Sie sind auf Seite 1von 17

Result.Math.

55 (2009), 231247
c 2009 Birkhauser Verlag Basel/Switzerland
1422-6383/030231-17, published online September 2, 2009
DOI 10.1007/s00025-009-0417-6
Results in Mathematics
Coincidences of Centers of Plane Quadrilaterals
Abdullah Al-Sharif, Mowaaq Hajja and Panagiotis T. Krasopoulos
Abstract. We consider the three natural centers of mass associated with a
general quadrilateral together with the FermatTorricelli center and explore
the degree of regularity implied by the coincidence of two or more of these
four centers. We then extend this investigation to include cyclic quadrilaterals,
where we add the circumcenter to the list, and circumscriptible quadrilaterals,
where we add the incenter.
Mathematics Subject Classication (2000). Primary 51M04; Secondary 51M15.
Keywords. Bicentric quadrilateral; centroid; center of mass; circumscriptible
quadrilateral; cyclic quadrilateral; isometry; isosceles trapezoid; kite; ortho-
diagonal quadrilateral; Pithots theorem; Spieker center; symmetry.
1. Introduction
It is easy to see that if any two of the traditional centers of a triangle coincide,
then it is equilateral; see [41, Exercise 1, p. 37] and [31, pp. 7879]. This still holds
for many other pairs of centers as seen in [19], and fails for other rather articial
pairs as seen in [1]. Analogues for higher dimensional simplices are explored in
[14, 15, 22, 23] and [20], where the degree of regularity implied by the coincidence
of two or more centers is investigated. No similar work seems to have been done
for (planar) polygons.
In this paper, we consider the same problem for (planar) quadrilaterals. We
take a convex quadrilateral ABCD, and we explore the degree of regularity implied
by the coincidence of two or more of its four natural centers. These are (i) the 0-
centroid G
0
= the center of mass of four equal masses placed at the vertices, (ii) the
1-centroid G
1
= the center of mass of four rods (of equal uniform density) placed
over the sides, (iii) the 2-centroid G
2
= the center of mass of a lamina (of uniform
density) placed over the whole gure, and (iv) the FermatTorricelli center F =
the point whose distances from the vertices have a minimal sum. One may refer to
G
0
, G
1
, G
2
as the vertex, edge, and face centroids; see [41, 10.7, p. 167]. As for F,
The rst and second named authors are supported by a research grant from Yarmouk University.
232 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
B A
x
C D
z
y w
Figure 1. Cyclic Quadrilaterals with G
1
= G
2
are (rectangles
and) trapezoids dened by y = w = x +z.
D
C
B
A




Figure 2. Circumscriptible Quadrilaterals with G
1
= G
2
are
(rhombi and) kites dened by = = +.
it is easy to see that it is the intersection of the diagonals; see for example [11, VII,
5.5, p. 359] or [30, Reformulation 1.2 (II), p. 58].
If ABCD is a parallelogram with digonals intersecting at E, then all the
centers above (and in fact any center that respects isometries) coincide with E.
This is because reection about E is an isometry of ABCD with E as its only xed
point. In Section 3, we prove that none of the 5 coincidences G
0
= G
1
, G
0
= G
2
,
G
0
= F, G
1
= F, G
2
= F can occur except for a parallelogram. However, this
is not the case for G
1
= G
2
. In fact, degenerate quadrilaterals ABCD for which
G
1
= G
2
can be constructed from an equilateral triangle ABC by taking D to
be any point on AC. Then G
1
and G
2
coincide with the centroid of ABC. Non-
degenerate examples are shown in Figures 1 and 2. Note that if D is an endpoint
(resp. the midpoint) of side CA of an equilateral triangle ABC, then ABCD
is a limiting member in the family shown in Figure 1 (resp., Figure 2). If D is
anywhere else on AC, then ABCD is expected to be a limiting member in a third,
yet to be discovered, family in which G
1
= G
2
. It is relevant to add here that for
a triangle, G
1
, known as the Spieker center, coincides with G
2
only if the triangle
is equilateral [5], and that G
0
and G
2
(resp. G
d
) coincide for all triangles (resp.
d-simplices) [9].
Vol. 55 (2009) Centers of Quadrilaterals 233
F = O
A B
C
D
a
b
c
d
x
y
z
w
Figure 3. Dening x, y, z, w, a, b, c, d.
F = O
A B
C
D


a
b
c
d
Figure 4. Dening , , , .
The families in Figures 1 and 2 are discovered in Sections 4 and 5 where co-
incidences of the afore-mentioned and other centers are investigated for cyclic and
circumscriptible quadrilaterals. Here, ABCD is said to be (i) cyclic if it admits
a circumcircle (i.e., a circle that passes through the vertices), and (ii) circum-
scriptible if it admits an incircle (i.e., a circle that touches the sides internally).
