THE NEUBERG CUBIC IN LOCUS PROBLEMS

ZVONKO ČERIN

Herrn Professor F. Schipp zum 60. Geburtstag gewidmet

Abstract. In this paper we continue the exploration of various locus problems whose so-
lutions involve the Neuberg cubic of the scalene triangle in the plane. We use analytical
geometry and the complex numbers to show that the Neuberg equation describes the es-
sential part of the locus in many geometric constructions. In this way we discover new
characteristics of the Neuberg cubic that has been extensively studied recently.

1. Introduction

Let ABC be a scalene triangle in the plane. The author has considered in [5] numerous locus
problems whose solutions involve the circular cubic N which Neuberg [18] calls the 21-point
cubic and which is known today as the Neuberg cubic of the triangle ABC. It is evident
from the extensive list of references on this curve given below that the Neuberg cubic has
attracted a lot of attention lately. The present paper is yet another such contribution. It
adds more than two dozens of new instances when the Neuberg cubic appears in various
geometric constructions. Most results utilise the notion of the homology for triangles but
there are also those that use the concurrence of lines and the concept of the power of a point
with respect to a circle.
Our proofs use the analytical geometry of complex numbers. This choice leads to the simplest
expressions and appears to be the most natural for our search for the Neuberg cubic. It is
suitable for implementation on computers. In fact, our results are all discovered with the
help of a computer (PC Pentium 200 MHz, 64 MB RAM) and the software Maple V (version
4). We leave out many details because Maple V (or any other package with symbolic algebra
computation capability) performs all factorisations and simplifications easily.
The paper is organised as follows. After the introduction we describe our notation and give
basics on the use of complex numbers in geometry. In the remaining sections we present and
prove some new results of our search for the Neuberg cubic that all give new characterisations
of this remarkable curve by various geometric constructions or locus problems. The section
titles are chosen to suggest the method of recognition.
Of course, since our results are characterisations of the same curve, in some cases one can
show easily that one method follows from the other(s). Observations of this kind and other
comments on possible extensions and special cases are included in remarks.

1991 Mathematics Subject Classification. Primary 51N20, 51M04, Secondary 14A25, 14Q05.
Key words and phrases. triangle, concurrent lines, Neuberg cubic, homologous triangles, reflections, pedal,
antipedal, isogonal conjugacy, inversion, locus.
1
2 ZVONKO ČERIN

2. Complex Numbers in Geometry

In this paper we shall use complex numbers in proofs because they provide simple expressions
and arguments. There are several books, for example [16], [9], [23], [12], [26], and [20], that
give introductions into the method which we utilise.
A point P in the Gauss plane is represented by a complex number. This number is called
the affix of P . It is customary to denote the affix of a point P with the corresponding small
Latin letter p and to identify a point and its affix. The complex conjugate of p is denoted
p̄. This rule has an important exception in that the vertices A, B, and C of the reference
triangle are represented by numbers u, v, and w on the unit circle. The letters a, b, and
c are reserved for the lengths of sides of ABC. Hence, the circumcentre O of ABC is the
origin. The affix of O is number 0 (zero) and complex conjugates of u, v, and w are u−1 ,
v −1 , and w−1 .
Let ϕ and ψ denote the first and the second cyclic permutation on triples of letters. For
example, ϕ(a) = b, ψ(a) = c, ϕ(u x) = v y, and ψ(u x) = w z. Finally, if f is an expression,
Sf and Pf replace f + ϕ(f ) + ψ(f ) and f ϕ(f ) ψ(f ). The expressions ϕ(f ) and ψ(f ) are
called relatives of f .
Most interesting points, curves,... associated with the triangle ABC are expressions that in-
volve symmetric functions of u, v, and w that we denote as follows: µ = u v w, σ = u + v + w,
τ = v w + w u + u v, σa = −u + v + w, σb = ϕ(σa ), σc = ψ(σa ), µa = v w, µb = w u, µc = u v,
τa = −v w + w u + u v, τb = ϕ(τa ), τc = ψ(τa ), δa = v − w, δb = ϕ(δa ), δc = ψ(δa ), ζa = v + w,
ζb = ϕ(ζa ), ζc = ψ(ζa ). For each k ≥ 2, σk , σka , σkb , and σkc are derived from σ, σa , σb , and σc
with the substitution u = uk , v = v k , w = wk . In a similar fashion we can define analogous
expressions using letters τ , µ, δ, and ζ.
Let us close this section on preliminaries with a few words on analytic geometry that we shall
use and on triangle notation. Any of the books mentioned above contains more than enough
information on basic constructions (line through two points, perpendicular and parallel to
a line through a point, condition for concurrence of three lines, condition for collinearity of
three points) that are needed to follow our arguments. As a convenience for the reader we
repeat them here.
In geometry lines are important so that we have the special notation [m, n] for the set of all
points P that satisfy the equation m̄ p − m p̄ + n = 0, where n is purely imaginary.
Let X, Y , and Z be three points with affixes x, y, and z and let ` be a line [f, h] in the
plane. Then the line XY is [x − y, x ȳ − y x̄], the parallel to ` through X is [f, f x̄ − f¯ x]
and the perpendicular to ` through X is [−f, −f x̄ + f¯ x].
The conditions for points X, Y , and Z to be collinear and for lines [m, n], [p, q], and [s, t] to
be concurrent are S x̄ (y − z) = 0 and S m (p̄ t − s̄ q) = 0. If X, Y , and Z are not collinear,
they determine the unique circle k(X, Y, Z) which goes through them.
The centroid, the circumcentre, and the centre of the nine-point circle of the triangle XY Z
are (S x)/3, (S x x̄ (y − z))/(S x̄ (y − z)), and (S x̄ (y 2 − z 2 ))/(2 S x̄ (y − z)).
Let P and Q be points and let ` be a line. Then pa(P, `), pe(P, `), pr(P, `), re(P, `), and
re(P, Q) denote the parallel to ` through P , the perpendicular to ` through P , the projection
of P onto `, the reflection of P in `, and the reflection of P at Q, respectively.
 THE NEUBERG CUBIC IN LOCUS PROBLEMS 3

