Journal for Geometry and Graphics

Volume 24 (2020), No. 1, 9–28.

Constructions in the Locus of Isogonal

Conjugates for a Quadrilateral
Daniel Hu
Los Altos High School, 201 Almond Avenue, Los Altos, California, USA
email: daniel.b.hu@gmail.com

Abstract. Given fixed distinct points A, B, C, D, we examine properties of

and constructions involving the cubic plane curve locus of points X for which
(XA, XC), (XB, XD) are isogonal. We establish a necessary and sufficient con-
dition for all cubics C ∈ R2 such that there exist corresponding A, B, C, D ∈ C for
which C is this exact locus.
Key Words: isogonal cubic, focal cubic, isogonal conjugate, excellent point, spiral
center of quadrangle, Miquel point
MSC 2010: 51N20, 51N35, 51N15

1. Introduction
In this paper, we characterize the locus of all points P with an isogonal conjugate in a
given quadrilateral ABCD. This turns out to be a cubic plane curve, which we will call
the isogonal cubic of ABCD. The isogonal cubic is a well-established figure in geometry.
However, its properties are often considered with respect to the base quadrilateral ABCD,
without considering the isogonal cubic as an individual curve, and often neglecting degenerate
cases of ABCD.
The first half of the paper is dedicated to discovering geometric properties of non-
degenerate isogonal cubics, and also providing constructions on the isogonal cubic with
straightedge and compass. This sets up the second half, which characterizes all possible
non-degenerate cubics C ∈ RP2 such that there exist A, B, C, D ∈ R2 for which C is the isogo-
nal cubic of ABCD. We also establish the notion of the spiral center and isogonal conjugation
purely with respect to a valid cubic C.
The following is the main result we aim to prove throughout this paper:
Theorem 1. Let C be a non-degenerate cubic in R2 , and let C0 denote its embedding in CP2 .
Then the following two conditions are equivalent:
(1) There exist distinct A, B, C, D ∈ C such that C is the isogonal cubic of ABCD.
(2) The circular points at infinity [8] I, J lie on C0 , and the tangents at I, J meet on C0 .
In particular, the assertion that (2) is a sufficient condition requires special care in R2
and CP2 .

ISSN 1433-8157/$ 2.50 c 2020 Heldermann Verlag

10 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

1.1. Definitions and conventions

Definition 1 (Isogonality in RP2 ). For points P, A, B, C, D ∈ R2 , pairs of lines (P A, P C),
(P B, P D) are called isogonal if they share the same pair of angle bisectors (Figure 1). If P
is a real point at infinity while A, B, C, D remain in R2 , we slightly modify our definition of
isogonality to mean that for any line ` intersecting P A, P B, P C, P D at points E, F, G, H ∈
R2 , directed lengths EF and HG will be equal.

Figure 1: Definition of isogonality of the pairs (P A, P C) and (P B, P D)

The definition for points at infinity is equivalent to the midline of parallel lines P A, P C
being the same as the midline of parallel lines P B, P D, which complies with the idea of angles
as a conceptual measure of “distance” between two lines.
Definition 2 (Quadrilateral conventions and isogonal conjugates). We use the term “quadri-
lateral” throughout this paper to refer to possibly self-intersecting quadrilaterals, whose ver-
tices are distinct but possibly collinear. Points P, Q are called isogonal conjugates in a (pos-
sibly self-intersecting) polygon A1 A2 . . . An if and only if (Ai P, Ai Q), (Ai Ai−1 , Ai Ai+1 ) are
isogonal for all i, where indices are cycled mod n (Figure 2).
We will exclusively work in directed angles. For points X, Y , the notation XY denotes
the line XY if X and Y are distinct, while XY denotes the tangent at X if X ≡ Y and the
context of the curve containing X is clear (usually the isogonal cubic). The notation (XY Z)
denotes the circumcircle of XY Z provided X, Y, Z are distinct.
Definition 3 (Notation for intersection). We will let S ∩ T denote the intersection of sets of
points S and T , which is unique when S and T are distinct lines. When S is a cubic and T
is a line XY such that X, Y ∈ S, we will use the notation XY ∩ S to denote
• If either X or Y is a singular point, whichever one of X, Y is singular
• If X, Y are distinct and XY is not tangent to S, the third intersection of XY with S
• If X, Y are distinct and XY is tangent to S, the tangency point of XY with S
• If X, Y are not distinct and X is not an inflection point of S, the intersection of the
tangent to S at X with S
• If X, Y are not distinct and X is an inflection point, the point X
These are essentially equivalent to the third intersection of XY with C counting multiplicity.
We begin with this well-known characterization of all points with isogonal conjugates:
Theorem 2. For fixed distinct points A, B, C, D ∈ R2 not all collinear, a point P ∈ RP2 is
called excellent if (P A, P C) and (P B, P D) are isogonal. Then P is excellent if and only if it
has an isogonal conjugate in ABCD.
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 11

Figure 2: Definition of isogonal conjugates Figure 3: Collinear case of Theorem 1.5

Most proofs for this fact do not address the case when three of A, B, C, D are collinear,
so we will provide the full proof of this lemma for the sake of rigor.
Proof. The first case is when, without loss of generality, B, C, D are collinear. In this case,
we need to prove the following: For triangle ABC and D ∈ BC and point P , the isogonal
conjugate Q of P satisfies that BC is a bisector of angle ∠P DQ if and only if (P A, P D),
(P B, P C) are isogonal.
This is, in turn, equivalent to the following: For isogonal conjugates P, Q in ABC, if Q0A
be the reflection of Q over BC, then (P A, P Q0A ), (P B, P C) are isogonal. To prove this,
let P, Q have pedal triangles PA PB PC , QA QB QC respectively; by [2], these share the same
circumcircle ω centered at the midpoint of P Q. Let P PA meet ω at RA 6= PA ; P RA QA Q0A is
a parallelogram, so (Figure 3)

∠Q0A P C = ∠Q0A P PA + ∠PA P C = ∠QA RA PA + 90◦ − ∠P CB

= ∠QA PC PA + 90◦ − ∠P CB = ∠QA PC PB + ∠BPC PA + 90◦ − ∠P CB
= ∠QA QB QC + ∠BP PA + 90◦ − ∠P CB
= ∠QA QB Q + ∠QQB QC + 90◦ − ∠CBP + 90◦ − ∠P CB
= ∠BCQ + ∠QAB − ∠CBP − ∠P CB
= ∠P CA + ∠CAP + ∠BP C = ∠BP C + ∠CP A = ∠BP A

as desired.
Next, we prove this fact when no three of A, B, C, D are collinear. While the proof for
the general case is well-known, we will provide it for the sake of completion.
Lemma 1. For quadrilateral ABCD and point P , let the projections of P onto AB, BC,
CD, DA be E, F, G, H. Then EF GH is cyclic if and only if ∠AP B = ∠DP C.
Proof. With the various cyclic quadrilaterals,

∠EF G + ∠GHE = ∠EF P + ∠P F G + ∠GHP + ∠P HE

= ∠EBP + ∠P CG + ∠GDP + ∠P AE = ∠AP B + ∠CP D,

which directly implies the desired statement (Figure 4). 

