinates (l, m, n), then the point R where AP meets BC has co-ordinates
(0, m, n). Lines through a point on BC parallel to AB have equation z k(x + y + z) = 0 and for
the particular line through R we must set k = n/(m + n). The line therefore has equation nx + ny
mz = 0 and this meets CA, y = 0, at the point E'(m, 0, n). Similarly the line through R parallel to
CA meets AB at F(n, m, 0). The remaining points may now be obtained by cyclic change and
have co-ordinates F'(l, n, 0), D'(0, m, l), D(0, l, n), E(l, 0, m).

Note that AF/FB = CE'/E'A, BD/DC = AF'/F'B and CE/EA = BD'/D'C, equations that should be
compared to those pertaining in the construction of the triplicate ratio circle.

The equation of the conic S passing through the six points D, E, F, D', E', F' may now be
obtained and is
mnx2 + nly2 + lmz2 (l2 + mn)yz (m2 + nl)zx (n2 + lm)xy = 0. (2.1)

2
The working is not given as the result may be checked by substitution.

It may also be checked that AD', BE', CF' are concurrent at the Cevian point Q'(1/n, 1/l, 1/m) and
that AD, BE, CF are concurrent at the Cevian point Q(1/m, 1/n, 1/l). Result (i) of the introduction
is generally true, each of the triangles DEF, D'E'F' having area
2lmn[ABC]/{(m + n)(n + l)(l + m)}, (2.2)

where [ABC] is the area of ABC. It is also easy to show that DQ divides EF in the same ratio as
F'Q' divides D'E', that EQ divides FD in the same ratio as D'Q' divides E'F' and that FQ divides
DE in the same ratio as E'Q' divides F'D'. Thus Q is situated in triangle DEF as Q' is situated in
triangle F'E'D'.

3. The circle
From Equation (2.1) and [1] the equations to be satisfied for S to be a circle are
l2 + mn + nl + lm = a2,
m2 + mn + nl + lm = b2, (3.1)
n2 + mn + nl + lm = c2.

the solutions of which are proportional to


l = b2c2 + c2a2 + a2b2,
m = b2c2 c2a2 + a2b2, (3.2)
n = b2c2 + c2a2 a2b2.

The x-co-ordinate of the centre of the circle may be calculated using the formula in [1] and is
a2(a4(b4 4b2c2 + c4) a2(b2 + c2)(b4 5b2c2 + c4) + b2c2(b4 4b2c2 + c4)).
Note that the co-ordinates of P are proportional to (1/b2 + 1/c2 1/a2, 1/c2 + 1/a2 1/b2, 1/a2 +
1/b2 1/c2), the midpoint M of the two Brocard points is (1/b2 + 1/c2, 1/c2 + 1/a2, 1/a2 + 1/b2),
the isotomic conjugate Kt of the symmedian point are (1/a2, 1/b2, 1/c2). These are normalized to
the same amount and P = 2M Kt, and hence P is the image of a 180o rotation of Kt about M.

The equation of the line KtP is


(x y)/c2 + (z x)/b2 + (y z)/a2 = 0, (3.3)
and evidently this line contains the centroid G(1, 1, 1).

In fact, as David Monk informed me, P may alternatively be obtained from Kt by means of a
homothety about G with ratio 2. He also pointed out this means that Kt is the Brocard midpoint
of the antimedial triangle.

3
It is also easy to show that D divides EF in the same ratio as F'' divides D'E', that E divides
FD in the same ratio as D'' divides E'F' and that F divides DE in the same ratio as E''
divides F'D'. Thus is situated in triangle DEF as ' is situated in triangle F'E'D'.

4. The perspectives

Incredibly there are eight different perspectives, one each corresponding to the following
hexagons F'FDD'EE', FF'DD'EE' and others formed by exchanging the roles of D and D' and of
E and E'. Only the first of these is shown in the Figure. It corresponds to the ordering l > m > n
or in the case of the circle of a > b > c. There are eight orderings of the values of l, m, n
corresponding to the eight hexagons. If two of these magnitudes are equal, then one of the points
is at a midpoint of BC, CA or AB, but it does not invalidate the analysis or the conclusion. If all
three of l, m, n are equal the hexagon degenerates and in that case there is nothing of interest. We
give the analysis only for the hexagon F'FDD'EE'. However, in the subsequent Figure we show
the eight vertices of perspective, as the large squares, where three lines intersect,

4
The equation of F'E' is
nx ly mz = 0. (4.1)
The equation of DF is
mx ny + lz = 0. (4.2)
The equation of ED' is
m2x + l2y lmz = 0. (4.3)

The co-ordinates of the vertices of the triangle formed by these three lines are A'(l(mn l2),
2lm2, m(l2 + mn)), B'(2l2m, m(nl m2), l(nl + m2), C'(l2 + mn, m2 + nl, n2 lm). The three lines
AA', BB', CC' meet at the point with co-ordinates (1/(m2 + nl), 1/(l2 + mn), 1/2lm), which is the
required vertex of perspective, marked as J in the diagram. The Cabri diagram shows that the
eight vertices of perspective themselves form figures that are in multiple perspective from the
vertices and from points on the sides of ABC. Cabri also indicates that three of the vertices of
perspective are collinear with the points Q and Q'.

5
5. When P lies at the symmedian point

In this case l = a2, m = b2, n = c2 and Q and Q' become the Brocard points and ' respectively.
The Figure shows that the seven point circle may be constructed if anything in a simpler fashion
than the standard method using the triplicate ratio circle.

Besides the circumcentre O, the symmedian point K and the Brocard points and ', the points
L, M, N are the intersections of pairs of Cevians of the points and '. To be precise L =
BE^CF', M = CF^AD' and N = AD^BE'.

Analytic details are left to the reader who may check all these points lie on the seven point circle
whose equation is
b2c2x2 + c2a2y2 + a2b2 z2 = a4yz + b4zx + c4xy. (5.1)

6
Reference
1. C.J.Bradley, The Algebra of Geometry, Highperception, Bath, 2007.

Flat 4,
Terrill Court,
12-14 Apsley Road
BRISTOL BS8 2SP.

7
The Symmedian point and the Polar Line
C J Bradley
ARTICLE 23

T D

A
N S
W M
V
K
B
U C
X

F
L
E

R
1. Introduction

To construct the figure above start with a triangle ABC and its symmedian point K. Next
draw AK, BK, CK to meet the circumcircle at L, M, N respectively. Then draw the
tangents at L, M, N to creating the triangle DEF, where D is the intersection of the
tangents at M and N etc. Next draw the tangents at A, B, C to creating the triangle RST,
where R is the intersection of the tangents at B and C. The next step is to draw BF and CE
to intersect at U, with V, W similarly defined. The following results now hold:

(i) D, A, K, U, L, R are collinear, as are E, B, K, V, M, S and F, C, K, W, N, T;


(ii) The six points D, E, F, R, S, T lie on a conic ;
(iii)The following lines meet at X: TAS, NM, WV, BC, ELF and the tangents to at D
and R; the following lines meet at Y: RBT, LN, UW, CA, FMD and the tangents to
E and S; and the following lines meet at Z: SCR, LM, UV, AB, END and the

1
The Symmedian point and the Polar Line
C J Bradley
tangents to at F and T.
(iv) X, Y, Z are collinear and lie on the polar line of K with respect to both and .

Note also that there are two triangles ABC and LMN inscribed in and two triangles RST
and DEF inscribed in . Two related porisms are thus created, the first being the Brocard
porism with a circle as circumconic and the Brocard ellipse (not drawn) as inconic, the
second being a related Brocard porism with the circle as inconic and the conic as
circumconic. K is the symmedian point of the triangles in the first porism, and Gergonnes
point in the second porism. Lines through K drawn from a vertex of any triangle in either
porism meet the opposite side of that triangle at the point of contact of that side with the
inconic. See the figure in which L, M, N are such points for triangles in the second porism.
See Bradley and Smith [1] for an account of the Brocard porism.

In subsequent sections we prove these results using areal co-ordinates throughout. For a
description of these and how to use these see Bradley [2].

2. The points L, M, N

We take ABC to be triangle of reference. K has co-ordinates (a2, b2, c2). The equation of
AK is
b2z = c2y. (2.1)

This meets the circumcircle , with equation

a2yz + b2zx + c2xy = 0, (2.2)

at the point L with co-ordinates (a2, 2b2, 2c2). Similarly M has co-ordinates (2a2, b2, 2c2)
and N has co-ordinates (2a2, 2b2, c2).

3. The points R, S, T, D, E, F

The equation of the tangent at A is


b2z + c2y = 0. (3.1)

and the equations of the tangents at B and C may be obtained by cyclic change of a, b, c
and x, y, z. The tangents at B and C meet at the point R with co-ordinates (a2, b2, c2).
Similarly S has co-ordinates (a2, b2, c2) and T has co-ordinates (a2, b2, c2). Note that R
lies on AK with Equation (1). Similarly S, T lie on BK, CK respectively.

The equation of the tangent at L (a2, 2b2, 2c2) is

4x/a2 + y2/b2 + z/c2 = 0. (3.2)

The tangents at M, N follow by cyclic change of a, b, c and x, y, z. These two tangents


meet at the point D with co-ordinates (5a2, b2, c2). Similarly the point E has co-
2
The Symmedian point and the Polar Line
C J Bradley
ordinates ( a2, 5b2, c2) and the point F has co-ordinates ( a2, b2, 5c2). Note that D, E,
F lie on AK, BK, CK respectively.

To determine the equation of the conic passing through the six points D, E, F, R, S, T
one forms a 6 x 6 determinant with x2, y2, z2, yz, zx, xy as the elements of row 1 and then
in the next 5 rows one puts the values of these entries for 5 of the points. On putting this
determinant equal to zero, the equation of the conic through the 5 points emerges. One may
then check that it passes through the 6th point. In this case we find that has equation

x2/a4 + y2/b4 + z2/c4 + 3yz/(b2c2) + 3zx/(c2a2) + 3xy/(a2b2) = 0. (3.3)

4. The points X, Y, Z, the polar line and the line MN

The line EF has Equation (3.2), as it coincides with the tangent at L. Similarly the line TS
has Equation (3.1), as it coincides with the tangent at A. We define X as the intersection of
these two lines, which therefore has co-ordinates (0, b2, c2). Similarly Y has co-ordinates
( a2, 0, c2) and Z has co-ordinates (a2, b2, 0).

It is clear that the line XYZ has equation


x/a2 + y/b2 + z/c2 = 0. (4.1)

Working out the polar of a point with respect to a conic involves the same algebra as
working out a tangent, so it is now possible to check that XYZ is the polar of K with
respect to both and .

The co-ordinates of M and N are given in Section 2. The equation of the line MN is
x/a2 = 2y/b2 + 2z/c2. (4.2)

Clearly MN passes through X.

5. The points U, V, W and the lines through X

The equation of the line EC is


5b2x + a2y = 0, (5.1)

and the equation of the line FB is


5c2x + a2z = 0. (5.2)

These two lines meet at the point U with co-ordinates ( a2, 5b2, 5c2). Similarly V and W
have co-ordinates (5a2, b2, 5c2) and (5a2, 5b2, c2) respectively.

The line VW therefore has equation


4x/a2 = 5y/b2 + 5z/c2. (5.3)

3
The Symmedian point and the Polar Line
C J Bradley
Again it is clear that VW passes through X. Note also that the tangent at L, with Equation
(3.2), also passes through X.

Finally the tangent at D to has equation


2x/a2 + 5y/b2 + 5z/c2 = 0. (5.4)

This line evidently passes through X. Since D and R lie on a line through K and the tangent
at D passes through X, it follows that the tangent at R also passes through X.

References
1. C. J. Bradley & G. C. Smith, accepted for publication.
2. C. J. Bradley, Challenges in Geometry, Oxford, 2005.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
ARTICLE 24
The Thirteen point Circle

Christopher J Bradley

A1

B2 C1

N2
M2
L2 M P
L
O
K B3 R
Q
C3 M1 '
N1 N L
1
A3
B B1 C2 C

A2

1. Introduction

In this article we give an account of the properties of the coaxal system of circles passing
through the two Brocard points and ' and having the Brocard axis as line of centres. This
coaxal system contains four circles of particular importance, the seven point circle S and the
three circles A', B' and C', where A, B and C are the vertices of the triangle. The
circumcircle is not a member of this coaxal system, nor is the coaxal system orthogonal to the
coaxal system containing the Apollonius circles.

We now introduce our notation. The centres of the four circles just mentioned are (i) the
midpoint of OK, where O is the circumcentre and K the symmedian point, (ii) P, (iii) Q, and (iv)
R. The circle A' meets AB at A1 and CA at A2. The circle B' meets BC at B1 and AB at
B2. The circle C' meets CA at C1 and BC at C2. A1 meets S at N1, B1 meets S at L1, C1

1
meets S at M1, A2' meets S at M2, B2' meets S at N2 and C2' meets S at L2. These additional
points on S convert it from being the seven point circle to the thirteen point circle of the title.

We establish the following results:


(i) P, Q, R lie on the Brocard axis;
(ii) BCB1 C2 meets B2C1 at a point A3 on the Brocard axis, CAC1A2 meets C2A1 at a point
B3 on the Brocard axis and ABA1B2 meets A2B1 at a point C3 on the Brocard axis;
(iii) A1B1C1 is a straight line passing through ' and A2B2C2 is a straight line passing
through and these two lines intersect at K, the symmedian point;
(iv) M2 and N1' pass through A3, N2 and L1' pass through B3, and L2 and M1'
pass through C3.

These results are illustrated in the Figure generated by CABRI software and the computer algebra
package DERIVE was used to check the algebra for which areal co-ordinates are used
throughout.

2. The seven point circle and the Brocard axis

The Brocard axis passes through O and K and putting (x, y, z) and their co-ordinates as rows of a
determinant we find its equation to be
=0 (2.1)

In areal co-ordinates the equation of any circle may be put in the form
, (2.2)

where u, v, w are constants to be determined, see Bradley [1, 2]. Putting in the co-ordinates of
K(a2, b2. c2), (1/b2, 1/c2, 1/a2), '(1/c2, 1/a2, 1/b2) we obtain three equations to determine u, v, w
and after some simplifications we obtain the equation of the seven point circle in the form
. (2.3)

3. The three circles and their intersections with the sides of the triangle

We obtain the equations of the circles A', B' and C' using the same method and their
equations turn out to be
= 0, (3.1)
(3 .2)

(3.3)

2
respectively. The intersections of these circles with the sides of ABC are now easily found and
are:
A1(a2(a2 b2), b2(a2 c2) , 0), A2(a2(a2 c2), 0, c2(a2 b2)), B1(0, b2(b2 c2), c2(b2 a2)),
B2(a2(b2 c2), b2(b2 a2), 0), C1(a2(c2 b2), 0, c2(c2 a2)), C2(0, b2(c2 a2), c2(c2 b2))

Since the perpendicular bisector of ' is the Brocard axis it is clear that the centres P, Q, R of
the circles A', B' and C' all lie on the Brocard axis.

4. The lines A1B1C1 and A2B2C2

Using the usual determinantal method the equation of the line A1B1C1 is
(4.1)
2 2 2 2 2 2
and it may be checked that K(a , b , c ) and '(1/c , 1/a , 1/b ) both lie on this line. Similarly the
equation of the line A2B2C2 is
(4.2)
2 2 2 2 2 2
And it may be checked that K(a , b , c ) and (1/b , 1/c , 1/a ) both lie on this line. Note that the
two lines intersect at K on the Brocard axis.

5. The points A3, B3, C3

The equation of C1B2 is


, (5.1)

and the equation of B1C2 is x = 0. These meet at the point A3(0, b2(a2 b2), c2(a2 c2)), a point
that clearly lies on the Brocard axis with Equation (2.1). Similarly B3 and C3 have co-ordinates
B3(a2(b2 a2), 0, c2(b2 c2)) and C3(a2(c2 a2), b2(c2 b2), 0) both of which also lie on the
Brocard axis.

6. The six points L1, L2, M1, M2, N1, N2

For the record L = B^C', M = C^ A', N = A^B' have co-ordinates L(a2, c2, b2), M(c2,
b2, a2), N(b2, a2, c2) and together with O, K, , ' form the seven points of the seven point
(Brocard) circle. As stated in Section 1 the six additional points to provide the thirteen in the
Figure are such that A1 meets S at N1, B1 meets S at L1, C1 meets S at M1, A2' meets S at
M2, B2' meets S at N2 and C2' meets S at L2. After some algebra we find their co-ordinates to
be L1(a2b2, b2(c2 + a2 b2), b4 + (a2 b2)(c2 + a2)), M1(c4 + (b2 c2)(a2 + b2), b2c2, c2(a2+ b2
c2)), N1(a2(b2 + c2 a2), a4 + (c2 a2)(b2 + c2), c2a2), L2(c2a2, a4 + (a2 c2)(b2 c2), c2(a2 + b2
c2)), M2(a2(b2 + c2 a2), a2b2, b4 + (b2 a2)(c2 a2)), N2(c4 + (c2 b2)(a2 b2), b2(c2 + a2 b2),
b2c2).

3
7. More lines through A3, B3, C3

We give the analysis to show that M2^N1' = A3, the pairs of lines that intersect in B3 and C3
then follow by cyclic change of letters A, B, C and L, M, N.

The equation of M2 is
(7.1)

The equation of N1' is


(7.2)

These two lines intersect at A3(0, b2(a2 b2), c2(a2 c2)).

References
1. C. J. Bradley, Challenges in Geometry, Oxford, 2005.
2. C. J. Bradley, The Algebra of Geometry, Highperception, Bath, 2007.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 25
When quadrangles are completely in perspective
Christopher J Bradley

1. Introduction

In this article we establish a sufficient condition for when a pair of quadrangles have
what may be appropriately called a Desargues axis of perspective. It is also shown
how the condition is automatically satisfied when corresponding pairs of vertices of
the two quadrangles lie on a conic and are in involution by means of perspectivity.
The ideas are then formulated and outlined to provide a guide as to what happens for
polygons with an arbitrary number of sides. The analysis is carried out using
projective co-ordinates.

EF^UV
CD^RS

FG^VW

AC^PR

AD^PS
EG^UW F To AB^PQ

A
W R
BC^QR
Q
U
D
S X E
To BD^QS B

G
P

To V

Figure 1

We define two types of perspective in dealing with the polygons A1A2An and
B1B2Bn. Partial perspective is defined to mean that A1B1, A2B2, ..., AnBn are
concurrent at a point X. Complete perspective is defined to mean that all the points
AjAk^BjBk, j, k = 1 to n, j k are collinear. So complete perspective is when an axis
exists that may be termed a Desargues axis of perspective.

V
1
2. Triangles

Theorem 1 (Desargues theorem)


When n = 3 and we are dealing with triangles, then Partial perspective Complete
perspective.

Proofs are given in all good geometry books, such as Pedoe [1].

3. Sufficient condition for quadrangles

Theorem 2
Let ABCD and PQRS be in partial perspective with vertex X, so that P lies on AX, Q
on BX, R on CX, and S on DX. If, in addition, AC^BD, PR^QS are collinear with X,
then ABCD and PQRS are in complete perspective.

What the extra condition means is that if one corresponding pair of diagonal points E
= AC^BD and U = PR^QS are such that EU also passes through X, and if ABCD and
PQRS are also in direct perspective with vertex X, then they are in complete
perspective.

Proof
In the real projective plane we take a co-ordinate system such that A is ( 1, 1, 1), B is
(1, 1, 1), C is (1, 1, 1), D is (1, 1, 1); the diagonal point triangle is then the triangle
of reference and in particular E = AC^BD is (0, 1, 0). Let X be (f, g, h). Then P has co-
ordinates (f , g + , h + ) for some value of , Q has co-ordinates (f + , g , h +
) for some value of , R has co-ordinates (f + , g + , h ) for some value of and
S has co-ordinates (f , g , h ) for some value of .

For the intersection of PR with EX we want a linear combination of P and R satisfying


hx = fz. This is easily found to be
P + R = (f( + ), g( + ) + 2 , h( + )).
Similarly the intersection of QS with EX is
Q + S = (f( + ), g( + ) + 2 , h( + )).
We require these to be the same point, the condition for which is
2 /( + ) = 2 /( + )
or
+ + + = 0. (3.1)
Consider now the axis of perspective of triangles ABC and PQR. One point on it is the
intersection of AB (x + y = 0) and PQ. This is P Q = ( , + , ). Another is
the intersection of BC (y + z = 0) and QR, which is Q R = ( , , + ). The
line joining these is easily calculated to be
( + )x + ( + )y + ( + )z = 0. (3.2)

2
From Desargues theorem for the triangle, CA^RP must lie on this line. We must show
that AD^PS also lies on this line. AD has equation y = z and the intersection is P S =
( , + , + ). So the condition, from equation (2), is
( + )( ) + ( + )( + ) + ( + )( + ) = 0,
and this reduces to Equation (1). Similarly BD^QS, CD^RS lie on the axis of
perspective.

Theorem 3
Under the same conditions as Theorem 2, the diagonal point triangles EFG of ABCD
and UVW of PQRS are in perspective, with vertex of perspective X and with the same
axis of perspective as the quadrangles themselves.

Proof
The co-ordinates of F are (0, 0, 1). By working analogous to that in Theorem 2,
equation (2) ensures that the intersection of PQ and FX is the same point as the
intersection of RS with FX, and this shows that AB^CD, X and PQ^RS are collinear.
Similarly AD^BC, X and PS^QR are collinear. This shows that triangles EFG and
UVW are in perspective with vertex X. It is straightforward, though tedious, to show
that the axis of perspective of the two diagonal point triangles is the same axis of
perspective as the quadrangles themselves and a proof is omitted.

Figure 1 illustrates Theorems 2 and 3.

An alternative theorem to Theorem 2 is possible. If one starts with triangles ABC and
PQR in perspective, so that what will become the Desargues axis is located. Then D
can be placed anywhere and the position of S may be fixed, not by the conditions of
Theorem 2, but by requiring AD^PS and BD^QS both lie on the Desargues axis. It
then follows that CD^RS automatically lies on the Desargues axis and the
quadrangles are in complete perspective, as before. It follows, since D, S, X are then
collinear, that for quadrangles complete perspective implies direct perspective.

4. Polygons

It is now possible to see how to generalise the result for quadrangles to polygons with
an arbitrary number of sides. As before direct perspective plus a number of other
conditions ensure complete perspective. To get the idea let us consider pentagons
ABCDE and PQRST. Let AP, BQ, CR, DS, ET be concurrent at X. Fix D and S as in
Theorem 2 so that quadrangles ABCD and PQRS are in complete perspective. Then E
and T are selected in similar fashion so that AC^BE, X, PR^QT are collinear. Then it
follows that ABCDE and PQRST are in complete perspective. Essentially one builds
up the pentagon from the two pairs of complete quadrangles ABCD, PQRS and ABCE,
PQRT. Admittedly this is merely a guide to what happens and does not constitute a
proof, though the figure that follows is convincing.
3
Figure 2

Figure 2 shows two pentagons in complete perspective. It shows how the points S and
T are selected in order to make 1, X, 1 collinear, where 1 = AC^BD and 1' = PR^QS
and 2, X, 2 collinear, where 2 = AC^BE and 2' = PR^QT. Note that when this is done
it is automatic that DE^ST lies on the Desargues axis.

5. Quadrangles inscribed in a conic


We now consider a special case when two quadrangles in direct perspective are
automatically in complete perspective.

Theorem 4
Let ABCD and PQRS be two distinct quadrangles in direct perspective with vertex Y,
and suppose all the vertices lie on a conic not passing through Y, and then the
quadrangles are in complete perspective.

Co-ordinate Proof
We show that AC^BD, Y, PR^QS are collinear, so that Theorem 2 applies.

Let the conic have equation y2 = zx, with points U on having co-ordinates given
parametrically as (u2, u, 1). Then the equation of UV is x + uvz = (u + v)y. Suppose
4
now that Y has co-ordinates (0, 1, 0), where Y does not lie on , then points U, V at
opposite ends of chords through Y are in involution by means of the perspectivity
through Y, given analytically by v = u. Let A, B, C, D have parameters a, b, c, d, then
P, Q, R, S have parameters a, b, c, d.

The equation of AC is x (a + c)y + acz = 0 and that of BD is x (b + d)y + bdz = 0


and AC^BD has co-ordinates (ac(b + d) bd(a + c), ac bd, a + c b d). Similarly
PR^QS has co-ordinates ( ac(b + d) + bd(a + c), ac bd, a c + b + d). Their
collinearity with Y is now immediate.

Two points need mentioning. The first is that in the real projective plane Y is external
to the conic in that not all lines through Y intersect the conic in real points. However,
in the complex plane all lines through Y intersect the conic in two points. So it is
perfectly general to take Y as a general point of the plane. The second observation is
that if ac = bd then the two points above coincide. This is perfectly possible, but does
not invalidate the theorem as we may then take two other corresponding diagonal
points, and not all pairs of diagonal points can coincide unless the quadrangles
coincide, which we have assumed not to be the case.

Synthetic proof
From the inscribed quadrangle ADPS the intersection AD^PS lies on the polar of Y and
similarly for the other five intersections.

Since A and P may be interchanged and also B and Q, C and R, and D and S it is also
the case that such intersections as AB^PQ, AC^PR, BC^QR, BD^QS, CD^RS also lie
on the polar of Y.
For the purposes of clarity and because of available space only a few of the
intersections are shown in Figure 3.

5
Figure 3

Theorem 5
In the same configuration as Theorem 4, let AYP meet BD and QS at A and P , let
CYR meet BD and QS at C and R , let BYQ meet AC and PR at B and Q and let DYS
meet AC and PR at D and S . Then A , B , C , D , P , Q , R , S lie on a conic and the
polar of Y with respect to this conic is the same Desargues axis as for the quadrangles
ABCD, PQRS.

Proof
The equation of BD is x (b + d)y + bdz = 0 and the equation of AYP is x = a2z and
these meet at A , with co-ordinates (a2(b + d), a2 + bd, (b + d)). Likewise P has co-
ordinates ( a2(b + d), a2 + bd, (b + d)). The co-ordinates of C and R follow by

6
replacing a2 by c2. The other four points now follow with a, c replacing b, d and b2, d2
replacing a2, c2.

The equation of the conic that passes through all eight points may now be verified to
be
((a+ c)2 + (b + d)2)x2 (a + c)2(b + d)2y2 + (b2c2d2 + a2c2d2 + a2b2d2 + a2b2c2
+ 2abcd(ac + bd))z2 ((a2 + c2)(b2 + d2) 4abcd)zx = 0. (5.1)
Furthermore the fact that xy and yz terms in equation (3) are absent means that the
polar of Y with respect to this conic has equation y = 0, which is also the polar of Y
with respect to the original conic y2 = zx.

Corollary
It now follows that the quadrangles A B C D and P Q R S are in perspective with
vertex Y and since they lie on a conic are in complete perspective. Thus intersections
such as A B ^P Q also lie on the same Desargues axis of perspective. See Figure 3 as
illustration of this.

It is also the case that polygons of an arbitrary number of sides that are in direct
perspective and whose vertices lie on a conic are in complete perspective, though the
proof is beyond the intention of this article.

6. Cross ratio
Consider the quadrangle ABCD inscribed in the conic with equation y2 = zx, where A,
B, C, D have parameters a, b, c, d then {A, C; B, D} = {a, c; b, d}. If the quadrangle
with vertices P, Q, R, S is defined to be in perspective through Y, as in Theorems 4 and
5, then its parameters are a, b, c, d respectively and since { a, c; b, d} =
{a, c; b, d}, it follows that {A, C; B, D} = {P, R; Q, S}. Since the point Y is at our
disposal we have proved the following theorem:

Theorem 6
Two quadrangles ABCD and PQRS in direct perspective on a conic are such that
{A, C; B, D} = {P, R; Q, S}.

Reference
1. Dan Pedoe, Geometry, A Comprehensive Course, Cambridge, 1970.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

7
Article 26
A circle concentric with the incircle
Christopher J Bradley

Figure 1
1. Introduction

Let ABC be a triangle and P a variable point. Draw AP, BP, CP to meet BC, CA, AB
respectively in U, V, W. Draw the circles UPV, VPW, WPU. These meet the sides in 6 points
labelled Q, R, S, T, E, F in the diagram.

1
Then it turns out that these 6 points always lie on a conic. Furthermore, this conic is a circle if,
and only if, P is Gergonnes point for ABC, and then the centre of the circle is the incentre I.
Now let L, M, N be the centres of circles VPW, WPU, UPV respectively. We prove that triangles
ABC and LMN are in perspective with vertex of perspective I. These properties are displayed in
Figure 1.

We prove also that the perspector (the Desarguess axis of perspective) is parallel to the polar of
Gergonnes point with respect to the incircle. In the proof of these results we use areal co-
ordinates, a description of whose definition and properties may be found in Bradley [1].

2. The six points

Take P to have co-ordinates (l, m, n), then U, V, W have co-ordinates (0, m, n), (l, 0, n) and (l,
m, 0) respectively. Suppose the equation of the circle VPW is
a2yz + b2zx + c2 xy (x + y + z)(ux + vy + wz) = 0. (2.1)

As shown in [1] all circles have an equation of this form for variable u, v, w. Putting in the co-
ordinates of V, P, W we can now find the values of u, v, w particular for this circle. They are
u = {mn(l(b2(l + m) + c2(l + n)) a2(l + m)(l + n))}/{l(l + m)(l + n)(l + m + n)}, (2.2)
2 2 2
v = {a n(l + n) l(b n c (l + n))}/{(l + n)(l + m + n)}, (2.3)
2 2 2
w = {a m(l + m) + l(b (l +m) c m)}/{(l + m)(l + m + n)}. (2.4)

Equations (2.2) to (2.4) may now be substituted in equation (2.1) to obtain the equation of circle
VPW, which is
mnx2(a2(l + m)(l + n) l(b2(l + m) + c2(l + n))) ly2(l + m)(a2n(l + n) l(b2n c2(l + n)))
+ lz2(l + n)(a2m(l + m) + l(b2(l + m) c2m)) + l2yz(l(b2(l + m) + c2(l + n)) a2(l + m)(l + n))
xy(a2n(l + m)(l m)(l + n) l(b2n(l + m)(l m) + c2(l + n)(l(m + n) + m2)))
+ zx(a2m(l + m)(l + n)(l n) l(b2(l + m)(l(m + n) + n2) + c2m(l + n)(l n))) = 0. (2.5)

This circle meets CA at the point Q with co-ordinates


((l + n)(a2m(l + m) + l(b2(l + m) c2m)), 0, m(l(b2(l + m) + c2(l + n)) a2(l + m)(l + n))).
From similar working the point R where this circle meets AB has co-ordinates
((l + m)(l(b2n c2(l + n)) a2n( l + n)), n(a2(l + m)(l + n) l(b2(l + m) + c2(l + n))), 0).

The hard way of proceeding from here would be to obtain the co-ordinates of the four points S,
T, E, F by cyclic change of letters and then show that all six points lie on a circle if, and only, if
(l, m, n) are the co-ordinates of Gergonnes point. It is easy to see that the analytic work involved
in proceeding in this way would be very difficult. The reason it was possible to be wise before
the event is that by good fortune we discovered that P was Gergonnes point by guesswork, using
CABRI software. That all six points lie on a conic whatever the choice of P made it likely that

2
some choice of the position of P, and no other, would result in the conic being a circle. This
allows the following easier approach.

The values of l, m, n for P to be Gergonnes point are


l = (c + a b)(a + b c),
m = (a + b c)(b + c a), (2.6)
n = (b + c a)(c + a b).

With these values the co-ordinates of Q are ((a + b c)(2a2 a(4b + c) + 2b2 bc c2), 0, (b + c
a)(2a2 a(b + c) (b c)2), and by cyclic change of letters the circle WPU meets AB at the
point T with co-ordinates ((b c a)(a2 + a(b 2c) 2b2 + bc + c2), (a b c)(a2 + a (b + c)
2(b2 2bc + c2)), 0) and the circle UPV meets BC at E with co-ordinates (0, (c a b)(a2 + a(c
2b) + b2 + bc 2c2), (a b + c)(2a2 a(b + 4c) b2 bc + 2c2)).