The centers of the circumcircle and incircle are called the circumcenter and incenter
and are denoted by C and I, respectively. Quadrilaterals that are both cyclic and
circumscriptible are said to be bicentric and treated in Section 6. Basic properties
of these types of quadrilaterals can be found in [41, Chapter 10, pp. 146170].
2. Notation and preliminaries
Throughout this paper, ABCD is a convex non-degenerate quadrilateral lying in
some Euclidean plane whose origin O lies at the intersection of the diagonals. As
shown in Figure 3, the lengths of the sides AB, BC, CD, DA are x, y, z, w,
and the lengths of the semi-diagonals OA, OB, OC, OD are a, b, c, d. We set
s = x +y +z +w. If ABCD admits an incircle, then , , , are the lengths of
the tangents from A, B, C, D to the incircle, as shown in Figure 4. The centers
F = O, G
0
, G
1
, G
2
, I, C are as dened in Section 1.
234 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
Points in the plane are thought of as position vectors. Thus A = OA,
AB = AB, etc. Here AB denotes the length of the line segment AB. The
scalar product of X and Y is denoted by X Y . Areas are denoted by [ ].
We shall freely use the obvious facts that
A+B +C +D = O A+C = B +D = O. (2.1)
ABCD is a parallelogram A+B +C +D = O. (2.2)
A+C = O a = c . (2.3)
Letting t = cos AFB, we shall also freely use the trivial facts that
A A = a
2
, B B = b
2
, C C = c
2
, D D = d
2
,
A B = abt , A C = ac , A D = adt ,
B C = bct , B D = bd , C D = cdt . (2.4)
Theorem 2.1. Letting s = x +y +z +w, the centers G
0
, G
1
, G
2
are given by
G
0
=
A+B +C +D
4
(2.5)
G
1
=
(x +w)A+ (y +z)C + (x +y)B + (z +w)D
2s
(2.6)
G
2
=
A+B +C +D
3
. (2.7)
Proof. (2.5) is trivial, and (2.6) is obtained by replacing the side AB by two equal
masses of magnitudes proportional to x placed at the vertices A and B and doing
the same for the other sides.
For (2.7), replace triangles ABD, CBD by masses of magnitudes [ABD],
[CBD] placed at their centroids, i.e., at the points (A+B+D)/3, (C+B+D)/3.
Since [ABD] : [CBD] = a : c, it follows that
G
2
=
1
a +c
_
a(A+B +D)
3
+
c(C +B +D)
3
_
=
1
a +c
_
a(A+B +C +D)
3
+
c(A+B +C +D)
3
_
, by (2.3)
=
A+B +C +D
3
, as claimed .
We end this section with a simple lemma that will be used in Sections 4 and 5.
Lemma 2.2. If AC and BD are two coplanar non-parallel line segments, and if X,
Y are points in their plane, then
X = Y XA
2
XC
2
= Y A
2
Y C
2
and
XB
2
XD
2
= Y B
2
Y D
2
. (2.8)
Proof. The locus of a point X moving in such a way that |XA|
2
|XC|
2
is constant
is a line perpendicular to AC. Thus if the right hand side of (2.8) holds, then X
Vol. 55 (2009) Centers of Quadrilaterals 235
F
G0
G2
A
B
C D
X
Y
Z
W P
Q
Figure 5. Locations of F, G
0
, G
2
.
and Y lie on a line perpendicular to AC. Similarly, they lie on a line perpendicular
to BD. Therefore they coincide with the point of intersection of these two lines.
Remark 2.3. If X, Y , Z, W, P, Q are the midpoints of AB, BC, CD, DA, AC,
BD, respectively, then it is obvious that G
0
is the intersection of the diagonals of
the parallelogram XY ZW (commonly known as the Varignons parallelogram of
ABCD). It is also the midpoint of PQ. The centroids G
1
and G
2
also have simple
Euclidean constructions; see [29, 34]. It is also known that F, G
0
, G
2
are collinear
with G
0
dividing FG
2
in the ratio 3:1; see [27, 39].
3. Coincidence of centers of a general quadrilateral
Throughout this section, reference is made to a convex non-degenerate quadrilat-
eral ABCD. We adhere to the notation in Section 2.
Theorem 3.1 follows immediately from (2.2) and Theorem 2.1. So does The-
orem 3.2 which has also appeared in [39] and [27]. These are used in the proofs of
the remaining theorems. Theorem 3.4 is the trickiest.
Theorem 3.1. If any two of F, G
0
, G
2
coincide, then ABCD is a parallelogram.