12 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

Back to the main problem, drop perpendiculars E, F, G, H from P to AB, BC, CD, DA.
Let AB meet CD at X, and AD meet BC at Y . First, we prove that if isogonal conjugate
then ∠AP B = ∠DP C. Let P have isogonal conjugate P 0 in ABCD. Then P 0 is the isogonal
conjugate of P in both Y AB and XAD. Drop from P 0 perpendiculars E 0 , F 0 , G0 , H 0 ; by [2]
on Y AB, EF HE 0 F 0 H 0 is cyclic, and on XAD we get EGHE 0 G0 H 0 is cyclic. In other words,
F, F 0 , G, G0 all lie on (EE 0 HH 0 ), implying that EF GH is cyclic, hence ∠AP B = ∠DP C as
desired.
Now, we prove that if ∠AP B = ∠DP C then it has an isogonal conjugate P 0 . Then EF GH
is cyclic; let its circumcircle meet AB, BC, CD, DA at E 0 , F 0 , G0 , H 0 (Figure 4). By [2], the
perpendiculars to AB, BC, CD at E 0 , F 0 , G0 concur at a single point P 0 , the isogonal conjugate
of P in XBC. Analogously, the perpendiculars to AB, CD, DA at E 0 , G0 , H 0 concur at the
isogonal conjugate of P in XAD. In other words, P 0 H 0 ⊥ DA and is the isogonal conjugate
of P in both XAD and XBC, implying that P 0 is the desired isogonal conjugate of P in
ABCD.

Figure 4: Lemma 1 – projections onto Figure 5: Locus of excellent points in

sides of quadrilateral a parallelogram

1.2. Degenerate cases

One case where the locus of isogonal conjugates becomes degenerate is when A, B, C, D are
collinear on a line `. In this case, the locus becomes the line ` along with the circle centered
on ` whose inversion swaps A with C and B with D, if this circle exists.
Another case is when ABCD is a parallelogram, where we have the following characteri-
zation:

Theorem 3. If ABCD is a parallelogram, the locus of excellent points is the line of infinity,
along with the circumhyperbola passing through the points of infinity on the two angle bisectors
of ∠ABC (Figure 5).

Proof. By our extension of isogonality to RP2 , the line of infinity is part of this locus. Then
for all points P ∈ R2 , by the dual of Desargues’ involution theorem ([5], 133), (P A, P C),
(P B, P D) are isogonal if and only if angle AP C has the same angle bisectors as the pair of
lines through P parallel to AB and AD. Thus if P1 and P2 are the points of infinity along
with these angle bisectors of ∠BAD, we essentially need to find the locus P ∈ R2 for which
the angle bisectors of AP C are parallel to `1 and `2 .
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 13
We claim that this locus is the hyperbola H centered at the midpoint M of AC passing
through P1 , P2 , A, C. For any point P ∈ H, H becomes the circumrectangular hyperbola of
triangle P AC centered at M , which is the isogonal conjugate of the perpendicular bisector of
AC wrt P AC. The isogonal conjugates of P1 , P2 in P AC then become the two arc midpoints
of AC in (P AC), so P1 , P2 are indeed the points of infinity along the angle bisectors of P AC.
For the other direction, take any point P such that ∠AP C has angle bisectors passing through
P1 , P2 . Then the isogonal conjugate H0 of the perpendicular bisector of AC wrt P AC will
also pass through P1 , P2 , implying that H ≡ H0 , so P ∈ H, as desired. It is now clear that
B, D ∈ H, which completes the proof.
For the rest of the paper, we will assume quadrilateral ABCD does not fall under either
of these cases. In particular, A, B, C, D are not all collinear, and the midpoints of AC, BD
are distinct.

2. Preliminary lemmas
Up until Section 6, we will work in RP2 . All angles are directed mod 180◦ .
The following provides another well-known characterization of isogonal conjugates.

Definition 4. Let P be the Miquel Point of ABCD; we will also call the Miquel Point the
spiral center, short for the center of spiral similarity. For any point Y , call the unique point
Y 0 for which P is the spiral center of AY CY 0 the spiral inverse of Y .
Theorem 4. The spiral inverse X 0 of an excellent point X is also the isogonal conjugate of X.

Figure 6: Spiral inverses equal isogonal conjugates

Proof. Let E = AD ∩ BC and F = AB ∩ CD. If no three of A, B, C, D are collinear, we

have no problems, and otherwise we will assume without loss of generality that A, B, D are
collinear. Either way, the following relation is true:

∠DX 0 C = ∠DX 0 P + ∠P X 0 C = ∠XBP + ∠P AX = ∠AP B + ∠BXA = ∠DEC + ∠CXD.

Note that if A, B, D are collinear, then we would have B ≡ E and D ≡ F , but the above
angle chase would still hold. Similarly, ∠EX 0 C = ∠EDC + ∠CXE, implying that X, X 0 are
isogonal conjugates in CDE.
14 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

If no three of A, B, C, D are collinear (Figure 6), then analogously, X, X 0 are isogonal conju-
gates in BCF , implying X, X 0 are isogonal conjugates in ABCD, so we are done. Otherwise,
under our assumption that A, B, D are collinear, X, X 0 will be isogonal conjugates in BCD.
Since X is excellent, this implies X, X 0 are isogonal conjugates in ABCD, as desired.
The following proof reveals that the locus of excellent points X for a given quadrangle is
cubic (Figure 7).