It is now possible to find the equation of the circle containing the points Q, T and E. There is
no need to give the working as the reader may verify that the three points with co-ordinates just
given lie on the conic whose equation we now give, and also to check, by means of Section
2.3.10 of [1] that this conic is a circle. The equation is
(b + c a)2(2b2 + 2c2 a2 4bc ab ca)(b2+ c2 2a2 2bc + ab + ca)x2 + +
+ (a + b c)(c + a b){5a4 11a3(b + c) + a2(3b2 + 3c2 + 20bc) + 7a(b3 b2c bc2 + c3)
4b4 2b3c + 12b2c2 2bc3 4c4}yz + + . = 0 (2.7)

Here the terms in y2 and z2 follow from those of x2 by cyclic change of a, b, c. Also the terms in
zx, xy follow from those of yz by means of cyclic change of a, b, c.

The point R has co-ordinates ((b c a)(2a2 a(b + 4c) b2 bc + 2c2), (a b c)(2a2 a(b +
c) (b c)2), 0) and it may now be checked that R lies on the circle with Equation (2.7).
Similarly, by symmetry, S and F lie on , which therefore contains all six points Q, R, S, T, E, F.

3. The circle VPW and its centre L

The circle VPW, for P having general co-ordinates (l, m, n), has Equation (2.5). Putting the
values (2.6) for Gergonnes point into Equation (2.5) we obtain its equation as
ux2 + vy2 + wz2 + 2fyz + 2gzx + 2hxy = 0, (3.1)

where u = (b + c a)2(2a2 a(b + c) (b c)2), v = (c + a b)2(2a2 a(b + 4c) b2 bc +


2c2), w = (a + b c)2(2a2 a(4b + c) + 2b2 bc c2), 2f = (a + b c)(c + a b)( 2a2 + a(b + c)
+ (b c)2), 2g = (b + c a)(a + b c)( 4a2 + a(5b + 2c) b2 bc + 2c2), 2h = (c + a b)(b + c
a)(4a2 a(2b + 5c) 2b2 + bc + c2).

3
Using Equation (2.9) of Section (2.3.6) of [1], we find the co-ordinates of the centre L of this
circle to be
L(2a3 7a2(b + c) + 2a(b c)2 + 3(b + c)(b2 2bc + c2), b(5a2 4a(b + c) (b c)2),
c(5a2 4a(b + c) (b c)2)). (3. 2)

It is immediate that AL passes through I(a, b, c). Similarly BM and CN pass through I and hence
triangles ABC and LMN are in perspective with perspector the incentre of ABC. The co-
ordinates of M and N follow from cyclic change of a, b, c and x, y, z and are hence
M(a(5b2 4b(c + a) (c a)2), 2b3 7b2(c + a) + 2b(c a)2 + 3(c + a)(c2 2ca + a2), c(5b2
4b(c + a) (c a)2)), (3. 3)
2 2 2 2 3 2 2
N(a(5c 4c(a + b) (a b) ), b(5c 4c(a + b) (a b) ), 2c 7c (a + b) + 2c(a b) + 3(a +
b)(a2 2ab + b2)). (3.4)

4. The perspector is parallel to the polar of Ge with respect to the incircle

The equation of the incircle of ABC is known to be


(b + c a)2 x2 + (c + a b)2y2 + (a + b c)2z2 2(a + b c)(c + a b)yz 2(b + c a)(a + b
c)zx 2(c + a b)(b + c a)xy = 0. (4.1)
The equation of the polar of Gergonnes point with respect to this circle is easily shown to be
(b + c a)x + (c + a b)y + (a + b c)z = 0. (4.2)

It is a straightforward but technically difficult piece of analysis to find the equation of MN and
its intersection U' with the line BC. U' is, of course, on the perspector. Its co-ordinates are
(0, (a + b c)(a2 + a(2b 4c) b2 4bc + 5c2), (c + a b)(a2 + a(4b 2c) 5b2 + 4bc + c2))
(4.3)
Lines parallel to the polar of Gergonnes point, with equation (4.2), have equations of the form
(b + c a)x + (c + a b)y + (a + b c)z + k(x + y + z) = 0. (4.4)

The line with Equation (4.4) contains U' if, and only if,
k = 3(a3 a2(b + c) a(b c)2 + (b + c)(b c)2)/{2(a2 2a(b + c) + (b c)2)}. (4.5)

Inserting the value of k from (4.5) into Equation (4.4) we obtain the line parallel to the polar of
Gergonnes point through U' to have equation
(b + c a)(5a2 4a(b + c) (b c)2)x + (c + a b)(5b2 4b(c + a) (c a)2)y +
(a + b c)(5c2 4c(a + b) (a b)2)z = 0. (4.6)

The symmetry of this equation under the cyclic permutations a, b, c and x, y, z shows that it not
only contains U' = BC^MN, but also V' = CA^NL and W' = AB^LM and is therefore the
perspector. It is left as an exercise to show that this line passes through the midpoint of Ge and
the inverse of Ge with respect to the incircle.

4
Reference
1. C. J. Bradley, The Algebra of Geometry, Highperception, Bath (2007).

Flat 4,
Terrill Court
12-14, Apsley Road
BRISTOL BS8 2SP

5
Porisms with a circular circumconic
C J Bradley and G C Smith

1. Introduction

1
1. Introduction

This article is concerned with triangle porisms. Whenever there is a single triangle that
circumscribes an inconic and is inscribed in a circumconic, it is known that there are an infinite
number of such triangles. See, for example, Maxwell [1]. We are concerned here with porisms in
which the circumconic is the circumcircle of all the triangles involved. If the inconic intersects
the circumcircle we call such porism an incomplete porism and there are points on the
circumconic that cannot entertain a triangle vertex. Those in which a vertex may appear at any
point of the circumcircle we call a complete porism. The complete porisms are those encountered
most frequently and include Poncelets porism, when the inconic is the incircle, and Brocards
porism, when the inconic is Brocards ellipse. See Bradley and Smith [2] for an account of this
porism, which has the remarkable property that any two triangles in the porism are in triple
reverse perspective. The result we prove in this article is if we take the orthocentre of all the
triangles in a case when the circumconic is the circumcircle, then the path traced out by the
orthocentre is circular (or linear). For a complete porism it is an entire circle, but for an
incomplete porism it is a circular arc (or segment). It is when the inconic is a parabola that the
locus is part of a straight line. Moreover we are able to identify the circle as the common Hagge
circle [3] of the triangles involved or in the case of an incomplete porism part of such a circle.
Analysis is carried out using areal co-ordinates. In what follows we refer repeatedly to Figure 1,
introducing points on the diagram as they occur in the text.

2. Key points of the analysis

We start with a triangle ABC as triangle of reference and its circumcircle, which we refer to as .
The orthocentre is H(u, v, w), where u : v : w = tan A : tan B : tan C so that the equation of the
circumcircle is
u(v + w)yz + v(w + u)zx + w(u + v)xy = 0. (2.1)

The co-ordinates of O, the centre of are (v + w, w + u, u + v) and the co-ordinates of the nine-
point centre are (2u + v + w, 2v + w + u, 2w + u + v). A parameter system on may be set up so
that a point R, with parameter r, has co-ordinates ( u(v + w)r(1 r), v(w + u)(1 r), w(u + v)r),
as may be checked by substitution in Equation (2.1).

The porism is created by choosing a point I to act as the centre of an inconic. (Note it can be
inscribed or escribed.) We take the equation of the inconic to be
l2x2 + m2y2 + n2z2 2mnyz 2nlzx 2lmxy = 0. (2.2)

This meets the side BC, x = 0, where (my nz)2= 0, the double root indicating tangency, and the
point of tangency being (0, 1/m, 1/n). The Brianchon point (the Cevian point for which the feet
of the Cevians are the points of tangency) for this inconic has co-ordinates (1/l, 1/m, 1/n).The

2
polar of the point with co-ordinates (m + n, n + l, l + m) with respect to the inconic is x + y + z =
0, the line at infinity, showing the co-ordinates of I are (m + n, n + l, l + m).

We introduce a point h lying on OI such that OI = Ih. The normalized x-co-ordinate of h is given
by (2Ix Ox) = (2(m + n)u (l m n)(v + w))/{(2(u + v + w)(l + m + n)}. We complete the
parallelogram OhHPg, so that Pg has x-co-ordinate Ox + Hx hx = l/(l + m + n). Its isogonal
conjugate P has x-co-ordinate u(v + w)/l. This is a key step, for in a Hagge circle construction
this figure is always a parallelogram.

The plan is to show that the Hagge circle of all the triangles in the porism with respect to P
coincide and have centre h, so that in particular, since the orthocentre of a triangle lies on all its
Hagge circles, the locus of the orthocentre of all triangles in the porism is all or part of this
Hagge circle.

3. Analysis of the special Hagge circle of triangle ABC generated by P

The point Q, such that HhQ is a straight line and Hh = hQ, will lie on this Hagge circle and has
co-ordinates Q(m + n l, n + l m, l + m n)/(l + m + n).We show in due course that the
equation of the Hagge circle is
(l + m + n)(u(v + w)yz + v(w + u)zx + u(v + w)xy) 2(x + y + z)(vwlx + wumy + uvnz) = 0.(3.1)

It may be checked by substituting their co-ordinates in Equation (3.1) that H and Q do indeed lie
on this circle.

The points X, Y, Z which are the intersections of AH, BH, CH with the proposed Hagge circle are
X((v + w)(m + n l), 2vl, 2wl), Y(2um, (w + u)(n + l m), 2wm), Z(2un, 2vn, (u + v)(l + m n)).

The equation of AP is vnz(w + u) = wmy(u + v) and this meets the circumcircle at the point D
with co-ordinates D( u(v + w), v(w + u)(m + n)/m, w(u + v)(m + n)/n).
Likewise E and F have co-ordinates E(u(v + w)(n + l)/l, v(w + u), w(u + v)(n + l)/n),
F(u(v + w)(l + m)/l, v(w + u)(l + m)/m, w(u + v)).

Now the circle HBC has equation


u(v + w)yz + v(w + u)zx + w(u + v)xy = 2vwx(x + y + z) (3.2)

This meets the proposed Hagge circle at U with co-ordinates


U(u(v + w), ((m + n)w + (n m)u)v/m, ((m + n)v + (m n)u)w/n). Similarly V and W have co-
ordinates V((n + l)w +(n l)v)u/l, v(w + u),((n + l)u +(l n)v)w/n)
W((l + m)v +(m l)w)u/l, ((l + m)u + (l m)w)v/m, w(u + v)).

3
The co-ordinates of D and U are normalised to the same amount, so the displacement DU is
proportional to ((v + w), v, w). The displacement BC is (0, 1, 1), and the perpendicularity
condition in areals is satisfied for these displacements. For those unfamiliar with this condition
it is that (p, q, r) and (e, f, g) are perpendicular displacements if, and only if,
u(v + w)(qg + rf) + v(w + u)(pg + re) + w(u + v)(pf + qe) = 0.
See Bradley [4] p.180. Clearly the midpoint of DU lies on BC. It follows that D and U are
reflections of one another in BC. Similar considerations hold for the pairs of points E, V and F,
W relative to the sides CA, AB respectively. Thus the circle with equation (3) is the Hagge circle
of P with respect to triangle ABC.

As an aside it is interesting to consider the transformation, that we call the Hagge map. This
takes (A, B, C) to (X, Y, Z). As it is known that this is an indirect similarity through the point P it
is a matrix whose columns contain the normalised co-ordinates of X, Y, Z respectively. It has
been checked by DERIVE that this matrix does indeed leave P invariant and maps triangle DEF
on to triangle UVW. This provides a double check on the analysis of this section.

4. Analysis for an arbitrary triangle in the porism

We now come to analyse an arbitrary triangle RST in the porism, where these points have
parameters r, s, t respectively. The equation of ST is
x/{u(v+w)} + sty/{v(w+u)} + (1 s)(1 t)z/{w(u+v)} = 0. (4.1)

ST is a tangent to the inconic if, and only if,


lu(v + w) + mv(w + u)/(st) + nw(u + v)/{(1 s)(1 t)}= 0,
this being the condition for equal roots when Equation (4.1) is substituted in Equation (2.2).

If we fix r then we know that s, t are the roots of the equation


lu(v + w) + mv(w + u)/(rx) + nw(u + v)/{(1 r)(1 x)}= 0.

Rearranging this gives


(s + t) = ( r2lu(v+w) + r(lu(v + w) + mv(w + u) + nw(u + v)) mv(w + u))/{r(1 r)lu(v + w)}
and
st = mv(w + u)/{rlu(v + w)}.

These equations may be compactly re-expressed as


rst = mv(w + u)/{lu(v + w)} and (1 r)(1 s)(1 t) = nw(u + v)/{lu(v + w)}. (4.2)

Essentially these equations say that triangle RST is in the porism, if, and only if, these two
equations hold. Equation (4.1) may be written, using Equations (4.2) as
lr(r 1)x + m(1 r)y + rnz = 0. (4.3)

4
A displacement (, , ) along ST (that is, ST is proportional to (, , )) is therefore given by
= m(1 r) nr, = nr lr(r 1), = lr(r 1) m(1 r). (4.4)

A rather tedious calculation shows that the line RP meets the circumcircle again at the point D'
with parameter
j = {m(l(r 1) n)}/{l(m(r 1) + nr)}. (4.5)

In terms of j the unnormalised co-ordinates of D' are (d, e, f) where


d = u(v + w)j(1 j) , e = v(w + u)(1 j), f = w(u + v)j. (4.6)

We suppose that the reflection of D' in ST is U'(g, p, q) and we further suppose that U' is
normalised with the same normalisation constant as D', so that
g + p + q = d + e + f. (4.7)

The midpoint (g + d, p + e, q + f) of D'U' lies on ST so we have from Equation (4.3)


lr(g + d)(r 1) + m(p + e)(1 r) + nr(q + f) = 0. (4.8)

The perpendicularity condition in this case is


u(v + w)((q f) + (p e)) + v(w + u)((q f) + (g d))
+ w(u + v)((p e) + (g d)) = 0. (4.9)

Values of , , in terms of l, m, n, r from Equations (4.4) may be substituted in Equation (4.9).


Equations (4.5) and (4.6) may also be used in Equations (4.7) (4.9), so that one is left with
three equations for g, p, q in terms of l, m, n, r, u, v, w. Using DERIVE we find the
(unnormalised) solution to be
g = mnu(l2r(r 1)(v + w) + l(v w)(m(r 1) + nr) + mn(v + w)), (4.10)
p = nlv(lr(m(r 1)(u w) nr(u + w)) + m(m(r 1)(u + w) + nr(w u))), (4.11)
q = lmw(l(r 1)(m(r 1)(u + v) + nr(v u)) n(m(r 1)(u v) nr(u + v)))(4.12)

It may now be determined by direct substitution that for any value of r the point U'(g, p, q) lies
on the circle with Equation (3.1). By symmetry points V', W', similarly defined also lie on this
circle. But the points U', V', W' define a Hagge circle, which contains H', the orthocentre of
triangle RST.

We have therefore proved that the locus of the orthocentres of the triangles in the porism is a
circle or part of a circle, this circle being the Hagge circle of P with respect to each and every
triangle in the porism. Consequently the porism is defined either by specifying the circumcircle,
one triangle in the porism, say ABC, and the incentre I of an inconic. Or it is defined by
specifying the point P instead of the point I. In the latter event the triangles in the porism may be

5
obtained as follows. Select a point H' on the Hagge circle of P with respect to triangle ABC.
Obtain Pg', the point that is going to be the isogonal conjugate of P with respect to the triangle
RST that has orthocentre H'. This is done by using the known fact that H'Pg'Oh is a
parallelogram. The position of R may now be determined from the fact that angle ORP = angle
H'RPg'. The positions of S and T then follow from the fact that ST is the perpendicular bisector
of H.

Reference
1. E. A. Maxwell, The methods of plane projective geometry based on the use of general
homogeneous coordinates. Cambridge University Press, 1946.
2. C.J. Bradley & G.C.Smith, Stationary Porisms, Accepted for publication.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

6
Article 28
Some Circles in a Cyclic Quadrilateral

Christopher J Bradley

E
A
T

B
P

Q
X
W O
U
Y R
Z

S
C

V
F
D

Figure

1. Introduction

In the figure ABCD is a cyclic quadrilateral, Points T, U, V, W are the midpoints of AB, BC.
CD, DA respectively and O is the centre, X is the midpoint of AC and Z is the midpoint of BD.
The diagonals AC and BD meet at Y. Points P, Q, R, S are the midpoints of OA, OB, OC, OD

1
respectively, so that PQRS is a cyclic quadrilateral similar to ABCD and half the size. The
following results now hold, several of which are immediate and others more interesting.

(i) Circles ATW, BUT, CVU, DWV all pass through O and touch the circle ABCD at its
vertices A, B, C, D respectively;
(ii) Circles ATW, CVU pass through X and circles BUT, DWV pass through Z;
(iii) Points O, X, Y, Z are concyclic;
(iv) Circles AYZ and COZ meet at Z and a point H lying on the circumcircle;
(v) Circles CYZ and AOZ meet at Z and a point F lying on the circumcircle;
(vi) Circles BXY and DOX meet at X and a point E lying on the circumcircle;
(vii) Circles DXY and BOX meet at X and a point G lying on the circumcircle.
(viii) Angle OXY = angle OZY = 90o.

In the following sections we prove these results using Cartesian co-ordinates.

2. Result (i)

Take ABCD as the unit circle with O the origin. Let A have co-ordinates (2a/(1 + a2), (1 a2)/(1
+ a2) and let B. C, D have similar co-ordinates with parameters b, c, d respectively. The co-
ordinates of T, the midpoint of AB, are therefore {(1 + ab)/((1 + a2)(1 + b2))}(a + b, 1 ab).
Similar expressions hold for the co-ordinates of U, V, W by making appropriate changes of
parameters.

The equation of the circle ATW may now be calculated and is


(1 + a2)(x2 + y2) 2ax (1 a2)y = 0. (2.1)
Note that this circle passes through the point O and has centre P, the midpoint of AO. Similarly
the circles BUT, CVU, DWV all pass through O and have centres at Q, R, S respectively. An
alternative argument for this result is that the four circles are the Miquel circles for the four
vertices A, B, C, D and must have a common point. And furthermore the quadrangle of centres
must be similar to the quadrangle ABCD, and since it derives from the midpoints of chords the
similarity is simply a reduction by a factor of 2 through the Miquel point O.

It is also easy to check that circle ATW touches circle ABCD at A.

3. Result (ii)

The circle CVU has an equation similar to Equation (2.1) with c replacing a. It follows trivially
that the circles ATW, CVU meet at the midpoint X of AC as well as at O. Similarly circles BUT
and DWV meet at the midpoint Z of BC as well as at O. The co-ordinates of X and Z are similar
to those of T but with appropriate changes of parameters.

2
4. Result (iii)

The point Y is the intersection of the diagonals AC and BD. The line AC has equation
(a + c)x + (1 ac)y = 1 + ac. (4.1)

The line BD is similar with b, d replacing a, c. These two lines meet at Y with co-ordinates (x,
y), where

x = 2(ac bd)/(abc + acd abd bcd + a b + c d), (4.2)


y = (abc + acd abd bcd a + b c + d)/(abc + acd abd bcd + a b + c d). (4.3)

Having obtained the co-ordinates of X, Y, Z we may now obtain the equation of circle XYZ and
this is found to be
(abc + acd abd bcd + a b + c d)(x2 + y2) + 2(bd ac)x +
(abc + acd abd bcd a + b c + d)y = 0. (4.4)

Clearly this circle passes through O.

5. Result (iv)

The circle AZY has equation


(abc + acd abd bcd + a b + c d)(x2 + y2) + (ab + ad 2ac + 2bd bc cd)x +
(abc 2abd + acd + b 2c + d)y (a c)(1 + bd) = 0. (5.1)

The circle COZ has equation


(bc2 2bcd + c2d b + 2c d)(x2 + y2) + 2x(bd c2) + (bc2 2bcd + c2d + b 2c + d) = 0. (5.2)

These circles meet at Z and the point H with co-ordinates (x, y), where
x = 2{(b 2c + d)(bc + cd 2bd)}/{b2(c2 4cd + 4d2 + 1) + 2b(c2d 2c(1 + d2) + d) +
c2(4 + d2) 4cd + d2}, (5.3)

y = {(b(c 2d + 1) + c(d 2) + d)(b(c 2d 1) + c(d + 2) d)}/ )}/{b2(c2 4cd + 4d2 + 1)


+ 2b(c2d 2c(1 + d2) + d) + c2(4 + d2) 4cd + d2}. (5.4)

It may now be verified that (x, y) given by Equations (5.3) and (5.4) satisfy x 2= y2 = 1 so that
circles AZY and COZ intersect at Z and a point H on the circumcircle of ABCD. Results (v),
(vi), (vii) now follow by symmetry.

6. Result (viii)

3
It is straightforward to show the co-ordinates of the centre K of circle OXYZ are half those
of point Y, which proves that O, K, Y are collinear and hence angles OZY and OXY are right
angles.

7. The cyclic pentagon

In Figure 2 we show a cyclic pentagon, in which the same construction is carried out for each of
the five component cyclic quadrilaterals. There are five diagonals whose midpoints are labelled
X1 X5 and whose intersections (other than A, B, C, D, E) are labelled Y1 Y5. There are thus 5
Miquel circles touching the circumcircle at its vertices and 5 more circles such as OX1Y2X2.
There would also be 40 circles intersecting in pairs on the circumcircle. For clarity we show only
one of these pairs in Figure 2, AY2X2^COX2 = H2.

4
A
B

H2 X5
Y1
Y5
X1

E O
Y4 Y2

X4 X2
X3
Y3

Figure 2

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

5
ARTICLE 29

The Miquel Circles for a Quadrilateral

Christopher J Bradley

D F
A T
U
Q H
P
L
E

W M R C
N

1. Introduction

Let ABCD be a quadrilateral in which no pair of opposite sides are parallel, and let AD and
BC meet at F and BA and CD meet at G. In triangle BFG, points V, W, J are taken on BF,
BG, FG respectively. Now let M be the Miquel point for this configuration, so that circles
JFV, JGW, BVW meet at M. Suppose now circle JFV meets ADF at T and circle JGW
meets GDC at U.

Now consider triangle FAB with T on FA, V on FB and W on AB. Since circles BVW and
FJV are known to meet at M, it follows that M is the Miquel point for this configuration

1
and it follows circle ATW also passes through M. Similarly by considering triangle GBC,
it follows that circle CVU passes through M. The following results now hold:
(i) Circle DTU passes through M;
(ii) Circles ATW and CVU meet at M and a point L lying on AC;
(iii)Circles BWV and DTU meet at M and a point N lying on BD;
(iv) If E is the intersection of the diagonals AC and BD, then L, E, M, N are concyclic.

Whereas it is true that results (i), (ii), (iii) can be proved by synthetic arguments, there does
not seem to be a pure argument of (iv). And unfortunately the analytic proof of (iv)
requires an analysis of everything else first and this means a technically difficult proof. We
use areal co-ordinates with B(1, 0, 0), F(0, 1, 0), G(0, 0, 1) so that BFG is triangle of
reference.

2. Points J, C, D, A, M

We take C to be the point C(m, 1 m, 0), A to be the point A(1 n, 0, n) and J to be the
point J(0, k, 1 k). The equation of CG is
(1 m)x = my, (2.1)

and the equation of the line AF is


(1 n)z = nx. (2.2)

CG and AF meet at D(m(1 n), (1 m)(1 n), mn). Rather than designate co-ordinates
for points such as T, U, V, W, it is easier to let M have co-ordinates (r, s, t) and to find T,
U, V, W in terms of r, s, t and other variables. Now since we are not dealing with
normalised co-ordinates or the ratio of lengths we can choose r + s + t to be whatever we
like, without altering the position of M. We therefore choose r + s + t = 1/k, so that J can
be made to alter its position without M moving. This simplification still allows M and J to
have independent positions.

3. Circles FJM, GJM and points T, U, V, W

In this section we give no working, but simply display the results as output by the algebraic
computer package DERIVE that we used, the aim being to prevent errors in typing. The
equation of the circle FJM is
x2 (a2t(s 1) + r(b2t + c2s)) x(y(a2t(s 1) + r(b2t c2(r + t))) + z(a2(r + st t) r(b2(r
+ s) c2s))) + a2rz(y(r + s + t 1) z) = 0. (3.1)

The circle GJM has equation


x2(a2s(r + s 1) r(b2t + c2s)) + x(z(a2s(r + s 1) + r(b2(r + s) c2s)) y(a2(r2 + r(t 1)
s2 + s) + r(b2t c2(r + t)))) a2ry(y(r + s + t 1) z) = 0. (3.2)

The co-ordinates of T are (x, y, z), where


x = (n 1)(a2(n(r2 + r(s + t 1) + st t) st + t) + r(n 1)(b2t c2( r + t))),
y = r(1 n)(b2(n(r + s + t) t) c2s) a2(n(r st + t) + st t), (3.3)
z = n(a2(n(r2 + r(s + t 1) + st t) st + t) + r(n 1)(b2t c2(r + t))).

2
The co-ordinates of V are (x, y, z), where
x = a2t(1 s) r(b2t c2(r + t)),
y = a2t(s 1) + r(b2t + c2s), (3.4)
z = 0.

The co-ordinates of W are (x, y, z), where


x = a2s(r + s 1) + r(b2(r + s) c2s),
y = 0, (3.5)
z = r(b2t + c2s) a2s(r + s 1).

The co-ordinates of U are (x, y, z), where


x = m(a2(m(r + s)(s 1) + r) + mr(b2(r + s) c2s)),
y = (m 1)(a2(m(r + s)(s 1) + r) + mr(b2(r + s) c2s)), (3.6)
z = a2(m(r2 + r(2s + t 1) + s(s 1)) r(r + s + t 1)) mr(b2t + c2(m(r + s + t) r t)).

4. The circles through A, B, C, D

In Section 3 we obtained points T, U, V, W lying on the lines AD, DC, CB, BA


respectively. These points cannot be independent of one another, for if one places points on
the sides at random then the four circles ATW, BWV, CVU, DUT do not pass through a
common point. In other words the circles FJM, GJM form a recipe for obtaining four
points T, U, V, W having the property that circles ATW, BWV, CVU, DUT share a
common point M. In this way, for any quadrilateral without a pair of parallel sides, it is
shown how to construct a Miquel point and the four associated circles.

In this section we find the equations of the four circles and check that they do share the
common point M with circles FJM and GJM.

The equation of circle BWV is

x(y(a2t(s 1) + r(b2t + c2s)) + z(r(b2t + c2s) a2s(r + s 1))) + y2(a2t(s 1) + r (b2t c2(r +
t))) + yz(a2(r2 + rt + (1 s)(s t)) r(b2(r + s t) + c2(r s + t))) z2(a2s(r + s 1) + r(b2(r
+ s) c2s)) = 0. (4.1)

The equation of circle DUT is

x2(a2(m(n(rs + (s 1)(s t)) + t(s 1)) + t(n 1)(s 1)) r(b2t + c2s)(m(2n 1) n + 1)) +
x(z(a2(m(n(r(2s 1) + (s 1)(2s t)) + (1 s)(r + s t)) + (n 1)(r + t(s 1))) +
r(b2(m(n(2r +2s t) r s + t) (n 1)(r + s)) c2s(m(3n 2) n + 1))) y(a2(m(n(r2 +
r(t 1) + (1 s)(s 2t)) 2t(s 1)) + t(1 s)(n 1)) + r(b2t(m(3n 2) n + 1) + c2((n
1)(r + t) m(n(2r s + 2t) r + s t))))) my2(a2(n(r2 + r(s + t 1) + t(s 1)) t(s 1)) +
r(n 1)(b2t c2(r + t))) + yz(a2(m(n(rs + (s 1)(s t)) + (1 s)(r + s t)) r(n 1)(r + s +
t 1)) + mr(n 1)(b2(r + s t) + c2(r s + t))) +z2(n 1)(a2(m(r + s)(s 1) + r) + mr(b2(r +
s) c2s)) = 0. (4.2)

The equation of circle ATW is

3
nx2 (a2s(r + s 1) r(b2 t + c2 s)) + x(z(a2s(r + s 1)(2n 1) + r(b2(n(r + s t) + t) + c2s(1
2n))) y(a2(n(r 2 + r(t 1) + (1 s)(s t)) t(s 1)) + r(b2 t(2n 1) c2 (n(r s + t) +
s)))) y2(a2 (n(r2 + r(s + t 1) + t(s 1)) t(s 1)) + r(n 1)(b2 t c2 (r + t))) yz(a2(n(r2
+ r(t 1) + (1 s)(s t)) r2 rt + (s 1)(s t)) + r(1 n)(b2 (r + s t) + c2 (r s +
t))) + z2 (n 1)(a2 s(r + s 1) + r(b2 (r + s) c2 s)) = 0. (4.3)

The equation of circle CUV is

x2 (m 1)(a2 t(s 1) + r(b2 t + c2s)) + x(y(a2 t(s 1)(2m 1) + r(b2t(2m 1) c2(m(r s +


t) r t))) z(a2 (m(r + s t)(s 1) + r + t(s 1)) + r(b2 (m(r + s t) r s) + c2 s(1
2m)))) + my2(a2t(s 1) + r(b2t c2(r + t))) yz(a2(m(r + s t)(s 1) r(r + s + t 1)) +
mr(b2(r + s t) + c2(r s + t))) z2(a2(m(r + s)(s 1) + r) + mr(b2(r + s) c2s)) = 0. (4.4)

It may now be checked that all four circles pass through M(r, s, t).

5. The points N, L, E and the circle LENM

The point N lies on BD and is the intersection of circles DUT and BVW. Its co-ordinates
are (x, y, z), where

x = a2(m2 (n2 (r2 + r(t s) + 2(1 s)(s t)) n(r2 + rt + (1 s)(s 3t)) + t(s 1)) + m(1
n)(n(r2 + rt + (1 s)(s 3t)) 2t(s 1)) + t(n 1)2 (s 1)) r(b2(m(n(r + s t) + t) + t(n
1)) + c2(m(n(r s + t) r t) (n 1)(r + t)))(m(2n 1) n + 1),

y = (m 1)(n 1)(a2(m(n(rs + (s 1)(s t)) + t(s 1)) + t(n 1)(s 1)) r(b2t + c2s)(m(2n
1) n + 1)), (5.1)

z = mn(a2(m(n(rs + (s 1)(s t)) + t(s 1)) + t(n 1)(s 1)) r(b2t + c2s)(m(2n 1) n +
1)).

The point L lies on AC and is the intersection of circles ATW and CUV. Its co-ordinates
are (x, y, z), where

x = a2(m2(s 1)(n(r + s + t) t) + m(n2(r2 + r(2s + t 1) + s2 + s(t 1) t) n(r2 + r(2s + t


2) + s2 + s(3t 1) 3t) + 2t(s 1)) n2(r2 + r(s + t 1) + t(s 1)) + n(r2 + r(s + t 1) +
2t(s 1)) t(s 1)) + r(m + n 1)(b2(m(n(r + s + t) t) t(n 1)) + c2((n 1)(r + t)
m(n(r + s + t) r t))),
y = (1 m)(a2(m(s 1)(n(r + s + t) t) + n(r t(s 1)) + t(s 1)) + r(m + n 1)(b2(n(r + s
+ t) t) c2s)), (5.2)
z = n(r(m + n 1)(b2t + c2(m(r + s + t) r t)) a2(m(n(r2 + r(2s + t 1) + (s 1)(s + t))
t(s 1)) n(r2 + r(s + t 1) + t(s 1)) + t(s 1))).