For a point P not on the circumcircle of a triangle XY Z, let ig(P, XY Z) be the isogonal
conjugate of P with respect to XY Z. This point is the intersection of lines which make
equal angles with internal angle bisectors as do the lines XP , Y P , and ZP .
Let G, O, H, F , K, Iv , and Iu be the centroid, the circumcentre, the orthocentre, the centre
of the nine-point circle, the symmedian or Grebe-Lemoine point, the first isogonic point, and
the second isogonic point of the base triangle ABC.
We shall need triangles Ax Bx Cx , where the index x is either e, r, t, u, and v. In order to
describe Ax Bx Cx it suffices to give description of the vertex Ax because Bx and Cx are its
relatives. The point Ax is the excentre on AI, the reflection re(A, BC), the intersection of
tangents to the circumcircle at B and C, and the apexes of equilateral triangles constructed
inwards and outwards on BC, respectively. Ae Be Ce is the excentral, At Bt Ct the tangential,
and Ar Br Cr the three images triangle of ABC.
Triangles X1 Y1 Z1 and X2 Y2 Z2 are homologous if lines X1 X2 , Y1 Y2 , and Z1 Z2 are concurrent.

3. Homology of Triangles — Circumcentres

Among the oldest known methods of recognising the Neuberg cubic are the following two
theorems which use the condition that two families of triangles are families of homologous
triangles. In this and the next five sections we consider some other uses of this method for
recognition of the Neuberg cubic N . Division into sections reflects different ways of defining
families of variable triangles.
Let W0 be the complement of the union of sidelines of the base triangle ABC in the plane.
For a point P in the plane, let Oα , Oβ , and Oγ denote the circumcentres of the triangles
BCP , CAP , and ABP . Neuberg [18] has first proved the following result. As a convenience
to the reader we shall give easy proofs of this and the next theorem using complex numbers.
Theorem 3.1. The locus of all points P in W0 such that ABC is homologous to Oα Oβ Oγ
is the intersection with W0 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. The Oα is µa M/na , while Oβ and Oγ are its relatives, where na and M are equations
p + µa p̄ − ζa and p p̄ − 1 of the sideline BC and of the circumcircle. The line AOα has
the form [ f /na , g/(u na ) ], where f = u p + µ p̄ − u p p̄ − τa and g = M (u2 − µa ), while lines
BOβ and COγ are its relatives. The triangles ABC and Oα Oβ Oγ are homologous if and only
if M N P δa na−1 = 0, where N = τ p2 p̄ − µ σ p̄2 p + µ τ p̄2 − σ p2 + σ2 p − τ2 p̄ is the equation
of the Neuberg cubic [16]. ƒ

The following result is proved on page 199 of [16]. It was well known to readers of Mathesis
(see [13]) and was mentioned again recently in [21].
Theorem 3.2. The Neuberg cubic of ABC is the locus of all points P such that ABC is
homologous to the triangle on the reflections of P in the sidelines of ABC.

Proof. The reflection Rα of the point P in the side BC is ζa − µa p̄ and the line ARα is
[ µa p̄ − σa , u−1 µa p̄ − µ−1
a up + µ
−1
ζa (u2 − µa ) ]. Hence, triangles ABC and Rα Rβ Rγ are
homologous if and only if µ−1 2 N P δa = 0. ƒ
4 ZVONKO ČERIN

Our first theorem is similar to the Theorem 3.1. We just replace a point P with its isogonal
conjugate Q with respect to ABC. Let W1 denote the complement of the circumcircle of
ABC in the plane.
Theorem 3.3. The locus of all points P in W1 such that ABC is homologous to the triangle
on the circumcentres of BCQ, CAQ, and ABQ is the intersection with W1 of the Neuberg
cubic of ABC.

Proof. The affix of the isogonal conjugate Q is (p + τ p̄ − µ p̄2 − σ)/M and the circumcentre
Sa of BCQ is na /M . The circumcentres Sb and Sc of CAQ and ABQ have analogous affixes.
The triangles ABC and
Sa Sb Sc are homologous if and only if M −2 N P δa /u2 = 0. ƒ

In the next theorems, we shall replace circumcentres of triangles BCQ, CAQ, and ABQ
with the circumcentres of Br Cr Q, Cr Ar Q, and Ar Br Q, where Ar , Br , and Cr are vertices of
the three images triangle of ABC.
The locus of all points P whose isogonal conjugates lie on the sideline Br Cr of Ar Br Cr is
a conic ∆a . Let W3 denote the complement in the plane of the union of the circumcircle of
ABC, of the conic ∆a , and of two other related conics ∆b and ∆c .
Theorem 3.4. The locus of all points P in W3 such that ABC is homologous to the triangle
on the circumcentres of the triangles Br Cr Q, Cr Ar Q, and Ar Br Q is the intersection with
W3 of the union of the sidelines and the Neuberg cubic of ABC and a quartic which goes
through the vertices of ABC.