Figure 7: Locus of excellent points: the isogonal cubic

Proof. Proving that the locus is cubic amounts to examining the equation
d−x
. d−x c−x
. c−x
=
a−x a−x b−x b−x
in the complex plane ([1], 6.1). Expanding this gives the desired third-degree equation in
x. Note that the coefficients of 3rd degree coefficients x2 x, xx2 in the expansion are both
zero if and only if a + c = b + d, which corroborates our earlier finding that if ABCD is a
parallelogram, then the locus would be degenerate.
Definition 5. Denote by C the cubic which is the locus of all excellent points X. We will
sometimes call C the “isogonal cubic” throughout this paper.
For the rest of this paper, we will assume that C is non-degenerate. Now, we may also
recall the following well-known fact.
Theorem 5 (Isogonal conjugate at infinity). Let M, N be the midpoints of AC, BD. Then
the isogonal conjugate of P is the point of infinity along M N .
Proof. Consider the parabola P tangent to the sides of ABCD. Considering P as an inscribed
parabola of ABC, by [12], the focus of P lies on (ABC) and the directrix passes through the
orthocenter of ABC. Applying the same logic to triangles BCD, CDA, DAB, we conclude
that the focus of P is the Miquel Point P , and its directrix is the Gauss-Bodenmiller line
([11]), which is perpendicular to M N .
It is well known ([2]) that for any conic with foci X1 , X2 and any point X for which tangents
from X exist, XX1 and XX2 are isogonal in the angle formed by the tangents from X to the
conic. Applying this to X ≡ A and conic P, we conclude that AP and the perpendicular from
A to the directrix are isogonal in ∠BAD. Similar relations with B, C, D imply the desired
result.
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 15
Theorem 6. Consider two pairs (X, X 0 ), (Y, Y 0 ) of isogonal conjugates. Then A, C are isog-
onal conjugates in XY X 0 Y 0 .
Proof. Since (AX, AX 0 ), (AY, AY 0 ) are isogonal, A, B, C, D are excellent in XY X 0 Y 0 . Since
A, C are spiral inverses in XY X 0 Y 0 , they are isogonal conjugates, as desired.
The following corollary immediately follows.
Corollary 1. For isogonal conjugates X, X 0 and excellent point Y , (Y X, Y X 0 ) and (Y A, Y C)
are isogonal.
This directly implies the following critical characterization:
Corollary 2 (Generalization of isogonal cubic). Consider two pairs (X, X 0 ), (Y, Y 0 ) of isog-
onal conjugates. Then the isogonal cubic of ABCD is the isogonal cubic of XY X 0 Y 0 . Fur-
thermore, any pair of isogonal conjugates (K, L) in ABCD are also isogonal conjugates in
XY X 0 Y 0 .
Proof. By Corollary 1, for any point Z, if pairs of lines (ZA, ZC), (ZB, ZD) are isogonal,
then lines (ZX, ZX 0 ), (ZY, ZY 0 ) are also isogonal, so ABCD and XY X 0 Y 0 indeed share the
same isogonal cubic. The second part then directly follows from Corollary 1.

Figure 8: Quadrilateral completeness

Thus the following is true by the dual of Desargues’ Involution Theorem on XY X 0 Y 0 :

Corollary 3 (Quadrilateral Completeness). For two pairs (X, X 0 ), (Y, Y 0 ) of isogonal conju-
gates, XY ∩ X 0 Y 0 and XY 0 ∩ X 0 Y lie on C (Figure 8).
We now illustrate the relationship between isogonality and inconics.
Theorem 7. Let ABCD have inconic ω and isogonal conjugates X, X 0 . Then the tangents
to ω from X and X 0 intersect at two pairs of isogonal conjugates.
Proof. Denote by IJKL the quadrilateral formed by these two pairs of tangents, such that
IJ, JK, KL, LI are tangent to ω. Since ω is an inconic of XIX 0 K and the tangents from A
to ω (AB and AD) are isogonal in ∠XAX 0 , by the dual of Desargues’ Involution on XIX 0 K
with respect to point A, (AI, AK) are also isogonal in ∠BAD. Similar arguments imply
(I, K), (J, L) are isogonal conjugates as desired.
Corollary 4. For excellent point X and isogonal conjugates Y, Y 0 , the line XY 0 is tangent
to the inconic of lines AB, BC, CD, DA, XY .
16 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

3. Relationship of inconics with excellent points

Consider any excellent point X. Let ω be the inconic of AB, BC, CD, DA, P X. It is well-
known that the center of ω lies on M N ([9]).
By Corollary 4, since P and the point of infinity P∞ along M N are isogonal conjugates,
the line `1 through X parallel to M N must also be tangent to ω. Reflect `1 over the center
of ω to get the second tangent `2 from the point of infinity ∞M N along M N to ω; let `2 meet
the second tangent from P to ω (other than P X) at X 0 (Figure 9).
Applying the proof of Corollary 7 on isogonal conjugates P and P∞ , points X and X 0
must also be isogonal conjugates. Let P X meet `2 at Y , and let P X 0 meet `1 at Y 0 ; then
Y, Y 0 are isogonal conjugates as well. We immediately get the following two corollaries:

Figure 9: Tangents to an inconic

Theorem 8. The midpoint of any two isogonal conjugates lies on M N .

Theorem 9. For X ∈ C and P∞ the point of infinity along M N , let Y = P X ∩ C and
X 0 = P∞ Y ∩ C. Then X, X 0 are isogonal conjugates.
We may also note that the midpoint of XY lies on M N . Since `1 k `2 , the bisectors of
∠P XP∞ , ∠P Y P∞ are parallel (perpendicular) to each other, with the perpendicular pairs of
bisectors intersecting on M N . Thus, the following is a direct result.
Theorem 10 (Parallel bisectors). If X, Y lie on C with P ∈ XY , then the midpoint of XY
lies on M N , and the bisectors of ∠AXC and ∠AY C are parallel to each other.
This produces a neat construction: if we are given an excellent point X, we may construct
an excellent point Y such that the bisectors of ∠AXC and ∠AY C are parallel to each other.
By Corollary 10, this is done by letting lines P X and M N intersect at a point O and setting
Y to be the reflection of X over O (Figure 10).

4. Constructing elements of the isogonal cubic

We begin by noting that since C has real coefficients, for any two non-singular points X and
Y on C in RP2 , line XY is either tangent to C at X or Y , or XY intersects C at a third point
in RP2 .
We begin with a well-known general lemma about cubics.
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 17

