b

Forum Geometricorum
Volume 6 (2006) 263–268. b b
FORUM GEOM
ISSN 1534-1178

A Note on the Barycentric Square Roots

of Kiepert Perspectors

Khoa Lu Nguyen

Abstract. Let P be an interior point of a given triangle ABC. We prove that

the orthocenter of the cevian triangle of the barycentric square root of P lies on
the Euler line of ABC if and only if P lies on the Kiepert hyperbola.

1. Introduction
In a recent Mathlinks message, the present author proposed the following prob-
lem.
Theorem 1. Given an acute √ triangle ABC with orthocenter H, the orthocenter
H
of the cevian triangle of H, the barycentric square root of H, lies on the
Euler line of triangle ABC.

√ O
H

H H

B C

Figure 1.

Paul Yiu has subsequently discovered the following generalization.

Theorem 2. The locus of point√P for which the orthocenter of the cevian triangle
of the barycentric square root P lies on the Euler line is the part of the Kiepert
hyperbola which lies inside triangle ABC.
Publication Date: October 30, 2006. Communicating Editor: Paul Yiu.
The author is grateful to Professor Yiu for his generalization of the problem and his help in the
preparation of this paper.
264 K. L. Nguyen

The barycentric square root is defined only for interior points. This is the reason
why we restrict to acute angled triangles in Theorem 1 and to the interior points on
the Kiepert hyperola in Theorem 2. It is enough to prove Theorem 2.

2. Trilinear polars
Let A
B
C
be the cevian triangle of P , and A1 , B1 , C1 be respectively the inter-
sections of B
C
and BC, C
A
and CA, A
B
and AB. By Desargues’ theorem,
the three points A1 , B1 , C1 lie on a line P , the trilinear polar of P .
A

C
B

P
C
A1 B A

B1
C1

Figure 2.

If P has homogeneous barycentric coordinates (u : v : w), then the trilinear

polar is the line
x y z
P : + + = 0.
u v w
For the orthocenter H = (SBC : SCA : SAB ), the trilinear polar
H : SA x + SB y + SC z = 0.
is also called the orthic axis.
Proposition 3. The orthic axis is perpendicular to the Euler line.
This proposition is very well known. It follows easily, for example, from the
fact that the orthic axis H is the radical axis of the circumcircle and the nine-point
circle. See, for example, [2, §§5.4,5].
The trilinear polar P and the orthic axis H intersect at the point
(u(SB v − SC w) : v(SC w − SA u) : w(SA u − SB v)).
In particular, P and H are parallel, i.e., their intersection is a point at infinity if
and only if
u(SB v − SC w) + v(SC w − SA u) + w(SA u − SB v) = 0.
Equivalently,
(SB − SC )vw + (SC − SA )wu + (SA − SB )uv = 0. (1)
A note on the barycentric square roots of Kiepert perspectors 265

Note that this equation defines the Kiepert hyperbola. Points on the Kiepert hyper-
bola are called Kiepert perspectors.
Proposition 4. The trilinear polar P is parallel to the orthic axis if and only if P
is a Kiepert perspector.

3. The barycentric square root of a point

Let P be a point inside triangle ABC, with homogeneous barycentric coordi-
nates (u : v : w). We may √ assume u, v, w > 0, and define √ the barycentric square
√ √
root of P to be the point P with barycentric coordinates√ ( u : v : w).
Paul Yiu [2] has given the following construction of P .
(1) Construct the circle CA with BC as diameter.
(2) Construct the perpendicular to BC at the trace A
of P to intersect CA at X
.


√ of angle BX C to intersect BC at X.
(3) Construct the bisector
Then X is the trace of P √ on BC. Similar constructions on the other two sides
give the traces
√ Y and Z of P on CA and AB respectively. The barycentric
square root P is the common point of AX, BY , CZ.
Proposition 5. If the trilinear polar P intersects BC at A1 , then
A1 X 2 = A1 B · A1 C.
Proof. Let M is the midpoint of BC. Since A1 , A
divide B, C harmonically, we
have M B 2 = M C 2 = M A1 · M A
(Newton’s theorem). Thus, M X
2 = M A1 ·
M A
. It follows that triangles M X
A1 and M A
X
are similar, and ∠M X
A1 =
∠M A
X
= 90◦ . This means that A1 X
is tangent at X
to the circle with diameter
BC. Hence, A1 X
2 = A1 B · A1 C.