Proof. From the proof of previous theorem we know the affix of Q and since the affixes of
Br , and Cr are τb /v and τc /w, we can find the affix of the circumcentre Sa of Br Cr Q. The
circumcentres Sb and Sc of Cr Ar Q and Ar Br Q have analogous affixes. The triangles ABC
and
Sa Sb Sc are homologous if and only if K M −2 N P δa na /(u La ) = 0, where K and La denote ex-
pressions 2 (τ2 p3 p̄ + µ2 σ2 p p̄3 ) + (4 τ 3 − σ 2 τ 2 + 4 µ σ 3 − 15 µ σ τ + 12 µ2 ) p2 p̄2 +(4 µ − σ τ )(p3 + µ2 p̄3 ) + (8
−τ 2 − 2 µ σ)(2 p2 − σ p) + (σ τ 2 − µ σ 2 − 2 µ τ )(2 µ p̄2 − τ p̄) + 3 (σ 2 τ 2 − 2 µ σ 3 − 2 τ 3 + 7 µ σ
τ − 12 µ2 ) p p̄ + 2 (µ σ 3 − 6 µ σ τ + 8 µ2 ) and τa p2 + (ζ2a u2 − ζa (ζ2a − µa ) u + µa ζ2a ) p p̄ − u
(ζ2a + u ζa ) p + µ µa σa p̄2 − µa (µa ζa + u ζ2a ) p̄ + (v 2 + µb )(w2 + µc ). Notice that K = 0 is
the equation of a quartic which goes through the vertices of ABC while La = 0 is the equa-
tion of the conic ∆a . ƒ

Instead of the homology with ABC, the following result looks at the homology of Oα Oβ Oγ
with the triangle on circumcentres of Br Cr P , Cr Ar P , and Ar Br P . Let Wr0 denote the
complement of the union of sidelines of triangles ABC and Ar Br Cr in the plane.
Theorem 3.5. The locus of all points P in Wr0 such that the triangle Oα Oβ Oγ on cir-
cumcentres of BCP , CAP , and ABP is homologous to the triangle on the circumcentres of
the triangles Br Cr P , Cr Ar P , and Ar Br P is the intersection with Wr0 of the union of the
circumcircle of Ar Br Cr , the circumcircle of ABC, and the Neuberg cubic of ABC.

Proof. Since the affixes of Ar , Br , and Cr are τa /u, τb /v, and τc /w, the circumcentre Oar of the
triangle Br Cr P is ( µ τa p p̄ + 2 µa (µa − u2 ) p − σa τb τc )Br Cr (P )−1 , where Br Cr (P ) denotes
 THE NEUBERG CUBIC IN LOCUS PROBLEMS 5

the value at P of the equation of the line Br Cr . Now we can determine the line Oα Oar ,
find the lines Oβ Obr and Oγ Ocr using the usual substitutions, and discover that these lines
are concurrent if and only if k(Ar , Br , Cr )(P ) M N P u−1 δa3 Br Cr (P )−1 BC(P )−1 = 0, where
k(Ar , Br , Cr )(P ) is the value at P of the equation of the circumcircle of Ar Br Cr . ƒ
Remark 1. We can get analogous results to the above two theorems by replacing Ar Br Cr
with either Au Bu Cu or Av Bv Cv .

4. Homology of Triangles — Antipedal Triangles

The common feature of results in this section is that the homology of the antipedal triangle
of a variable point P with triangles on circumcentres is used to recognise the Neuberg cubic.
Theorem 4.1. The locus of all points P in W0 such that the antipedal triangle P α P β P γ
of P with respect to ABC is homologous to the triangle on the circumcentres of BCHα ,
CAHβ , and ABHγ , where Hα , Hβ , and Hγ are orthocentres of BCP , CAP , and ABP , is
the intersection with W0 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. The point Hα has affix (p2 − µa p p̄ + µa ζa p̄ − ζ2a )/na so that the affix of the circum-
centre Sa of BCHα is (ζa p + µa ζa p̄ − µa p p̄ − ζ2a − µa )/na . The lines P β Sb and P γ Sc are
relatives of P α Sa . These three lines are concurrent if and only if 2 M N P δa /(u na ) = 0. ƒ
Remark 2. Let Rα denote the reflection of a point P at the sideline BC. The triangles BCHα
and BCRα have the same circumcentre so that in the above theorem the orthocentres Hα ,
Hβ , and Hγ could be replaces with the reflections Rα , Rβ , and Rγ of P at the sidelines of
ABC.