Figure 10: Angles with parallel bisectors

Lemma 2. From any point X on general non-degenerate cubic C, there are at most 4 points
Y ∈ C other than X for which XY intersects C at Y with multiplicity 2.
Proof. Note that if X is singular, there are no such points Y , or else line XY would intersect
C at both X and Y with multiplicity 2, yielding 2 + 2 = 4 total intersections. Assuming that
X is non-singular now, consider the embedding of C in CP2 with equation F (x, y, z) = 0. For
any point Y = (p : q : r) on C, Y is either a singular point, or the equation of the tangent at
Y is given by
∂F ∂F ∂F
(p, q, r) · x + (p, q, r) · y + (p, q, r) · z = 0
∂x ∂y ∂z
We want X = (x0 : y0 : z0 ) to satisfy the above equation, so fixing X gives us an equation in
p, q, r with degree 3 − 1 = 2. Let g(x, y, z) denote the expression
∂F ∂F ∂F
(x, y, z) · x0 + (x, y, z) · y0 + (x, y, z) · z0
∂x ∂y ∂z
Regardless of whether Y is a singular point of C or XY is tangent to C at Y , all such points
Y will be solutions to the cubic F (x, y, z) = 0 and the conic g(x, y, z) = 0, which by Bezout’s
Theorem ([4], Section 5.3) gives at most 3(3 − 1) total solutions.
Note that X itself also satisfies both equations; we now claim that X is actually a solution
with multiplicity at least 2. Let ` be the tangent to C at X; then ` has equation
∂F ∂F ∂F
(x0 , y0 , z0 ) · x + (x0 , y0 , z0 ) · y + (x0 , y0 , z0 ) · z = 0
∂x ∂y ∂z
To prove that X is a solution to both F and g with multiplicity 2, we use the fact that
IX (F, g) ≥ mX (F )mX (g), where IX denotes the multiplicity of the intersection of curves F
and g at X, and mX (F ), mX (g) denoting the multiplicity of point P on curves F, g ([4],
Section 3.3).
If X is a singular point of F , then IX ≥ 2 as desired. Otherwise, equality holds iff the tangents
at X to F and g are distinct. Thus it suffices to show that ` is tangent to the conic H formed
by g. To prove this, the tangent to H at X is given by equation Ax + By + Cz = 0, where
∂2F ∂2F ∂2F
A= (x 0 , y0 , z0 ) · x 0 + (x 0 , y0 , z0 ) · y 0 + (x0 , y0 , z0 ) · z0
∂x2 ∂y 2 ∂z 2
18 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

and B, C are defined similarly. By Euler’s Homogeneous Function Theorem ([10]),

∂F ∂2F ∂2F ∂2F
2· (x0 , y0 , z0 ) = (x 0 , y0 , z0 ) · x 0 + (x 0 , y0 , z0 ) · y 0 + (x0 , y0 , z0 ) · z0
∂x ∂x2 ∂y 2 ∂z 2
which implies that
∂F ∂F ∂F
A=2· (x0 , y0 , z0 ), B =2· (x0 , y0 , z0 ), C =2· (x0 , y0 , z0 )
∂x ∂y ∂z
so the tangent to H at X indeed has the same equation as `, as desired.
Thus X is a solution to C and H with multiplicity at least 2, so there are at most 3(3−1)−2 = 4
such points Y , as desired.
The following lemma also better characterizes C.
Lemma 3. In RP2 , C contains exactly one point at infinity.
Proof. The embedding of C in CP2 will contain the circular points at infinity I, J by virtue
of being isogonal conjugates, so back in RP2 there can only be one real point at infinity. On
the other hand, given ABCD the point of infinity along the Newton-Gauss line will lie on C,
so there is exactly one.
To better establish tangencies in C, we first need to examine singular points.
Theorem 11 (Singular points on the isogonal cubic). A point I ∈ C is a singular point if
and only if the isogonal conjugate of I is itself.
Proof. We remind our readers of our assumption in Section 1 that C is not degenerate.
First, we prove that if I is its own isogonal conjugate, then it is a singular point. Assume the
contrary, that I is not singular; then there are at most five non-singular points X ∈ C such
that XI is tangent to C at either X or I. For all points X such that XI is not tangent to C,
line XI will intersect C at a point Y 6= I, X. By Corollary 1 this means that the line through
I, X, Y bisects angles ∠AXC, ∠AY C.
In particular, this means line CX is the reflection of line AX over line XY , and line CY is
the reflection of line AY over line XY , which implies that C is the reflection of A over XY .
Thus XI is the perpendicular bisector of AC. But line XI rotates around I as we vary X
along the cubic, contradicting the uniqueness of the perpendicular bisector of AC, the desired
contradiction.
Next, we prove that if I is a singular point, then the isogonal conjugate of I is itself. Assume
the contrary; then let J 6= I be the isogonal conjugate of I. Choose any isogonal conjugates
K, L distinct from I, J (though we can set K ≡ A etc). By Corollary 3, X = KI ∩ LJ and
Y = KJ ∩ LI will lie on C. Since I is a singular point, KI will not intersect C at a point
other than K or I, so X is either the same as K or I.
If X ≡ I, then I, L, J are collinear. But since I is singular, line ILJ intersects C at I with
multiplicity 2, so our assumption that L 6= I, J implies that I ≡ J, the desired contradiction.
Thus we must have X ≡ K, so K, L, J are collinear. Thus we conclude J lies on line KL for
any isogonal conjugates K, L. Choosing another pair (R, S) of isogonal conjugates such that
no three of K, L, R, S are collinear, let T = KR ∩ LS and U = KS ∩ LR; by Corollary 3, T
and U are isogonal conjugates in C, so KL, RS, T U concur at a single point J.
Consider a conic H passing through K, L, R, S but not tangent to line T U . Then the pole
of line T U in H is the intersection of KL and RS, which is precisely J, which lies on line
T U . For the pole of T U in H to lie on T U itself, T U must be tangent to H at J, the desired
contradiction.
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 19

Figure 11: Tangents to the cubic

In this case where C contains a singular point, it becomes a strophoid. The paper [6] by
H. Stachel explores this specific case in detail.
Now, we will construct the tangent to C at any non-singular point X as follows (Figure 11).
Theorem 12 (Tangent to isogonal cubic). For isogonal conjugates X, X 0 , the isogonal ` to
XX 0 in ∠AXC is tangent to C at X.
Proof. As we move a point Y ∈ C with isogonal conjugate Y 0 , XY and XY 0 are isogonal in
∠AXC. Therefore, as Y approaches X 0 , Y 0 will approach X, eventually letting XY 0 intersect
C with multiplicity 2 at X.
In particular, the tangent to C at the point of infinity along M N is given by the unique
(by Corollary 3) asymptote of C.
Theorem 13. Let the tangents to C at isogonal conjugates X, X 0 meet at Y , and let XX 0
meet C at Z 6= X, X 0 . Then Y, Z are isogonal conjugates.
Proof. Let Z ∗ be the isogonal conjugate of Z. By Corollary 6 (XZ, XZ ∗ ) are isogonal in
∠AXC, and (X 0 Z, X 0 Z ∗ ) are isogonal in ∠AX 0 C, so Z ∗ ≡ Y , as desired.
Lemma 4. The bisectors of ∠AZC are perpendicular and parallel to XX 0 .
Proof. Examining isogonal conjugates (A, C), (X, X 0 ), this follows from Corollary 1.
Theorem 14. Let P Z meet C at W 6= Z. Then W X = W X 0 .
Proof. By Corollary 10, the bisectors of XW X 0 are perpendicular and parallel to XX 0 , which
gives the desired result.
Corollary 5. If we denote P by 0 on the conic, then W = X + X 0 under cubic addition ([4],
Proposition 5.6.4). Thus, the cubic sum of any two isogonal conjugates X, X 0 is equidistant
from X, X 0 .
We thus obtain the following construction, if we desire to find all pairs of isogonal conju-
gates (Y, Y 0 ) such that Y Y 0 passes through a given excellent point X.
Theorem 15. For X ∈ C, let distinct Y, Z 6= X lie on C such that X, Y, Z are collinear and
XY bisects ∠AXC. Then Y, Z are isogonal conjugates (Figure 12).
20 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