The point E is the intersection of the diagonals AC and BD and has co-ordinates (x, y, z),
where
x = 2n(1 n),
y = (1 m)(1 n), (5.3)
z = mn.
4
It may now be verified that points L, E, N, M all lie on the circle with equation

x2(a2(m(n(rs + s2 s(t + 1) + t) + t(s 1)) + t(n 1)(s 1)) r(b2t + c2s)(m(2n 1) n +


1)) + x(z(a2(m(n(r(3s 1) + 3s2 s(t + 3) + t) + r(1 2s) + (1 s)(2s t)) + (n 1)(r + t(s
1))) + r(b2(m(2n(r + s t) r s + 2t) (n 1)(r + s)) c2s(m(4n 3) n + 1)))
y(a2(m(n(r2 + r(t 1) s2 + s(3t + 1) 3t) 3t(s 1)) + t(1 s)(n 1)) + r(b2t(m(4n 3)
n + 1) + c2((n 1)(r + t) m(2n(r s + t) r + 2s t))))) my2(a2(n(r2 + r(s + t 1) +
2t(s 1)) 2t(s 1)) + 2r(n 1)(b2 t c2(r + t))) yz(a2(m(n(r2 + r(t s) + 2(1 s)(s t))
r2 + r(s t 1) +2(s 1)(s t)) + r(n 1)(r + s + t 1)) + 2mr(1 n)(b2(r + s t) + c2(r
s + t))) + z2(n 1)(a2(m(r(2s 1) + 2s(s 1)) + r) + 2mr(b2 (r + s) c2 s)) = 0. (5.4)

6. Other quadrilaterals

The same construction of the Miquel point is valid with minor simplifications for
quadrilaterals that have one or two pairs of opposite sides parallel. For a trapezium with
AD parallel to BC, F recedes to infinity and the circle FJM becomes a line. For a
parallelogram both F and G recede to infinity and both circles FJM and GJM become lines.
The result is that TMV and WMU become straight lines.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

5
Article 30
More on Circular Perspective
Christopher J Bradley

A
F
Y
E

C
B U
D

V
Figure

1. Introduction

1
When the vertices of two triangles ABC and UVW (with no vertices of ABC on the sides of
UVW and vice versa) are such that circles AVW, BWU, CUV share a common point, then
circles UBC, VCA, WAB also share a common point and the circles are said to be in circular
perspective. If furthermore circles AWU, BUV, CVW share a common point then circles AUV,
BVW, CWU also share a common point; and consequently circles UCA, VAB, WBC share a
common point and circles UAB, VBC, WCA also share a common point. Then triangles ABC
and UVW are said to be in triple circular perspective. See Article 19.

A singular case occurs when the vertices U, V, W lie on the sides BC, CA, AB respectively and
all of the above holds, except that circles UBC, VCA, WAB are actually straight lines (to be
thought of as having the complex point in common). An even more singular case occurs when
UVW is a transversal of triangle ABC and in this article we provide an instance of this singular
case when triple circular perspective holds.

The case we investigate is when U is the intersection of the tangent at A to the circumcircle of
triangle ABC with the side BC, with V and W similarly defined on the tangents at B and C and
on the sides CA and AB respectively. See the figure above. The surprise is that the five finite
points of intersection lie, three of them on the circumcircle of ABC (the points D, E, F in the
figure) and two on the line UVW (the points Y, Z in the figure). That one of the three points
should be on the circumcircle is because the Miquel point of circles AVW, BWU, CUV is bound
to do so, but the other two are not Miquel points and hence the surprise.

Analytic proofs are provided in the following sections, using areal co-ordinates with ABC as
triangle of reference.

2. The tangents and the points U, V, W

The equation of the tangent at A is c2y + b2z = 0 and this meets BC, x = 0, at the point U with co-
ordinates (0, b2, c2). Similarly V has co-ordinates ( a2, 0, c2) and W has co-ordinates (a2, b2,
0). U is, of course, the harmonic conjugate of the point L on BC, with respect to B and C, where
the line AK meets BC, K being the symmedian point of ABC.

The equation of the line UVW is easily seen to be


x/a2 + y/b2 + z/c2 = 0 (2.1)

and hence UVW is a transversal of triangle ABC.

3. Circles ABV, BCW, CAU and the point Y

The equation of the circle ABV is

2
a2yz + b2zx + c2xy + (ux + vy + wz)(x + y + z) = 0, (3.1)

where u, v, w are to be chosen so that A, B and V lie on the curve. A short analysis gives u = 0, v
= 0 and w = a2b2/(c2 a2). Its intersection with the line having Equation (2.1) is a point with co-
ordinates (x, y, z), where
x = a2(a2 b2),
y = b2(b2 c2), (3.2)
2 2 2
z = c (c a ).

Since these co-ordinates are invariant under the cyclic change of x, y, z and a, b, c, this point also
lies on circles BCW and CAU and is therefore the point Y in the figure.

4. Circles ABU, BCV, CAW and the point Z

A similar analysis to that of Section 3 gives u = 0, v = 0, w = a2b2/(c2 b2) for the circle ABU.
The intersection of circle ABU and the line with Equation (2.1) is a point with co-ordinates (x, y,
z), where
x = a2(a2 c2),
y = b2(b2 a2), (4.1)
2 2 2
z = c (c b ).

Again since these co-ordinates are invariant under the cyclic change of x, y, z and a, b, c, this
point also lies on circles BCV and CAW and is therefore the point Z in the figure.

5. Circles AUV, BVW, CWU and the point F

The circle AUV has Equation (3 .1) with u = 0, v = a2c2(a2 b2)/{(a2 c2)(b2 c2)}, w = a2b2/(c2
a2) and this circle cuts the circumcircle with equation
a2yz + b2zx + c2xy = 0, (5.1)

at A and the point with co-ordinates (x, y, z), where


x = a2(a2 b2)(b2 c2),
y = b2(b2 c2)(c2 a2), (5.2)
z = c2(c2 a2)(a2 b2).

Again since these co-ordinates are invariant under the cyclic change of x, y, z and a, b, c, this
point also lies on circles BVW and CWU and is therefore the point F in the figure.

6. The points D and E

3
Further analysis shows that circles AVW, BWU, CUV meet at the point D on the circumcircle
with co-ordinates (x, y, z), where
x = a2(a2 b2)(c2 a2),
y = b2(b2 c2)(a2 b2), (6.1)
2 2 2 2 2
z = c (c a )(b c ).

Likewise, circles AWU, BUV, CVW meet at the point E on the circumcircle with co-ordinates
(x, y, z), where
x = a2(b2 c2)(c2 a2),
y = b2(c2 a2)(a2 b2), (6.2)
2 2 2 2 2
z = c (a b )(b c ).

In conclusion it should be mentioned that not all transversals show triple circular perspective.

Acknowledgement
My thanks to David Monk who has communicated with me and informed me of a second case of
triple circular perspective with a transversal when U is the intersection of BC with the bisector of
the exterior angle at A, with V and W similarly defined.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 31
More cases of Circular Perspective

Christopher J Bradley

A
M

bH
G
aH

' H
cH L
B C

Figure

1
Commentary

This article should be read in conjunction with Article 19. There it is proved, but not explicitly,
that circles AB, CA', BCH are concurrent at aH, circles BC, AB', CAH are concurrent at
bH and circles CA, BC', ABH are concurrent at cH.

It follows that triangles ABC and H' are in triple circular perspective.

This means, additionally, that circles AH, BH', C' are concurrent at a point L, circles
AH', B', CH are concurrent at a point M and circles A', BH, CH' are concurrent at
a point N. See the figure above.

2
ABC and O' are in Triple Circular Perspective

M
A

R

N
O
P
Q
'
B

Figure

3
1. Introduction

We now establish a very interesting result, namely that triangles ABC and 'O are in triple
circular perspective, but not only that, it is also the case that the six points of concurrence lie
three on the seven-point circle (which is the circumcircle of triangle 'O) and three on the
circumcircle of ABC. One is led to wonder if this configuration is essentially unique for the six
points of concurrence to lie three by three on the two circumcircles. In the following analysis we
use areal co-ordinates with triangle ABC as triangle of reference.

2. The circles ABO, BC, CA' and the points P, Q, R

The general equation of a circle is


a2yz + b2zx + c2xy + (ux + vy + wz)(x + y + z) = 0, (2.1)

where u, v, w are constants to be determined. One then inserts the co-ordinates of the three
points on the given circle to provide equations for u, v, w. When this is done for circle ABO we
find that u = 0, v = 0 and w = a2b2/(a2 + b2 c2) and hence its equation is
a2b2z2 a2(c2 a2)yz + b2(b2 c2)zx + c2(a2 + b2 c2)xy = 0. (2.2)

Using the same method we find the equation of circle BC is


b2x2 + a2yz + (c2 b2)xy = 0. (2.3)

These circles meet at a point R with co-ordinates (a2, b2, a2 + b2 c2). See the figure above.

Now the equation of CA' is


a2y2 + b2zx + (c2 a2)xy = 0. (2.4)

It may now be verified that R lies also on this circle.

It follows now by cyclic change of letters that circles BCO, CA, AB', all pass through the
point P with co-ordinates (b2 + c2 a2, b2, c2) and that circles CAO, AB, BC' all pass through
the point Q with co-ordinates (a2, c2 + a2 b2, c2).

The equation of the seven-point circle is


b2c2x2 + c2a2y2 + a2b2z2 a4yz b4zx c4xy = 0. (2.5)

It may now be checked that P, Q, R all lie on the seven-point circle.

3. Circles OC, 'A, 'OB and the points L, M, N

4
Using the same method as in Section 2 we find the equation of circle OC to be
b2c2(b2 c2)x2 c2a2(a2 b2)y2 + a2(a4 a2(b2 c2) + c4)yz + a2b2(a2 b2)zx
+ c2(b2 c2)2xy = 0. (3.1)
Also the equation of circle 'A is
c2a2(a2 b2)y2 + a2b2(c2 a2)z2 a2(a2 b2)(c2 a2)yz + b2c2(a2 b2)zx
b2c2(c2 a2)xy = 0. (3.2)
These two circles meet at the point L with co-ordinates (a (a b )(c a ), b (c a )(b c2),
2 2 2 2 2 2 2 2 2

c2(b2 c2)(a2 b2)).

Now the equation of the circle 'OB is


b2c2(b2 c2)x2 a2b2(c2 a2)z2 a2(a4 a2(b2 + c2) + b4)yz b2(b2 c2)2zx
+ c2a2(c2 a2)xy = 0. (3.3)
It may now be checked that L lies also on this circle.

The equations of the circles OA, 'B, 'OC may now be written down with cyclic change of
letters and verified that they all pass through the point M with co-ordinates (a2(c2 a2)(b2 c2),
b2(b2 c2)(a2 b2), c2(a2 b2)(c2 a2)). Similarly the circles OB, 'C, 'OA all pass through
the point N with co-ordinates (a2(b2 c2)(a2 b2), b2(a2 b2)(c2 a2), c2(c2 a2)(b2 c2)).

Finally it may be verified that L, M, N all lie on the circumcircle of ABC, whose equation is
a2yz + b2zx + c2xy = 0. (3.4)

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

5
ARTICLE 32
The GH Disc and another case of Triple Circular Perspective

Christopher J Bradley


N V
U G

H
'
W

B
L C

1. Introduction

1
This article should be read after Articles 19, 30, 31, in which circular perspectivity is defined and
where particular emphasis is given to triple circular perspectivity and certain examples of this
property. In this article we take three points U, V, W on the orthocentroidal circle S and show
that triangle UVW is in triple circular perspective with ABC. Circles BUC, CVA, AWB meet at
the orthocentre H, circles BVC, CWA, AUB meet at the first Brocard point and circles BWC,
CUA, AVB meet at the second Brocard point '. Also circles AVW, BWU, CUV meet at a point
L on the circumcircle of ABC, its counterpart H lying on the circumcircle of UVW (the
orthocentroidal circle). In the above the point U may be defined as follows. Draw AH to meet S
again at X, then draw XK to meet S again at U. Here K is the symmedian point. V, W are
similarly defined. Alternatively U, V, W are the points on the Hagge circle of K (which
coincides with S) such that U, V, W are the reflections of D, E, F in BC, CA, AB respectively,
where D, E, F are the intersections of AK, BK, CK with the circumcircle of ABC.

2. The orthocentroidal circle and the points U, V, W

The equation of the orthocentroidal circle S, see Bradley and Smith [1], is
(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2 a2yz b2zx c2xy = 0. (2.1)

The equation of the line AH is


(c2 + a2 b2)y = (a2 + b2 c2)z. (2.2)

This meets S at the point X with co-ordinates (a2, a2 + b2 c2, c2 + a2 b2). The symmedian point
K has co-ordinates (a2, b2, c2) and so XK has equation
(b2 + c2 a2)(b2 c2)x + a2(a2 b2)y + a2(c2 a2)z = 0. (2.3)

This meets the circle S again at U with co-ordinates (a2, b2 + c2 a2, b2 + c2 a2). Similarly V
has co-ordinates (c2 + a2 b2, b2, c2 + a2 b2) and W has the co-ordinates (a2 + b2 c2, a2 + b2
c2, c2). (It will be seen these points are those called aH, bH, cH in Article 19.)

3. Circles BUC, CVA, AWB all pass through H

The equation of the circle BUC is


(b2 + c2 a2)x2 a2yz + (c2 a2)zx (a2 b2)xy = 0. (3.1)

It may now be checked that this circle passes through H and by cyclic change of x, y, z and a, b,
c it follows that circles CVA and AWB also pass through H.

4. Circles BVC, CWA, AUB all pass through

The equation of circle BVC is

2
b2x2 a2yz + (b2 c2)xy = 0 (4.1)

and the equation of CWA is


c2y2 + (c2 a2)yz b2zx = 0. (4.2)

These circles meet at the point (1/b2, 1/c2, 1/a2).

Similarly circles BWC. CUA, AVB meet at the point '(1/c2, 1/a2, 1/b2).

5. Circles BWU, CUV, AVW meet on the circumcircle of ABC

The equation of the circle BWU turns out to be


c2(b2 c2)(b2 + c2 a2)x2 + a2(a2 b2)(a2 + b2 c2)z2 + c2a2(b2 c2)yz
(a6 a4(b2 + c2) a2(b4 3b2c2 + c4) + (b2 + c2)(b2 c2)2)zx + c2a2(a2 b2)xy = 0. (5.1)

The equations of circles CUV and AVW may now be written down by cyclic change of x, y, z
and a, b, c.

These circles all meet at L on the circumcircle of ABC with co-ordinates (x, y, z), where
x = a2/((b2 c2)(b2 + c2 a2)),
y = b2/((c2 a2)(c2 + a2 b2)), (5.2)
z = c2/((a2 b2)(a2 + b2 c2)).

Circles BUV, CVW, AWU meet at M and circles BVW, AWU, AUV meet at N, see the figure.

Reference
1. C.J.Bradley and G.C.Smith, The locations of triangle centres, Forum. Geom., 6 (2006)
57-70.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

3
4
ARTICLE 33

29 Circles

Christopher J Bradley

V
O

R
W

U '
L

B C

Q
M

Figure

1
1. Introduction

This article should be read in conjunction with Article 19, and 30 32. It involves two
cases of triple circular perspective. The first is between triangle ABC and triangle 'O,
where O is the circumcentre of ABC and , ' are respectively the first and second
Brocard points of ABC. The resulting six points of concurrence are labelled L, M, N, U, V,
W. The second is between ABC and triangle UVW and the sixpoints of concurrence are ,
', O, P, Q, R. Interestingly U, V, W lie on the 7-point circle defined by O, and ' and
all of L, M, N, P, Q, R lie on the circumcircle of ABC. It is also the case that U, V, W lie
on BK, CK, AK respectively, where K is the symmedian point of ABC, which lies itself on
the 7-point circle diametrically opposite O. This interaction between the circumcircle and
the 7-point circle is especially interesting. See the figure above. Note also that U, V, W do
not coincide with any of the usual 7 points on the 7-point circle and are therefore additional
points of some significance. The analysis of the configuration just described is carried out
in subsequent sections using areal co-ordinates with triangle ABC as triangle of reference.

2. The points L, M, N

The equation of any circle in areal co-ordinates is


a2yz + b2zx + c2xy + (ux + vy + wz)(x + y + z) = 0. (2.1)

To obtain the equation of the circle C' one puts into Equation (2.1) the co-ordinates of
C, , and ' in turn, solve for u, v, w and then re-inset their values back into the equation,
yielding the result. It turns out to be
b2c2(c2 a2)x2 c2a2(b2 c2)y2 a2b2(c2 a2)yz + a2b2(b2 c2)zx
+ c2(c2 a2)(b2 c2)xy = 0. (2.2)

This circle meets the circumcircle of ABC again at the point L with co-ordinates (x, y, z),
where
x = a2(a2 b2)(b2 c2),
y = b2(a2 b2)(c2 a2), (2.3)
z = c2(c2 a2)(b2 c2).

The equation of circle A'O is


c2a2(b2 c2)y2 a2b2(a2 b2)z2 + a2(a2 b2)2yz b2c2(b2 c2)zx
+ c2(a4 a2c2 c2(b2 c2))xy = 0. (2.4)

It may now be checked from Equations (2.3) and (2.4) that L also lies on this circle.
Symmetry considerations now imply that L also lies on circle BO.

We next determine the co-ordinates of the point M. The equation of circle A' is
c2a2(a2 b2)y2 a2b2(c2 a2)z2 + a2(a2 b2)(c2 a2)yz b2c2(c2 a2)zx
+ b2c2(c2 a2)xy = 0. (2.5)

This circle meets the circumcircle again at the point M with co-ordinates (x, y, z), where
x = a2(a2 b2)(c2 a2),
y = b2(b2 c2)(c2 a2), (2.6)
2
z = c2(a2 b2)(b2 c2).

The equation of circle CO is


b2c2(b2 c2)x2 c2a2(a2 b2)y2 + a2(a4 a2(b2 + c2) + c4)yz + a2b2(a2 b2)zx
+ c2(b2 c2)2xy = 0. (2.7)

It may be checked from Equations (2.6) and (2.7) that M also lies on this circle and
similarly on circle B'O.

We next determine the co-ordinates of the point N. The equation of circle B' is
b2c2(a2 b2)x2 a2b2(b2 c2)z2 c2a2(a2 b2)yz b2(b2 c2)(a2 b2)zx
+ c2a2(b2 c2)xy = 0. (2.8)

This circle meets the circumcircle at the point N with co-ordinates (x, y, z), where
x = a2(b2 c2)(c2 a2),
y = b2(a2 b2)(b2 c2), (2.9)
z = c2(a2 b2)(c2 a2).

The equation of circle AO is


c2a2(c2 a2)y2 a2b2(b2 c2)z2 + a2(c2 a2)2yz + b2(a4 a2b2 + b2(b2 c2))zx
+ b2c2(b2 c2)xy = 0. (2.10)

It may be checked from Equations (2.9) and (2.10) that N lies on this circle too and by
symmetry considerations also on circle C'O.

3. The points U, V, W

We first consider circle AB whose equation is found to be


a2z2 + (a2 b2)zx c2xy = 0. (3.1)

Now the equation of the 7-point circle is


b2c2x2 + c2a2y2 + a2b2c2 a4yz b4zx c4xy = 0, (3.2)

Circle AB meets the 7-point circle at the point U with co-ordinates (a2, c2 + a2 b2, c2).

Circle AOC is found to have equation


c2a2y2 a2(a2 b2)yz + b2(c2 + a2 b2)zx c2(b2 c2)xy = 0. (3.3)

It may now be checked that this circle passes through the point U, and similarly circle
B'C passes through U.

The equation of BC is
b2x2 a2yz +(b2 c2)xy = 0. (3.4)

From Equations (3.2) and (3.4) this meets the 7-point circle at the point V with co-
ordinates (a2, b2, a2 + b2 c2).

3
The equation of circle AOB is
a2b2z2 + a2(c2 a2)yz b2(b2 c2)zx c2(a2 + b2 c2)xy = 0. (3.5)
It may now be checked that this circle passes through V and similarly circle C'A passes
through V.

Finally the point W with co-ordinates (b2 + c2 a2, b2, c2) lies on the 7-point circle and
circle CA with equation
c2y2 +(c2 a2)yz b2zx = 0. (3.6)

It may now be confirmed that W also lies on circles BOC and A'B.

4. The points P, Q, R

The equation of circle AVW is


c2(b2 c2)y2 b2(c2 a2)z2 (c2 a2)2yz b2(b2 c2)zx + c2(c2 a2)xy = 0. (4.1)

This meets the circumcircle again at the point P with co-ordinates (x, y, z), where
x = a2(b2 c2)(c2 a2),
y = b2(a2 b2)(c2 a2), (4.2)
z = c2(a2 b2)(b2 c2).

The equation of circle CUV is


b2(a2 b2)x2 a2(b2 c2)y2 a2(a2 b2)yz + b2(b2 c2)zx (b2 c2)2xy = 0. (4.3)

It may now be checked that P lies on this circle and similarly also on circle BWU.

The circle AUV has equation


c2a2(c2 a2)y2 a2b2(a2 b2)z2 a2(a2 b2)(c2 a2)yz + b2(a2b2 b4 + b2c2 c4)zx
+ c2(a2c2 b4 + b2c2 c4)xy = 0. (4.4)

This circle meets the circumcircle again at the point Q with co-ordinates (x, y, z), where
x = a2(a2 b2)(c2 a2),
y = b2(a2 b2)(b2 c2), (4.5)
z = c2(b2 c2)(c2 a2).

It may now be checked that Q also lies on circles BVW and CWU.

In similar fashion it may be shown that circles AWU, BUV, CVW all lie on the point R on
the circumcircle with co-ordinates (x, y, z), where
x = a2(a2 b2(b2 c2),
y = b2 (b2 c2)(c2 a2), (4.6)
2 2 2 2 2
z = c (a b )(c a ).

5. Concluding remark

Note that AVK, BWK, CUK are all straight lines, where K is the symmedian point, the
first of these lines having equation b2x = a2y and so on. This is reminiscent of the fact that
4
AaHG, BbHG, CcHG are straight lines, see Article 19. It seems likely that other cases
when ABC and UVW are in triple circular perspective are such that a point P exists on
circle UVW such that AUP, BVP, CWP are straight lines.

Flat 4
Terrill Court,
12-14 Apsley Road
BRISTOL BS8 2SP

5
Article 34

Some Circles in the Cyclic Quadrilateral configuration


Christopher J Bradley

A Q

D G O

R C

Figure
1. Introduction

When ABCD is a cyclic quadrilateral, centre O, and BA^CD = E, BC^AD = F and AC^BD
= G, then circles ABF, CDF, BCE, ADE meet at a point T on EF. The perpendicular from
T to EF passes through G and O. The centres of the four circles, O and T lie on a circle.
These facts are established using Cartesian co-ordinates. These are not new results, though
until now I did not know that the last mentioned circle passes through T as well as the five
circle centres. See the figure above.

2. Preliminary analysis and the diagonal points E, F, G, and T

We take ABCD to be the unit circle, origin the centre O, and A with co-ordinates (2a/(1 +
a2), (1 a2)/(1 + a2)). B, C, D have similar co-ordinates but with parameters b, c, d instead
of a. The chord AB has equation
(a + b)x + (1 ab)y = (1 + ab), (2.1)
Other chords have similar equations with appropriate change of parameters.

The chords AB and CD meet at E with co-ordinates (x, y), where


x = 2(ab cd)/(abc + abd acd bcd + a + b c d),
y = (abc + abd acd bcd a b + c + d)/( abc + abd acd bcd + a + b c d). (2.2)

F = BC^AD has similar co-ordinates to those in Equation (2.1), but with a and c
interchanged. G = AC^BD has similar co-ordinates to those in Equation (2.1), but with b
and c interchanged.

The equation of EF is found to be


2(ac bd)x (abc abd +acd bcd a + b c +d)y
abc + abd acd + bcd a + b c + d) = 0. (2.3)

The perpendicular to EF through O has equation


2(ac bd)y + (abc abd +acd bcd a + b c +d)x = 0. (2.4)

It may now be checked that this line passes through G. Indeed it is well known that O is
the orthocentre of triangle EFG.

The lines with Equations (2.3) and (2.4) meet at a point T with co-ordinates (m/k, n/k),
where
m = 2(a2c(b(c d) + cd + 1) a(b2d(c d) + b(c2d + c(d2 + 1) + d) c(c d)) + bd(b(cd +
1) c + d)),

n = (a(b(c d) + cd + 1) b(cd + 1) + c d)(a(b(c d) + cd 1) + b(1 cd) c + d)(2.5)

k = a2(b2(c d)2 + 2b(c d)(cd 1) + c2(d2 + 4) 2cd + 1) 2a(b2(c d)(cd 1) + b(c2 +


1)(d2 + 1) + (c d)(cd 1)) + b2(c2d2 2cd + 4d2 + 1) + 2b(c d) (cd 1) + (c d)2.

3. The four circles through T

Every circle may be expressed in the form


x2 + y2 + 2gx + 2fy + k = 0, (3.1)

and since this may be written in the form


(x + g)2 + (y + f)2 = g2 + f2 k, (3.2)

it represents a circle, centre ( g, f) and radius (g2 + f2 k).

In order to obtain the equation of a circle through three given points one can put the co-
ordinates of those points in turn into Equation (3.1), solve the three resulting equations for
f, g, k and insert those values back into the equation. To specify a circle one may write
down its equation or specify its values of f, g, k. The latter is useful if the centre is what is
required.

The equation of circle ABF turns out to be


(abc + bcd abd acd a + b + c d)(x2 + y2) + 2(a + b)(d c)x + 2(c d)(ab 1)y
+ abc abd + acd bcd + a b + c d = 0. (3.3)

It may now be verified that this circle passes through the point T.

Similarly the circles CDF, BCE, ADE all pass through T.

4. The six point circle PQRSOT

The points P, Q, R, S are the centres of circles ABF, BCE, CDF, ADE respectively. Their
co-ordinates are:

For P: ((a + b)(c d), (1 ab)(c d))/(abc abd acd + bcd a + b + c d); (4.1)

For Q: ((b + c)(a d), (1 bc)(a d))/(abc + abd acd bcd + a + b c d); (4.2)

For R: ((b a)(c + d), (b a)(1 cd))/(abc abd acd + bcd a + b + c d); (4.3)

For S: ((a + d)(b c), (1 ad)(b c))/abc + abd acd bcd + a + b c d); (4.4)

The equation of the circle that contains all these points and also O and T is
(abc + abd acd bcd + a + b c d)(abc abd acd + bcd a + b +c d)(x2 + y2)
+ (a2(b2(c d)(cd 1) + b(1 + c2)(1 + d2) +(c d)(cd 1)) a(b2 + 1)(c2 + 1)(d2 + 1) +
b2(c d)(cd 1) + b(c2 + 1)(d2 + 1) + (c d)(cd 1))x + (a2(b2(c + d)(c d) + c2(d2 + 2) +
1) b2(c2d2 + 2d2 + 1) + c2 d2)y = 0. (4.5)

Flat 4
Terrill Court
12-14, Apsley Road,
BRISTOL BS8 2SP
ARTICLE 35
Where Seven Circles meet
Christopher J Bradley

W Y

G
U
X P
H
Z
V
C
B Q

Figure

1
1. Introduction

The orthocentroidal circle S of triangle ABC on GH as diameter possesses an unusual property.


If you draw circles BHC, CHA, AHB to meet S again at points X, Y, Z and circles BGC, CGA,
AGB to meet S again at U, V, W then the following property holds: circles AYX, BZX, CXY,
AVW, BWU, CUV all pass through a point Q on the circumcircle of ABC. The points X, Y, Z
are known to be the same points on S as those points where the medians AG, BG, CG
respectively meet S.

2. The circle S and the points X, Y, Z, and circles AYZ, BZX, CXY

The equation of the orthocentroidal circle S, see Bradley and Smith [1], is
(b2 + c2 a2)x2 + (c2 + a2 b2)y2 + (a2 + b2 c2)z2 a2yz b2zx c2xy = 0. (2.1)

The co-ordinates of X, Y, Z, where AG, BG, CG meet the orthocentroidal circle are X(a2, b2 + c2
a2, b2 + c2 a2), Y(c2 + a2 b2, b2, c2 + a2 b2), Z(a2 + b2 c2, a2 + b2 c2, c2).

Now that these co-ordinates are known the equation of circle AYZ may be computed and is
c2(c2 + a2 b2)(c2 a2)y2 b2(a2 + b2 c2)(a2 b2)z2 + (a6 a4(b2 + c2) a2(b4 3b2c2 + c4) +
(b2 c2)(b4 c4))yz b2c2(c2 a2)zx + b2c2(a2 b2)xy = 0. (2.2)

The equations of circles BZX and CXY may now be written down by cyclic change of a, b, c and
x, y, z.

3. The point Q where circles AYZ, BZX, CXY meet

The circle AYZ with Equation (2.2) meets the circumcircle at A and again at Q with co-ordinates
(x, y, z), where
x = a2/{(b2 + c2 a2)(b2 c2)},
y = b2/{(c2 + a2 b2)(c2 a2)}, (3.1)
2 2 2 2 2 2
z = c /{(a + b c )(a b )}.

The symmetry of these expressions ensures that circles BZX, CXY also pass through this point.

4. The points U, V, W, and circles AVW, BWU, CUV

The co-ordinates of U, V, W, where circles BGC, CGA, AGB meet the orthocentroidal circle are
found to be, for U (x, y, z) where
x = 3a2(a2 + b2 c2)(b2 + c2 a2),
y = 3a6 + a4(b2 c2) + a2(b2 3c2)(c2 b2) + (b2 + c2)(b2 c2)2, (4.1)

2
z = (c2 + a2 b2)(3a4 + 2a2(b2 c2) (b2 c2)(b2 + c2))

and those for V and W follow by cyclic change of a, b, c and x, y, z.

The equation of circle CUV can now be obtained and is


b2(a2 + b2 + c2)(b2 + c2 a2)(c2 b2)x2 + a2(a2 + b2 + c2)(c2 + a2 b2)(a2 c2)y2
a2(2a6 2a4(b2 + c2) a2(b4 5b2c2 + 2c4) + (b2 c2)(3b4 + b2c2 2c4))yz
+ b2(a2 + b2 c2)(3a4 + a2(c2 4b2) + 2(b4 c4))zx
(a2 + b2 c2)(a6 a4(b2 + c2) a2(b4 3b2c2 + c4) + (b4 c4)(b2 c2))xy = 0. (4.2)

It may now be checked that Q lies on this circle and by symmetry lies also on the circles AVW,
BWU.

CABRI indicates that given any circle in the plane of a triangle ABC (other than the
circumcircle) the there is one and only one pair of diametrically opposite points D, E lying on it
such that circles BDC, CDA, ADB, BEC, CEA, AEB meet in points X, Y, Z, U, V, W such
that circles AYZ, BZX. CXY, AVW, BWU, CUV have a common point Q lying on the
circumcircle. We can offer no proof of this result, but it may be regarded as a safe conjecture.

Reference
1. C.J.Bradley and G.C.Smith, The locations of triangle centres, Forum. Geom., 6 (2006)
57-70.

Flat 4
Terrill Court,
12-14 Apsley Road
Clifton
BRISTOL BS8 2SP

3
Article 36
Where 7 Circles meet Part 2
Christopher J Bradley

U
V
E P
Q
W

C
B X

T
Y

Figure 1

1. Introduction

This article should be read as the last of a sequence of articles, starting with Article 19 and
continuing through 30 33 and 35, on circular perspective. It deals with case when a triangle

1
ABC is in circular perspective with a pair of triangles both of which have vertices on a circle.
What happens is that if a circle is drawn, other than the circumcircle S and not through any
vertex, and arbitrary points D and E are chosen on it, then if circles BCD, CAD, ABD are drawn
to meet again at X, Y, Z respectively and if circles BCE, CAE, ABE are drawn to meet at U,
V, W respectively, then the following happens: circles AYZ, BZX, CXY concur at a point T on
S and circles AVW, BWU, CUV meet at a point T' on S. However, if E is chosen in a particular
way then T and T' coincide. In fact for each choice of D on , there is only one point E on with
the required property and when E is chosen appropriately, then UX, VY, WZ are concurrent.