Let W2 be the complement in the plane of the union of the three sidelines of ABC and the
three circles with sides of ABC as diameters.
Theorem 4.2. The locus of all points P in W2 such that the antipedal triangle P α P β P γ
of P with respect to ABC is homologous to the triangle on the circumcentres of BCOα ,
CAOβ , and ABOγ , where Oα , Oβ , and Oγ are circumcentres of BCP , CAP , and ABP , is
the intersection with W2 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. The point Oα has affix µa M/na so that the affix of the circumcentre Sa of BCOα is
µa (µa (p2 p̄2 − µa p̄2 − 1) − p2 + ζ2a − 2 U )/(na U ), where U = 2 µa p p̄ − ζa p − µa ζa p̄ + ζ2a
is the equation of the circle with BC as diameter. The lines P β Sb and P γ Sc are relatives of
P α Sa . These three lines concur if and only if 2 M N P δa (p − u)(u p̄ − 1)/(U na ) = 0. ƒ

5. Homology of Triangles — Orthocentres

Here we obtain the Neuberg cubic in homologies with triangles on the orthocentres of variable
triangles. The last result also uses the centres of the nine-point circles.
Theorem 5.1. The locus of all points P in W0 such that ABC is homologous to the triangle
on the orthocentres of the triangles OOβ Oγ , OOγ Oα , and OOα Oβ is the intersection with
W0 of the union of the circumcircle and the Neuberg cubic of ABC.
6 ZVONKO ČERIN

Proof. The orthocentre Ha of OOβ Oγ has affix u ζa M (p − u)/(nb nc ). Of course, the other
two orthocentres Hb and Hc have analogous affixes. Hence, the triangles ABC and Ha Hb Hc
are homologous if and only if M N P δa /(u na ) = 0. ƒ
Remark 3. We get an analogous result to the above theorem by replacing ABC with the
triangle Gα Gβ Gγ on centroids of BCP , CAP , and ABP .
Theorem 5.2. The locus of all points P in W0 such that ABC is homologous to the triangle
on the orthocentres of the triangles GGβ Gγ , GGγ Gα , and GGα Gβ is the intersection with
W0 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. The orthocentre Ha of GGβ Gγ has affix (2 µa p p̄ + u p + µ p̄ − τ − µa )/(3 na ). The

other two orthocentres Hb and Hc are relatives of Ha . It follows that the triangles ABC and
8
Ha Hb Hc are homologous if and only if 27 M N P δa /(u na ) = 0. ƒ
Remark 4. We get a similar result to the above theorem by replacing ABC with the triangle
Oα Oβ Oγ on circumcentres of BCP , CAP , and ABP .
Theorem 5.3. The Neuberg cubic of the triangle ABC is the locus of all points P such
that ABC is homologous to the triangle on either the centres of the nine-point circles or the
orthocentres of Rβ Rγ P , Rγ Rα P , and Rα Rβ P , where Rα , Rβ , and Rγ are reflections of P at
the sidelines BC, CA, and AB.

Proof. The orthocentre Ha of Rβ Rγ P is p + (τ − µa ) p̄ + ζa . The orthocentres Hb and Hc of

Rγ Rα P , and Rα Rβ P are relatives of Ha . The triangles ABC and Ha Hb Hc are homologous
if and only if µ2−1 N P δa = 0.
For the second part observe that the centre of nine-point circle of Rβ Rγ P is collinear with
the points A and Ha . ƒ

6. Homology of Triangles — Symmedian and Isogonic Points

The second theorem in this section is analogous to the following theorem which is an exercise
on page 200 of [16]. It was restated as the Superior Locus Problem by J. Tabov in [24] and
it was resolved by the author in [3] (see also [4]).
Theorem 6.1. The locus of all points P in W0 such that the Euler lines of the triangles
ABP , CAP , and BCP are concurrent (at the point on the Euler line of ABC) is the
intersection with W0 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. We know the circumcentre Oα of the triangle BCP and since its centroid Gα is
(p + ζa )/3 it follows that the Euler line Gα Oα of this triangle is
[ n−1 2 −1
a (p − 2 µa p p̄ + µa ζa p̄ + µa − ζ2a ), na M (p − µa p̄) ].

Hence, the Euler lines Gα Oα , Gβ Oβ , and Gγ Oγ concur if and only if M N µ−1 P δa n−1
a = 0.
Notice that these lines intersect on the Euler line GO of ABC. ƒ

Recall that the Brocard diameter or the Brocard axis are the names for the central line joining
the circumcentre with the symmedian point (or the Grebe-Lemoine point) of a triangle.
 THE NEUBERG CUBIC IN LOCUS PROBLEMS 7

Theorem 6.2. The locus of all points P in W0 such that the Brocard diameters of the
triangles ABP , CAP , and BCP are concurrent (at the point on the Brocard axis of ABC)
is the intersection with W0 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. The affix of Kα is (µa ζa p p̄ + 2 (µa − ζ2a ) p − 2 µ2a p̄ + µa ζa )/Ua , where the com-
plex number Ua is ((2 p − ζa )(2 µa p̄ − ζa ) − 3 δa2 )/2 and thus is never zero. The affix of
Oα is µa M/na , so that the triangles Kα Kβ Kγ and Oα Oβ Oγ are homologous if and only
if 8 M N P δa (p − u)(u p̄ − 1)/(na Ua ) = 0. Notice that the lines Oα Kα , Oβ Kβ , and Oγ Kγ
intersect on the Brocard axis KO of ABC. ƒ