Figure 12: Isogonal conjugates collinear with a given point

Proof. Let Y 0 be the isogonal conjugate of Y ; then by Corollary 1, (XY, XY 0 ) and (XA, XC)
are isogonal. Since XY bisects ∠AXC, XY 0 and XY must be the same line, implying that
either Y ≡ Y 0 or Z ≡ Y 0 .
In the latter case we are done. In the former case, Y must be an incenter or excenter
of ABCD, so by Corollary 11 Y is a singular point. But X, Y, Z are collinear and distinct
despite line XY Z intersecting Y with multiplicity 2, the desired contradiction.
Note that this also gives us a construction of points Y on C such that ZY is tangent to C
at Y , where Z is a fixed point on C. This is done by letting X be the isogonal conjugate of
Z and intersecting the angle bisectors of ∠AXC with C. By the above, there will be up to
four such intersections X1 , X2 , X3 , X4 on C for which ZX1 , ZX2 , ZX3 , ZX4 are tangent to C
at X1 , X2 , X3 , X4 .

5. Constructing intersections with lines and circles

Theorem 16 (Line intersection). Consider excellent points X, Y . Denote by Z the intersec-
tion of the reflections of XY over the bisectors of ∠AXC and ∠AY C. Then the intersection
of XY with C other than X, Y is also the isogonal conjugate of Z. Furthermore, P XY Z is
cyclic.

Proof. Let W = XY ∩ C, and let W have isogonal conjugate W 0 . By Corollary 1, XW and

XW 0 are isogonal in ∠AXC, so line XW 0 is the reflection of XY over the bisectors of ∠AXC,
implying that W 0 ≡ Z, proving that XY ∩ C is indeed the isogonal conjugate of Z.
To prove P XY Z is cyclic, let X 0 , Y 0 be the isogonal conjugates of ABC. By Corollary 3,
XY ∩ C lies on X 0 Y 0 , hence W ∈ X 0 Y 0 . Then under spiral inversion, line X 0 Y 0 W is mapped
to the circumcircle of XY Z, which must pass through P , as desired.
As a direct corollary, we have the following well-known theorem:

Corollary 6 (Spiral center of isogonal conjugates lies on circumcircle). For isogonal conju-
gates (A, C), (B, D) in 4XY Z, the spiral center of ABCD lies on (XY Z).

We may remark that this provides a construction of the intersection of C with any line
XY , provided that X and Y lie on C themselves. Next, we characterize intersections of C
with circles.
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 21

Figure 13: Intersecting with circles

Theorem 17 (Circle intersection). Consider excellent points E, F, G with isogonal conjugates

E 0 , F 0 , G0 . Then (EF G) meets C at one other point which lies on (EF 0 G0 ), (E 0 F G0 ), (E 0 F 0 G).

Proof. There are two parts to this. First, we prove that if H is a point on C such that EF GH
is cyclic, then H lies on (E 0 F 0 G) (which would imply that it lies on (EF 0 G0 ), (E 0 F G0 ) by
symmetry). To prove this, by Corollary 6 with EGE 0 G0 , since H ∈ C, ∠F 0 GE 0 = ∠EGF =
∠EHF = ∠F 0 HE 0 .
Next, we prove that if H ≡ (EF G) ∩ (E 0 F 0 G), then ∠F 0 HE 0 = ∠F 0 GE 0 = ∠EGF = ∠EHF ,
which implies that E ∈ C as desired (Figure 13).

We may remark that this provides a construction for all points on (EF G) lying on C,
provided E, F, G lie on C themselves. The following is in fact true.

Theorem 18. All circles intersect C in the real plane at at most 4 points.

Proof. The circular points at infinity lie on C by virtue of being isogonal conjugates. The
result follows from Bezout’s Theorem, where curves of degree 2 and 3 meet for at most six
points in CP2 .

6. Characterizing the isogonal cubic

For this section, we will work in CP2 and let I, J denote the circular points at infinity. We
must first extend the definition of isogonality to CP2 as follows:

Definition 6. For distinct points P, A, B, C, D ∈ CP2 , call the two pairs of lines (P A, P B)
and (P C, P D) isogonal if and only if the three pairs of lines (P A, P B), (P C, P D), (P I, P J)
comprise a single involution, where I, J are the circular points at infinity.

One can check this complies with the angular definition of isogonality if P, A, B, C, D ∈ RP2 .

Corollary 7. For distinct points A, B, C, D such that neither of I, J lie on any of the lines
AB, BC, CD, DA, the locus of points X for which (XA, XB), (XC, XD) are isogonal is a
cubic (or curve of lesser degree) through A, B, C, D, I, J in CP2 .
22 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

Proof. For any six points A, B, C, D, E, F , the locus of points X for which (XA, XB),
(XC, XD), (XE, XF ) comprise a single involution is a cubic through A, B, C, D, E, F . Set-
ting E, F as the circular points at infinity gives the desired result.
Hence we will call a non-degenerate cubic C the “isogonal cubic” of quadrilateral ABCD
if it is the locus of all points X for which (XA, XC), (XB, XD) are isogonal (using the new
definition).

Corollary 8 (Loci of isogonality). If the locus of points X for which (XA, XC), (XB, XD)
are isogonal is a non-degenerate cubic, then neither I nor J cannot lie on any of the lines
AB, BC, CD, DA.

Proof. Assume the contrary, that WLOG I ∈ AB. Then for any point P on line AB, pairs
(XA, XC), (XB, XD), (XI, XJ) are part of a single degenerate involution. Thus the locus
of points X for which (XA, XC), (XB, XD) are isogonal includes line AB, contradicting the
proposition that the locus is a non-degenerate cubic.
In other words, if ABCD has a non-degenerate isogonal cubic C, then I and J will not
lie on AB, BC, CD, DA.
Now, the main result of this paper is the following:

Theorem 19 (Characterization of all isogonal cubics). Let C be a non-degenerate cubic in

CP2 containing circular points at infinity I and J at non-singular points. Then the following
two conditions are equivalent:
(1) There exist non-singular A, B, C, D ∈ C such that C is the isogonal cubic of ABCD.
(2) The tangents to C at I, J intersect each other on C.

We begin with the following direct result of Cayley-Bacharach ([7]).