What we do first is to consider a limiting case when the circle degenerates into a line m. This
situation is shown in Figure 1. The line m is a transversal of ABC and does not pass through any
vertex. The points X, Y, Z are the points where m meets BC, CA, AB respectively. Q is a fixed
point not at a vertex. It then follows that circles AYZ, BZX, CXY concur at the Miquel point T,
which in the case of a transversal is well known to lie on the circumcircle S. The points U, V, W
are now defined as the intersections of circles BCE, CAE, ABE with m. It now turns out that
circles AVW, BWU, CUV always concur at a point T' on S. But there is one and only one
position of E which results in the point T' coinciding with T. This turns out to be when the circles
XUQ, YVQ, ZWQ are coaxal. Interestingly E does not move from its favourable position on m
when the point Q is altered to another position.

The reader, by now, will no doubt have appreciated that the general problem is now proved by
inverting Figure 1 through the point Q. The fixed point Q is relabelled D, the coaxal circles
simply become the lines XU, YV, ZW. It is hoped that no confusion is caused by not putting
stars on the points in the inverted figure.

The analysis for the configuration of Figure 1 is carried out using areal co-ordinates with ABC as
triangle of reference and it is by no means straightforward. Some geometers might regard it as
horrendous, but my retort would then be to ask them to prove it otherwise.

2. The Miquel point T for triangle ABC and the points X, Y, Z

We take the transversal to have equation


lx + my + nz = 0. (2.1)

It follows that X, Y, Z have co-ordinates X(0, n, m), Y( n, 0, l), Z(m, l, 0). The circle AYZ
has equation
a2yz + b2zx + c2xy + (x + y + z)(ux + vy + wz) = 0, (2.2)

2
where u = 0, v = c2m/(l m), w= b2n/(l n). The line with equation ux + vy + wz = 0, with these
values of u, v, w is the equation of the common chord of circle AYZ and the circumcircle S and
therefore meets S at A and the point T with co-ordinates
x = a2mn(l m)(n l),
y = b2nl(l m)(m n), (2.3)
2
z = c lm(m n)(n l).

The symmetry of these co-ordinates under cyclic change of x, y, z and a, b, c and l, m, n shows
that T also lies on circles BZX, CXY and is therefore the Miquel point.

3. The points U, V, W when E is chosen at random on XYZ

In what follows it is best to parameterize points on the transversal XYZ, with Equation (2.1),
with a parameter s such that a general point on it has co-ordinates (smn, (1 s)nl, lm). We now
choose E to have parameter t so that E has co-ordinates (tmn, (1 t)nl, lm).

The circle EBC has an equation of the form of Equation (2.2) with v = w = 0 and
u = {l(a2l(t 1) t(b2m + c2n(t 1)))}/{t(l(m + n(t 1)) mnt)}. (3.1)

This circle meets XYZ at points with parameter s, where s = t (the point E) and at U with
s = {a2l(l(m + n(t 1)) mnt)}/{n(a2l(l m)(t 1) mt(b2(l m) + c2(n 1)))}. (3.2)

The circle EAB has an equation of the form of Equation (2.2), with u = v = 0 and
w = {n(a2l(t 1) t(b2m + c2n(t 1)))}/{mnt l(m + n(t 1))}. (3.3)

This circle meets XYZ at points with parameter s, where s = t (the point E) and at W with
s = {l(a2m(l n) + (n m)(b2m + c2n(t 1)))}/{c2n(l(m + n(t 1)) mnt)}. (3.4)

The circle ECA has an equation of the form of Equation (2.2), with u = w = 0 and
v = {m(a2l(t 1) t(b2m + c2n(t 1)))}/{(1 t)(l(m + n(t 1)) mnt)}. (3.5)

This circle meets XYZ at points with parameter s, where s = t(the point E) and at V with
s = {l(a2n(l m)(t 1) + (m n)(b2m + c2n(t 1)))}/{n(a2l(l m)(t 1) + b2mt(m 1) +
c2l(m n)(t 1))}. (3.6)

The co-ordinates of U, V, W now follow from Equations (3.2), (3.4) and (3.6).

4. The circles AVW, BWU, CUV and the point T'

3
We omit calculation of the equations of the circles AVW, BWU, CUV which though arduous is
straightforward. We record the co-ordinates of the point T' where they are concurrent on the
circumcircle. Its co-ordinates are (x, y, z), where
x = a2/{a2l(l(m n(t 1)2) + mnt(t 2)) +mnt2(b2(l m) + c2(n 1))}, (4.1)
2 2 2 2 2 2 2 2
y = b /{a ln(l m)(t 1) + b m(mnt l(m + n)(t 1))) + c ln(m n)(t 1) }, (4.2)
z = c2/{a2lm(l n) + b2lm(n m) + c2n(mnt2 l(m(2t 1) + n(t 1)2))}. (4.3)

For the points T and T' to coincide the ratios of their co-ordinates must be equal. This turns out
to be a quadratic equation in t with one root equal to l(n m)/n(l m). However this would put
the point E at infinity, which is not of interest as U, V, W would then coincide with X, Y, Z. This
leaves a unique value of t, which is
t = {l(a2(l(m + n) 2mn) + (n m)(b2m c2n))}/{n(a2l(l m) + b2m(m l) +
c2(l(2m n) mn))}. (4.4)

5. The positions of E, U, V, W when T and T' coincide

The value of t in Equation (4.4) immediately gives us the co-ordinates of E when it assumes its
special position. These are (x, y, z), where
x = a2(l(m + n) 2mn) + (n m)(b2m c2n),
y = a2l(n l) + b2(l(m 2n) + mn) + c2n(l n), (5.1)
2 2 2
z = a l(m l) + b m(l m) + c (mn l(2m n)).

The circle EAB has an equation of the form (2.2) with u = v = 0 and
w = {a4l2 2a2l(b2m + c2n) + b4m2 2b2c2mn + c4n2}/{2(a2l(l m) + b2m(m l)
+ c2(l(2m n) mn))}. (5.2)

This circle meets XYZ at E and again at the special position of W with co-ordinates (x, y, z),
where
x = m(a2l b2m + c2n),
y = l(a2l b2m c2n), (5.3)
2
z = 2lmc .

We now take the point Q to have co-ordinates Q(p, q, r). The circle QWZ has an equation of the
form (2.2), where u, v, w are given by the equations
a2qr + b2rp + c2pq + (p + q + r)(pu + qv + rw) = 0, (5.4)
2
(l m)(lv mu) c lm = 0, (5.5)
4 2 3 2 2 2 2 2 2 2
a (c l m + l (l m)(lv mu)) 2a l(b m(c lm + (l m)(lv mu)) + c (c lmn + l (nv m(v + w))
+ lm2(u + w) m2nu)) + b4m2(c2lm + (l m)(lv mu)) 2b2c2m(c2lmn + l2(m(v + u) nv)
lm2(u + w) + m2nu) + c6lmn2 + c4(l(2m n) mn)(l(2mw nv) mnu) = 0. (5.6)

4
The circle EBC has the form of Equation (2.2) with v = w = 0 and
u = {a4l2 2a2l(b2m + c2n) + b4m2 2b2c2mn + c4n2}/{2(a2(l(m + n) 2mn) +
(n m)(b2m c2n))}. (5.7)

This circle meets XYZ at E and again at the special position of U with co-ordinates (x, y, z),
where
x = 2a2mn,
y = n(a2l + b2m c2n), (5.8)
2 2 2
z = m(a l b m + c n).

The circle QUX has an equation of the form (2.2), where u, v, w are given by the equations
a2qr + b2rp + c2pq + (p + q + r)(pu + qv + rw) = 0, (5.9)
2
(m n)(mw nv) a mn = 0, (5.10)
a l mn a (2b lm n + 2c lmn + (2mn l(m + n))(l(mw + nv) 2mnu)) + a (b m c n)(b2m2n
6 2 4 2 2 2 2 2 2 2

c2mn2 2(l(m2w n2v) mn(m(u + w) n(u + v)))) + (m n)(b2m c2n)2(mw nv) = 0.(5.11)

In the same way it can be shown that circle ECA meets XYZ at E and again at the special
position of V with co-ordinates (x, y, z), where
x = n(a2l + b2m c2n),
y = 2b2nl, (5.12)
2 2 2
z = l( a l + b m + c n).

Once the values of u, v, w have been obtained from Equations (5.4) (5.6) and (5.9) (5.11) the
equation of the common chord of the circles QWZ and QUX may be obtained as
(ux + vy + wz)QWZ = (ux + vy + wz)QUX (5.13)

In the same way the equation of the common chord of circles QWZ and QVY may be obtained
and is found to coincide with the common chord having Equation (5.13). This proves that the
condition for E to have special position on XYZ in order for T and T' to coincide the circles
QUX, QVY, QWZ must be coaxal.

Inverting with respect to Q proves the more general theorem when the eight points D, E, X, Y, Z,
U, V, W lie on a circle and the condition then for T and T' to coincide is that UX, VY, WZ are
concurrent. See Figure 2 below

5
A

Z
U
V

E
P
D
Y X
B C
W
T

Figure 2

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

6
7
Article 37
More Circles in the Cyclic Quadrilateral Configuration
Christopher J Bradley
F

Y
B

S W G
Q
X
O

E
Z
C

Figure

1
1. Introduction

In the figure ABCD is a cyclic quadrilateral, centre O, with AB^CD = E, AD^BC = F, AC^BD =
G so that EFG is the diagonal point triangle. The midpoints of AB, BC, CD, DA, EF are P, Q, R,
S, T respectively. Circles ASP, BPQ, CQR, DRS are drawn and they all pass through O. The
circles ASP, CQR meet at O and Y, circles BPQ, DRS meet at O and Z. It is shown that circle
OYZ passes through G and has centre W, the midpoint of OG. It follows that angles OYG, OZG
are right angles. The lines PR, QS meet at a point X. We show that X is the midpoint of YZ and
that points T, X, Y, Z are collinear. In the sections that follow we use Cartesian co-ordinates with
circle ABCD as unit circle.

2. The four circles through O

The point A has co-ordinates (2a/(1 + a2),(1 a2)/(1 + a2)) and B, C, D have similar co-ordinates
but with parameters b, c, d rather than a. The point P, the midpoint of AB, has co-ordinates (x,
y), where
x = {(a + b)(1 + ab)}/{(1 + a2)(1 + b2)}, (2.1)
2 2
y = {(1 ab)(1 + ab)}/{(1 + a )(1 + b )}. (2.2)
Points Q, R, S have similar co-ordinates but with parameters chosen appropriately. The equation
of the circles ASP is
(1 + a2)(x2 + y2) 2ax (1 a2)y = 0. (2.3)
Circles BPQ, CQR, DRS have similar equations, but with parameters b, c, d rather than a. It is
easy to see that these circles all pass through O, and indeed that their radii are and they lie on
AO, BO, CO, DO as diameters.

3. The points Y, Z and the circle OYZG

The circles ASP and CQR meet at O and at a point Y with co-ordinates as in Equations (2.1) and
(2.2) but with c instead of b. This means that Y is the midpoint of AC. Similarly the circles BPQ
and DRS meet at O and at Z, the midpoint of BD.

The circle OYZ may now be computed and has equation


(abc abd + acd bcd + a b + c d)(x2 + y2) + 2(bd ac)x
+ (abc abd + acd bcd a + b c + d)y = 0. (3.1)

The lines AC and BD meet at the point G with co-ordinates (x, y), where
x = 2(ac bd)/(abc abd + acd bcd + a b + c d), (3.2)
y = (abc abd + acd bcd a + b c + d)/(abc abd + acd bcd + a b + c d). (3.3)

2
It may now be checked that G lies on the circle OYZ and indeed that the centre W of the circle
OYZG is the midpoint of OG. It follows that angles OYG and OZG are each 90o.

4. The midpoint X of YZ is also the intersection of the diagonals of parallelogram


PQRS

From the co-ordinates of Y and Z the co-ordinates of X, the midpoint of YZ may be obtained and
they are (x, y), where
x = a/{2(1 + a2)}, (4.1)
2
y = 1 + 1/{2(1 + a )}. (4.2)

In Equations (4.1) and (4.2) the sum is taken over a, b, c, d. The co-ordinates of P are given by
Equations (2.1) and (2.2) and those of R are similar, but with c, d instead of a, b. The midpoint of
PR may now be computed and it is found to have the same co-ordinates as X. Similarly X is also
the midpoint of QS.

5. The points E, F, T and the collinearity of T, Y, Z

The point E is the intersection of AB and CD and has co-ordinates (x, y), where
x = {2(ab cd)}/{abc + abd acd bcd + a + b c d}, (5.1)
y = {abc + abd acd bcd a b + c + d}/{abc + abd acd bcd + a + b c d}. (5.2)

The point F is the intersection of BC and AD and has co-ordinates (x, y), where
x = {2(bc ad)}/{abc abd acd + bcd a + b + c d}, (5.3)
y = {abc abd acd + bcd + a b c + d}/{abc abd acd + bcd a + b + c d}. (5.4)

The co-ordinates of Y and Z are known, so the equation of YZ is


(a2b2c2 a2b2d2 + a2c2d2 b2c2d2 + 2a2c2 2b2d2 + a2 + c2 b2 d2)x + (abcd(abc abd + acd
bcd) a2b2c + a2b2d + a2bc2 + a2bd2 + a2c2d a2cd2 ab2c2 ab2d2 ac2d2 + b2c2d b2cd2 +
bc2d2 + a2b a2c + a2d ab2 ac2 ad2 b2c + b2d + bc2 + bd2 + c2d cd2 a + b c + d)y + (1
+ ac)(1 + bd)(abc abd bcd + acd +a b + c d) = 0. (5.5)

It may now be shown that T, the midpoint of EF lies on this line.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

3
4
Article 38
Eight points on a line and seven circles through a point
Christopher J Bradley

A
U
T

Z
W
X V

D
E

B C

Figure

1. Introduction

Triangle ABC and a transversal are given and any point D (not on the sides of ABC) is chosen
on the transversal. Circles BCD, CAD, ABD meet the transversal again at points X, Y, Z. Circles
AYZ, BZX, CXY are now drawn and they are found to concur on the circumcircle ABC at a
point T. The process is repeated with a variable point E intead of D, yielding three further points
U, V, W on the transversal. Circles AVW, BWU, CUV are now drawn and, in general, they meet
at a point S on the circumcircle, distinct from T. However, it is shown there are two positions of

1
E when S coincides with T. One is obviously when E coincides with D and the other is at a point
related to E by a formula that we discover. Bearing in mind the results of Articles 19 and 30-32
we may say that circular perspectives are found between ABC and the singular circles XYZ and
UVW that have a common vertex of perspective. The analysis of the situation described above
and illustrated in the figure is developed using areal co-ordinates with ABC as triangle of
reference.

2. The point D, the circle BCD and the point X

We take the transversal to have equation lx + my + nz = 0. Points on this line may be


parameterized by p, with x = mnp, y = nl(1 p), z = lm. We take D to have parameter t, so its
co-ordinates are D(mnt, nl(1 t), lm).

The equation of any circle using areal co-ordinates may be put in the form
a2yz + b2zx + c2xy + (ux + vy + wz)(x + y + z) = 0. (2.1)

In order to find the equation of any particular circle through three given points, one substitutes
the co-ordinates of the points into Equation (2.1) to provide three equations for u, v, w.

For the circle BCD we find v = w = 0 and


u = (a2l2(t 1) b2lmt c2nlt(t 1))/(lmt + nlt(t 1) mnt2). (2.2)

The circle BCD now meets the transversal at a point with parameter s given by s = t (that is at D)
and the point X with parameter
s = a2(l2m + nl2(t 1) lmnt)/(a2ml(l m)(t 1) b2mn(l m)t c2mn(n l)t). (2.3)

The co-ordinates of X may therefore be taken as (x, y, z), where


x = a2(lm + nl(t 1) mnt),
y = (a2l(l n) + b2n(l m)t + c2n(n l)t), (2.4)
z = (a2l(l m)(t 1) b2m(l m)t c2m(n l)t.

3. The circle CAD and the point Y

Following the same approach as in section 2 we obtain for circle CAD the values u = 0, w = 0,
v = (a2lm(t 1) b2m2t c2mnt(t 1))/{(1 t)(lm + nl(t 1) mnt)}, (3.1)

and the co-ordinates of Y as (x, y, z), where


x = a2n(l m)(t 1) + b2m(m n) + c2n(m n)(t 1),
y = b2(mnt lm nl(t 1)), (3.2)
z = a2l(l m)(1 t) + b2m(l m)t + c2l(m n)(1 t).

2
4. The circle ABD and the point Z

In the same way we obtain for circle ABD the values u = 0, v = 0,


w = (a2nl(t 1) b2mnt c2n2t(t 1))/(mnt lm nl(t 1)), (4.1)

and the co-ordinates of Z as (x, y, z), where


x = a2m(l n) + b2m(n m) + c2n(n m)(t 1),
y = a2l(n l) + b2l(m n) + c2n(l n)t, (4.2)
z = c2(mnt lm nl(t 1)).

5. The circle AYZ and its intersection T with the circumcircle

Finding u, v, w for the circle AYZ is similar only technically more complicated. We find u = 0,
v = c2a2nl(l m)(t 1)2 + b2c2m2nt2 b2c2lm2 b2c2lmn(t2 1) + c4nl(m n)(t 1)2 all
divided by l(1 t)(a2(l m)(l n) + b2(n m)(l m) + c2(n m)(n l)), (5.1)
w = a b lm(l n) + b lm(n m) + b c mn t 2b c lmnt + b c lmn + b c ln (t 1)2 all
2 2 4 2 2 2 2 2 2 2 2 2 2 2

divided by l(1 t)(a2(l m)(l n) + b2(l m)(n m) + c2(l n)(m n)). (5.2)

Since the equation of the circumcircle is


a2yz + b2zx + c2xy = 0, (5.3)

it follows that the common chord of circles AYZ and ABC has equation vy + wz = 0, where v
and w are given by Equations (5.1) and (5.2). Solving simultaneously we find the common points
of these two circles are A and a point T with co-ordinates (x, y, z), where
x = a2(a2lm(l n) + b2lm(n m) + c2mn2t2 c2lmn(2t 1) + c2n2l(t 1)2)
.(a2nl(l m)(t 1)2 + b2m2nt2 b2lm2 b2lmn(t 1)2 + c2nl(m n)(t 1)2)),
y = b2(a2lm(l n) + b2lm(n m) + c2mn2t2 c2lmn(2t 1) + c2n2l(t 1)2)
.(a2l2m a2nl2(t 1)2 + a2lmnt(t 2) + b2mn(l m)t2 + c2mn(n l)t2), (5.4)
2 2 2 2 2 2 2 2 2 2 2
z = c (a l m a nl (t 1) + a lmnt(t 2) + b mn(l m)t + c mn(n l)t )
.(a2nl(l m)(t 1)2 + b2m2nt2 b2lm2 b2lmn(t 1)2 + c2nl(m n)(t 1)2)).

It now follows from the general theory of circular perspective that circles BZX and CXY also
pass through thus point.

6. What happens when another point E is chosen and the work repeated

If we now choose a point E with parameter s, instead of D, the work follows as before, but with s
replacing t, and Equations (5.4) produce a distinct point S on the circumcircle. However, if the
ratios of the co-ordinates of S and T are equal it is found that there are four solutions for s, two of
which are complex, s = t, of course, when E coincides with D, but there is another point E on the

3
transversal when circles AVW, BWU, CUV all pass through T. The value of s for which this
occurs is
s = (b2mt c2nt a2l(t 2))/(a2l b2m + c2n(1 2t)). (6.1)

See the figure again, in which the point Y is off the paper.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 39
The 60o, 75o, 45o Triangle and its Euler line
Christopher J Bradley

75
R

O
60
45
Q A
C

Figure

1. Introduction

Let ABC be a 60o, 75o, 45o triangle and suppose that the Euler line OH, where O is the
circumcentre and H the orthocentre, meets the sides BC, CA, AB at P, Q, R respectively. We
prove that PR = RQ and RH = OQ. It is also the case that AQ = AR so that angle BPR = 15o.

2. The sides of the triangle and u, v, w

1
The sides are proportional to the sines of the angles of the triangle and may therefore be taken as
a = 3, b = (6 + 2), c = 2. In the sections that follow we use the co-ordinates of H(u, v,
w), where
u = 1/(b2 + c2 a2), v = 1/(c2 + a2 b2), w = 1/(a2 + b2 c2). (2.1)

3. The Euler line and the points P, Q, R

The Euler line passes through H(u, v, w) and O(v + w, w + u, u + v) and so has the equation
(v w)x + (w u)y + (u v)z = 0. (3.1)

The Euler line meets BC, x = 0, at P(0, 1 + (1/3)3, (1/3)3). It meets CA, y = 0 at Q(1
(1/3)3, 0, (1/3)3), and it meets AB, z = 0 at R( (1/6)3, (1/6)3 + , 0). It is clear that PR
= RQ, since xP + xQ = 2xR.

4. O, H and RH = OQ

The values of u, v, w given in Equation (2.1), which when normalized, are the co-ordinates of
H(2 3, (1/3)3, (2/3)3 1) and those of O are (3 , (1/6)3, 1 (1/3)3).

It now follows that xH + xO = xQ + xR so that RH = OQ.

From the co-ordinates of Q and R and the side lengths b and c it immediately follows that
triangle RQA is equilateral and hence that angle BPR = 15o.

5. The circumcircle of ABC, Circle CQH and the point U

The equation of the circumcircle, using u, v, w, is


u(v + w)yz + v(w + u)zx + w(u + v)xy = 0. (5.1)

This becomes
(93 15)yz + (3 1)zx + (63 10)xy = 0. (5.2)

The equation of the circle CQH is


(3 1)x2 + (33 193)y2 + 2(3 23)yz + 2(3 2)zx + 2(7 43)xy = 0. (5.3)

These two circles meet at C and again at a point U with co-ordinates


(1/13)(7 53), (11//3)3 + 7, (4/3)3 1).

The point U is a key point in this configuration.

2
6. The circles BRH, COQ and the points S and T

The equation of the circle BRH is


2(3 2)x2 + 2(53 9)z2 + (9 53)yz + (113 19)zx + 2(7 43)xy = 0. (6.1)

It may now be verified that this circle passes through U. Also it meets BC at B and the point S
with co-ordinates S(0, 2/3, 1/3). That is, where SC = 2BS. This is a pleasing and unexpected
result.

The circle COQ has equation


(3 1)x2 + (9 53)y2 + 2(53 9)yz + 2(3 2)zx + 2(33 5)xy = 0. (6.2)

It may now be verified that this circle passes through the point S. It also meets the circumcircle
at a point T with co-ordinates (1/6)(3 33, 3 + 3, 23). The points T, S, H turn out to be
collinear.

U
B
75
T R

S
H

45 60
Q A
C

Figure

7. Circles AOU and PUQ pass through S

The equation of circle AOU is


3
2(7 43)y2 2(3 1)z2 + (173 29)yz + (3 1)zx + 2(53 8)xy = 0. (7.1)

It may now be verified that this circle passes through S.

The equation of circle PUQ is


(3 1)x2 + (33 5)y2 + 4(3 2)z2 + 2(7 43)yz + 2(23 3)zx + 6(2 3)xy = 0. (7.2)

And this circle also passes through S.

8. Circle ABH passes through P

The equation of circle ABH is


2(3 23)z2 + (53 9)yz + (5 33)zx + 2(33 5)xy = 0. (8.1)

It may now be verified that this circle passes through P.

Finally it may now be verified that circles ABH and PUQ touch at P.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 40
The 120o, 45o, 15o Triangle and its Euler Line
Christopher J Bradley

30
30
15
15 15

75 O

30
15

P
105
60
S
U 60

90
T
60 120
R 30
120 A
135 45
30 B

30 15
15

Figure
H

1
1. Introduction

We find that the triangle with angles A, B, C equal to 120o, 45o, 15o respectively together with its
Euler line OH, where O is the circumcentre and H the orthocentre, exhibits a number of
interesting results. We define P, Q, R to be the points OH^BC, OH^CA, OH^AB respectively.
We prove in what follows the following results:
(i) PQ = QR;
(ii) OQ = RH;
(iii)Triangle AQR is equilateral;
(iv) If U is the fourth vertex of parallelogram RAQU, then U lies on the circumcircle of ABC;
(v) Circle PAC passes through H;
(vi) Circle CBH passes through O;
(vii)Circle COQ passes through U;
(viii)Circle COQ meets BC again at a point S such that CS = 2SB;
(ix)Circle OBR passes through U;
(x) Circle APR passes through U and S.

As a consequence of these results and angle-chasing angles between various lines are as shown
in the figure above. In the following areal-co-ordinates are used with u, v, w instead of a, b, c,
where
u = 1//(b2 + c2 a2), v = 1/(c2 + a2 b2), w = 1/(a2 + b2 c2) (1.1)
and ABC as triangle of reference.

2. Values of a, b, c and the co-ordinates of H and O

The sides of ABC are proportional to the Sines of 120o, 45o, 15o so we set a = 3, b = 2 and
c = (6 2) Then, from Equations (1.1) we find, after normalization so that u + v + w = 1,
that
u = 2 + 3, v = 1 (23)/3, w = (3)/3. (2.1)
These are now the normalized co-ordinates of the orthocentre, H.

The co-ordinates of the circumcentre O are now (v + w, w + u, u + v) = (x, y, z) so that


x = (1 + 3), y = 1 + (1/3)3, z = (1/6)(3 + 3). (2.2)

3. The Euler line and P, Q, R

The equation of the Euler line through O, H is known to be


(v w)x + (w u)y + (u v)z = 0, (3.1)

which in this case is

2
(1 + 3)x + (4 + 23)y (5 + 33)z = 0. (3.2)

The co-ordinates of P, Q, R, where the Euler line meets BC, CA, AB respectively are
P(0, (1/3)3, 1 (1/3)3), Q((1/6)(3 + 3), 0, (1/6)(3 3)), R(1 + (1/3)3, (1/3)3, 0). (3.3)

4. Results (i) (iv) and the point U

From the co-ordinates of P, Q, R it may be seen that P + R = 2Q, so that PQ = QR. Similarly O +
H = R + Q, so that OQ = RH.

From the co-ordinates of A, Q, R we find AQ = AR = (1/12)(32 6) so that triangle AQR is


equilateral. Since U is defined as the fourth vertex of the parallelogram RAQU, its co-ordinates
are given by U = Q + R A. that is
U((1/2)(1 + 3), (1/3)3, (1/6)(3 3)). (4.1)

It may now be checked that U lies on the circumcircle of triangle ABC, whose equation is
3(1 + 3)yz + 2(1 + 3)zx + (3 1)xy = 0. (4.2)

5. Results (v) (viii) and the point S

Circle PAC has equation


23y2 (3 + 3)yz 2(1 + 3)zx + (1 + 3)xy = 0. (5.1)

It may now be checked that H lies on this circle.

Circle CBH has equation


2x2 + 3(1 + 3)yz + 2(2 + 3)zx + (1 + 3)xy = 0. (5.2)

It may now be checked that O lies on this circle.

Circle COQ has equation


2x2 + (3 + 3)y2 2(3 + 3)yz 2(2 + 3)zx + 2(1 + 3)xy = 0. (5.3)

It may now be checked that U lies on this circle. Also the circle COQ meets BC, x = 0, at the
point S with co-ordinates (0, 2/3, 1/3) so that CS = 2BS, a rather unexpected trisection of the side
BC.

6. Results (ix) and (x)

Circle BOR has equation

3
(3 1)x2 (3 + 3)z2 + 2(3 + 3)yz + 2(1 + 3)zx + 2xy = 0. (6.1)

It may now be checked that U lies on this circle.

Circle APR has equation


2y2 + 2(1 + 3)z2 (5 + 3)yz 4zx + (3 1)xy = 0. (6.2)

It may now be checked that both U and S lie on this circle.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 41
The 105o, 60o, 15o Triangle and its Euler Line and Circles
Christopher J Bradley

30
15

15

60
U 120 R

45
P

60 105 60
Q A
C

Figure

1. Introduction

1
In the above figure we display the 105o, 60o, 15o triangle and its Euler line OH, where O is the
circumcentre and H is the orthocentre, together with several circles, the square UOAH and
several other features. In the following sections we prove the following results for this
configuration:
(i) If the Euler line meets the sides BC, CA, AB respectively at points P, Q, R then HQ =
RO;
(ii) The co-ordinates of U are obtained, where AP = PU;
(iii) U lies on the circumcircle ABC;
(iv) Circle BCH passes through O;
(v) UOAH is a square;
(vi) Triangle BUO is equilateral.

Areal co-ordinates are used throughout with ABC as triangle of reference. The figure also
contains other details and some angles, which can easily be established from the data and the
above results.

2. The sides a, b, c and the co-ordinates of H and O

The side lengths a, b, c may be chosen equal to Sin A, Sin B, Sin C respectively, and so
a = 3, b = (6 2), c = (6 + 2). (2.1)

The co-ordinates of H are (u, v, w), where u = 1/(b2 + c2 a2), v = 1/(c2 + a2 b2), w = 1/(a2 + b2
c2) and after normalization these are
u = 1, v = 1 (2/3)3, w = 1 + (2/3)3. (2.2)

The co-ordinates of O are


(v + w, w + u, u + v) = (1, (1/3)3, (1/3)3). (2.3)

The equation of the Euler line OH may now be determined and is


2x (3 + 1)y + (3 1)z = 0. (2.4)

3. The co-ordinates of P, Q, R and results (i) and (ii)

Solving Equation (2.4) with x = 0, we obtain P, whose co-ordinates are therefore


x = 0, y = (1/6)(3 3), z = (1/6)(3 + 3). (3.1)

Solving Equation (2.4) with y = 0, we obtain Q, whose co-ordinates are therefore


x = (1/3)3, y = 0, z = 1 + (1/3)3. (3.2)

Solving Equation (2.4) with z = 0, we obtain R, whose co-ordinates are therefore

2
x = (1/3)3, y = 1 (1/3)3, z = 0. (3.3)

Since the co-ordinates of H, Q, R, O satisfy H + O = Q + R, HQ = RO, which is result (i).

If U is on AP and is such that P is the midpoint of AU, then the co-ordinates of U satisfy U = 2P
A and are therefore
x = 1, y = 1 (1/3)3, z = 1 + (1/3)3. (3.4)

4. Results (iii) and (iv)

The equation of the circumcircle ABC is


u(v + w)yz + v(w + u)zx + w(u + v)xy = 0, (4.1)

which becomes
3yz + (2 3)zx + (2 + 3)xy = 0. (4.2)

It may now be checked that U lies on the circumcircle.

The equation of the circle BCH is


x2 3yz + (3 1)zx (3 + 1)xy = 0. (4.3)

It may now be checked that O lies on this circle. This is always the case when angle A = 60o.

5. Results (v) and (vi)

We already know that P is the midpoint of both OH and AU, so UOAH is a parallelogram. We
next show that AO = AH. The displacement AO is (0, (1/3)3, (1/3)3) and the displacement
AH is ( 2, 1 (2/3)3, 1 + (2/3)3). Now given a displacement d = (f, g, h), the square of its
length is known to be d2 = a2gh b2hf c2fg. From equation (2.1) we have
a2 = , b2 = (2 3), c2 = (2 + 3). (5.1)

It soon follows that AO2 = AH2 = , so UOAH is a rhombus. Finally the displacement HO = (2,
3 1, 3 1), so that HO2 = . Thus angle OAH = 90o and UOAH is a square.

It is now easy to show BO2 = BU2 = and hence triangle BOU is equilateral.

Other features indicated by Cabri II plus are that U is the centre of a circle passing through R
and Q; R is the centre of a circle passing through U, A and Q; and R is the centre of a circle
passing through B and H It soon follows that RUQ is an equilateral triangle as well as RQA.