X

A1 B A
X M C

Figure 3.
266 K. L. Nguyen

To complete the proof it is enough to show that A1 X = A1 X
, i.e., triangle

A1 XX
is isosceles. This follows easly from
∠A1 X
X =∠A1 X
B + ∠BX
X
=∠X
CB + ∠XX
C
=∠X
XA1 .

Corollary 6. If X1 is the intersection of Y Z and BC, then A1 is the midpoint of
XX1 .
Proof. If X1 is the intersection of Y Z and BC, then X, X1 divide B, C har-
monically. The circle through X, X1 , and with center on BC is orthogonal to the
circle CA . By Proposition 5, this has center A1 , which is therefore the midpoint of
XX1 . 

4. Proof of Theorem 2
Let P be an interior point
√ of triangle ABC, and XY Z the cevian triangle of its
barycentric square root P .
Proposition 7. If H
is the orthocenter of XY Z, then the line OH
is perpendic-
ular to the trilinear polar P .
Proof. Consider the orthic triangle DEF of XY Z. Since DEXY , EF Y Z, and
F DZX are cyclic, and the common chords DX, EY , F Z intersect at H
, H
is
the radical center of the three circles, and
H
D · H
X = H
E · H
Y = H
F · H
Z. (2)
Consider the circles ξA , ξB , ξC , with diameters XX1 , Y Y1 , ZZ1 . These three
√
circles are coaxial; they are the generalized Apollonian circles of the point P .
See [3]. As shown in the previous section, their centers are the points A1 , B1 , C1
on the trilinear polar P . See Figure 4.
Now, since D, E, F lie on the circles ξA , ξB , ξC respectively, it follows from
(2) that H
has equal powers with respect to the three circles. It is therefore on the
radical axis of the three circles.
We show that the circumcenter O of triangle ABC also has the same power with
respect to these circles. Indeed, the power of O with respect to the circle ξA is
A1 O2 − A1 X 2 = OA21 − R2 − A1 X 2 + R2 = A1 B · A1 C − A1 X 2 + R2 = R2
by Proposition 5. The same is true for the circles ξB and ξC . Therefore, O also lies
on the radical axis of the three circles. It follows that the line OH
is the radical
axis of the three circles, and is perpendicular to the line P which contains their
centers. 
The orthocenter H
of XY Z lies on the Euler line of triangle ABC if and only
if the trilinear polar P is parallel to the Euler line, and hence parallel to the orthic
axis by Proposition 3. By Proposition 4, this is the case precisely when P lies on
the Kiepert hyperbola. This completes the proof of Theorem 2.
A note on the barycentric square roots of Kiepert perspectors 267

B1

D
Z √
P

P O
H

E F
X1 A1 B X C

C1

Z1

Figure 4.

√
Theorem 8. The orthocenter of the cevian triangle of P lies on the Brocard axis
if and only if P is an interior point on the Jerabek hyperbola.

Proof. The Brocard axis OK is orthogonal to the Lemoine axis. The locus of
points whose trilinear polars are parallel to the Brocard axis is the Jerabek hyper-
bola. 

5. Coordinates
In homogeneous barycentric coordinates, the orthocenter of the cevian triangle
of (u : v : w) is the point
268 K. L. Nguyen

    
 2    
1 1 1 1 1 1 1 1 1 1
SB + + SC + −SA + + SB − 2 + SC − 2
w u u v v w u2 w u2 v

: ··· : ··· .

2 2 2
 to the square root of the orthocenter, with (u : v : w ) =
 Applying this
1 1 1
SA : SB : SC , we obtain
 



a2 SA · SABC + SBC a2 SA : · · · : · · ·  ,

cyclic

which is the point H

in Theorem 1.
More generally, if P is the Kiepert perspector
 
1 1 1
K(θ) = : : ,
SA + Sθ SB + Sθ SC + Sθ
√
the orthocenter of the cevian triangle of P is the point


a2 SA (SA + Sθ )(SB + Sθ )(SC + Sθ )





+SBC a2 SA + Sθ + a2 Sθ SA SA + Sθ : · · · : · · ·  .
cyclic cyclic

References
[1] R. A. Johnson, Advanced Euclidean Geometry, 1925, Dover reprint.
[2] P. Yiu, Introduction to the Geometry of the Triangle, Florida Atlantic University lecture notes,
2001.
[3] P. Yiu, Generalized Apollonian circles, Journal of Geometry and Graphics, 8 (2004) 225–230.

Khoa Lu Nguyen: Massachusetts Institute of Technology, student, 77 Massachusetts Avenue,

Cambridge, MA, 02139, USA
E-mail address: treegoner@yahoo.com