Lemma 5 (Cubics containing complete quadrilateral). For P, Q on non-degenerate cubic C,

consider T ∈ C and let U = P T ∩ C, V = QT ∩ C such that P, Q, T, U, V are non-singular.
Then P V ∩ QU ∈ C iff P P ∩ QQ ∈ C.

Proof. Let X = P P ∩ QQ, Y = P V ∩ QU . Applying the Cayley-Bacharach Theorem on

triples of lines (XP P, QT V, QU Y ), (XQQ, P T U, P V Y ) completes both directions.

Lemma 6 (Locus of involution). Consider distinct points A, B, C, D, E, F in general position

and G = AC ∩ BD, H = AD ∩ BC, I = AE ∩ BF, J = AF ∩ BE, such that none of the
ten points are the circular points at infinity. Then there is a unique cubic C through these
ten points. Furthermore, for every P ∈ C, we have (P A, P B), (P C, P D), (P E, P F ),
(P G, P H), and (P I, P J) are part of a fixed involution (Figure 14).

Proof. By the dual of Desargues’ Involution Theorem, the locus C of all points P for
which (P A, P B), (P C, P D), (P E, P F ) are part of a single involution is a cubic through
A, B, C, D, E, F, G, H, I, J. Thus, there exists a cubic through these 10 points. Since
A, B, C, D, E, F are in general position, no four of the 10 constructed points are collinear.
Since C passes through these 10 fixed points, the cubic through these 10 points must be
unique, as desired.
We are finally set up to prove the main result.
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 23

Figure 14: Two complete quadrilaterals

Theorem 20 (Characterization in CP2 , condition (2) ⇒ (1)). Let C be a non-degenerate cubic

through I, J such that I and J are non-singular, and II intersects JJ at a point X on C.
Then there exist non-singular points A, B, C, D ∈ C apart from I, J such that C is the isogonal
cubic of ABCD.
Proof. Choose any point A ∈ C. Let B = IA ∩ C, D = JA ∩ C; by Corollary 5, ID and
JB meet a point C on C. Construct four points A0 , B 0 , C 0 , D0 ∈ C distinct from A, B, C, D
analogously, where I = A0 B 0 ∩ C 0 D0 and J = A0 D0 ∩ B 0 C 0 . We may select A, A0 such that
none of A, A0 , B, B 0 , C, C 0 , D, D0 are singular.
By Corollary 6, C is the locus of points P for which (P I, P J), (P A, P C), (P A0 , P C 0 ) are part
of a single involution. But since this involution concerns the circular points at infinity, it
follows that C is the locus for which (P A, P C), (P A0 , P C 0 ) are isogonal. We are done by
taking quadrilateral AA0 CC 0 .
Theorem 21 (Characterization in CP2 , condition (1) ⇒ (2)). Let C be the non-degenerate
isogonal cubic of ABCD where A, B, C, D are non-singular. Suppose that I, J lie on C at
non-singular points distinct from A, B, C, D. Then II ∩ JJ ∈ C.
Proof. Let X = AI ∩CJ and Y = AJ ∩CI; then X, Y are non-singular. Note that (XA, XC)
and (XI, XJ) are the same pair of lines, so (XA, XC), (XB, XD), (XI, XJ) form an invo-
lution. (If X is the same point as either A, C, I, or J, we instead use the tangent to C at
X when necessary.) So by definition, X ∈ C; similarly, Y ∈ C. By Corollary 5, this implies
II ∩ JJ ∈ C as desired.
Returning to R2 , we derive the complete characterization of all non-degenerate isogonal
cubics:
Theorem 22 (Characterization of all isogonal cubics in R2 ). Let C be a non-degenerate
cubic in R2 , and let C0 denote its embedding in CP2 . Then the following two conditions are
equivalent:
(1) There exist distinct A, B, C, D ∈ C such that C is the isogonal cubic of ABCD.
(2) The circular points at infinity I, J lie on C0 , and the tangents to C0 at I, J intersect
each other on C0 .
24 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

Proof. The only aspects of the proof we need to modify for R2 are to prove that:
1. Under the conditions of (1), if C is the isogonal cubic of ABCD where A, B, C, D are
distinct, then A, B, C, D cannot be singular points of C0 .
2. Under the conditions of (2), if any cubic C in R2 satisfies that its embedding C0 in CP2
passes through I and J, then I and J are not singular.
3. Under the conditions of (2), for A ∈ C such that K = AI ∩C0 , L = AJ ∩C0 , A0 = IL∩JK
all lie on distinct points of C0 , then the point A0 will be contained in C as well.
Let C have Cartesian equation ax3 + bx2 y + cxy 2 + dy 3 + G(x, y) = 0, where G is a second-
degree polynomial in x, y. Then a, b, c, d must be real, and since C is not degenerate, they
cannot all be zero. Thus C0 has equation F (x, y, z) = 0 where F (x, y, z) = ax3 + bx2 y + cxy 2 +
dy 3 + zP (x, y, z), where P (x, y, z) is a second-degree homogeneous polynomial in x, y, z.
Proof. For (a), assume that C is the isogonal cubic of ABCD. Note that A is a singular point
in C iff it is a singular point in C0 , because both are equivalent to ∂F
∂x
(A) = ∂F
∂y
(A) = ∂F
∂z
(A) = 0,
2 2
the same equation in both RP and CP . We just need to show that A is not a singular point
in R2 .
By Corollary 11, A is singular if and only if A is the isogonal conjugate of itself in ABCD.
But the isogonal conjugate of A is C, and since A and C are distinct, this cannot happen.
Therefore, A and similarly B, C, D cannot be singular points of C0 . This proves part (a). 
Proof. For (b), we consider general cubic C which contains I, J. Plugging in I = (1 : i : 0)
yields equation a + bi − c − di = 0, which implies that a = c and b = d because a, b, c, d
are all real. Thus C has equation (x2 + y 2 )(ax + by) + zP (x, y, z). We get ∂F ∂x
(1 : i : 0) =
2 2
3ax + 2xby + ay .
Assume, for the sake of contradiction, that I is a singular point. We require ∂F ∂x
= 0 for
(1 : i : 0), which rearranges to 2a + 2bi = 0. Since a, b are real, this implies that a = b = 0,
so a, b, c, d are all zero - the desired contradiction. This proves part (b). 