3
Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 42
A Cyclic Quadrilateral, 29 Points and 33 Lines

Christopher J Bradley

ac

To ac

ab

T
Sc B S
ESd
Y G U W V
ad bc
F Sa P
Sb Q C
R

D
cd
bd

To Z

To H

Z
1. Introduction

In the Figure ABCD is a cyclic quadrilateral, Sa, Sb, Sc, Sd are the symmedian points of triangles
BCD, ACD, ABD, ABC respectively, ab is the intersection of the tangents to circle ABCD at A

1
and B, with ac, ad, bc, bd, cd similarly defined. By the definition of a symmedian point Sa is the
point of concurrency of C bd, B cd and D bc and similarly for Sb, Sc, Sd.

A catalogue of the other 15 points in the figure is:


P = AB^CD, Z = AD^BC, U = AC^BD (so UZP is the diagonal point triangle of ABCD), Q =
ASc^BSd, R = ASd^CSb, S = BSc^DSa, T = CSa^DSb, V = CSd^BSa, W = CSa^BSd, X = ASb^CSd,
Y = ASb^DSc, E = BSc^DSa, F = DSa^CSb, G = DSb^ASc, H = BSa^DSc.

It is proved in the following sections that these 29 points lie severally on 33 lines as follows:
ABP, CDP, SaSbP, ScSdP, ad bc P, ac bd P are 6 lines concurrent at P;
ACU, BDU, SaScU, SbSdU, ad bc U, ab cd U are 6 lines concurrent at U;
ADZ, BCZ, SbScZ, SaSdZ, ab cd Z, ac bd Z are 6 lines concurrent at Z;
The diagonal line ZP contains the 6 additional points H, S, R, X, bd, ac;
The diagonal line PU contains the 6 additional points Y, G, W, V, bc, ad;
The diagonal line UZ contains the 6 additional points E, Q, F, T, cd, ab;
There are 12 additional lines all of the same type containing 5 points each:
ASbXYcd, BScESad, CSdVXab, DSaFSbc, ASdERbc, BSaVHcd, CSbFRad, DScYHab, AScGQbd,
BSdWQac, CSaWTbd, DSbGTac.

Each point considered lies on at least three lines. Proof s are established using Cartesian co-
ordinates, making full use of symmetry.

2. The Symmedian Points

We take ABCD to be the unit circle with equation x2 + y2 = 1 and A to have co-ordinates
A(2a/(1 + a2), (1 a2)/(1 + a2)), with B, C, D having similar co-ordinates, but with parameters b,
c, d instead of a. It is well known that chord AB has equation
(a + b)x + (1 ab)y = (1 + ab), (2.1)

and that the tangent at A has equation


2ax + (1 a2)y = (1 + a2). (2.2)

Other chords and tangents of ABCD may be written down with suitable change of parameters.
The tangents at A and B meet at the point ab with co-ordinates ((a + b)/(1 + ab), (1 ab)/(1 +
ab)). The co-ordinates of ac, ad, bc, bd, cd may also be written down by suitable change of
parameters.

The line ad B has equation


2(ad b2)x + (b2(a + d) 2abd + a + d 2b)y + b2(a + d) 2abd a d + 2b = 0. (2.3)

2
The line bd A has the form of Equation (2.3) but with a and b interchanged. As is known these
lines meet at the symmedian point Sc of triangle ABD, which therefore has co-ordinates (x, y),
where
x = (1/k)(a2(b + d) + a(b2 6bd + d2) + bd(b + d)),
y = (1/k)(a2(b2 bd + d2 1) a(b2d + bd2 b d) + b2d2 b2 + bd d2), (2.4)
2 2 2 2 2 2 2 2 2
k = a (b bd + d + 1) a(b d + bd + b + d) + b d + b bd + d .

The co-ordinates of other symmedian points may be obtained by appropriate change of


parameters.

3. The diagonal point triangle

The chords AB and CD meet at the point P with co-ordinates 0028, y), where
x = (2/h)(ab cd),
y = (1/h)(abc + abd acd bcd a b + c + d), (2.5)
h = (abc + abd acd bcd + a + b c d).

When the equation of the line ScSd has been obtained it may be checked that it passes through P.
Similarly SaSb passes through P. It follows by symmetry that ABCD and SaSbScSd have the same
diagonal point triangle PUZ, where U = AC^BD and Z = AD^BC. The co-ordinates of U and Z
follow from Equation (2.5) by suitable change of parameters.

The equation of the diagonal line ZP may now be obtained and is


2(ac bd)x (abc abd + acd bcd a + b c + d)y + ( abc + abd acd + bcd a + b c + d)
= 0. (2.6)
The equations of the other diagonal lines may now be obtained by suitable change of parameters.

4. Other points on ZP

It may now be checked that both bd and ac lie on ZP.

The equation of A cd is
2(a2 cd)x ((a2 + 1)(c + d) 2a(1 + cd))y ((a2 + 1)(c + d) + 2a(cd 1)) = 0. (4.1)

It may be checked that Sb lies on this line.

The equation of C ab may be found from Equation (4.1) by interchanging a and c and also b and
d. The lines A cd and C ab meet at the point X with co-ordinates (x, y), where
x = (1/k)(a2(c + d) + a(c2 3c(b + d) + bd) + bc(c + d)),
y = (1/k)(a2(b(c d) c2 + 1) + a(b + c)(cd 1) + bd(1 c2) + c(c d)), (4.2)

3
k = a2(b(c d) c2 1) + a(b + c)(cd + 1) bd(1 + c2) + c(c d).

It may now be checked that X lies on ZP. The disposition of the remaining points Q, R, S, T, V,
W, Y, E, F, G, H on the diagonal lines now follows by suitable change of parameters.

Flat 4
Terrill Court,
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 43
The Altitudes and Radii of a triangle and its circumcircle
Christopher J Bradley

ca ab
R
cb
V
O P
N
bc H
Q
ba ac

B C

1. Introduction

In the figure a scalene triangle ABC is shown, with its altitudes AH, BH, CH and the radii AO,
BO, CO of its circumcircle. The point ab is the intersection of AH and BO, with similar
definitions of points ac, bc, ba, ca, cb. The point P is the midpoint of the points ab, ac lying on
the altitude AH. Points Q and R are similarly defined on BH and CH respectively. In this article
we show that circle PQR is the circle on OH as diameter and circles ab bc ca and ba ac cb both
pass through H. The first circle has centre N, the nine-point centre, and the centres of the other
two circles are collinear with N and equidistant from N. Cartesian co-ordinates are used to
establish these results, with circle ABC as the unit circle.

2. Lines AH, BH, CH, AO, BO, CO and the point H

The vertex A is taken to have co-ordinates (2a/(1 + a2), (1 a2/(1 + a2)) and B and C have similar
co-ordinates with parameters b, c rather than a. The chord BC has equation
(b + c)x + (1 bc)y = (1 + bc). (2.1)

The line perpendicular to this through A is the line AH, which has equation
(1 + a2)(1 bc)x (1 + a2)(b + c)y = a2(b + c) + 2a(1 bc) b c. (2.2)

BH and CH have similar equations with appropriate cyclic change of co-ordinates. Their
intersection H has co-ordinates (x, y), where
x = (2/k)(a2(b + c)(1 + bc) + a(1 + b2)(1 + c2) + (1 + bc)(b + c)),
y = (1/k)(a2(b2(3c2 + 1) + c2 1) +b2(c2 1) c2 3), (2.3)
where
k = (1 + a2)(1 + b2)(1 + c2). (2.4)

The equation of AO is (1 a2)x = 2ay, with similar equations for BO and CO with parameters b
and c respectively, rather than a.

3. The points ab and ac, the points P, Q, R

The points ab and ac are the intersections with AH of the radii OA and OB respectively. Thus ab
has co-ordinates (x, y), where x = 2bh, y = (1 b2)h, where
h = ((a2 1)(b + c) + 2a(1 bc))/((1 + a2)(1 + b2)(b c)). (3.1)

The point ac has similar co-ordinates, but with b and c exchanged. The midpoint P of ab and ac
has co-ordinates (f(1 bc), f(b + c)), where
f = ((a2 1)(b + c) + 2a(1 bc))/((1 + a2)(1 + b2(1 + c2)). (3.2)

Point Q, R on BH, CH have similar co-ordinates with appropriately chosen cycic changes of a, b,
c.

4. The circle PQR


The equation of the circle PQR may now be obtained and has equation
(1 + a2)(1 + b2)(1 + c2)(x2 + y2) 2((1 + a2) (1 + bc)(b + c) + a(1 + b2)(1 + c2))x
+ (a2(b2(3c2 + 1) + c2 1) +b2(c2 1) c2 3)y = 0. (4.1)

It is obvious that this circle passes through O and it may be checked that it also passes through H.
Indeed from Equations (2.3) and (2.4) it may be observed that its centre is midway between O
and H, the nine-point centre N.

5. The circle ab bc ca

The co-ordinates of the three points ab, bc, ca follow from Equation (3.1) and so the equation of
the circle through these points may be calculated. The working is technically difficult, and the
equation is very lengthy, and we relied, as in most of the previous Articles, on the algebra
computer package DERIVE, and the geometry computer package CABRI II plus for drawings and
for forecasting results. The equation of the circle is
(1 + a2(1 + b2) (1 + c2)(a b)(b c)(c a)(x2 + y2) + 2(a4(b4c2 + b3c(1 c2) + b2(1 + c2)2 +
bc(1 c2) c2) a3(1 + c2)(b4c + b3(1 + c2) + 2b2c b(1 + c2) + c) + a2(b4(c4 + 2c2 1) + 2b3c(1
c2) + 2b2(c2 + 1)(c2 1) + 2bc(1 c2) + c4 2c2 1) + a(1 + c2)(b4c b3(1 + c2) + 2b2c + b(1 +
c2) + c) + b4c2 + b3c(1 c2) b2(1 + c2)2 + bc(1 c2) c2)x + 2(a4(b4c 2b3c2 + b2c (1 + c2) + b(1
+ c4) + c3) + a3(b4 2b2c2 + 1)(1 + c2) a2(2b4c3 b3(1 c2)2 + 2b2c(1 + c2) b(c4 2c2 + 1) +
2c) + a(1 + c2)(b4c2 2b2 + c2) + b4c + b3(1 + c4) + b2c(1 + c2) 2bc2 + c3)y + (a(b2 2bc 1) +
b2c + 2b c)(a(2bc c2 + 1) + b(1 c2) 2c)(a2(b + c) + 2a(1 bc) b c) = 0. (5.1)

The circle ba ac cb has a similar equation with, say, a and c interchanged. It has been checked
that both these circles pass through H, but the other assertions about their centres has not been
checked, though Cabri is so accurate there is no doubt about them.

Flat 4
Terrill Court
12-14 Apsley Road
Clifton
BRISTOL BS8 2SP
Article 44

A natural Enlargement and a resulting Collineation

Christopher J Bradley

A'

feu A

Feu

k
X
g G K H
h

B C

B'

C'

1. Introduction

Given a triangle ABC, draw the angle bisectors to meet its circumcircle and at those three points
draw the tangents to form a triangle A'B'C', which, since the three points are the midpoints of the
minor arcs BC, CA, AB, is a triangle similar to ABC. In fact it turns out that triangle A'B'C' is
an enlargement from a point X, the enlargement factor being R/r, the circumcircle of ABC being
the incircle of A'B'C'. In this short article we find the point X, which has the property that if J is
any triangle centre of ABC and j the corresponding triangle centre in A'B'C' then jJX is a straight
line. In the Figure above the lines hHX, gGX, kKX are shown. Remarkably it turns out that the
Feuerbach point Feu lies on hHX, so that in any triangle Feu, X and H are collinear. The Figure
illustrates this and displays the nine-point circles and incircles which are tangent at feu and Feu.

Of course, one can produce an enlargement of a triangle about any point you want, but as this
particular enlargement arises naturally the point X is of special significance, and we call it the
centre of centres. It is actually X56, the Exsimilicenter (or External Similitude Centre) [1] and so
this collineation is already known. For this reason this article is kept very brief.

2. The equations of the sides of A'B'C'

The equation of the line joining the circumcentre O and the midpoint L of BC is (b2 c2)x + a2(y
z) = 0. This meets the circumcircle at the point with co-ordinates ( a2, b(b + c), c(b + c)). The
tangent to the circumcircle at this point is the line B'C' with equation
(b + c)2x + a2(y + z) = 0. (2.1)

The equations of C'A' and A'B' follow by cyclic change of a, b, c. These lines intersect at The
point A' with co-ordinates (a(a2 + ab + 2bc + ca), b2(a + b c), c2(c + a b)).

3. The co-ordinates of X56, the Exsimilicenter

The equation of AA' is therefore


c2(c + a b)y = b2(a + b c)z. (3.1)

The three lines AA', BB', CC' meet at X56, the Exsimilicenter (or External Similitude Centre),
whose co-ordinates are (a2/(b + c a), b2/(c + a b), c2/(a + b c)).

4. The Collineation X4X11X56

The co-ordinates of Feuerbachs point Feu or X11 are ((b + c a)(b c)2, (c + a b)(c a)2, (a +
b c)(a b)2) and it is now easily shown that the orthocentre X4, Feuerbachs point X11 and X56
are collinear.

Reference
1. C. Kimberling, Major Centers of Triangles," Amer. Math. Monthly 104 (1997) 431-438.
Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP
Article 45
On the Circles defining the Brocard Points

Christopher J Bradley

Z'

Y' R
X
O V P

' H
X'
Z Q Y

B C

Figure 1

1. Introduction

1
Since writing Article 43 a number of further results have emerged in connection with its
configuration. However, this article may be read independently, without recourse to any previous
results. I am extremely grateful to David Monk [1] for pointing out the connection between the
configuration in Article 43 and the circles defining the Brocard points.

In Figure 1 we display a triangle ABC with circumcentre O and orthocentre H. The points X =
CH^BO, Y = AH^CO, Z = BH^AO. The points X' = BH^CO, Y' = CH^AO, Z' = AH^BO. The
point P is the midpoint of YZ', Q is the midpoint of ZX' and R is the midpoint of XY'.

We prove the following results:


(i) Circle PQR has OH as a diameter;
(ii) Circle XYZ passes through H and , the first Brocard point;
(iii) Circle X'Y'Z' passes through H and ', the second Brocard point;
(iv) Lines XX', YY', ZZ' are concurrent at a point V.
(v) Triangle XYZ is similar to triangle CAB and triangle X'Y'Z' is similar to triangle
BCA (where ordering of letters is important).

In Figure 2 we display, in addition, circles BCX, CAY, ABZ, BCX', CAY', ABZ' and we prove
the following results:
(vi) Circles BCX, CAY, ABZ all pass through ;
(vii) Circles BCX', CAY', ABZ' all pass through '.
Thus these circles are the defining circles of the Brocard points. So, circle ABZ touches BC at B,
circle BCX touches CA at C, and circle CAY touches AB at A. And circle ABZ' touches CA at
A, circle BCX' touches AB at B and circle CAY' touches BC at C.

In Figure 3 we display, instead of the six circles in Figure 2, the following six circles: AYZ,
BZX, CXY, AY'Z', BZ'X', CX'Y' and we prove the following result:
(viii) All these six circles pass through the same point D lying on the circumference of the
circumcircle of ABC.

The point D is Tarrys point, associated with the Brocard configuration, and has the following
additional properties:
(ix) AH^BD = E lies on circle CXY, BH^CD = F lies on circle AYZ, CH^AD = G lies on
circle BZX, BH^AD = T lies on circle CX'Y', CH^BD = U lies on circle AY'Z' and
AH^CD = S lies on circle BZ'X'.

Apart from Results (i) and (v), the above results are proved using areal co-ordinates with ABC as
triangle of reference.

2. The points P, Q, R, X, Y, Z, X', Y', Z' and the circle PQR

2
Result (i)
BO and OC are equally inclined to AH, so triangle OYZ' is isosceles. OP is therefore at right
angles to AH. Similarly OQ is at right angles to BH and OR is at right angles to CH. It follows
that P, Q, R lie on the circle with OH as diameter.

The co-ordinates of X, Y, Z, X', Y', Z'

The equation of AH is
(c2 + a2 b2)y = (a2 + b2 c2)z (2.1)

and the equation of BO is


c2(a2 + b2 c2)x = a2(b2 + c2 a2)z. (2.2)

These lines meet at the point Z' with co-ordinates (x, y, z), where
x = a2(b2 + c2 a2)(c2 + a2 b2),
y = c2(a2 + b2 c2)2, (2.3)
z = c2(a2 + b2 c2)(c2 + a2 b2).

The co-ordinates of points X' and Y' follow from those of Z' by cyclic change of both a, b, c and
x, y, z.

The equation of CO is
b2(c2 + a2 b2)x = a2(b2 + c2 a2)y. (2.4)

This line meets AH, with Equation (2.1) at the point Y with co-ordinates (x, y, z), where
x = a2(b2 + c2 a2)(a2 + b2 c2),
y = b2(c2 + a2 b2)(a2 + b2 c2), (2.5)
z = b2(c2 + a2 b2)2.

The co-ordinates of points Z and X follow from those of Y by cyclic change of both a, b, c and x,
y, z.

3. The circle XYZH and result (ii)

From the co-ordinates of X, Y, Z|, see Equation (2.5) we may obtain the equation of the circle
XYZ, which is
b2(b2 + c2 a2)(c2 + a2 b2)x2 + c2(c2 + a2 b2)(a2 + b2 c2)y2 + a2(a2 + b2 c2)(b2 + c2 a2)z2
+ (b2 + c2 a2)(a2b2 + b2c2 2c2a2 b4)yz + (c2 + a2 b2)(b2c2 + c2a2 2a2b2 c4)zx
+ (a2 + b2 c2)(c2a2 + a2b2 2b2c2 a4)xy = 0. (3.1)

3
It may now be checked that both H and (1/b2, 1/c2, 1/a2) lie on this circle.

4. The circle X'Y'Z'H' and result (iii)

From the co-ordinates of X', Y', Z', see Equation (2.3) we may obtain the equation of the circle
X'Y'Z', which is
c2(a2 + b2 c2)(b2 + c2 a2)x2 + a2(b2 + c2 a2)(c2 + a2 b2)y2 + b2(c2 + a2 b2)(a2 +b2 c2)z2
+ (a2 + b2 c2)(a2b2 + b2c2 2c2a2 b4)yz + (b2 + c2 a2)(b2c2 + c2a2 2a2b2 c4)zx
+ (c2 + a2 b2)(c2a2 + a2b2 2b2c2 a4)xy = 0. (4.1)

It may now be checked that both H and '(1/c2, 1/a2, 1/b2) lie on this circle.

5. Result (iv)

If one considers the point V with co-ordinates (x, y, z), where


x = a2(b2 + c2 a2)(a4b2 + a4c2 2a2b4 + 4a2b2c2 2a2c4 + b6 b4c2 b2c4 + c6), (5.1)
and with y, z obtained by cyclic change of a, b, c, then it may now be verified that X, X', V are
concyclic, that Y, Y', V are concyclic, and that Z, Z', V are concyclic.

6. Result (v)

It is a simple piece of angle chasing that result (v) holds, namely triangle XYZ is similar to
triangle CAB and triangle X'Y'Z' is similar to triangle BCA (where ordering of letters is
important). In fact, it is true for any circle passing through H that the intersections of the circle
with the altitudes through H form a triangle similar to ABC.

7. The six circles like AZB

The equation of the circle AZB is easily calculated and is


a2z2 + (a2 b2)zx c2xy = 0. (7.1)

The exciting fact about this circle are that it is tangent to BC at B, and is therefore one of the
defining circles of a Brocard point. And indeed it is easy to check that it passes through ,
whose areal co-ordinates are (1/b2, 1/c2, 1/a2).

In similar fashion it may be shown that circles BXC, CYA both pass through , and that circles
AZ'B, BX'C, CY'A all pass through '(1/c2, 1/a2, 1/b2).

4
A

Z'
Y' R

X
O P
V

' H
X'
Q Y
B Z
C

Figure 2

8. The six circles like AYZ

The equation of the circle AYZ may now be obtained and is


c2(a2 + b2 c2)(a4 a2b2 c2b2 + c4)y2 c2(a2 + b2 c2)(a4 a2c2 b2c2+ b4)z2
+ (c2 b2)(a6 2a4b2 a4c2 + a2b4 + a2c4 + c2b4 c6)yz (a6c2 + a4(b4 2c4) + a2(c2 b2)(2b4
+ b2c2 + c4) + b4(b2 c2)2)zx + c2(a6 a4b2 + a2(b4 c4) b2(b2 c2)2)xy = 0. (8.1)

5
G

D
A

U
Z'

R S
Y'
X T
O V P

' H F
X' Y
Z Q

B C

Figure 3

This circle meets the circumcircle at A and the point D with co-ordinates (x, y, z), where
x = 1/(b4 + c4 a2(b2 + c2)),
y = 1/(c4 + a4 b2(c2 + a2)), (8.2)

z = 1/(a4 + b4 c2(a2 + b2).

This point lies similarly on the five circles BZX, CXY, AY'Z', BZ'X', and CX'Y'.

The point D is Tarrys point, which is diametrically opposite the Steiner point, but in this
configuration it has a further interesting property. See result (ix) above, which we now establish.

The line AD has equation


(c4 + a4 b2(c2 + a2))y = (a4 + b4 c2(a2 + b2))z. (8.3)

6
The line AD meets AH with Equation (2.1) at the point G with co-ordinates (x, y, z), where
x = (b2 c2 a2)(a4 + b4 c2(a2 + b2)),
y = (a2 b2 c2)(a4 + b4 c2(a2 + b2)), (8.4)
2 2 2 4 4 2 2 2
z = (a b c )(c + a b (c + a )).

It may now be checked that G lies on circle BZX, which has an equation similar to Equation
(8.1) with cyclic change of a, b, c and x, y, z.

Similarly it may be proved that AH^BD = E lies on circle CXY, BH^CD = F lies on circle AYZ,
BH^AD = T lies on circle CX'Y', CH^BD = U lies on circle AY'Z' and AH^CD = S lies on circle
BZ'X'.

Reference
1. D. Monk, private communication.

Flat 4,
Terrill Court,
12-14, Apsley Road,
BRISTOL BS8 2SP.

7
A Mean of two Cevian Points and the Construction of Triangle Centres
Christopher J Bradley

P
D'
R
E'
E D
Q

F'

B C
Figure

1. Introduction

The idea of a triangle centre is well known. Over 3000 such points have been investigated and
tabulated [1]. There are, of course, an infinity of triangle centres because every point on a line
joining any two of them is also a triangle centre. For example, the Euler line collapses to the
centroid when a triangle becomes equilateral. So the nine-point centre is a triangle centre, and
one knows this without the need to work out its co-ordinates. Obviously there has to be some
additional reason as to what actually qualifies as a triangle centre to make it worth tabulating.
Presumably it must involve a construction of some kind relating it to the triangle or to other
triangle centres that makes it significant or acceptable as an additional triangle centre. The aim of
this short article is to show that this is a very difficult task. My understanding is that Clark
Kimberling decides when a point is significant. And the reason is not presumably that it lies on a
line joining any two of them! What we do is to choose two points, produce a significant
construction to provide a third point, which is a triangle centre if the two originally chosen ones
are. Does the third point qualify or not? If the two chosen points are the centroid and the
orthocentre the third point turns out to be the symmedian point. Does this convince one that the
third point should qualify? If it does, remember there are already about 3000 triangle centres, so
this would only add about 4.5 million.

2. The construction

We choose any two points P and Q, not on the sides of the triangle and not so as to be on the
same line through a vertex. Let P have co-ordinates (p, q, r) and Q have co-ordinates (u, v, w).
AP has equation qz = ry and BQ has equation wx = uz. These lines meet at the point X(ur, wq,
wr). BP^CQ = Y(up, vp, ru) and CP^AQ = Z(vp, vq, wq). Next we interchange P and Q and
form X' = AQ^BP with co-ordinates (wp, vr, wr), BQ^CP = Y'(up, uq, wp), CQ^AP = Z'(uq, vq,
vr).

The equation of XX' may now be calculated and is found to be


wr(wq vr)x + wr(wp ur)y + (r2uv w2pq)z = 0. (2.1)

The equation of YY' may be obtained from this by cyclic change of x, y, z and p, q, r and u, v, w.

We now obtain the point of intersection R with co-ordinates


x = up(wq + vr),
y = vq(wp + ur), (2.2)
z = wr(vp + uq).

See the figure above. It is easy to see that if v, w u and q, r p , so that P, Q are triangle centres,
then R is also a triangle centre. In particular if P is the centroid then R has co-ordinates (u(v +
w), v(w + u), w(u + v)), and further if Q is the orthocentre, then R is the symmedian point.

Reference
1. Clark Kimberling, "Major Centers of Triangles," Amer. Math. Monthly 104 (1997)
431-438.

Flat 4
Terrill Court
12-14, Apsley Road,
BRISTOL BS8 2SP
Article 47
A Cevian point and its Six Harmonics

Christopher Bradley

C3
A

B3

N C2
E
R
A3 W F
B2
P M
D S
G
V A2
U T
K J

B C1 B1 C L'
L

A1

Figure

1. Introduction

Let P be a Cevian point of a triangle ABC and L, M, N the feet of the Cevians on BC, CA, AB
respectively. Define the point B1 to be the harmonic conjugate of B with respect to L and C and
let C1 be the harmonic conjugate of C with respect to B and L. Define points C2 and A2 on CA

1
and A3 and B3 on AB by a similar process involving harmonic conjugates. See the figure above.
Now draw lines AB1, AC1, BC2, BA2, CA3 and CB3. The following additional Cevian points are
thereby created:
(i) R with Cevians ARL, BRC2, CRB3;
(ii) S with Cevians ASB1, BSM, CSB3;
(iii) T with Cevians ATB1, BTA2, CTN;
(iv) U with Cevians AUL, BUA2, CUA3;
(v) V with Cevians AVC1, BVM, CVA3;
(vi) W with Cevians AWC1, BWC2, CWN.

These six Cevian points we refer to as the harmonics of the initial Cevian point P.

Six additional points D, E, F, G, J, K are also defined as follows:


(i) D lying on BC2, A2A3, NL;
(ii) E lying on AC1, B3B1, MN;
(iii) F lying on AB1, C1C2, MN;
(iv) G lying on CB3, A2A3, LM;
(v) J lying on CA3, , B3B1, LM;
(vi) K lying on BA2, C1C2, NL.

Lines A3C1 and A2B1 are drawn and meet at A1. Similarly points B2, C3 are defined. The
following results now hold:
(i) A conic passes through A, B2, C, A1, B, C3;
(ii) A conic passes through R, S, T, L, V, W;
(iii) A conic passes through R, S, T, U, V, N;
(iv) A conic passes through R, M, T, U, V, W.

Also it is found that


(i) R and U lie on APL;
(ii) S and V lie on BPM;
(iii) T and W lie on CPN.
(iv) A conic passes through D, E, F, G, J, K

In proving these assertions areal co-ordinates are used with ABC as triangle of reference and
though full proofs would be prohibitively long and rather tiresome, what we have done is to give
a catalogue of (i) sets of co-ordinates of each point; (ii) equations of all lines and conics.
Checking results is then left to the reader.

2. Points B1, C1, C2, A2, A3, B3

2
Suppose P has co-ordinates (l, m, n). Then L has co-ordinates (0, m, n), M has co-ordinates (l, 0,
n) and N has co-ordinates (l, m, 0). The conditions that {BB1; LC} = {CC1, LB} = 1 mean that
B1 has co-ordinates (0, m, 2n) and C1 has co-ordinates (0, 2m, n). Similarly C2 has co-ordinates
(2l, 0, n) and A2 has co-ordinates (l, 0, 2n). And again A3 has co-ordinates (l, 2m, 0) and B3 has
co-ordinates (2l, m, 0).

3. The Cevians

AL, BM, CN: These have equations ny = mz, nx = lz, mx = ly respectively;


AB1, AC1: These have equations 2ny = mz, ny = 2mz respectively;
BC2, BA2: nx = 2lz, 2nx = lz respectively;
CA3, CB3: 2mx = ly, mx = 2ly respectively.

4. The points A1, B2, C3 and the conic A1BC3AB2C

A1 has co-ordinates ( l, 2m, 2z), B2 has co-ordinates (2l, m, 2n) and C3 has co-ordinates (2l,
2m, n). The conic A1BC3AB2C has equation
lyz + mzx + nxy = 0. (4.1)

Note that when P is the centroid this conic is the outer Steiner ellipse and when P is the
symmedian point it is the circumcircle of ABC.

5. The points R, V, T, U, S, W

R has co-ordinates (2l, m, n), V has co-ordinates (l, 2m, n) and T has co-ordinates (l, m, 2n);
U has co-ordinates (l, 2m, 2n), S has co-ordinates (2l, m, 2n) and W has co-ordinates (2l, 2m, n).

6. The conics RSTLVW, VWRMTU, TUVNRS

The conic RSTLVW has equation


x2/l2 + 2y2/m2 + 2z2/n2 4yz/mn zx/nl xy/lm = 0. (6.1)

The conic VWRMTU has equation


2x2/l2 + y2/m2 + 2z2/n2 yz/mn 4zx/nl xy/lm = 0. (6.2)

The conic TUVNRS has equation


2x2/l2 + 2y2/m2 + z2/n2 yz/mn zx/nl 4xy/lm = 0. (6.3)

7. The points D, E, F, G, J, K

3
E'

It is found that D has co-ordinates (2l, 3m, n), that J has co-ordinates (l, 2m, 3n) and that F has
co-ordinates (3l, m, 2n). Also it is found that E has co-ordinates 3l, 2m, n), that K has co-
ordinates (l, 3m, 2n) and that G has co-ordinates (2l, m, 3n).

Finally it is found that the conic DEFGJK has equation


11x2/l2 + 11y2/m2 + 11z2/n2 14yz/mn 14zx/nl 14xy/lm = 0. (7.1)

It may be shown that B3C2, A3A2, MN are concurrent at L'(0, m, n), the harmonic conjugate of
L with respect to B and C. Similarly M' and N' on CA and AB respectively are harmonic
conjugates of M and N respectively with respect to C and A and to A and B.

If the points L, M, N are not the feet of Cevians, but are just the vertices of some arbitrary
triangle, then many of the above results still hold. See the diagram below.

A
C3

B3
E''
F
P C2
A3 X
W E
Y U
F'' B2
Q
A2
R
VZ

B C1 D B1 D'' D'
C

A1

Figure

4
Flat 4,
Terrill Court,
12-14, Apsley Road,
BRISTOL BS8 2SP

5
Symmedian Point and its Harmonics

Christopher J Bradley

To N'
C3
A

B3

N C2
M1 X
F
N1
A3 P B2
L2 W E M
V
K L3
Q U R A2
Y
N2 Z
M3
D
B C1 L B1 C L'

A1
To M'

Figure

1. Introduction

In the Figure above K is the Symmedian point and L, M, N are the feet of the Cevians through K.
Lines MN, NL, LM are drawn. Points B1, C1 on BC are the harmonic conjugates of B in LC and
C in BL and points C2, A2 and A3, B3 are defined similarly on CA and AB respectively. Lines
AB1, AC1, BC2, BA2, CA3, CB3 are drawn. Points P = AL^MN, Q = BM^NL, R = CN^LM are
shown, as well as U= BA2^CA3, V = CB3^AB1, W = AC1^ BC2.

1
It turns out that conics can be drawn passing through PQRWVL, PQRUWM, PQRVUN. Points
X= B3A2^A3C2, Y = C1B3^B1A3, Z = A2C1^C2B1 are shown as well as D = A3B1^C1A2, E =
B1C2^A2B3, F = C2A3^B3C1. It is then shown that conics may be drawn through DEFXYZ and
C1B1A2C2B3A3. Points A1, B2, C3 are now located as A3C1^B1A2, B1A2^C2B3, C2B3^A3C1
respectively. It is then shown that A1, B2, C3 lie on the circumcircle of triangle ABC. Finally
points M1 = AC1^MN, N2 = BA2^NL, L3 = CB3^LM, N1 = AB1^MN, L2 = BC2^NL, M3 =
CA3^LM and a conic is shown to pass through L2L3M3M1N1N2. Working is carried out using
areal co-ordinates with ABC as triangle of reference. Results are given, but not working as
points, lines, and conics are so numerous.