The second theorem in this section is similar to the Theorem 3.1. In it we replace the
circumcentres with isogonic points. Let W5 be the complement in the plane of the apexes
Au , Bu , and Cu of equilateral triangles built towards inside on the sides of ABC.
Theorem 6.3. The locus of all points P in W5 such that ABC is homologous to the triangle
Iuα Iuβ Iuγ on the second isogonic points of BCP , CAP , and ABP is the intersection with
W5 of the union of the equilateral hyperbola through Au , Bu , and Cu with the centre at the
first isogonic point Iv of ABC and the Neuberg cubic of ABC.
√
Proof. The point Iuα is (U η + V )/(X η + Y ), where η denotes − 21 + i 23 (the cube root of
unity), the letter U is an abbreviation for µa ζa p p̄ + 2 (µa − ζ2a ) p − 2 µ2a p̄ + µa ζa , the letter
V for v µa p p̄ − δa p2 + (µa − ζ2a ) p − µa (µa − ζ2a ) p̄ + w µa + δ3a , and the letters X and Y
for 2 µa p p̄ − ζa p − 2 µa ζa p̄ + 4 µa − ζ2a and µa p p̄ − (δa + v) p − µa (δa + v) p̄ + v 2 + 2 w δa .
The other two second isogonic points Iuβ and Iuγ are relatives of Iuα . It follows that the
triangles ABC and Iuα Iuβ Iuγ are homologous if and only if Hu N P δa /(u2 (X η + Y )) = 0,
where Hu = 0 is the equation (in p) of the hyperbola from the statement of the theorem. In
order to see that X η + Y = 0 only when p = m, where m is the affix of Au , it suffices to
note that the value of X η + Y at m + n is equal (1 + 2 η) n n̄ v w. ƒ
Remark 5. Of course, there is a dual result to the above theorem with first isogonic points
of BCP , CAP , and ABP . The hyperbola of the locus has its centre at the second isogonic
point of ABC.

7. Homology of Triangles — Reflections

In this section we use homology with triangles whose vertices are reflections in appropriate
lines. Let W4 be the complement in the plane of the vertices A, B, and C of the triangle
ABC.
Theorem 7.1. The locus of all points P in W4 such that ABC is homologous to the triangle
on reflections in sidelines of ABC of inversions of A, B, and C with respect to the circles
k(B, C, P ), k(C, A, P ), and k(A, B, P ) is the intersection with W4 of the union of the
sidelines, the circumcircle, and the Neuberg cubic of ABC.

Proof. The inversion of A with respect to the circle k(B, C, P ) is (τa M − u na )/(u M − na )
and its reflection Ta in BC is (µ M − τa na )/(µa M − u na ). The other two reflections Tb
and Tc are relatives of Ta . The triangles ABC and Ta Tb Tc are homologous if and only if
M N P δa3 na /(u2 (u M − na )(µa M − u na )) = 0. From this our theorem follows immediately
provided one observes that up to a constant µa M − u na is a complex conjugate of u M − na
and both are zero only at the affixes of B and C. ƒ
8 ZVONKO ČERIN

Theorem 7.2. The locus of all points P in W4 such that ABC is homologous to the triangle
on reflections in sidelines of the extriangle Ae Be Ce of inversions of A, B, and C with respect
to the circles k(B, C, P ), k(C, A, P ), and k(A, B, P ) is the intersection with W4 of the
union of the circumcircle and the Neuberg cubic of ABC.

Proof. In this proof, in order to avoid the appearance of square roots, we shall assume
that the vertices A, B, and C have affixes u2 , v 2 , and w2 for some unimodular numbers
u, v, and w. The reflection Ta in Be Ce of the inversion of A with respect to the circle
k(B, C, P ) is (U M − u4 c2 (na ))/c2 (µa M − u na ), where U = µa (u4 + (µa − ζ2a ) u2 + µ2a )
and c2 performs the substitution u → u2 , v → v 2 , w → w2 . The other two reflections Tb
and Tc are relatives of Ta . The triangles ABC and Ta Tb Tc are homologous if and only if
M 4 c2 (N ) P c2 (δa )3 /(u2 c2 (u M − na ) c2 (µa M − u na )) = 0. ƒ
Theorem 7.3. If ABC has no right angle, then the locus of all points P in W4 such that
the tangential triangle At Bt Ct is homologous to the triangle on reflections in BC, CA, and
AB of the second intersections of lines AP , BP , and CP with the circumcircle of ABC is
the intersection with W4 of the union of the sidelines and the Neuberg cubic of ABC.

Proof. Since the affix of At is 2 µa /ζa , the affix of the second intersection Sa of AP with
the circumcircle of ABC is (u − p)/(u p̄ − 1), and the affix of the reflection Ta of Sa in BC
is (ζa p + µ p̄ − τ )/(p − u), the triangles At Bt Ct and Ta Tb Tc are homologous if and only if
2 N P δa na /(u ζa (p − u)(u p̄ − 1)) = 0. ƒ
Theorem 7.4. The locus of all points P in W4 such that the pedal triangle Pα Pβ Pγ of P
with respect to ABC is
homologous to the triangle on reflections in Pβ Pγ , Pγ Pα , and Pα Pβ of P is the intersection
with W4 of the union of the sidelines, the circumcircle, and the Neuberg cubic of ABC.

Proof. Since the affix of Pα is (p − µa p̄ + ζa )/2 and the affix of the reflection Ta of P in
Pβ Pγ is (p2 − u ζa p p̄ + ζa p − µ u p̄2 + u (τ + µa ) p̄ − ζb ζc )/(2 (p − u)), the triangles Pα Pβ Pγ
and Ta Tb Tc are homologous if and only if 16 1
M 2 N P δa na /(u2 (p − u)(u p̄ − 1)) = 0. ƒ
Remark 6. Since the triangle Rα Rβ Rγ on reflections of a point P in sides of ABC is homo-
thetic to the pedal triangle Pα Pβ Pγ from P , the above theorem holds also for Rα Rβ Rγ in
place of Pα Pβ Pγ .
Theorem 7.5. The locus of all points P in W0 such that the antipedal triangle P α P β P γ of
P with respect to ABC is
homologous to the triangle on reflections in P β P γ , P γ P α , and P α P β of P is the intersection
with W0 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. The affix of P α is (µa p p̄ − p2 + ζa p − 2 µa )/na and the affix of the reflection Ta of P
in P β P γ is 2 u − p, so that the triangles P α P β P γ and Ta Tb Tc are homologous if and only if
16 M N P δa /(u na ) = 0. ƒ