Proof. For (c), it suffices to show that A0 ∈ R2 . Note that I, J, A, K, L, A0 all lie on C0 , which
has all real coefficients. Now, K, L do not lie on the line of infinity, or else A would lie on the
line of infinity, which would imply C0 containing four points on a line and thus be degenerate,
a contradiction. Thus K and L are contained in C2 and thus can be expressed in Cartesian
coordinates (kx , ky ) and (lx , ly ) respectively. Since A ∈ R2 , note that K and L cannot lie in R2
— otherwise, the lines AK and AL would be entirely contained in RP2 , and would therefore
never intersect the line of infinity at complex points I and J.
From part (b), C must have equation of the form ax3 + bx2 y + axy 2 + by 3 + G(x, y) = 0 where
a, b and the coefficients of G are real numbers. For K to satisfy this equation, the point
K 0
whose Cartesian coordinates are the complex conjugates of K — that is, K 0 = kx , ky in
Cartesian coordinates — must also satisfy this equation, and thus lie on C0 . Having started
with A, I, K collinear, we now claim that A, J, K 0 are collinear. It suffices to show that

1 −i 0 1 i 0
ax ay 1 = 0 given that ax ay 1 = 0
kx ky 1 kx ky 1

Letting kx = p + qi and ky = r + si for p, q, r, s ∈ R, the second determinant equation yields

1 i 0
0 = ax ay 1 = −r − si − q + pi + ay − iax
p + qi r + si 1
 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 25
where (ax , ay ) are the Cartesian coordinates of A. Equating the real and imaginary parts
yields ay = q + r, ax = p − s. Similarly, the first determinant equation gives us

1 −i 0
0= ax ay 1 = −r + si − q − pi + ay + iax
p − qi r − si 1

and equating the real and imaginary parts yields ay = q + r, ax = p − s - the exact same
conditions. Therefore, given A, I, K are collinear, we indeed conclude A, J, K 0 are collinear.
In other words, K 0 is the unique intersection of C0 with AJ, hence K 0 ≡ L. Thus A0 =
IK 0 ∩ JK, and A0 will not be a point at infinity (otherwise K will also be a point at infinity).
Hence, in quadrilateral AKA0 K 0 ∈ C2 , we derive that AK meets A0 K 0 at a point of infinity,
and AK 0 meets A0 K at a point of infinity, so complex segments AA0 and KK 0 share the same
midpoint M . Letting A have Cartesian coordinates (m, n) in C2 , this means that M has
Cartesian coordinates
a + m a + n  
x y kx + kx ky + ky
, = ,
2 2 2 2

But kx +2 kx is just the real part of kx , so the coordinates of M are real as well. Hence m and
n are real, so A0 = (m, n) lies in R2 , proving part (c). 

With (a), (b), (c) proven, for the sake of completion we will show how this fully finishes
our characterization. For the direction (1) ⇒ (2), we start with ABCD, and by (a) none of
A, B, C, D are singular points. Then the result directly follows from Corollary 21.
For the direction (2) ⇒ (1), we start with circular points at infinity I, J lying on C0 , which
by (b) implies that I, J are not singular points. Assuming that the tangents to C0 at I and
J intersect each other on C, we can choose any point A ∈ C and letting K = AI ∩ C0 ,
L = AJ ∩ C0 , and A0 = IL ∩ JK where A0 ∈ C0 , and then choose another point B ∈ C and
define B 0 ∈ C0 the same way, such that all points formed by these intersections are distinct.
By Corollary 20, C0 will be the isogonal cubic of ABA0 B 0 . In addition, (c) implies that A0 and
B 0 will in fact lie in R2 as well. Therefore, ABA0 B 0 is fully contained in R2 , so C is indeed
the isogonal cubic of ABA0 B 0 . This completes the solution.

7. Uniqueness in the isogonal cubic

With this algebraic characterization of all isogonal cubics in R2 in mind, in this section, we
prove that given an isogonal cubic C ∈ RP2 , there is only one possible spiral center P , and
for any X ∈ C, there is only one possible point that could be the isogonal conjugate of X.
Theorem 23 (Uniqueness of the spiral center). Consider non-degenerate C ∈ RP2 such that
there exist A, B, C, D ∈ R2 for which C is the isogonal cubic of ABCD. Let ABCD have
spiral center P . Let C0 denote the embedding of C in CP2 . Then P I and P J are respectively
tangent to C0 at I and J.
Proof. Assume, for the sake of contradiction, that P I is not tangent to C0 at I; then P J
cannot be tangent to C0 at J either, so by part (c) of Corollary 22, P I and P J intersect C0 at
K, L ∈ C2 respectively, distinct from I, J, P , and IL and JK intersect C0 at Q ∈ R2 . Then
by Corollary 20, C is the non-degenerate isogonal cubic of the three quadrilaterals ABCD,
AP CQ, BP DQ.
26 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

By Corollary 3, there is one point of infinity P∞ ∈ C, which is the point of infinity along the
Newton-Gauss lines of ABCD, AP CQ, BP DQ. Let AP meet C at E; then by Corollary 9,
C, E, P∞ are collinear. Since C is the isogonal cubic of AP CQ, it follows that AP ∩ CQ ∈ C,
so in fact Q = CE ∩ C. Since C, E lie in R2 they are distinct from P∞ .
If C, E, P∞ are distinct, then Q ≡ P∞ , contradicting Q ∈ R2 , as desired. So line CE intersects
C with multiplicity 2. We assumed that ABCD ∈ R2 , so we cannot have C ≡ P∞ , hence
either C ≡ E or E ≡ P∞ . In either case, we cannot have Q ≡ P∞ else Q ∈ R2 is contradicted;
thus Q ≡ C. But considering A, P are the respective isogonal conjugates of C, Q in AP CQ,
so this implies A ≡ P . Now, the isogonal conjugates of A, P in ABCD are C, P∞ , which
implies that C ≡ P∞ — the desired contradiction.
In other words, a given non-generate isogonal cubic can only have one possible spiral
center — we may now call this the spiral center of a given isogonal cubic C. The point P is
called the singular focal point of C, conventionally defined as the intersection of the tangents
to C at I and J. This leads to the following result, allowing us to define isogonal conjugation
on any given isogonal cubic without having to construct a base quadrilateral ABCD:

Theorem 24 (Uniqueness of the isogonal conjugate). Consider non-degenerate C ∈ RP2 such

that there exist A, B, C, D ∈ R2 for which C is the isogonal cubic of ABCD. Then for any
point X ∈ C, there is only one possible point X 0 ∈ C which could be the isogonal conjugate of
X in ABCD.

Proof. Let P be the spiral center of C, and let P∞ be the point of infinity of C. Consider any
X ∈ C. If X ≡ P , its isogonal conjugate is P∞ , and vice versa.
If X is neither P nor P∞ , let Y = P X ∩ C and X 0 = P∞ Y ∩ C. Then by Corollary 9, X 0 is
the isogonal conjugate of X in ABCD no matter which ABCD we choose. Since P is fixed,
X 0 depends only on X, as desired.
Therefore, given any non-degenerate isogonal cubic C ∈ R2 and any point X ∈ C, the
spiral center P and the isogonal conjugate of X with respect to C are well-defined. Thus,
we may now revisit our constructions of intersections and tangents, this time with a general
isogonal cubic.