2. Points K, L, M, N, B1, C1, A2, C2, B3, A3

The co-ordinates of these points are K(a2, b2, c2), L(0, b2, c2), M(a2, 0, c2), N(a2, b2, 0). The
harmonic conjugate points have co-ordinates B1(0, b2, 2c2), C1(0, 2b2, c2), A2(a2 0, 2c2), C2(2a2,
0, c2), B3(2a2, b2, 0), A3(a2, 2b2, 0).

Equations of various lines involving these points are:


LM x/a2, + y/b2 z/c2 = 0, MN y/b2 + z/c2 y/b2 = 0, NL z/c2 + x/a2 y/b2.
AL c2y = b2z, BM a2z = c2x, CN b2x = a2y.
AB1 2c2y = b2z, BC2 2a2z = c2x, CA3 2b2 x = a2y, AC1 c2y = 2b2z, BA2 a2z = 2c2x, CB3 b2x =
2a2y.

The equation of the conic C1B1A2C2B3A3 is


2x2/a4 + 2y2/b4 + 2z2/c4 5yz/b2c2 5zx/c2a2 5xy/a2b2 = 0. (2.1)

3. Points P, Q, R, U, V, W and conics PQRLVW, PQRMWU, PQRNUV

The co-ordinates of the points are


P = AL^MN (2a2, b2, c2), Q = BM^NL (a2, 2b2, c2), R = (a2, b2, 2c2),
U = BA2^CA3 (a2, 2b2, 2c2), V = CB3^AB1 (2a2, b2, 2c2), W = AC1^BC2 (2a2, 2b2, c2).

The equation of the conic PQRLVW is


x2/a4 + 2y2/b4 + 2z2/c4 4yz/b2c2 zx/c2a2 xy/a2b2 = 0. (3.1)

The equation of the conic PQRMWU is


2x2/a4 + y2/b4 + 2z2/c4 yz/b2c2 4zx/c2a2 xy/a2b2 = 0. (3.2)

The equation of the conic PQRNUV is


2x2/a4 + 2y2/b4 + z2/c4 yz/b2c2 zx/c2a2 4xy/a2b2 = 0. (3.3)

2
4. Points X, Y, Z, D, E, F and the conic through these points

The co-ordinates of the points are


X = B3A2^A3C2 (5a2, 2b2, 2c2), Y = C1B3^B1A3 (2a2, 5b2, 2c2), Z = A2C1^C2B1 (2a2, 2b2, 5c2),
D = A3B1^C1A2 (a2, 4b2, 4c2), E = B1C2^ A2B3 (4a2, b2, 4c2), F = C2A3^B3C1 (4a2, 4b2, c2).

The equation of the conic XYZDEF is


8x2/a4 + 8y2/b4 + 8z2/c4 11yz/b2c2 11zx/c2a2 11xy/a2b2 = 0. (4.1)

5. The points A1, B2, C3

A1 = A3C1^B1A2,B2 = B1A2^C2B3, C3 = C2B3^A3C1. These points have co-ordinates ( a2, 2b2,


2c2), (2a2, b2, 2c2), (2a2, 2b2, c2) respectively. It may be checked that these points lie on the
circumcircle of triangle ABC.

6. The points M1, N2, L3, N1, M2, L3 and the conic through these points

These points have co-ordinates


M1 (3a2, 2b2, c2), N2 (a2, 3b2, 2c2), L3 (2a2, b2, 3c2), N1 (3a2, b2, 2c2), L2 (2a2, 3b2, c2), and M3 (a2,
2b2, 3c2).

It may be shown that N1 is the harmonic conjugate of N with respect to P and M and M1 the
harmonic conjugate of M with respect to N and P, and similarly for other point pairs on NL and
LM.

The equation of the conic through M1, N2, L3, N1, M2, L3 is
11x2/a4 + 11y2/b4 + 11z2/c4 14yz/b2c2 14zx/c2a2 14xy/a2b2 = 0. (6.1)

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

3
4
Perpendicular Bisectors and Angle Bisectors

Christopher Bradley

X Z''

W
Y' Y
O
M

J N
Q U
G X''

V Z I K
Z'
X' Y''
'

B L C

Figure 1

1. Introduction

1
In triangle ABC the incentre is I and the circumcentre is O. The angle bisectors AI, BI, CI are
drawn, as well as the perpendicular bisectors of the sides OL, OM, ON. Of the nine intersections
of these pairs of three lines, three, namely AI^OL, BI^OM, CI^ON are on the circumcircle (at
the midpoints of the minor arcs). The other six are labeled as follows AI^OM = Y, BI^ON = Z,
CI^OL = X, OL^BI = X', OM^CI = Y', ON^AI = Z'. Circles XYZ and X'Y'Z' are drawn, and for
reasons that will shortly become clear the 7-point circle is also drawn.

The following results now hold:


(i) Circles XYZ and X'Y'Z' both pass through I, and they pass through the Brocard
points and ' respectively;
(ii) Their second point of intersection J lies on the 7-point circle;
(iii) The line IJ passes through K, the symmedian point of ABC;
(iv) Triangles YZX and Z'X'Y' are similar;
(v) Lines YX', ZY', XZ' are concurrent at Q, the Spieker centre of ABC;
(vi) If U, V, W are the midpoints of YZ', ZX', XY' respectively then circle UVW has OI
as a diameter and passes through J;
(vii) Circles AYC, BZA, CXB pass through the Brocard point and circles AY'C, BZ'A,
CX'B pass through the Brocard point ';
(viii) Circle AYC and BZ''A meet at a point X'' lying on the 7-pt circle, circle BZA
and CX''B meet at a point Y'' lying on the 7-pt circle and circle CXB and AY''C
meet at a point Z'' also lying on the 7-pt circle;
(ix) Circles YZZ', ZXX', XYY' have a common point S' lying on circle X'Y'Z';
(x) Circles X'Y'X, Y'Z'Y, Z'X'Z have a common point S lying on circle XYZ;
(xi) Circles AYZ, BZX, CXY have a common point D lying on the circumcircle, circles
AY'Z', BZ'X', CX'Y' have a common point D' lying on the circumcircle and circles
AY''Z'', BZ''X'', CX''Y'' have a common point D'' lying on the circumcircle.
(xii) Circles BX''C, CY''A, AZ''B all pass through O.

The line of centres of circles XYZ and X'Y'Z' is, of course perpendicular to IJ through its
midpoint, but the centres do not appear to be significant. Results (i) to (viii) are illustrated in
Figure 1 and results (ix) to (xii) are illustrated in Figure 2. Results are proved in the following
sections using areal-co-ordinates with ABC as triangle of reference.

2. The angle bisectors and perpendicular bisectors

The equations of the angle bisectors AI, BI, CI are respectively cy = bz, az = cx, bx = ay.

The equation of the perpendicular bisector of BC is the line OL, with equation
(b4 c4 a2b2 + a2c2)x + a2(b2 + c2 a2)(y z) = 0. (2.1)

2
The equations of the perpendicular bisectors of CA, AB may be obtained from Equation (2.1) by
cyclic change of x, y, z and a, b, c,.

Z''
W Y'
O Y
S' M
J N
Q X''
G U
S
V Z I K
Z'
Y''
X' '
D'

B L C

D D''

F ig ure 2

3. The six points X, Y, Z, X', Y', Z'

3
The co-ordinates of these six points are as follows:
Y(bc + c2 a2, b2, bc), Z(ca, ca + a2 b2, c2), X(a2, ab, ab + b2 c2),
X'(a2, ca + c2 b2, ca), Y'(ab, b2, ab + a2 c2), Z'(bc + b2 a2, bc, c2).

4. The circles XYZ and X'Y'Z' and their common chord IJ

The circle XYZ has equation


b2cx2 + c2ay2 +a2bz2 a(ca + a2 c2)yz b(ab + b2 a2)zx c(bc + c2 b2)xy = 0. (4.1)

It may now be checked that I(a, b, c) lies on this circle and also that it passes through the Brocard
point (1/b2, 1/c2, 1/a2).

The circle X'Y'Z' has equation


bc2x2 + ca2y2 + ab2z2 a(ab + a2 b2)yz b(bc + b2 c2)zx c(ca + c2 a2)xy = 0. (4.2)

It may now be checked that I(a, b, c) lies on this circle and also that it passes through the Brocard
point '(1/c2, 1/a2, 1/b2).

The second point of intersection of the two circles is J, which has co-ordinates
x = a(a4 a3(b + c) a2bc + 2abc(b + c) bc(b2 + c2)),
y = b(b4 b3(c + a) b2ca + 2abc(c + a) ca(c2 + a2)), (4.3)
z = c(c4 c3(a + b) c2ab + 2abc(a + b) ab(a2 + b2).

It may now be checked that J lies on the 7-point circle with equation
b2c2x2 + c2a2y2 + a2b2z2 a4yz b4zx c4xy = 0. (4.4)

The equation of IJ is
bc(b c)x + ca(c a)y + ab(a b)z = 0. (4.5)

It is immediate that the line IJ passes through the symmedian point K(a2, b2, c2).

5. The similarity of triangles YZX and Z'X'Y'

The circles YZX and Z'X'Y' intersect at I and J, and YZ', ZX', XY' all pass through I. It follows
from Wood [1], that triangle Z'X'Y' is similar to YZX by means of a direct similarity composed
of a rotation about J followed by an enlargement.

6. The concurrence of YX', ZY', XZ' at the Spieker centre Q

The equation of YX' is

4
b(b c)x + a(a c)y (a2 + b2 ab c2)z = 0. (6.1)

The equations of ZY' and XZ' follow from Equation (6.1) by cyclic change of x, y, z and a, b, c.
These three lines all pass through Q(b + c, c + a, a + b), which is the Spieker centre lying on line
IG, where G is the centroid of ABC and IG = 2 GQ.

7. Points U, V, W and properties of circle UVW

U is the midpoint of YZ' and so has co-ordinates U((b + c)2 2a2, b(b + c), c(b + c)). Similarly V
has co-ordinates V(a(c + a), (c + a)2 2b2, c(c + a)) and W has co-ordinates W(a(a + b), b(a + b),
(a + b)2 2c2).

The equation of the circle UVW may now be obtained and it has equation
bc(b + c)x2 ca(c + a)y2 + ab(a + b)z2 a(2a2 b2 c2 + a(b + c))yz
b(2b2 c2 a2 + b(c + a))zx c(2c2 a2 b2 + c(a + b))xy = 0. (7.1)

It now follows that this circle passes through O and I and has centre the midpoint of OI and
furthermore passes through J.

8. Circles YZZ', ZXX' and XYY' and the point S' lying on circle X'Y'Z'

The equation of the circle YZZ' is


b2c2x2 + c2(c(a + b) a2)y2 + bc(a2 + b2 ca)z2 + (a4 a2(b2 + bc + 2c2) + ac2(c b) +
bc(b2 + c2))yz + b(a2(b + c) ac2 b2(b + c))zx + c3(a b c)xy = 0. (8.1)

Circles ZXX' and XYY' have equations that may be written down from Equation (8.1) by cyclic
change of x, y, z and a, b, c.

It may now be verified that these circles have a common point S' with co-ordinates (x, y, z),
where
x = a(a3(b + c) a2(b + c) abc(b + 2c) + b2c(b c)),
y = b(a2(b + c)2 a(b3 bc2 + c3) b2c(b c)), (8.2)
3 2 2 2 2 2 2
z = c(a b a (b + bc + c ) + ac(c 2b ) bc (b c)).

It may now be checked that S' lies on circle X'Y'Z'. It follows from Result 2 in the Appendix that
circles X'Y'Z, Y'Z'X, Z'X'Y have a common point S lying on circle XYZ.

9. The connection with the Brocard points

Circle AYC has equation

5
c2y2 + (c2 a2)yz b2zx = 0. (9.1)

It may now be checked that this circle passes through the Brocard point (1/b2, 1/c2, 1/a2).
Similarly circles BZA and CXB pass through . These three circles are, of course, the circles
that are often chosen to define . Their appearance in the context of the angle bisectors is mildly
surprising.

Circle AY'C has equation


a2y2 b2zx + (a2 c2)xy = 0. (9.2)
2 2 2
This circle passes through the other Brocard point ' (1/c , 1/a , 1/b ), as do the circles BZ'A and
CX'B.

The equation of the circle BZ'A is


b2z2 + (b2 a2)yz c2xy = 0. (9.3)

Circles AYC and BZ''A meet at X'' with co-ordinates (b2 + c2 a2, b2, c2). Similarly
circles BZA and CX''B meet at Y' with co-ordinates (a2, c2 + a2 b2, c2) and circles CXB
and AY''C meet at Z'' with co-ordinates (a2, b2 a2 + b2 c2). Points X'', Y'', Z'' are the vertices
of the second Brocard triangle and lie on the 7-point circle with Equation (4.4).

10. The points D, D', D''

It has already been established that circles AYC, BZA, CXB pass through , as does circle
XYZ. It follows from Result 2 of the Appendix that circles AYZ, BZX, CXY meet at a point D
on the circumcircle of ABC. Similarly circles AY'C, BZ'A, CX'B pass through ', as does circle
X'Y'Z'. It follows again from Result 2 of the Appendix that circles AY'Z', BZ'X', CX'Y' meet at a
point D' on the circumcircle of ABC.

The equation of the circle CX''Y'' is


b2c2(b2 c2)x2 c2a2(c2 a2)y2 a2(a4 a2b2 a2c2 + b4)yz b2(a4 a2b2 + b4 b2c2)zx
+ c2(a2 c2)(b2 c2)xy = 0. (10.1)

Circles AY''Z'' and BZ''X'' have similar equations that may be written down by cyclic change of
x, y, z and a, b, c. These circles meet at the point D'' with co-ordinates (x, y, z), where
x = a2(a2 b2)(c2 a2),
y = b2(b2 c2)(a2 b2), (10.2)
2 2 2 2 2
z = c (c a )(b c ).

It may now be checked that this point lies on the circumcircle of ABC.

6
The equation of circle AY''C is
c2a2y2 + a2(b2 a2)yz b2(c2 + a2 b2)zx + c2(b2 c2) xy = 0. (10.3)

Circles BZ''A and CX''B have similar equations to Equation (10.3) and may be written down by
cyclic change of x, y, z and a, b, c.

It may now be checked that all these circles pass through the circumcentre O.

11. Another case with a similar configuration

It may be asked whether there are other examples of pairs of three lines whose intersections have
such interesting properties. And the answer to this question is that the author knows of one other
case. This is when one has the three altitudes AH, BH, CH and the three radii of the circumcircle
AO, BO, CO. Points are now defined with X'= BH^CO, Y' = CH^AO, Z' = AH^BO, X =
CH^BO, Y = AH^CO, Z = BH^AO. It then follows that circles XYZ, XY'Z', YZ'X', ZX'Y' all
pass through the Brocard point and the circles X'Y'Z', X'YZ, Y'ZX, Z'XY all pass through the
Brocard point '. It is also the case that circles XYZ and X'Y'Z' meet at H and again at a point J
of the 7-point circle. See Figure 3.

7
A

Z'
Y'

J
X
O
X'
' H

Y
Z

B C

Figure 3

Proofs of all these results appear in Article 45 of Bradley [2].

Appendix

8
Some Observations on Pairs of Circles ABC and PQR having an intertwining property

We say that circles ABC and PQR intertwine if circles ABR, BCP, CAQ pass through the same
point S.

Result 1 The intertwining property is symmetric

Thus, if circles BCP, CAQ, ABR have a common point S, then circles QRA, RPB, PQC have a
common point D.

Proof
If S is the point of concurrence of circles BCP, CAQ and ABR, then invert with respect to S and
the configuration becomes one in which (using stars for inverted points) P* lies on B*C*, Q* lies
on C*A* and R* lies on A*B*. It then follows that circles Q*R*A*, R*P*B*, P*Q*C* are
concurrent at the point D*, the Miquel point. The inverse image D of the point D* is now the
point of concurrence of circles QRA, RPB and PQC.

Result 2 If in Result 1 the point S lies on circle PQR, then the point D lies on circle ABC

Proof
See Result 1. If S now lies on circle PQR, after inverting with respect to S, the point D* is the
point of intersection of the four circumcircles of the four triangles formed by the lines P*Q*R*,
P*B*C*, Q*C*A*, R*A*B*. The inverse image D of the point D* now lies on circle ABC. See
the Figure above.

Result 3 The intertwining property is transitive

Thus if ABC and PQR intertwine and ABC and QRP also intertwine, then ABC and RPQ also
intertwine. This result is not required in this article, but is added for the sake of completeness.
This time, after inversion, the result is still not immediate, and the details are left to the reader. A
proof appears in Article 19, by the Author [2] on the Internet, entitled Circular Perspective.

9
A

Q*
A*

B* S
P*
P
R C*
B C

D*
R*

Acknowledgement
I am grateful to David Monk [3] for pointing out the extension of Result 1 above, recorded as
Result 2 and also for many useful exchanges about Section 11.

10
References
1. F.E. Wood, Amer. Math. Monthly 36:2 (1929) 67-73.
2. C.J.Bradley, Article 19 ,http://people.bath.ac.uk/masgcs/bradley.html
3. D. Monk, private communication.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

11
Three Triangles in Mutual Triple Perspective
Christopher Bradley

T
To T'

A
Z
Y
L Y' Q
P'
K'
N' S'
K To X
H'
Z'
M' W
V
S
D G' NG
X' H M
W' P
L' C
B E R' R

Figure

1. Introduction

Given a triangle ABC and P a point not on its sides, the point P', the isotomic conjugate of P is
selected and then the following points are constructed: L = AP^BP', M = BP^CP', N = CP^AP',
L' = AP'^BP, M' = BP'^CP, N' = CP'^AP. It then turns out that triangles ABC, LMN, L'M'N' are
in mutual triple perspective. Of the nine perspectors, three are at P, three are at P' one at W, one
at W' and one at V. Here W is the perspector of triangles ABC and MNL, W' is the perspector of
triangles ABC and M'N'L' and V the perspector of triangles LMN and L'M'N'. Of the nine
perspectrices three coincide with DEF and one each with XYZ, X'Y'Z', RST, R'S'T', GHK and
G'H'K'. These latter six perspectrices have a common point Q. W' turns out to be the isotomic
conjugate of W. Also L and L', M and M', N and N' are isogonal conjugate pairs. The Figure
1
above displays this information. Points on the perspectrices are defined in later sections, when
appropriate. The analysis is carried out using areal co-ordinates with ABC as triangle of
reference with the help of DERIVE and the drawing is by CABRI.

2. The points P, P', L, L', M, M', N, N'

Suppose P has co-ordinates (p, q, r), then P' has co-ordinates (qr, rp, pq). L = AP^BP' and since
AP has equation qz = ry and BP' has equation rz = px, the co-ordinates of L are (r2, pq, rp).
Similarly L' = AP'^BP and since AP' has equation qy = rz and BP has equation pz = rx, the co-
ordinates of L' are (pq, r2, qr). Note that L'is the isotomic conjugate of L. The co-ordinates of M,
N may be written down from those of L by cyclic change of x, y, z and p, q, r. Likewise those of
M', N' may be written down by cyclic change from those of L'. Thus we have M(pq, p2, qr),
N(rp, qr, q2), M'(pr, qr, p2) and N'(q2, pq, rp). Note that M', N' are the isogonal conjugates of M,
N respectively.

3. The equation of the lines MN, NL, LM, M'N', N'L', L'M'

From the co-ordinates of points in Section 2 we find the lines have equations:
MN: q2(r2 p2)x + pq(q2 r2)y + rp(p2 q2)z = 0; (3.1)
NL: pq(q2 r2)x + r2(p2 q2)y + qr(r2 p2)z = 0; (3.2)
LM: rp(p2 q2)x +qr(r2 p2)y + p2(q2 r2)z = 0; (3.3)
M'N': pq(r2 p2)x + p2(q2 r2)y + qr(p2 q2)z = 0; (3.4)
N'L': rp(q2 r2)x + qr(p2 q2)y + q2(r2 p2)z = 0; (3.5)
L'M': r2(p2 q2)x + pq(r2 p2)y + rp(q2 r2)z = 0. (3.6)

4. The points D, E, F, V, X, Y, Z, X', Y', Z', Q and the lines DEF, XYZ, X'Y'Z'

It is now time to start studying the perspectives. First take triangles LMN and L'M'N'. The
equation of LL' is
rp(q2 r2)x qr(r2 p2)y + (r4 p2q2)z = 0. (4.1)

The equations of MM' and NN' may be obtained from this by cyclic change of x, y, z and p, q, r.
It may now be checked that these three lines all pass through the point V with co-ordinates
(p(q2 + r2), q((r2 + p2), r(p2 + q2)). The lines MN and M'N meet at the point D with co-ordinates
(p(q2 r2), q(r2 p2), 0). The lines NL and N'L' meet at the point E with co-ordinates (0, q(r2
p2), r(p2 q2)) and the lines LM and L'M' meet at the point F with co-ordinates (p(q2 r2), 0,
r(p2 q2)). The line DEF has equation
x/p(q2 r2) + y/q(r2 p2) + z/r(p2 q2)) = 0. (4.2)

2
Next we take the triangles ABC and LMN. They are clearly in perspective (by construction) at
the point P. Now MN and BC meet at X(0, r(p2 q2), q(q2 r2)), NL and CA meet at Y( r(r2
p2), 0, p(q2 r2)) and LM meets AB at Z(q(r2 p2), p(p2 q2), 0). The perspectrix XYZ has
equation
px/(r2 p2) + qy/(p2 q2) + rz/(q2 r2) = 0. (4.3)

Next we take the triangles ABC and X'Y'Z'. These are clearly in perspective (by construction) at
the point P'. Now BC and M'N' meet at X' with co-ordinates (0, qr(p2 q2), p2(q2 r2)), CA and
N'L' meet at Y' with co-ordinates ( q2(r2 p2), 0, rp(q2 r2)) and AB and L'M' meet at Z' with
co-ordinates (pq(r2 p2), r2(p2 q2), 0). The perspectrix X'Y'Z' has equation
r2px/(r2 p2) + p2qy/(p2 q2) + q2rz/(q2 r2) = 0. (4.4)

The perspectrices XYZ and X'Y'Z' meet at the point Q with co-ordinates (1/p(q2 r2), 1/q(r2
p2), 1/r(p2 q2)). The point Q is significant as it lies on other perspectrices to be determined.

5. The perspectrices RST and R'S'T'

We next consider triangles ABC and NLM which have perspector P'. LM meets BC at R with
co-ordinates (0, p2(q2 r2), qr(r2 p2). MN meets CA at S with co-ordinates ( rp(p2 q2), 0,
q2(r2 p2)) and NL meets AB at T with co-ordinates (r2(p2 q2), pq(q2 r2), 0). The perspectrix
RST has equation
pq2x/(p2 q2) + qr2y/(q2 r2) + rp2z/(r2 p2) = 0. (5.1)

It may now be checked that RST passes through Q.

We next consider triangles ABC and N'M'L' which have perspector P. L'M' meets BC at R' with
co-ordinates (0, r(q2 r2), q(r2 p2)). M'N' meets CA at S' with co-ordinates ( r(p2 q2), 0,
p(r2 p2)) and N'L' meets AB at T' with co-ordinates (q(p2 q2), p(q2 r2), 0). The perspectrix
R'S'T' has equation
px/(p2 q2) + qy/(q2 r2) + rz/(r2 p2) = 0. (5.2)

It may now be checked that R'S'T' passes through Q.

6. The other two perspectors W and W'

These arise from triangles ABC and MNL, and ABC and M'N'L' respectively and have co-
ordinates W(r2p, p2q, q2r) and W'(q2p, r2q, p2r). It is interesting to note that W and W' are
isotomic conjugates.

The line WW' has equation

3
qr(p4 q2r2)x + rp(q4 r2p2)y + pq(r4 p2q2)z = 0. (6.1)

It is noteworthy that WW' passes through V, like other pairs of isotomic conjugates L, L' and M,
M' and N, N'.

As can be seen from the figure the perspectrices of these perspectives both coincide with DEF.

7. The final pair of perspectives

These arise from the pair LMN and M'N'L' with perspector P' and the pair LMN and N'L'M' with
perspector P. Sides MN and N'L' meet at the point G with co-ordinates (pq(r2p4 + p2q4 + q2r4
3p2q2r2), p4q4 + q4r4 + r4p4 p4q2r2 q4r2p2 r4p2q2, qr(p4q2 + q4r2 + r4p2 3p2q2r2)). The co-
ordinates of points H, K may be written down from those of G by cyclic change of x, y, z and p,
q, r. Sides MN and L'M' meet at the point G' with co-ordinates (pqr(p4 + q4 + r4 p2q2 q2r2
r2p2), r(r2p4 + p2q4 + q2r4 3p2q2r2), q(p4q2 + q4r2 + r4p2 3p2q2r2)). The co-ordinates of points
H', K' may be written down from those of G' by cyclic change of x, y, z and p, q, r. The equations
of lines GHK and G'H'K' are too complicated to record, but DERIVE has checked that both these
perspectrices also pass through Q.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
When one Conic produces two more
Christopher J Bradley

U Y M
D
N
Q

R F

O
Z W
V

B E P X C

Figure

1. Introduction

In the Figure a triangle ABC is shown and O is a point not on its sides. Lines BO, CO are drawn
and points M and N on these lines are marked. A conic ABCMN is now drawn meeting the line
AO at L. Further points are defined as follows LM^BC = X, MN^CA = Y, NL^AB = Z, MN^AB
= D, NL^BC = E, LM^CA = F, AL^BC = P, BM^CA = Q, CN^AB = R, AL^MN = U, BM^NL
= V and CN^LM = W. The following results now hold: (i) A conic may be drawn through
XYZDEF; (ii) A conic may be drawn through PQRUVW; (iii) Triangles XYZ and UVW are in
perspective. It is, of course trivial that triangles PQR and UVW are also in perspective. Analysis
is carried out using projective co-ordinates with ABC as triangle of reference and O as unit point
(1, 1, 1).

2. The conic ABCLMN

The points L, M, by construction have co-ordinates of the form L(f, 1, 1) and M(1, g, 1). The
conic through A, B, C, L, M has equation
f(1 g)yz + g(1 f)zx + (fg 1)xy = 0. (2.1)

If the point N(1, 1, h) lies on the conic then


h = (fg 1)/(2fg f g). (2.2)

Consequently we may take N to have co-ordinates N(2fg f g, 2fg f g, fg 1).

3. Lines LM, MN, NL and points X, Y, Z, L, M, N

The co-ordinates of L, M, N are given in Section 2 and from them the equations of the lines LM,
MN, NL turn out to be
LM: (1 g)x + (1 f)y + (fg 1)z = 0; (3.1)
MN: f(g 1)x + (f 1)y (2fg f g)z = 0; (3.2)
NL: (g 1)x + g(f 1)y (2fg f g)z = 0. (3.3)

Co-ordinates of points now follow easily and are:


X = BC^LM (0, fg 1, f 1);
Y = CA^MN (2fg f g, 0, f(g 1);
Z = AB^NL (g(f 1), 1 g, 0);
D = AB^MN (f 1, f(1 g), 0);
E = BC^NL (0, 2fg f g, g(f 1));
F = CA^LM (fg 1, 0, g 1);

It may now be checked that triangles XYZ and DEF are in perspective with vertex O. Some hard
work (by DERIVE) shows the conic XYZDEF has equation
f(g 1)2x2 + g(f 1)2y2 + (fg 1)(2fg f g)z2 + (1 f)(f(g2 + 2g 1) 2g)yz
+ (1 g)(g(f2 + 2f 1) 2f)zx + (f 1)(g 1)(1 + fg)xy = 0. (3.4)

The fact that a conic passes through these six points has an easier pure proof. The triangles
ABC and LMN are in perspective so three of the nine points of intersection lie on a line (of
degree one in x, y, z), the Desargues axis of perspective. The nine points being the intersection
of two triangles must create a degenerate cubic, so the remaining six points of intersection X, Y,
Z, D, E, F must lie on a conic.
4. The points P, Q, R, U, V, W and the conic passing through these points

The points P, Q, R obviously have co-ordinates P(0, 1, 1), Q(1, 0, 1), R(1, 1, 0). The remaining
three points have co-ordinates as follows:
U = MN^y = z (2f 1, f, f);
V = NL^z = x (g, 2g 1, g);
W = LM^x = y (fg 1, fg 1, f + g 2).

It may now be checked that these six points lie on the conic with equation
f(1 g)x2 + g(1 f)y2 + (fg 1)z2 +(1 g)yz + (1 f)zx + (2fg f g)xy = 0. (4.1)

Triangles PQR and UVW are in perspective with vertex O.

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP
Article 52

Three triangles in mutual Triple Reverse Perspective


Christopher J Bradley

A1 B2 C3 P
A2 B3 C1 Q
A3 B1 C2 R AQ BP CR 2
N' AR BQ CP 3
AP BR CQ 1

P1 Q2 R3 A
P2 Q3 R1 B
W P3 Q1 R2 C

X''
M'
V
N
Z
R
F' 1
Z' Z''
2 E
Y X
P Y'' X'
Q
3 V'

D U'
B D' L C L' U
Y'

To
E'
To W' M

1. Introduction

The three triangles concerned are triangles ABC, PQR and 132 (all having anticlockwise
labelling). The figure is interesting in the sense that P, Q, R are the perspectors of triangles ABC
and 123, that 1, 2, 3 are the perspectors of triangles ABC and QPR and that A, B, C are the
perspectors of triangles PQR and 123. (Apologies to those who feel that labels 2 and 3 ought to

1
have been interchanged, but that would have meant a rewriting of the analysis and that might
well have introduced errors.)

Note how the figure is drawn in order to provide total accuracy. First triangle ABC and points P
and Q were chosen. Then lines AP, BP, CP, AQ, BQ, CQ were drawn. Now point 1 was defined
as AP^CQ and point 2 was defined as AQ^BP. Then point 3 was defined as CP^BQ. Lines B1,
C2, A3 are then found to be concurrent at the point R.

The method of construction ensures the mutual triple reverse perspectives mentioned above. No
further analysis is actually needed to establish this, but in view of intended applications it
becomes necessary to set up the analysis and to have recorded the co-ordinates of all points and
the equations of the nine perspectrices. Areal co-ordinates are used with ABC as triangle of
reference.

2. Points P, Q, R, 1, 2, 3

We take P to have co-ordinates (p, q, r) and Q to have co-ordinates (l, m, n). Then lines AP, BP,
CP have equations ry = qz, rx = pz and qx = py respectively. LinesAQ, BQ, CQ have equations
ny = mz, nx = lz and mx = ly respectively.

AP meets CQ at point 1(lq, mq, mr). BP meets AQ at point 2(np, mr, nr). CP meets BQ at point
3(lp, lq, np).

B1 has equation mrx = lqz, A3 has equation npy = lqz and C2 has equation mrx = npy. R is the
point of concurrence of these lines and has co-ordinates (lqnp, lqmr, mrnp).