8. Homology of Triangles — Isogonal Conjugacy

Here we encounter the Neuberg cubic in homologies with triangles whose vertices are isogonal
conjugates of various points with respect to appropriate variable triangles.
 THE NEUBERG CUBIC IN LOCUS PROBLEMS 9

Theorem 8.1. The union of the circumcircle and the Neuberg cubic of the triangle ABC is
the locus of all points P such that the pedal triangle Pα Pβ Pγ of the point P with respect to
ABC is homologous to the triangle on isogonal conjugates of Pα , Pβ , and Pγ with respect to
triangles P Pβ Pγ , P Pγ Pα , and P Pα Pβ .

Proof. The vertex Pα has affix (p − µa p̄ + ζa )/2 while the isogonal conjugate Ta of Pα with
respect to the triangle P Pβ Pγ has affix (2 p − u M )/2. Hence, the triangles Pα Pβ Pγ and
1
Ta Tb Tc are homologous if and only if 16 M 2 N P δa /u2 = 0. ƒ
Remark 7. The above theorem holds also for the triangle on reflections of a point P in sides
of ABC instead of the pedal triangle Pα Pβ Pγ .
Theorem 8.2. The locus of all points P in W0 such that the triangle Oα Oβ Oγ on the
circumcentres of triangles BCP , CAP , and ABP is homologous to the triangle on the
isogonal conjugates of Oα , Oβ , and Oγ with respect to triangles P Oβ Oγ , P Oγ Oα , and P Oα Oβ
is the intersection with W0 of the union of the circumcircle and the Neuberg cubic of ABC.

Proof. The vertex Oα has affix µa M/na while the isogonal conjugate Ta of Oα with respect
to the triangle P Oβ Oγ has affix u M/(u p̄ − 1). It follows that the triangles Oα Oβ Oγ and
Ta Tb Tc are homologous if and only if M 4 N P δa /(na (p − u)(u p̄ − 1)) = 0. ƒ
Theorem 8.3. The locus of all points P in W4 such that the triangle Sα Sβ Sγ on the second
intersections of lines AP , BP , and CP with the circumcircle of ABC is homologous to the
triangle on the isogonal conjugates of Sα , Sβ , and Sγ with respect to triangles P Sβ Sγ , P Sγ Sα ,
and P Sα Sβ is the intersection with W4 of the union of the sidelines, the circumcircle, and
the Neuberg cubic of ABC.

Proof. The complex number (u − p)/(u p̄ − 1) is the affix of the vertex Sα . On the other
hand, (ζa p2 p̄ + µ p p̄2 − p2 − (τ + µa ) p p̄ + u p + µa )/((u − p)(v p̄ − 1)(w p̄ − 1)) is the affix
of the isogonal conjugate Ta of Sα with respect to the triangle P Sβ Sγ . Hence, the triangles
Sα Sβ Sγ and Ta Tb Tc are homologous if and only if M 6 N P δa na /((p − u)3 (u p̄ − 1)3 ) = 0. ƒ
Theorem 8.4. The locus of all points P in W1 such that the triangle Hα Hβ Hγ on the
orthocentres of triangles BCP , CAP , and ABP is homologous to the triangle on isogonal
conjugates of A, B, and C with respect to those triangles is the intersection with W1 of the
union of the sidelines and the Neuberg cubic of ABC.

Proof. Since ig(A, BCP ) is (p2 + u ζa p p̄ − σ p − µ p̄ + µa )/(u M ) and the vertex Hα has
the affix (p2 − µa p p̄ + µa ζa p̄ − ζ2a )/na it is easy to check that the triangles Hα Hβ Hγ and
ig(A, BCP )ig(B, CAP )ig(C, ABP ) are homologous iff M −3 N P δa na u−3 = 0. ƒ

9. Concurrent Parallels

Results in this section use the condition that three lines are concurrent. However, these lines
are not lines joining corresponding vertices of two triangles as in previous sections but are
parallels to lines.
Theorem 9.1. The Neuberg cubic of the triangle ABC is the locus of all points P such that
the parallels through A, B, and C to the Euler lines of triangles P Pβ Pγ , P Pγ Pα , and P Pα Pβ
formed by P and the vertices of its pedal triangle are concurrent.
10 ZVONKO ČERIN

Proof. The parallel pa(A, Ga Oa ) with the Euler line Ga Oa of P Pβ Pγ through the vertex A
is the line [ (u ζa p̄ − p − σa )/2, (µ σ p̄ − τ p + ζa (u2 − µa ))/(2 µ)]. The other two parallels
pa(B, Gb Ob ) and pa(C, Gc Oc ) are relatives of pa(A, Ga Oa ). The condition for these lines to
concur is 81 N P δa u−2 = 0. ƒ
Remark 8. The above theorem is true also when parallels to the Euler lines of P Pβ Pγ ,
P Pγ Pα , and P Pα Pβ are drawn through vertices of either the pedal triangle of P with respect
to ABC or the triangle on reflections of P in sidelines of ABC.
Theorem 9.2. The locus of all points P in W0 such that the parallels through the vertices
P α , P β , and P γ of its antipedal triangle to the Euler lines of triangles P P β P γ , P P γ P α , and
P P α P β are concurrent is the intersection with W0 of the union of
the circumcircle and the Neuberg cubic of ABC.