Theorem 25 (Tangents to the Isogonal Cubic). For non-singular X ∈ C, let X 0 be its isogonal
conjugate. Let ` be the isogonal of XX 0 wrt lines XP, XP∞ . Then ` is tangent to C at X.

Figure 15: Construction of the isogonal conjugate in a general cubic

 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral 27
Theorem 26 (Line intersections in the isogonal cubic). For distinct X, Y ∈ C, let `X be the
isogonal of XY wrt lines XP, XP∞ ; define `Y analogously. Then XY ∩ C is the isogonal
conjugate of `X ∩ `Y .
We note that with these results, the isogonal cubic is now equivalent to a focal curve, as
mentioned by F. Dingeldey in [3, Sect. 61]. It is also stated here that C is the locus of the
foci of all conics tangent to AB, BC, CD, DA.

8. Algebraic characterization in the cartesian plane

To conclude the paper, we present a purely algebraic characterization of all possible isogonal
cubics in R2 for the sake of completion.
Theorem 27. A non-degenerate cubic C ∈ R2 is an isogonal cubic of some quadrilateral
ABCD if and only if it has the form f (x, y) = f (p, q), where

f (x, y) = Ax3 + Bx2 y + Axy 2 + By 3 + Cx2 + Dxy + Ey 2 + F x + Gy

such that all coefficients are real and (A, B) 6= (0, 0), and
AE − AC − BD BC − AD − BE
p= , q= .
2(A2 + B 2 ) 2(A2 + B 2 )

Furthermore, the spiral center of C is (p, q), and the unique real asymptote of C is given by

(A3 + AB 2 )x + (A2 B + B 3 )y + (A2 E − ABD + B 2 C) = 0.

Proof. Let the embedding C0 of C in CP2 have equation g(x, y, z) = 0, where

g(x, y, z) = Ax3 + Bx2 y + Axy 2 + By 3 + Cx2 z + Dxyz + Ey 2 z + F xz 2 + Gyz 2 + Hz 3

where the equality of the coefficients of x3 with xy 2 and x2 y with y 3 is given by part (b) of
Corollary 22. Let g denote the left-hand side of the above equation. We compute
∂g ∂g
= 3Ax2 + 2Bxy + 2F xz + Ay 2 + Dyz + F z 2 , = 3By 2 + 2Axy + 2Eyz + Bx2 + Dxz + Gz 2
∂x ∂y
∂g
= 3Hz 2 + 2F xz + 2Gyz + Cx2 + Dxy + Ey 2
∂z
Plugging in the partial derivatives for (1 : i : 0), the tangent to C0 at (1 : i : 0) has equation

(2A + 2Bi)x + (−2B + 2Ai)y + (C + Di − E)z = 0

and similarly the tangent to C0 at (1 : −i : 0) has equation

(2A − 2Bi)x + (−2B − 2Ai)y + (C − Di − E)z = 0

The spiral center P of C is then given by the solution to these two equations. Solving yields

(x : y : z) = AE − AC − BD : BC − AD − BE : 2(A2 + B 2 )

Since A and B are not both 0, converting back to Cartesian coordinates implies that P indeed
has coordinates given by (p, q). In Cartesian coordinates, the value

Ax3 + Bx2 y + Axy 2 + By 3 + Cx2 + Dxy + Ey 2 + F x + Gy

28 D. Hu: Constructions in the Locus of Isogonal Conjugates for a Quadrilateral

must be a constant, particularly −H. Plugging in (p, q) immediately gives the equation for C
to be f (x, y) = f (p, q) as desired.
To determine the asymptote, we find that points of infinity on C0 are given by
0 = Ax3 + Bx2 y + Axy 2 + By 3 = (x2 + y 2 )(Ax + By)
so the real point of infinity is given by P∞ = (B : −A : 0). Plugging this into the equations
for the partial derivatives yields that the tangent to C0 at P∞ , and by extension the unique
real asymptote of C, indeed takes the above equation. This completes the proof.

Acknowledgements
The author would like to acknowledge and thank to Anant Mudgal for providing some
of the necessary theory in complex projective geometry, helping to formalize the analytic
characterizations, and checking over the solutions. The author would also like to give special
thanks to Michael Diao for checking the paper, suggesting additions to the content, and
helping to format the document and diagrams. Finally, the author would like to thank the
Editorial Board for their detailed comments and suggestions on the paper.

References
[1] Evan Chen: Euclidean Geometry in Mathematical Olympiads. Mathematical Association of
America, Washington, DC, 2016.
[2] Evan Chen: Three Properties of Isogonal Conjugates. Power Overwhelming, Wordpress,
30 November 2014, https://usamo.wordpress.com/2014/11/30/three-properties-of-isogo
nal-conjugates/.
[3] F. Dingeldey: Kegelschnitte und Kegelschnittsysteme. Encyklopädie der math. Wiss. III.2.1,
no. C1, 1903, pp. 1–16.
[4] W. Fulton: Algebraic Curves. Addison-Wesley, 1989, pp. 37–63.
[5] D.N. Lehmer: Elementary Course in Synthetic Projective Geometry. General Books, 2010,
pp. 88–103.
[6] H. Stachel: Strophoids are auto-isogonal cubics. G – Slovak Journal for Geometry and Graphics
12/24, 45–59 (2015).
[7] E.W. Weisstein: Cayley-Bacharach Theorem. Wolfram MathWorld, Wolfram Research, Inc.,
25 November 2019, http://mathworld.wolfram.com/Cayley-BacharachTheorem.html.
[8] E.W. Weisstein: Circular Point at Infinity. Wolfram MathWorld, Wolfram Research, Inc., 25
November 2019, http://mathworld.wolfram.com/CircularPointatInfinity.html.
[9] E.W. Weisstein: Complete Quadrilateral. Wolfram MathWorld, Wolfram Research, Inc., 25
November 2019, http://mathworld.wolfram.com/CompleteQuadrilateral.html.
[10] E.W. Weisstein: Euler’s Homogeneous Function Theorem. Wolfram MathWorld, Wol-
fram Research, Inc., 25 November 2019, http://mathworld.wolfram.com/EulersHomogeneous
FunctionTheorem.html.
[11] E.W. Weisstein: Gauss. Wolfram MathWorld, Wolfram Research, Inc., 25 November 2019,
http://mathworld.wolfram.com/Gauss-BodenmillerTheorem.html.
[12] E.W. Weisstein: Inconic. Wolfram MathWorld, Wolfram Research, Inc., 25 November 2019,
http://mathworld.wolfram.com/Inconic.html.

Received December 7, 2019; final form May 5, 2020