3. Lines 12, 23, 31 and Desargues axes LMN, DEF, UVW

Line 12 has equation


mr(mr nq)x + nr(lq mp)y + mq(np lr) z = 0. (3.1)

Line 23 has equation


nr(lq mp)x pn(lr np)y pl(nq mr)z = 0. (3.2)

Line 31 has equation


mq(lr np)x lp(mr nq)y lq(lq mp)z = 0. (3.3)

Now A1, B2, C3 meet at P and the Deasargues axis of this perspective is LMN, where L =
BC^23, M = CA^31, N = AB^12. L has co-ordinates (0, l(mr nq), n(lr np)). M has co-

2
ordinates (l(mp lq), 0, m(np lr)). N has co-ordinates (n(lq mp), m(nq mr), 0). The axis
LMN has equation
mx/(lq mp) + ny/(mr nq) + lz/(np lr) = 0. (3.4)

Triangles ABC and 231 are in perspective with vertex Q and the Desargues axis of perspective
is DEF, where D = BC^31, E = CA^12, F = AB^23. D has co-ordinates (0, q(lq mp), p(nq
mr)). E has co-ordinates (q(lr np), 0, r(mr nq)). F has co-ordinates (p(lr np), r(lq mp), 0).
The axis DEF has equation
rx/(lr np) + py/(mp lq) + qz/(nq mr) = 0. (3.5)

Triangles ABC and 312 are in perspective with vertex R and the Desargues axis of perspective
is UVW, where U = BC^12, V = CA^23, W = AB^31. U has co-ordinates (0, mq(lr np), nr(lq
mp)). V has co-ordinates (lp(mr nq), 0, nr(mp lq)). W has co-ordinates (lp(mr nq), mq(lr
np), 0). The axis UVW has equation
x/(lp(mr nq)) + y/(mq(np lr)) + z/(nr(lq mp)) = 0. (3.6)

4. Lines PQ, QR, RP and axes D'E'F', L'M'N', U'V'W'

The line PQ has equation


(nq mr)x + (lr np)y + (mp lq)z = 0. (4.1)

The line QR has equation


mnr(lq mp)x + lnp(mr nq)y + lmq(np lr)z = 0. (4.2)

The line RP has equation


mqr(lr np)x +pnr(mp lq)y + plq(nq mr)z = 0. (4.3)

Triangles ABC and QPR are in perspective with vertex 2 and the Desargues axis of perspective
is D'E'F' where D' = BC^PR, E' = CA^RQ, F' = AB^QP. D' has co-ordinates (0, lq(nq mr),
nr(lq mp)). E' has co-ordinates (lq(lr np), 0, nr(lq mp)). F' has co-ordinates (lr np, mr
nq, 0). The axis D'E'F' has equation
nrx/(lr np) + nry/(nq mr) + lqz/(mp lq) = 0. (4.4)

Triangles ABC and RQP are in perspective vertex 3 and the Desargues axis of perspective is
L'M'N', where L' = BC^QP, M' = CA^ PR, N' = AB^RQ. L' has co-ordinates (0, lq mp, lr np).
M' has co-ordinates (lp(mr nq), 0, mr(lr np)). N' has co-ordinates (lp(nq mr), mr(lq mp),
0). The axis L'M'N' has equation
mrx/(mr nq) + lpy/(lq mp) + lpz/(np lr) = 0. (4.5)

3
Triangles ABC and PRQ are in perspective with vertex 1 and the Desargues axis of perspective
is U'V'W', where U' = BC^RQ, V' = CA^QP, W' = AB^PR. U' has co-ordinates (0, mq(lr np),
np(mr nq)). V' has co-ordinates (lq mp, 0, nq mr). W' has co-ordinates (np(lq mp), mq(lr
np), 0). The axis U'V'W' has equation
mqx/(lq mp) + npy/(np lr) + mqz/(mr nq) = 0. (4.6)

5. Axes XYZ, X'Y'Z', X''Y''Z''

Triangles PQR and 123 are in perspective with vertex A and the Desargues axis of perspective
is XYZ, where X = QR^23, Y = RP^31, Z = PQ^12. X has co-ordinates (lp(lmr2 n2pq), lmr(lqr
+ mpr 2npq), npr(2lmr lnq mnp)). Y has co-ordinates (lp(m2pr lnq2), lmq(2mpr npq
lqr), mpr(lmr + mnp 2lnq)). Z has co-ordinates (l2qr mnp2, m(2lqr mpr npq), r(lmr + lnq
2mnp)). The axis XYZ has equation
mr((l2qr(mr + nq) + lp(m2r2 6mnqr + n2q2) + mnp2(mr + nq))x pr(l2(m2r2 mnqr + n2q2)
lmnp(mr + nq) + m2n2p2)y lm(l2q2r2 lpqr(mr + nq) + p2(m2r2 mnqr + n2q2))z = 0. (5.1)

Triangles PQR and 231 are in perspective with vertex B and the Desargues axis of perspective is
X'Y'Z', where X' = PQ^23, Y'= QR^31, Z' = RP^12. X' has co-ordinates (p(lmr + mnp 2lnq),
m2pr lnq2, n(2mpr npq lqr)). Y' has co-ordinates (lpq(lmr + lnq 2mnp), mq(l2qr mnp2),
mnp(2lqr mpr npq)). Z' has co-ordinates (npq(2lmr lnq mnp), mq(lmr2 n2pq), mnr(lqr +
mpr 2npq)). The axis X'Y'Z' has equation
mn(l2q2r2 lpqr(mr + nq) + p2(m2r2 mnqr + n2q2))x pn(l2qr(mr + nq) + lp(m2r2 6mnqr +
n2q2))y+ pq (l2(m2r2 mnqr + n2q2) lmnp(mr + nq) + m2n2p2)z = 0. (5.2)

Triangles PQR and 312 are in perspective with vertex C and the Desargues axis of perspective is
X''Y''Z'' where X'' = PR^32, Y'' = PQ^31, Z''= QR^12. X'' has co-ordinates (lnp(2lqr pmr
pnq), lqr(lmr + lnq 2mnp), nr(l2qr mnp2)). Y'' has co-ordinates (l(lqr + pmr 2npq), q(2lmr
lnq mnp, lmr2 n2pq). Z'' ha co-ordinates (lnq(lqr + npq 2mpr), mqr(2lnq lmr mnp),
nr(lnq2 m2pr). The axis X''Y''Z'' has equation
qr(l2(m2r2 mnqr + n2q2) lmnp(mr + nq) + m2n2p2)x + ln(l2q2r2 lpqr(mr + nq) + p2(m2r2
6mnqr + n2q2))y lq(l2qr(mr + nq) + lp(m2r2 6mnqr + n2q2) + mnp2(mr + nq))z = 0. (5.3)

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

4
ARTICLE 53

Hagge Circles touching at H

Christopher Bradley

R
4 Q
0 O
S
N 3
2
P

B C

1. Introduction

A construction is described that provides circumcircles to a pair of triangles, circles which touch
at the orthocentre of ABC, and are hence touching Hagge circles. Analysis is given proving the
1
results by means of areal co-ordinates, with (u, v, w) for H and ABC as reference triangle. The
triangles are enlargements /reductions of ABC, the centres of enlargement being vertices of
perspective. Rather curiously for each triangle and its circumcircle the centre of the circle, the
vertex of perspective and the nine-point centre N are collinear. This is, of course, related to
Peisers [1] result for Hagge circles. In the Figure above, triangle 123 is in perspective with ABC
through point 4, circle 123 having centre 0 and triangle PQR is in perspective with ABC through
point S, circle PQR having centre O and touching circle 123 at the point 3 (which coincides with
H).

The construction is as follows: Take 3 to be the orthocentre H, and draw a line parallel to BC
through 3 and choose a point 2 on this line. Point 1 is now fixed, being the intersection of lines
through 2 and 3 parallel to AB and AC respectively. Points P, Q, R are now found so that P =
B3^C2, Q = C1^A3, R = A2^B1. Results now are : (i) AP, BQ, CR are concurrent at a point
S;(ii) A1, B2, C3 are concurrent at a point 4; (iii) circles 123 and PQR touch at 3; (iv) N, S, O are
collinear; and (v) N, 0, 4 are collinear.

2. Preliminary analysis

Point 3 has co-ordinates (u, v, w) and as 3 is the orthocentre we know u, v, w, are proportional to
1/(b2 + c2 a2), 1/(c2 + a2 b2), 1/(a2 + b2 c2) respectively. To get the centre of enlargement we
write
4 = k3 + (1 k)C. (2.1)

Writing t = (1 k)/k and dividing Equation (2.1) by k the co-ordinates of 4 may be written as (u,
v, w + t). Since 21 is parallel to BA and 31 is parallel to CA, the co-ordinates of 2 are (u v t, z
= w + t) and those of 1 are (u t, v, w + t).

Using these co-ordinates we find the equations of lines needed to find P, Q, R are as follows:
A2: z(v t) = y(w + t); (2.2)
B1: x(w + t) = z(u t); (2.3)
B3: xw = zu; (2.4)
C2: x(v t) = yu; (2.5)
A3: zv = yw; (2.6)
C1: y(u t) = xv. (2.7)

The co-ordinates of P = B3^C2 are therefore (u, v t, w), the co-ordinates of Q = B3^C2 are
therefore (u t, v, w) and the co-ordinates of R = A2^B1 are (u t, v t, w + t).

3. The circles PQR and 123

2
The equation of any circle, using u, v, w as the co-ordinates of H, is
u(v + w)yz + v(w + u)zx + w(u + v)xy + (px + qy + rz)(x + y + z) = 0, (3.1)
where p, q, r are to be determined from the co-ordinates of three points on the circle.

Using the co-ordinates of P, Q, R we find the equation of the circle PQR is


vw(v + w)x2 + wu(w + u)y2 uv(2t u v)z2 u(t(v w) + 2vw)yz v(t(u w) +
2uw)zx + w(t(u + v) 2uv)xy = 0. (3.2)

Repeating the exercise using the co-ordinates of 1, 2, 3 we find the equation of the circle 123 is
vw(t + u + v + w)(v + w)x2 + wu(t + u + v + w)(w + u)y2 uv(tu + tv + 2tw u2 v2 2uv
uw vw)z2 u(tuv tuw + tv2 + 2tvw tw2+ 2vw(u + v + w))yz v(t(u2 + uv + 2wu) twv
tw2) + 2uw(u + v + w))zx + w(t(u2 + uw + v2 + vw) 2uv(u + v + w))xy = 0. (3.3)

The midpoint of 3R has co-ordinates


1/(2(u + v + w)(u + v + w t))(2u2 + 2u(v + w t) t(v + w), u(2v t) + 2v2 + 2v(w t)
tw, u(2w + t) + v(2w + t) + 2w2). (3.4)

It may now be checked that the co-ordinates of O, the centre of circle PQR are the same as those
of the midpoint of 3R, given by Equation (3.4).

The centre 0 of circle 123 may now be calculated and its co-ordinates are
(2u(u + v + w) t(v + w), 2v(u + v + w) t(w + u), t((u + v + 2w) + 2w(u + v + w))). (3.5)

It now follows that 0 lies on the line 3R. The conclusion from this is that circles 123 and PQR
touch at 3 and that 3R is the diameter of circle PQR.

4. Conclusions

It is easy to show that the point S = AP^BP^CP has co-ordinates (u t, v t, w). The nine-point
centre N has co-ordinates (2u + v + w, u + 2v + w, u + v + 2w). And the vanishing of certain
determinants establishes that 40N and OSN are straight lines.

Reference
1. A.M. Peiser, The Hagge circle of a triangle, Amer. Math. Monthly, 49 (1942) 524-527.

Flat 4,
Terrill Court,
12-14, Apsley Road,
BRISTOL BS8 2SP

3
4
Article 54
Four Triangle Conics

Christopher J Bradley

M2

N
N1
U
V I T
F
N2 P
J E
W Q M1
K
D S
R

B L1 L2 C

Figure
M

1. Introduction

1
A point P is taken in the plane of a triangle ABC and lines through P are drawn parallel to the
sides of a triangle, creating six points on the sides of the triangle (as in the construction of the
triplicate ratio circle when P is the symmedian point). These six points are found to lie on a conic
S. Another 15 points may be constructed, as described below, from points of intersection of the
lines joining these six points. 3 of these points lie on the polar of P with respect to S, and the
remaining 12 lie 6 by 6 on two other conics. When P is the symmedian point, these two conics
do not appear (as far as I am aware) in Brocard geometry and so do not appear to be related in
any way to the 7-point circle or the Brocard ellipse. The Figure above shows these results, which
will be clearer when the points in the figure have been carefully defined.

2. The six points L1, L2, M1, M2, N1, N2 on the sides of ABC

The equation of the line parallel to BC through P is (q + r)x = p(y + z), the equation of the line
through P parallel to CA is (r + p)y = q(z + x) and the equation of the line through P parallel to
AB is (p + q)z = r(x + y). From these equations we may find the co-ordinates of the six points on
the sides of the triangle. These are
L1: (0, p + q, r); L2 (0, q, r + p); M1 (p, 0, q + r); M2 (p + q, 0, r); N1(p + r, q, 0); N2( p, q + r, 0).

3. The conic through the 6 points

By substituting the co-ordinates of five of these points into the general equation of a conic and
solving for the ratios of the six coefficients we find the equation of the conic through the five
points is
qr(q + r)x2 + rp(r + p)y2 + pq(p + q)z2 p(p2 + pq + rp + 2qr)yz q(q2 + rp+ pq + 2rp)zx
r(r2 + pq + qr + 2pq)xy = 0. (3.1)

As a check it was then shown by substitution that the sixth point satisfied Equation (3.1).

4. Defining the points R, S, T, U, V, W and determining the conic through these six
points

The equations of the following six lines are:


M1N1: q(q + r)x = (r + p)(q + r)y + pqz ; N1L1: r(r +p)y = (p + q)(r + p)z + qrx ;
L1M1: p(p + q)z = (p + q)(q + r))x + rpy; M2N2: r(q + r)x = rpy + (p + q)(q + r)z;
N2L2: p(r + p)y = pqz + (q + r)(r + p)x; L2M2: q(p + q)z = qrx + (r + p)(p + q)y;

The six points are:


R = L2N2^L1M1 (p2, (p + q)(q + r), (q + r)(r + p);
S = L2M2^L1M1 (p2(p + q), q2(p + q), p(p + q)(q + r) + q2r);

2
T = M1N1^L2M2 ((p + q)(r + p), q2, (r + p)(q + r));
U = M1N1^M2N2 (q(q + r)(r + p) + r2p, q2(q + r), r2(q + r));
V = N1L1^M2N2 ((p + q)(r + p), (p + q)(q + r), r2);
W = N1L1^N2L2 (p2(r + p), r(r + p)(p + q) + p2q, r2(r + p)).

Using the same method as in Section 3 we find the equation of the conic RSTUVW is
(q + r)(p(q + r) + q2 + r2)x2 + (r + p)(q(r + p) + r2 + p2)y2 + (p + q)(r(p + q) + p2 + q2)z2
+ (p3 p2(q + r) 2p(q2 + qr + r2) 2qr(q + r))yz + (q3 q2(r + p) 2q(r2 + rp + p2) 2rp(r +
p))zx + (r3 r2(p + q) 2r(p2 + pq + q2) 2pq(p + q))xy = 0. (4.1)

5. The collinear points L, M, N and the point Q

The points L, M, N are defined as follows:


L = N1L1^L2M2 ( p(p + q)(r + p), qr(p + q), qr(r + p));
M = L1M1^M2N2 (rp(p + q), q(q + r)(p + q), rp(q + r));
N = M1N1^ N2L2 (pq(r + p), pq(q + r), r(q + r)(r + p)).

It may now be checked that these three points lie on the line with equation
qr(q + r)x + rp(r + p)y + pq(p + q)z = 0. (5.1)

AL, BM, CN evidently concur at the point Q (1/(q + r), 1/(r + p), 1/(p + q)).

6. The points D, E, F, I, J, K and the conic passing through these points

First three lines L2N1, M1N2, M2L1. Their equations are (in that order):
(r + p)y = q(z + x); (q + r)x = p(y + z); (p + q)z = r(x + y).

The six points have co-ordinates:


D = L2N1^L1M1 (p2, q(p + q), pq + qr + rp);
E = M1N2^L2M2 (p(p + q), q2, pq + qr + rp);
F = M2L1^M1N1 (pq + qr + rp, q2, r(q + r));
I = N1L2^M2N2 (pq + qr + rp, q(q + r), r2);
J = M1N2^N1L1 (p(r + p), pq + qr + rp, r2);
K = L1M2^N2L2 (p2, pq + qr + rp, r(r + p)).

A conic passes through these six points and has equation


(p2(q2 + r2)(q + r)2 + pqr(q3 + q2r + qr2 + r3) q3r3)x2 + ... + ...
+ (p4(q2 + qr + r2) p3(q3 + r3) p2qr(2q2 + 3qr + 2r2) 3pq2r2(q + r) 2q3r3)yz + ... + .... (6.1)

7. The 7-point conic

3
A

E'
Q M

W
X
N
L
F' W' P E

B D D' C

Figure 2

In Figure 2 we show the construction of the (triplicate ratio) conic and the (7-point) conic when
lines are drawn parallel to the sides of ABC through an arbitrary point P (rather than the
symmedian point K). As you would expect these conics have the same centre, and PXQ in the
Figure 2 is a diameter. W and W' are the analogues of the Brocard points and LMN the vertices
of the first Brocard triangle. (In the figure D replaces L1 etc.) A final interesting fact is that if P is
chosen to be the circumcentre O, then K lies on the 7-point conic, so that OK is a diameter in
both cases, and X in Figure 2 is then the same as the centre of the 7-point circle.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

4
5
Article 55
Perpendiculars to a Triangles Sides through its Vertices
Christopher J Bradley

L2

N
L3

E M
F P R
M1 Q O

N1
B C

D
Y
M3

N2

1. Introduction

1
The six points M1, N1, N2, L2, L3, M3 may be defined in three equivalent ways. First they are the
points where the perpendiculars to the sides through the vertices of triangle ABC meet the
opposite sides. For example L2 is the point on CA where the perpendicular to BC at B meets CA.
Secondly, if LMN is the triangle formed by the tangents to the circumcircle of triangle ABC and
XYZ is the triangle that has LMN as its medial triangle, then the six points are the points on the
sides of ABC where the (non-corresponding) sides of XYZ meet the sides of ABC. For example
L2 is the point common to XY and CA. Thirdly if one constructs the three circles through B, C
and C, A and A, B that cut the circumcircle orthogonally, the six points are the intersections of
these circles with the sides of ABC. For example, circle L2ABM1 is orthogonal to the
circumcircle of ABC and cuts CA at L2 and BC at M1. In this article these equivalences are
established and it is also shown that the six points lie on a conic. It is also shown that XA, YB,
ZC are concurrent at a point R and that XL, YM, ZN are concurrent at a point Q. Work is carried
out using areal co-ordinates with ABC as triangle of reference.

2. The triangle LMN and the lines through its vertices parallel to its opposite sides

The circumcircle S of triangle ABC has equation


a2yz + b2zx + c2xy = 0. (2.1)

It follows that the tangents to S at A, B, C are respectively b2z + c2y = 0, c2x + a2z = 0 and a2y +
b2x = 0. The point L is the intersection of the tangents at B and C and so has co-ordinates ( a2,
b2, c2). Similarly M has co-ordinates (a2, b2, c2) and N has co-ordinates (a2, b2, c2).

The equation of the line through L parallel to MN (the tangent at A) is


2b2c2x + c2(a2 + b2 c2)y + b2(c2 + a2 b2)z = 0. (2.2)

The equations of the lines through M parallel to NL and through N parallel to LM may now be
written down from Equation (2.2) by cyclic change of x, y, z and a, b, c.

3. The points X, Y, Z and the six points M3, N2, N1, L3, L2, M1

The line through L parallel to MN and the line through M parallel to NL meet at the point Z with
co-ordinates (x, y, z), where
x = a2(a4 + 2a2(b2 c2) (b2 c2)(3b2 + c2)),
y = b2(3a4 2a2(b2 + c2) (b2 c2)2), (3.1)
2 4 2 2 2 2 2 2
z = c (a 2a (b + c ) + (b c ) ).

The co-ordinates of points X and Y may now be written down from Equation (3.1) by cyclic
change of x, y, z and a, b, c.

2
We define the six key points as the intersections of the sides of triangle XYZ with the (non-
corresponding) sides of ABC. Thus YZ meets AB at M3 and CA at N2. The co-ordinates of M3
are therefore (c2 a2 b2, 2b2, 0) and the co-ordinates of N2 are (b2 c2 a2, 0, 2c2). In similar
fashion we may work out the co-ordinates of the other key points as N1(0, a2 b2 c2, 2c2),
L3(2a2, c2 a2 b2, 0), L2(2a2, 0, b2 c2 a2), M1(0, 2b2, a2 b2 c2).

4. The circles M3BCN2, N1CAL3 and L2ABM1 and the conic through the six key points

The equation of circle M3BCN2 is


2b2c2x2 a2(b2 + c2 a2)yz + b2(c2 + a2 b2)zx + c2(a2 + b2 c2)xy = 0. (4.1)

The equations of circles N1CAL3 and L2ABM1 may now be written down from Equation (4.1) by
cyclic change of x, y, z and a, b, c.

The equation of the conic M3N2N1L3L2M1 is


2x2/a2 + ... + ... + (a4 + b4 + c4 + 6b2c2 2c2a2 2a2b2)/(b2c2(b2 + c2 a2))yz + ... + ... = 0. (4.2)

It is not surprising that the six key points lie on a conic as they are defined in such a consistent
fashion.

5. AX, BY, CZ meet at a point R and XL, YM, ZN meet at a point Q

These are not difficult to check.

The x co-ordinate of R is
x = a2(a4 + b4 3c4 + 2b2c2 + 2c2a2 2a2b2)(a4 + c4 3b4 + 2b2c2 2c2a2 + 2a2b2). (5.1)
The y and z co-ordinates follow by cyclic change of a, b, c.

The x co-ordinate of Q is
x = a2(3a4 b4 c4 +2b2c2 2c2a2 2a2b2). (5.2)
The y and z co-ordinates follow by cyclic change of a, b, c.

6. The points D, E, F and their significance

The line CO meets the circumcircle again at the point F with co-ordinates (x, y, z), where
x = 2a2(b2 + c2 a2),
y = 2b2(c2 + a2 b2), (6.1)
z = (b2 + c2 a2)(c2 + a2 b2).

The co-ordinates of D, E may now be written down by cyclic change of x, y, z and a, b, c.

3
The equation of the line BF is
(c2 + a2 b2)(b2 + c2 a2)x + 2a2(b2 + c2 a2)z = 0. (6.2)

It may now be checked that the point L2 lies on this line. Note that since CO is a diameter of S,
the line BF is perpendicular to BC through B.

This establishes the third property of the six key points; that they lie on the six lines through the
vertices perpendicular to the sides.

7. Three final points

It may now be checked that


(i) The midpoints of YZ, ZX, XY are respectively the points L, M, N;
(ii) L is also the midpoint of M3N2 etc;
(iii)The tangent to circle M3BCN2 at C passes through O, the centre of S, and hence each of
the three circles is orthogonal to S.

Flat 4
Terrill Couret
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 56
The Cevian point Conic
Christopher Bradley

W
A
V

N
M
R
E
F
Z Q Y

P
S

D X C
B

Figure

1. Introduction

In triangle ABC let P be a Cevian point with D, E, F the feet of the Cevians on BC, CA, AB
respectively. Points L, M, N lie on AD, BE, CF respectively and are such that P is the midpoint
of AL, BM, CN. It is found that the points A, B, C, L, M, N lie on a conic which we call the
Cevian point conic of P. (When P lies at the centroid G this conic is the Steiner conic.) The
tangents at A, B, C to the Cevian point conic form a triangle UVW. The midpoints of the sides
BC, CA, AB are denoted by X, Y, Z respectively. The following results now hold: (i) DU, EV,
FW are concurrent at a point Q; (ii) LU, MV, NW are concurrent at a point R; (iii) P, Q, R are

1
collinear; (iv) AU, BV, CW are concurrent at a point S; (iv) LX, MY, NZ are concurrent at a
point T.

2. Co-ordinates of L, M, N and the equation of the Cevian point Conic

Since AP = PL, BP = PM, CP = PN and P has co-ordinates (l, m, n), say, the co-ordinates of L,
M, N are L(2l 1, 2m, 2n), M(2l, 2m 1, 2n), N(2l, 2m, 2n 1). Note that we take l + m + n = 1.
Then it is easily checked that the equation of the Cevian point conic is
l(1 2l)yz + m(1 2m)zx + n(1 2n)xy = 0. (2.1)

3. Points U, V, W, D, E, F, X, Y, Z

The equation of the tangent at A to the Cevian point conic is n(1 2n)y + m(1 2m)z = 0. The
tangents at B and C may be written down by cyclic change of x, y, z and l, m, n.

The tangents at B and C meet at U(l(1 2l), m(2m 1), n(2n 1)).Similarly the points V and W
have co-ordinates V(l(2l 1), m(1 2m), n(2n 1)) and W(l(2l 1), m(2m 1), n(1 2n)).

Clearly we have D(0, m, n), E(l, 0, n), F(l, m, 0), X(0, 1, 1), Y(1, 0, 1), Z(1, 1, 0).

4. DU, EV, FW are concurrent at Q

The equation of DU is
2mn(n m)x + l(1 2l)(ny mz) = 0. (4.1)

The equations of EV, FW follow by cyclic change of x, y, z and l, m, n.

It then follows that the point Q with co-ordinates (x, y, z), where
x = l(1 2l)(2l 2m 2n + 1),
y = m(1 2m)(2m 2n 2l + 1), (4.2)
z = n(1 2n)(2n 2l 2m, + 1),

lies on all three lines.

5. LU, MV, NW are concurrent at R

The equation of LU is
4mn(n m)x + n(1 2l)(2n + 2l 1)y m(1 2l)(2l + 2m 1)z = 0. (5.1)

The equations of MV, NW follow by cyclic change of x, y, z and l, m, n.

2
It then follows that the point R with co-ordinates (x, y, z), where
x = l(1 2l)(2l + 2m + 2n 8mn 1),
y = m(1 2m)(2l + 2m + 2n 8nl 1), (5.2)
z = n(1 2n)(2l + 2m + 2mn 8lm 1),

lies on all three lines.

It may now be checked that the three points P, Q, R are collinear.

6. AU, BV, CW are concurrent at S

The equation of AU is
n(1 2n)y = m(1 2m)z. (6.1)

The equations of BV, CW follow by cyclic change of x, y, z and l, m, n.

It then follows that the point S with co-ordinates (l(2l 1), m(2m 1), n(2n 1)) lies on all three
lines.

Note that when P is the symmedian point, then S is the circumcentre.

7. LX, MY, NZ are concurrent at T

The equation of LX is
2(n m)x + (2l 1)(y z) = 0. (7.1)

The equations of MY, NZ follow by cyclic change of x, y, z and l, m, n.

It then follows that the point T with co-ordinates (2l 1, 2m 1, 2n 1) lies on all three lines.

Below is the figure when P is the orthocentre, showing that even for an internal point, sometimes
the Cevian point conic is a hyperbola.

Note that when P is the symmedian point, T is the isotomic conjugate of the orthocentre.

3
S

To A
S

U
F
N Y
T
Z M
W
H
E

Q C
X D
B

Flat 4
Terrill Court
12-14 Apsley Road
BRISTOL BS8 2SP

4
Article 57
Perpendiculars to the Cevians at the Cevian Point
Christopher Bradley

B C L

FIGURE

1. Introduction

If ABC is a triangle and P a point not on the sides or extensions of the sides, then the lines AP,
BP, CP are called the Cevians of P. If now the perpendicular to AP at P is drawn and it meets BC
at L, and M and N are similarly defined, it turns out that L, M, N are collinear. In this note a
proof is given using areal co-ordinates with ABC as triangle of reference.

2. The displacements AP and PL

Let the co-ordinates of P be (l, m, n), where l + m + n = 1, and those of L be (0, u, 1 u). Then
the displacement AP is (f, g, h) = ( m n, m, n) and the displacement PL is (p, q, r) = (m + n
1, u m, 1 u n).

The condition for these displacements to be perpendicular is, see Bradley [1],
a2(gr + qh) + b2(ph + fr) + c2(fq + pg) = 0. (2.1)

This provides the value for u, which is


{a2m(2n 1) b2(m(2n 1) + 2n(n 1)) c2m(2m + 2n 1)}/{(b2 c2)(m + n) a2(m n)}
(2.2)
If M has co-ordinates (1 v, 0, v) and N has co-ordinates (w, 1 w, 0) then v and w may be
written down from equation (2.2) by cyclic change of a, b, c and l, m, n.

3. The collinearity of L, M, N

L, M, N are collinear if and only if


uv + vw + wu u v w + 1 = 0. (3.1)

Substitution of the values of u, v, w into the left hand side of equation (3.1) shows this to be true,
since there is a factor of (l + m + n 1) = 0.

Reference
1. C. J. Bradley, The Algebra of Geometry, Highperception, Bath, UK, (2007).

Flat 4,
Terrill Court,
12-14, Apsley Road,
BRISTOL BS8 2SP
Article 58
Additional results for the Miquel configuration
Christopher J Bradley

S2
N1

A
C1

N
Q
R P
W
M
C3
U
N2 C
L2 B L3
S L

M3

V S3
C2
S1

FIGURE
M1

1. Introduction

1
In the standard Miquel configuration ABC is a triangle with points L, M, N on BC, CA, AB
respectively. Then circles C1= AMN, C2 = BNL, C3 = CLM have a point P in common. When
LMN is a transversal then P lies on the circumcircle.
The additional construction here is that the perpendiculars from L, M, N are drawn to the sides
BC, CA, AB respectively. The one from L meets CA at M1 and AB at N1. The one from M meets
BC at L2 and AB at N2. The one from N meets BC at L3 and CA at M3. It is a simple matter of
angle-chasing to show that circles S1, S2, S3 may be drawn through M, N, N2, M3 and N, L, L3,
N1 and L, M, M1, L2 respectively. The following results now hold: (i) Circles C1, S2, S3 have a
point Q in common; (ii) Circles C2, S3, S1 have a point R in common; (iii) Circles C3, S1, S2 have
a point S in common. It is also obvious from angular considerations that if the three
perpendiculars form a triangle UVW then UVW is similar to ABC. When LMN is a transversal
the following additional results hold: (iv) Points N2, M3, L3, N1, M1, L2 lie on a conic; (v) AU,
BV, CW intersect at a point T, so that not only are UVW and ABC similar they are also in
perspective with LMN the Desargues axis of perspective: (vi) T lies on the circumcircles of both
ABC and UVW; (vii) If O1, O2, O3 are the centres of C1, C2, C3 respectively and O is the
circumcentre of ABC, then it is known that O, O1, O2, O3, P are concyclic, but this circle also
contains T and O1, O2, O3 lie on AT, BT, CT respectively, providing a Wood [1] configuration,
establishing immediately the known similarity of triangles ABC and O1O2O3.
We prove these results using areal co-ordinates with ABC as triangle of reference.

2. Co-ordinates of L, M, N
In the general case we take L(0, u, 1 u), M(1 v, 0, v), N(w, 1 w, 0). When LMN is the
transversal with equation x/l + y/m + z/n = 0, this means that
u = m/(m n), v = n/(n l), w = l/(l m). (2.1)

3. Co-ordinates of N1, L2, M3, M1, N2, L3


If (f, g, h) and (p, q, r) are two displacements at right-angles, it is known, see Bradley [2], that
a2(gr + hq) + b2(fr + hp) + c2(fq + gp) = 0. (3.1)
Suppose then that N1 has co-ordinates (s, 1 s, 0) then the displacement BC = (0, 1, 1) and the
displacement LN1 = s, 1 s u, u 1). Using these values for f, g, h, p, q, r we obtain
s = 2a2(1 u)/(c2 + a2 b2). (3.2)
It follows that N1 has co-ordinates (x, y, z), where
x = 2a2(1 u), y = 2a2u a2 b2 + c2, z = 0. (3.3)
It follows by cyclic change of x, y, z and a, b, c and u, v, w that L2 has co-ordinates (x, y, z),
where
x = 0, y = 2b2(1 v), z = 2b2v b2 c2 + a2. (3.4)
And again by cyclic change that M3 has co-ordinates (x, y, z), where

2
x = 2c2w c2 a2 + b2, y = 0, z = 2c2(1 w). (3.5)
If M1 has co-ordinates (1 t, 0 , t) then the displacement LM1 is (1 t, u, t 1 + u) and using
Equation (3.1) we find
t = {a2(1 2u) + b2 c2}/(a2 + b2 c2). (3.6)
It follows that M1 has co-ordinates (x, y, z), where
x = 2a2u, y = 0, z = a2(1 2u) + b2 c2. (3.7)
It now follows by cyclic change that the co-ordinates of N2 are (x, y, z), where
x = b2(1 2v) + c2 a2, y = 2b2v, z = 0. (3.8)
And finally, L3 has co-ordinates (x, y, z), where
x = 0, y = c2(1 2w) + a2 b2, z = 2c2w. (3.9).