Proof. As in the proof of the previous theorem, we first find the parallel pa(P α , Ga Oa ) with
the Euler line Ga Oa of P P β P γ through the vertex P α . This line has a rather complicated
polynomial of order five in p and p̄ as the second term. Of course, the other two parallels
pa(B, Gb Ob ) and pa(C, Gc Oc ) are relatives of pa(A, Ga Oa ). These lines concur if and only
if 48 M 2 N P δa (p − u)(u p̄ − 1)/na = 0. ƒ
Remark 9. The above theorem remains true when parallels to the Euler lines of P P β P γ ,
P P γ P α , and P P α P β are drawn through vertices of the triangle on the second intersections
of lines AP , BP , and CP with the circumcircle of ABC.

10. Characterisations with Power

Neuberg [18] noticed the following theorem which requires the notion of the power of a point
with respect to a circle that we recall now.
Let P be a point and k be a circle in the plane with the centre S and the radius r. Then
the power ω(P, k) of the point P with respect to the circle k is the number |P S|2 − r2 . For
points X and Y in the plane, let k(X, Y ) denote the circle with the centre at X which passes
through Y .
Theorem 10.1. The Neuberg cubic of ABC is the locus of all points P in the plane such
that the product of powers of the point P with respect to the circles k(A, B), k(B, C), and
k(C, A) is equal to the product of powers of the point P with respect to the circles k(A, C),
k(B, A), and k(C, B).

Proof. Let W = p p̄ − u−1 p − u p̄. Since W + µc−1 (ζ2 c − µc ) and W + µb−1 (ζ2 b − µb ) are the
powers ω(P, k(A, B)) and ω(P, k(A, C)), the difference P w(P, k(A, B)) − P w(P, k(A, C))
2 N P δa .
is equal to µ−1 ƒ

The above result uses circles determined by two points (the centre and a point on it). Much
more interesting is to consider powers with respect to circles which are given by three points.
For a point P and triangles U V W and XY Z, let P ω(U, k(P, Z, X)) − P ω(U, k(P, X, Y ))
be ν(P, U V W, XY Z).
Theorem 10.2. The locus of all points P in W0 such that ν(P, ABC, Oα Oβ Oγ ) = 0 is the
intersection with W0 of the union of the circumcircle and the Neuberg cubic of ABC, where
Oα , Oβ , and Oγ are circumcentres of triangles BCP , CAP , and ABP .
 THE NEUBERG CUBIC IN LOCUS PROBLEMS 11

Proof. Since the quotient ((τ − µb ) p p̄ − µa ζc p̄ − ζc p + µb + v 2 )(p − u)(u p̄ − 1)/(u na nc ) is

the power ω(A, k(P, Oγ , Oα )) and the power ω(A, k(P, Oα , Oβ )) is the analogous quotient
((τ − µc ) p p̄ − µa ζb p̄ − ζb p + µc + w2 )(p − u)(u p̄ − 1)/(u na nb ) and all the remaining four
powers which appear in ν(P, ABC, Oα Oβ Oγ ) are their relatives, ν(P, ABC, Oα Oβ Oγ ) = 0
is true if and only if M N P δa (p − u)(u p̄ − 1)/(u n2a ) = 0. ƒ
Remark 10. The above theorem remains true when ν(P, ABC, Oα Oβ Oγ ) = 0 is replaced by
the equation ν(O, ABC, Oα Oβ Oγ ) = 0, where O is the circumcentre of ABC.
Theorem 10.3. The Neuberg cubic of ABC is the locus of all points P in the plane such
that ν(P, ABC, Rα Rβ Rγ ) = 0, where Rα , Rβ , and Rγ are reflections of P in the sidelines
BC, CA, and AB, respectively.

Proof. Since (u p + v µc p̄ − µc p p̄ + µc − ζ2c )/µc and (u p + w µb p̄ − µb p p̄ + µb − ζ2b )/µb are

powers ω(A, k(P, Rγ , Rα )) and ω(A, k(P, Rα , Rβ )) and the other four powers which appear
in ν(P, ABC, Rα Rβ Rγ ) are their relatives, ν(P, ABC, Rα Rβ Rγ ) = 0 is true if and only if
N P δa u−2 = 0. ƒ

Let W6 be the complement of the union of the circumcircles of triangles BCO, CAO, and
ABO in the plane. If P is a point different from the circumcentre O of ABC, let R denote
the inversion of P with respect to the circumcircle of ABC.
Theorem 10.4. The intersection of the Neuberg cubic of ABC with W6 is the locus of all
points P in W6 such that ν(R, ABC, Rα Rβ Rγ ) = 0.