4. The Miquel Point


The Miquel point P is the common point of circles AMN, BNL, CLM. We first obtain the
equation of circle AMN. Every circle in areal co-ordinates has an equation of the form
a2yz + b2zx + c2xy + (px + qy + rz)(x + y + z) = 0. (4.1)
The constants p, q, r are evaluated from equations found by inserting the co-ordinates of A, M, N
in turn. The result is that circle AMN has equation
c2wy2 + b2(1 v)z2 (a2 b2(1 v) + c2w)yz b2vzx c2(1 w)xy = 0. (4.2)
Circles BNL, CLM may now be written down by cyclic change of x, y, z and a, b, c and u, v, w.
The Miquel point P is the common point of the three circles and has co-ordinates (x, y, z), where
x = a2(a2u(u 1) + b2(u 1)(v 1) + c2wu),
y = b2(a2uv + b2v(v 1) + c2(v 1)(w 1)), (4.3)
z = c2(a2(w 1)(u 1) + b2vw + c2w(w 1)).
When L, M, N are collinear, then from Equations (2.1) the Miquel point P has co-ordinates
x = {a2/(m n)}, y = {b2/(n l)}, z = {c2/(l m)}. (4.4)
It is easy to check that in this case P lies on the circumcircle of ABC.

5. The circles S1, S2, S3


The circle S1 is the circle passing through the points M, N, M3, N2. That a circle passes through
these points follows immediately from the construction because of the right angles. Its equation
is

3
2b2c2v(w 1)x2 + c2w(a2 + b2(2v 1) c2)y2 + b2(1 v)(a2 b2 + c2(1 2w))z2 (a4 + a2(b2(v
2) c2(w + 1)) + b4(1 v) b2c2(v(4w 1) 3w + 1) + c4w)yz b2(a2v b2v + c2(2(w 1)
v(4w 3)))zx + c2(a2(w 1) + b2(2v(2w 1) w + 1) + c2(1 w))xy = 0. (5.1)
The equations of S2 and S3 may now be written down by cyclic change of x, y, z and a, b, c and
u, v, w. It may now be verified that the point Q with co-ordinates (x, y, z) given next, lies on the
three circles C1, S2, S3.
x = a2(a4 + a2(b2(2v 2) 2c2w) + b4(1 2v) + b2c2(2v 2w)) + c4(2w 1))(a2u(u 1) +
b (u 1)(v 1) + c2wu),
2

y = b2(a2(2u 1) b2 + c2)(a4uv a2(b2v(u v + 1) + c2(u(v + 2(w 1)) + (v 1)(w


1))) + (1 v)(b4v b2c2(v w + 1) + c4(1 w))), (5.2)
z = c2(a2(2u 1) b2 + c2)(a4(u 1)(w 1) a2(b2(u(2v + w 1) + v(w 2) w + 1) +
c2(u w 1)(w 1)) + w(b4v b2c2(v w + 1) + c4(1 w))).
Points R and S with co-ordinates provided by cyclic change lie similarly on circles C2, S3, S1 and
C3, S1, S2 respectively.

6. The triangle UVW


We now obtain the equations of the lies through L, M, N perpendicular to the sides that form the
triangle UVW. The equation of the line LN1 is
(a2(2u 1) b2 + c2)x + 2a2(u 1)y + 2a2uz = 0. (6.1)
The equations of the lines ML2 and NM3 may be written down from Equation (6.1) by cyclic
change of x, y, z and a, b, c and u, v, w.
LN1 and ML2 meet at the point W with co-ordinates (x, y, z), where
x = 2a2(a2u + b2(u + 2(v 1)) c2u),
y = 2b2(a2(2u + v 1) + (b2 c2)(v 1)), (6.2)
z = a4(2u 1) 2a2(b2(u v) + c2(u 1)) + (c2 b2)(b2(2v 1) c2).
Points U and V have co-ordinates that may be written down from Equation (6.2) with cyclic
change of x, y, z and a, b, c and u, v, w.
It may be verified that W lies on C3 and U lies on C1 and V lies on C2.

7. The six points when L, M, N are collinear


These may be obtained from Section 3 with the aid of Equation (2.1) the co-ordinates are:
N1: ( 2a2n, a2(m + n) + (c2 b2)(m n), 0);
L2: (0, 2b2l, b2(n + l) + (a2 c2)(n l));
M3: (c2(l + m) + (b2 a2)(l m), 0, 2c2m);

4
N2: (b2(n + l) + (c2 a2)(l n), 2b2n, 0);
L3: (0, c2(l + m) + (a2 b2)(m l), 2c2l);
M1: ( 2a2m, 0, a2(m + n) + (b2 c2)(n m)). (7.1)
It may now be shown that these six points lie on the conic with equation
2b2c2lmn(a2(m + n) + (b2 c2)(n m))x2 + + + a2mn(a4(l m)(l n) 2a2l(b2(l m) + c2(l
n)) + b4(l m)(l + n) + 2b2c2(3l2 + mn) + c4(l + m)(l n))yz + + = 0. (7.2)

8. AU, BV, CW meet at T which lies on circles ABC and UVW & P also lies on circle
UVW
When L, M, N are collinear then, from Equations (6.2) and (2.1) the co-ordinates of W are (x, y,
z), where
x = 2a2(a2m(l n) + b2(l(m 2n) + mn) + c2m(n l)),
y = 2b2(a2(l(m + n) 2mn) + l(b2 c2)(n m)), (8.1)
z = a4(l n)(m + n) 2a2(b2(lm n2) + c2n(l n)) + (b2 c2)(m n)(b2(l + n) + c2(l n)).
The co-ordinates of U and V now follow by cyclic change of x, y, z and a, b, c and l, m, n.
It may now be shown that AU, BV, CW are concurrent at the point T with co-ordinates (x, y, z),
where
x = a2/(a2(l(m + n) 2mn) + l(b2 c2)(n m)),
y = b2/(a2m(l n) b2(l(m 2n) + mn) + c2m(n l)), (8.2)
z = c2/(a2n(l m) + b2n(m l) + c2(l(2m n) mn)).
It may now be checked that T lies on the circumcircle of ABC.
The circle UVW may now be obtained and has equation
2b2c2(m n)(a2(lm + nl 2mn) + (b2 c2)l(n m))x2 + + (a4(l m)(l n)(m n)
2a2(b2(l m)(l(m 2n) + n2) + c2(l n)(l(2m n) m2)) + b4(l m)(l(m 3n) + n(m + n)) +
2b2c2l(m n)(2l m n) + c4(l n)(l(3m n) m(m + n)))yz = 0. (8.3)
It may now be verified that T lies on circle ABC and that P also lies on circle UVW.
The Figure illustrating these results when L, M, N are collinear is shown below. It will be
observed that triangle UVW is similar to triangle ABC by means of a rotation of 90 o and
enlargement/reduction about P, since we have a Wood [1] configuration for similar triangles in
perspective.
It may also be shown that O1 lies on AT, O2 lies on BT and O3 lies on CT and that circle O1O2O3
passes through O, P and T. Thus, as is known, triangle O1O2O3 is similar to ABC and circles
ABC, UVW, O1O2O3 are coaxal. Details of the proofs are left to the reader.

5
N1

N O1
O

P
M W
L3 L2 O3
B C
S L
U Q
N2

T
O2

R
M1

V
M3

References
1. F. E. Wood, Similar-Perspective triangles, Amer. Math. Monthly, 36:2 (1929) 67-73.
2. C. J. Bradley, The Algebra of Geometry, Highperception, Bath, UK, 2007.

Flat 4, Terrill Court,


12-14, Apsley Road, BRISTOL BS8 2SP.

6
7
Article 59
7 Points on any Circle not through a Vertex
Christopher J Bradley

G
Z N
Y
A'

B L C

B'
P
C'

Figure

1. Introduction

The general result (of which a particular case is illustrated above and analysed below in later
sections) is as follows: Choose a point Q not on the sides or extensions of the sides of a triangle
(in the illustration the centroid G is chosen), then points L, M, N are chosen on the sides BC, CA,
AB so that Q is the Miquel point for circles AMN, BNL, CLM. Any appropriate choice of points

1
can be made. Any two points Y and Z are now chosen so that circle QYZ does not pass through a
vertex (in the Figure Y and Z are such that BY = BG and CZ = CG). The three Miquel
circles AMN, BNL, CLM now intersect circle QYZ again in points A', B', C' respectively. The
beautiful result is that AA', BB', CC' are concurrent at a point P, which lies on circle QYZ. In
what follows we establish the validity of the particular case, using areal co-ordinates. The
general case is too difficult for DERIVE. It is hoped that a pure proof will emerge.

2. The circle GYZ

The points Y and Z are chosen to have co-ordinates (1, 2, 1) and (1, 1, 2) respectively, so that
BY = BG and CZ = CG. The circle through these points and G(1, 1, 1) has equation
3(b2 + c2)x2 + (2a2 b2 + 2c2)y2 + (2a2 + 2b2 c2)z2 (8a2 b2 c2)yz
+ (2a2 7b2 + 2c2)zx + (2a2 + 2b2 7c2)xy = 0. (2.1)

3. The circle BLG and the points B' and N

The choice of L is open to us and we choose it to lie at the midpoint of BC with co-ordinates (0,
1, 1). The equation of BLG now follows and is
(a2 2b2 2c2)x2 3a2z2 + 3a2yz 2(a2 2b2 + c2)zx + (a2 2b2 + 4c2)xy = 0. (3.1)

The circles GYZ and BLG meet at G and again at the point B' with co-ordinates (x, y, z, where
x = b2 + c2 2a2,
y = 2c2 + 2a2 b2, (3.2)
z = 3c2.

The circle BGL meets AB again at the point N with co-ordinates (a2 + 4c2 2b2, 2(b2 + c2) a2,
0).

4. The circle ANG and the points A' and M

Now that N is known we can derive the equation of circle ANG, which is
(a2 2b2 + 4c2)y2 + (a2 + 4b2 2c2)z2 2(2a2 b2 c2)yz + (a2 2b2 2c2)(zx + xy) = 0. (4.1)

Circle ANG and GYZ meet at G and again at the point A' with co-ordinates (x, y, z), where
x = 4b2c2 b4 c4,
y = b2(b2 + c2), (4.2)
2 2 2
z = c (b + c ).

Circle ANG meets CA at M with co-ordinates (a2 2c2 + 4b2, 0, 2b2 + 2c2 a2).

2
5. The circle CMG and the point C'

Now that M is known we can derive the equation of circle CMG, which is
(a2 2b2 2c2)x2 3a2y2 + 3a2yz + (a2 + 4b2 2c2)zx 2(a2 + b2 2c2)xy = 0. (5.1)

This circle, by Miquel circle theory, must pass through L and this has been checked. It also
meets circle GYZ at the point C' with co-ordinates (b2 + c2 2a2, 3b2, 2a2 + 2b2 c2).

6. AA', BB', CC' meet at a point P lying on circle GYZ

It is now straightforward to show that AA', BB', CC' are concurrent at a point P whose co-
ordinates are (b2 + c2 2a2, 3b2, 3c2) and that P lies on circle GYZ.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP

3
Article 60
Centroid-centred Similar Ellipses
Christopher J Bradley

N V

H
K Z Y
G
M
W
R

B L X U C

Figure

1. Introduction

The three ellipses drawn in the Figure are (i) the ellipse centre G passing through A, B, C known
as the outer Steiner ellipse, (ii) the ellipse centre G touching the sides of ABC at its midpoints X,
Y, Z and (iii) the intermediate ellipse centre G passing through points a quarter of the way along
each side from each vertex, the points L, M, N, U, V, W. The co-ordinates in areals of some of
the key points are listed and it is noted that AU, BM, CG amongst others are Cevian lines. It is
also noted that if UG meets AB at H and MG meets AB at K, then HK is 1/5 AB. These ellipses
are, of course, produced by an affinity from similar circles in an equilateral triangle.

1
2. The equations of the conics

The outer Steiner ellipse is well known to have an equation


yz + zx + xy = 0. (2.1)
It is the circumscribing conic of minimum area. It meets the circumcircle at the Steiner point
whose co-ordinates are (1/(b2 c2), 1/(c2 a2), 1/(a2 b2)).

The inner Steiner ellipse is well known to have the equation


x2 + y2 + z2 2yz 2zx 2xy = 0. (2.2)
It is the inscribed conic of maximum area.

The inner and outer Steiner ellipses carry triangles that have an affinity with Poncelets porism.

The conic through the points L(0, 3, 1), M(1, 0, 3), N(3, 1, 0), U(0, 1, 3), V(3, 0, 1), W(1, 3, 0)
has equation
3x2 + 3y2 + 3z2 10yz 10zx 10xy = 0 (2.3)
For obvious reasons we call this the intermediate Steiner conic.

3. A few properties of the intermediate Steiner conic

The line UG has equation 3y = 2x + z, so it meets z = 0 at the point H(3, 2, 0) so that AH = 2/5
AB.
The line UG meets the conic again at the point with co-ordinates (8, 5, 1) and G(1, 1, 1) is the
midpoint of this segment. It follows by a sequence of such arguments that G is the centre of the
intermediate Steiner conic.
The line MG meets AB at a point K such that AK = 3/5 AB. It follows that HK = 1/5 AB and
that the midpoint of HK is the midpoint Z of AB.
The lines AU, BM, CG meet at the point R (1, 1, 3). There are two other similar Cevian points
with co-ordinates (3, 1, 1) and (1, 3, 1).
All three conics are similar as they arise from the same affinity.

Flat 4
Terrill Court
12-14, Apsley Road
BRISTOL BS8 2SP

2
Article 61
Concurrent lines in a Triangle with a Circle cutting the Sides
Christopher J Bradley

F To Z

A
To F
W

V
Q
H E
N
R G M
P K

X
B L U C
D

1. Introduction

If ABC is a triangle and P a Cevian point such that AP, BP, CP meet BC, CA, AB respectively in
L, M, N and then a circle is drawn through L, M, N to meet BC, CA, AB respectively at U, V,
W, then it has been known for a long time, see Bradley [1], that AU, BV, CW are concurrent at

1
another Cevian point Q. When this construction is carried out a number of other sets of three
lines are concurrent and sets of three points are collinear. In this article we exhibit some of these
in the Figure above and in sections that follow we establish their validity, using areal co-
ordinates with ABC as triangle of reference.

The results we establish are as follows: (i) If VW meets BC at X, NL meets CA at Y and UM


meets AB at Z then X, Y, Z are collinear; (ii) If NL meets UM at D, UM meets VW at E and VW
meets NL at F, then AD, BE, CF are concurrent at a point R; (iii) MN, BE and LV are concurrent
at a point G. Symmetry considerations now imply WU, MN, CF are concurrent at a point H and
WU, AD, LV are concurrent at a point K. Permutation of letters L, M, N and U, V, W and A, B,
C provide other collinearities and concurrences.

2. The circle through L, M, N and the points U, V, W

Let P be the point (l, m, n) so that L(0, m, n), M(l, 0, n) and N(l, m, 0). The equation of the circle
LMN is found to be
mn(a2mn(l + m)(n + l) b2nl(m + n)(l + m) c2lm(n + l)(m + n))x2 + ... + ...
(m + n)(a2mn(l + m)(n + l) + l(n m)(b2n(l + m) c2m(n + l)))yz ... ... = 0. (2.1)

This circle cuts BC, CA, AB at U, V, W whose co-ordinates are


U: (0, a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)), a2mn(l + m)(n + l) l(m + n)(b2n(l
+ m) c2m(n + l)));
V: (b2nl(m + n)(l + m) m(n + l)(c2l(m + n) a2n(l + m)), 0, b2nl(m + n)(l + m) + m(n + l)(c2l(m
+ n) a2n(l + m)));
W: (a2mn(l + m)(n + l) b2nl(l + m)(m + n) + c2lm(n + l)(m + n), a2mn(l + m)(n + l) + b2nl(l +
m)(m + n) + c2lm(n + l)(m + n), 0).

3. AU, BV, CW are concurrent at Q

The equation of AU is
(a2mn(l + m)(n + l) l(m + n)(b2n(l + m) c2m(n + l)))y = (a2mn(l + m)(n + l) + l(m +
n)(b2n(l + m) c2m(n + l)))z. (3.1)

The equation of BV is
(b2nl(m + n)(l + m) + m(n + l)(c2l(m + n) a2n(l + m)))x = (b2nl(m + n)(l + m) m(n +
l)(c2l(m + n) a2n(l + m)))z. (3.2)

AU and BV meet at the point Q with co-ordinates (x, y, z), where


x = 1/(l(m + n)(bn(l + m) + c2m(n + l)) a2mn(l + m)(n + l)),
y = 1/(a2mn(l + m)(n + l) l(m + n)(b2n(l + m) c2m(n + l))), (3.3)

2
z = 1/(a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l))).

It may now be checked that Q lies on the line CW.

4. The lines NL, UM, WV and the collinear points X, Y, Z

The equation of the line NL is


nly = mnx + lmz. (4.1)

The equation of the line UM is


n(a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)))x + l(a2mn(l + m)(n + l) l(m +
n)(b2n(l + m) c2m(n + l)))y + (a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)))z. (4.2)

The equation of the line WV is


(a2(mn(l2 + lm + mn + nl) b2nl(m2 + lm + mn + nl) c2lm(n2 + lm + mn + nl)))x +
(a2mn(l + m)(n + l) l(m + n)(b2n(l + m) c2m(n + l)))y + (a2mn(l + m)(n + l) + l(m + n)(b2n(l +
m) c2m(n + l)))z = 0. (4.3)

NL meets CA at the point Y( l, 0, n).

UM meets AB at the point Z(l(a2mn(l + m)(n + l) l(m + n)(b2n(l + m) c2m(n + l))), n(a2mn(l
+ m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l))), 0).

WV meets BC at the point X(0, a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)), (a2mn(l
+ m)(n + l) l(m + n)(b2n(l + m) c2m(n + l)))).

When the co-ordinates of X, Y, Z are put in the rows of a matrix, then the determinant of that
matrix, as has been checked, vanishes, showing the points to be collinear.

5. The points D, E, F and then triangle ABC is in perspective with triangle DEF

D is the intersection of NL and UM and has co-ordinates (x, y, z), where


x = l(l(m + n)2(b2n(l + m) c2m(n + l)) a2mn(l + m)(n + l)(m n)),
y = 2mn (a2mn (l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l))), (5.1)
z = n(m + n)(a2mn(l + m)(n + l) + l(n m)(b2n(l + m) c2m(n + l))).

E is the intersection of UM and WV and has co-ordinates (x, y, z), where


x = 2l(a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)))K,
y = (a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)))(a2mn(l3 + l2(m + 2n) + ln(2m +
n) + mn2) l(m + n)(b2n(l + m)(l n) + c2(n + l)2)), (5.2)

3
z = (n + l)(l(m + n)(b2n(l + m) + c2m(l n)) a2mn(l2 + lm nl mn))K,
K = l(m + n)(b2n(l + m) c2m(n + l)) a2mn(l + m)(n + l).

F is the intersection of WV and NL and has co-ordinates (x, y, z), where


x = l(m + n)(a2mn(l2 + lm + mn + nl) + l(n m)(b2n(l + m) c2m(n + l))),
y = m(n + l)(l(m + n)(b2n(l + m) + c2m(l n)) a2mn(l2 + lm nl mn)), (5.3)
z = n(l(m + n)(b n(l + m) + c m(l m)(l + n)) a mn(l + l (2m + n) + lm(m + 2n) + m2n)).
2 2 2 2 3 2

It may now be checked that the point R with co-ordinates (x, y, z) below lies on all three lines
AD, BE, CF.
x = 2l(a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)))H,
y = 2m(n + l)(a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l)))(l(m + n)(b2n(l +
m) + c2m(l n)) a2mn(l2 + lm nl mn)), (5.4)
2 2 2 2
z = (n + l)(l(m + n)(b n(l + m) + c m(l n)) a mn(l + lm nl mn))H,
H = (m + n)(a2mn(l + m)(n + l) + l(n m)(b2n(l + m) c2m(n + l))).

6. BE, MN, LV are concurrent at a point G

The line MN has equation


mnx = nly + lmz. (6.1)

The line BE has equation


((n + l)(l(m + n)(b2n(l + m) + c2m(l n)) a2mn(l2 + lm nl mn)))x = (2l(a2mn(l + m)(n
+ l) + l(m + n)(b2n(l + m) c2m(n + l))))z. (6.2)

The line LV has equation


(a2mn(l + m)(n + l) + b2nl(l + m)(m + n) c2lm(n + l)(m + n))(ny mz) mx(a2mn(l +
m)(n + l) b2nl(l + m)(m + n) c2lm(n + l)(m + n)) = 0. (6.3)

It may now be checked that these three lines all pass through the point G with co-ordinates (x, y,
z), where
x = 2nl(a2mn(l + m)(n + l) + l(m + n)(b2n(l + m) c2m(n + l))),
y = m(a2mn(n + l)(l2 +lm + mn + nl) l(m + n)(b2n(l + m)(l n) + c2m(n + l)2)), (6.4)
z = n(n + l)(l(m + n)(b2n(l + m) + c2m(l n)) a2mn(l2 + lm nl mn)).

As mentioned in Section 1 symmetry considerations now imply WU, MN, CF are concurrent at a
point H and WU, AD, LV are concurrent at a point K.

Flat 4, 12-14, Apsley Road,


BRISTOL BS8 2SP

4
Article 62
A Triangle with an Arbitrary Conic cutting its Sides
Christopher J Bradley

P
Z
R S

F A
N
V E
W M
G B L U C
X
D

1
1. Introduction

A conic is chosen that cuts the sides BC, CA, AB of a triangle ABC in points L, U; M, V; N, W
(in that order anticlockwise) respectively. The chords LW, MU, NV form a triangle DEF as
shown in the Figure. The following properties now hold: (i) AD, BE, CF are concurrent at a
point P; (ii) If NV meets BC at X, LW meets CA at Y, MU meets AB at Z then X, Y, Z are
collinear; (iii) NU, LV, AD are concurrent at a point Q; (iv) LV, MW, BE are concurrent at a
point R; (v) MW, NU, CF are concurrent at a point S; (vi) If MW meets BC at G, NU meets CA
at H, LV meets AB at K, then G, H, K are collinear.

These results are illustrated in the Figure and are established as true in later Sections using areal
co-ordinates with ABC as triangle of reference.

2. The choice of conic and the points L, M, N, U, V, W

The main difficulty with the work in this article is the choice of equation of a general conic. If
one of the usual forms is chosen than the points on the sides turn out to have square roots in
them, which are virtually impossible to handle algebraically. Once this difficulty is solved so that
the co-ordinates of the six points L, M, N, U, V, W have rational co-ordinates, then the results
follow fairly easily.

The equation of the conic chosen with a view to this end is


vw(f u)(g v)(h w)(gh vw)x2 + ... + ... (f2(gh + vw) fu(gw + vh + 2vw) + 2u2vw)yz
... ... = 0. (2.1)

The co-ordinates of the points on the sides may now be calculated and are:
L : (0, v(f u), hf wu); M : (fg uv, 0, w(g v)); N : (u(h w), gh vw, 0);
U : (0, fg uv, w(f u)), V : (u(g v), 0, gh vw); W : (hf wu, v(h w), 0). (2.2)

Carnots theorem (see Bradley [1]),


(BL/LC)(BU/UC)(CM/MA)(CV/VA)(AN/NB)(AW/WB) = 1 (2.3)

follows immediately

3. Points D, E, F and the perspective of ABC and DEF

The co-ordinates of the points D, E, F are D(f, v, w), E(u, g, w), F(u, v, h) so that AD, BE, CF
are concurrent at the point P(u, v, w).

4. The points X, Y, Z and their collinearity

2
The equation of the line DE is
w(v g)x + w(u f)y + (fg uv)z = 0. (4.1)

The equations of the lines EF and FD now follow by cyclic change of x, y, z and f, g, h and u, v,
w.

EF meets BC at X(0, g v, w h); Y is (u f, 0, h w), and Z is (f u, v g, 0). It follows that


X, Y, Z are collinear and the equation of XYZ is
x/(u f) + y/(v g) + z/(w h) = 0. (4.2)

5. LV, NU and AD are concurrent

The equation of LV is
v(f u)(gh vw)x + u(g v)(fh wu)y + uv(u f)(g v)z = 0. (5.1)

The equation of NU is
w(f u)(gh vw)x + wu(h w)(f u)y + u(h w)(uv fg)z = 0. (5.2)

These meet at the point Q(fu(g v)(w h), v(f u)(gh vw), w(f u)(gh vw)) and this
obviously lies on AD with equation wy = vz.

Similarly results (iv) and (v) hold.

6. The collinear points G, H, K

LV meets AB at K(u(v g)(fh wu), v(f u)(gh vw), 0). NU meets CA at H(u(w h)(fg
uv), 0, w(f u)(gh vw)). The equation of KH is
vw(f u)(gh vw)x + wu(g v)(hf wu)y + uv(h w)(fg uv)z = 0. (6.1)

The symmetry of this equation implies this line also passes through the point G = BC^MW.

Reference
1. C. J. Bradley, The Algebra of Geometry, Highperception, Bath, UK, (2007).

Flat 4,
Terrill Court,
12-14, Apsley Road,
BRISTOL BS8 2SP.

3
Article 63
A Cascade of Conics
Christopher J Bradley

F A
N
V
E

W M
P

B L U C

D
Figure

1. Introduction

Triangle ABC is given and a conic S is drawn through the vertices. A point P is chosen not on
the sides or their extensions and AP, BP, CP are drawn to meet S again at D, E, F. ABC and DEF
are therefore in perspective. DE is drawn to meet BC at U and CA at M. EF is drawn to meet
CA at V and AB at N. FD is drawn to meet AB at W and BC at L. It now turns out that a conic
may be drawn through the six points L, U, M, V, N, W. Furthermore triangles LMN and VWU
are in perspective with vertex P. It then follows by the same reasoning as before that the
intersections of non-corresponding sides of these two triangles lie on a conic. The process may

1
be repeatedly indefinitely, producing the cascade of conics in the title. The result is proved in the
following sections using homogeneous projective co-ordinates with ABC as triangle of reference
and P as unit point.

2. The conic S through A, B, C, D, E, F

Suppose the conic S through ABC has equation


fyz + gzx + hxy = 0. (2.1)

With P as unit point the line AP has equation y = z. It meets S at the point D with co-ordinates
( f, g + h, g + h). Similarly E and F have co-ordinates E(h + f, g, h + f), F(f + g, f + g, h ).

3. The points U, M, V, N, W, L

The equation of the line DE is


(g + h)x + (h + f)y hz = 0. (2.2)

DE meets BC at U(0, h, h + f) and DE meets CA at M(h, 0, g + h). Similarly V(f + g, 0, f), N(h +
f, , f, 0), W(g, g + h, 0) and L(0, f + g, g).

4. The conic through U, M, V, N, W, L

The conic may now be calculated and has equation


f(g + h)x2 + g(h + f)y2 + h(f + g)z2 (f2 + fg + hf + 2gh)yz (g2 + fg + gh + 2hf)zx
(h2 + hf + gh + 2fg) (2.3)

It may now be checked that WM, UN, VL all pass through P and so we once again have six
points on a conic whose vetices form two triangles in perspective and so the whole process may
be repeated and the cascade of conics is created.

Flat 4
Terrill Court
12-14, Apsley Road,
BRISTOL BS8 2SP

2
Article 64
Porism created by the Circumcircle and Triangles in Perspective

Christopher J Bradley

af M

A W
F ea

fb E U

P ce

L B
C
bd D

N dc

Figure

1. Introduction

1
Given a triangle ABC and its circumcircle S any point P is chosen that is not on its sides or the
extension of the sides and lines AP, BP, CP are drawn to intersect S again at D, E, F respectively.
The tangents at A, B, C meet EF, FD, DE at U, V, W respectively. The tangents at D, E, F meet
BC, CA, AB at L, M, N respectively. The tangents at A and F meet at af and points fb, bd, dc, ce,
ea are similarly defined. The following results now hold: (i) U, V, W are collinear; (ii) L, M, N
are collinear; (iii) points af, fb, bd, dc, ce, ea lie on a conic . This fact and the construction
leading up to it mean that the hexagon af fb bd dc ce ea has it vertices on and its sides touch S.
A hexagonal porism is thereby constructed. Furthermore (iv) af dc, fb ce, bd ea all pass through
P so the details of the construction may be repeated ad infinitum, first with as inconic and a
new conic as circumconic.

In the following Sections these results are established using areal co-ordinates with ABC as
triangle of reference.

2. The points D, E, F

Let P have co-ordinates P(l, m, n). The line AP has equation ny = mz and this meets the
circumcircle S with equation a2 yz + b2zx + c2xy = 0 at the point D with co-ordinates (x, y, z),
where
x = a2mn, y = m(b2n + c2m), z = n(b2n + c2m). (2.1)

The co-ordinates of E, F may now be obtained by cyclic change of x, y, z and a, b, c and l, m, n.

3. The points U, V, W and their collinearity

Given the co-ordinates of E, F from Section 2 we may obtain the equation of EF and this is
l(y(a2n + c2l) + z(a2m + b2l)) a2mnx = 0. (3.1)
2 2
The equation of the tangent to S at A is b z + c y = 0. These two lines meet at the point U with
co-ordinates U(l(c2m b2n), b2mn, c2mn) The co-ordinates of the points V, W may be obtained
from those of U, by cyclic change of x, y, z and a, b, c and l, m, n. The 3 x 3 determinant having
as its rows the co-ordinates of U, V, W vanishes proving that U, V, W are collinear. By
symmetry it follows (by reversing the roles of A, B, C and D, E, F) that L, M, N are also
collinear.

4. The points af, fb, bd, dc, ce, ea and the conic

The equation of the tangents at A, B, C are b2z + c2y = 0, c2x + a2z = 0, a2y + b2x = 0
respectively. The tangent at D has equation
(b2n + c2m)2x + a2b2n2y + c2a2m2z = 0. (4.1)

2
The tangents at E, F have equations that can be written down from Equation (4.1) by cyclic
change of x, y, z and a, b, c and l, m, n. It is now possible to obtain the co-ordinates of af, fb, bd,
dc, ce, ea, which are:
af : (a2m + 2b2l, b2m, c2m);
fb : (a2l, 2a2m + b2l, c2l);
bd: ( a2n, b2n + 2c2m, c2n);
dc: ( a2m, b2m, 2b2n + c2m);
ce : (a2l, b2l, 2a2n + c2l);
ea : (a2n + 2c2l, b2n, c2n).

It s now an extensive calculation to show these six points lie on the conic with equation
a4b2c2m2n2(b2n + c2m)x2 +... + ...+ a2(a6m2n2(b2n + c2m) + 2a4lmn(b4n2 + 2b2c2mn + c4m2)
+ a2l2(b6n3 + 4b4c2mn2 + 4b2c4m2n + c6m3) + b2c2l3(b4n2 + c4m2))yz + ... + ... = 0. (4.1)

The hexagon with the six points above has its vertices lying on the conic and its sides touch the
circumcircle S. Thus a porism of hexagons is created.

Furthermore, it can be checked that fb P ce are collinear as are af P dc and bd P ea. Thus
triangles af bd ce and fb dc ea have vertices on and are in perspective. It follows that the
construction may be repeated indefinitely.

See Article 28 for further information about porisms created by a pair of triangles in perspective.

Flat 4,
Terrill Court,
12-14 Apsley Road,
BRISTOL BS8 2SP.

3
Article 65
When I replaces K and Ge replaces H and Mi replaces O
Christopher J Bradley

A
W

T'
F

E'
T X'
Z S'
W V
M Y
F' L G Ge S
N' I E
Mi
N W'
Z'