Proof. Since (u p + v µa p̄ − µb − v 2 )(p − u)(u p̄ − 1)/(µc (ζb M − nb )) is ω(A, k(R, Rγ , Rα ))

and ω(A, k(R, Rα , Rβ )) is (u p + w µa p̄ − µc − w2 )(p − u)(u p̄ − 1)/(µb (ζc M − nc )) and all
the other four powers which appear in ν(R, ABC, Rα Rβ Rγ ) are their relatives, it follows that
ν(R, ABC, Rα Rβ Rγ ) = 0 is true if and only if N P δa (p − u)(u p̄ − 1)/(u2 (ζa M − na )) = 0.
From this our theorem follows immediately if we observe that ζa M − na = 0 is the equation
of the circle k(B, C, O) (or the sideline BC when the angle A is right). ƒ

When the angle A is right, let ka denote the sideline BC of ABC. Otherwise, we use ka
for a circle which passes through the points B and C and which has the lines joining these
points with the circumcentre of ABC as tangents. The (lines) circles kb and kc are defined
analogously. Let W7 be the complement in W1 of the union of ka , kb , and kc . For a point P
outside the circumcircle of ABC, let Q denote its isogonal conjugate with respect to ABC.
Theorem 10.5. The intersection of the union of the sidelines and the Neuberg cubic of
ABC with W7 is the locus of all points P in W7 such that ν(Q, ABC, Rα Rβ Rγ ) = 0.

Proof. Since ((τ − µc ) p p̄ − ζa p − µb ζa p̄ + µc + w2 )(p − u)(u p̄ − 1) na /(µ M (ζb M − 2 nb ))

is the power ω(A, k(R, Rγ , Rα )) and the power ω(A, k(R, Rα , Rβ )) is also the quotient
((τ − µb ) p p̄ − ζa p − µc ζa p̄ + µb + v 2 )(p − u)(u p̄ − 1) na /(µ M (ζc M − 2 nc )) and the other
powers which appear in ν(Q, ABC, Rα Rβ Rγ ) are their relatives, ν(Q, ABC, Rα Rβ Rγ ) = 0
if and only if M −2 N P δa na (p − u)(u p̄ − 1)/(u3 (ζa M − 2 na )) = 0. From this our theorem
follows provided one observes that ζa M − 2 na = 0 is the equation of the circle (line) ka . ƒ
Remark 11. Let ν0 (P, U V W, XY Z) = P ω(P, k(V, W, Y )) − P ω(P, k(V, W, Z)) for a point
P and triangles U V W and XY Z. It is interesting that in all results in this section replacing
12 ZVONKO ČERIN

the function ν with the above function ν0 the Neuberg cubic of ABC will again appear.
However, the exception sets are more complicated and the locus might include curves of
higher order.

References
[1] B. H. Brown, The 21-point cubic, Amer. Math. Monthly, 32 (1925), 110–115.
[2] B. H. Brown, Note on the preceding paper, Amer. Math. Monthly, 32 (1925), 247.
[3] Z. Čerin, Locus of intersections of Euler lines, (preprint).
[4] Z. Čerin, An extraordinary locus recognised, Math. and Inf. Quar. 6 (1996), 101–106.
[5] Z. Čerin, Locus Properties of the Neuberg Cubic, (preprint).
[6] H. S. M. Coxeter, Cubic Curves related to a Quadrangle, C. R. Math. Rep. Acad. Sci. Canada 15
(1993), 237–242.
[7] H. S. M. Coxeter, Some applications of trilinear coordinates, Linear Algebra and Its Appl. 226-228
(1995), 375–388.
[8] H. M. Cundy and C. F. Parry, Some cubic curves associated with a triangle, J. of Geom. 53 (1995),
41–66.
[9] R. Deaux, Introduction to the geometry of complex numbers, Ungar Publ. Co., New York, 1956.
[10] H. L. Dorwart, The Neuberg cubic: A nostalgic look, California Math. 3 (1978), 31–38.
[11] R. H. Eddy and J. B. Wilker, Plane mappings of isogonal-isotomic type, Soochow J. of Math. 18
(1992), 123–126.
[12] L. Hahn, Complex numbers and geometry, MAA, Washington, 1994.
[13] E. Hain, Question 653, Mathesis (1890), 22.
[14] T. W. Moore and J. H. Neelley, The circular cubic on twenty-one points of a triangle, Amer.
Math. Monthly 32 (1925), 241–246.
[15] F. Morley, Note on Neuberg cubic curve, Amer. Math. Monthly 32 (1925), 407–411.
[16] F. Morley and F. V. Morley, Inversive Geometry, Chelsea Publ. Co., New York, 1954.
[17] J. Neuberg, Memoire sur la tétraèdre, F. Hayez, Bruxelles, 1884.
[18] J. Neuberg, Sur la parabole de Kiepert, Annales de la Société scientifique de Bruxelles, (1909-1910),
1–11.
[19] J. Neuberg, Bibliographie du triangle et du tétraèdre, Mathesis, 38 (1924), 241.
[20] D. Pedoe, A course of geometry, Cambridge Univ. Press, Cambridge, 1970.
[21] G. M. Pinkernell, Cubic curves in the triangle plane, J. of Geom. 55 (1996), 141–161.
[22] P. Rubio, Cubic lines relative to a triangle, J. of Geom. 34 (1989), 152–171.
[23] H. Schwerdtfeger, Geometry of complex numbers, Oliver and Boyd, Toronto, 1962.
[24] J. Tabov, An extraordinary locus, Math. and Inf. Quar., 4 (1994), 70.
[25] O. M. Thalberg, Application of the theorem of residuation to the 21-point cubic, Amer. Math. Monthly
32 (1925), 412–414.
[26] I. M. Yaglom, Complex numbers in geometry, Academic Press, New York, 1968.
[27] P. Yff, Two families of cubics associated with a triangle, In Eves’ Circles (Ed.: Joby Milo Anthony),
MAA, Washington, 1993.

Kopernikova 7, 10010 Zagreb, Croatia

E-mail address: [email protected]