Theorem 3.2. If ABCD is not a parallelogram, then F, G
0
, G
2
are distinct,
collinear, and FG
0
: G
0
G
2
= 3 : 1.
Theorem 3.3. If G
1
= G
0
, then ABCD is a parallelogram.
Proof. Suppose G
1
= G
0
. By Theorem 2.1, we have
(x +w)A+ (y +z)C + (x +y)B + (z +w)D
2s
=
A+B +C +D
4
.
Therefore (x +w y z)(A C) = (x +y z w)(B D) = 0, by (2.1). Since
A = C and B = D, it follows that x + w = y + z and x + y = z + w. Therefore
x = z, y = w, and ABCD is a parallelogram.
Theorem 3.4. If G
1
= F, then ABCD is a parallelogram.
236 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
O
A
B
C D
X
a
b
c
d
x
y
z
w
Figure 6. Illustrating the Proof of Theorem 3.4.
Proof. Suppose G
1
= F(= O). It follows from Theorem 2.1, (2.1), and (2.3) that
(x +w)a = (y +z)c , (x +y)b = (z +w)d . (3.1)
Without loss of generality, suppose that
[BOA] [COD] , [AOD] [COB] . (3.2)
Then [BDA] [CAD]. Therefore the height of B on AD is less than or equal
to that of C. Hence the line from B parallel to AD crosses the segment CD; see
Figure 6. Similarly, the line from D parallel to AB crosses the segment CB. Thus
the point X of intersection of these lines lies inside the triangle DBC. Also, ABXD
is a parallelogram. By Proposition 21 of Book I of Euclids Elements [16, 17],
BX +XD BC +CD with equality if and only if X = C, i.e.,
x +w y +z with equality if and only if ABCD is a parallelogram. (3.3)
Also, it follows from (3.2) and the sine law for areas that ad bc and ab cd.
Multiplying, we obtain a c, with equality if and only if ad = bc and ab = cd, i.e.,
a = c and b = d. Thus a c with equality if and only if ABCD is a parallelogram.
From this and (3.3) it follows that (x + w)a (y + z)c with equality if and only
if ABCD is a parallelogram. By (3.1), (x +w)a = (y +z)c. Therefore ABCD is a
parallelogram, as claimed.
We now combine Theorems 3.1, 3.3, and 3.4 in the following theorem.
Theorem 3.5. If any two centers in {G
0
, G
1
, G
2
, F}, other than G
1
, G
2
, coincide for
a quadrilateral, then it is a parallelogram.
Remark 3.6. In Sections 4 and 5, we shall exhibit families of quadrilaterals that
are not parallelograms and that have G
1
= G
2
. These are described in Figures 1
and 2, and they do not exhaust all possibilities.
4. Coincidence of centers of a cyclic quadrilateral
In this section, we specialize to the case when ABCD is cyclic, and we denote its
circumcenter by C. We adhere to the notations in Section 2, and we emphasize the
assumption that the diagonals intersect at the origin O which coincides with F.
Vol. 55 (2009) Centers of Quadrilaterals 237
Note that a cyclic quadrilateral is completely determined by its side lengths,
which in turn can be any numbers x, y, z, w satisfying 0 < x, y, z, w < (x + y +
z +w)/2; see [33, page 8] for a proof and [35] for a rigorous treatment.
Clearly, cyclic parallelograms are rectangles and cyclic trapezoids are isosce-
les. Thus in the context of cyclic quadrilaterals, we talk about trapezoids when we
mean isosceles trapezoids. In other words, we freely use the fact that, for a cyclic
quadrilateral, the conditions w = y, a = b, c = d, AB CD are equivalent.
We also freely use the following fact that holds for a cyclic quadrilateral:
a : b : c : d = wx : xy : yz : zw (4.1)
In fact, a : c = wx : yz because each is equal to [BAD] : [BCD], and a : b = w : y
because triangles AOD and BOC are similar.
According to Theorem 3.2, the centers F, G
0
, G
2
are collinear with |FG
0
| :
|FG
2
| = 3 : 4. The line joining these centers will be denoted by L(F, G
0
, G
2
) or
simply by L. For a rectangle, L is a single point. For an isosceles trapezoid, L
is the bimedian joining the midpoints of the parallel sides, this being the axis of
symmetry consisting of all points left invariant under all isometries.
Theorem 4.1 characterizes those quadrilaterals for which G
1
lies on L. It is
used to answer, in Corollary 4.2, the question of when G
1
coincides with certain
centers. Theorem 4.3 and Corollary 4.5 do the same for C. Theorem 4.6 character-
izes those quadrilaterals for which G
1
= C. A summary of the results is given in
Theorem 4.7 at the end of the section.
Note that X L if and only if X is a scalar multiple of A+B +C +D.
Theorem 4.1. Let ABCD be a cyclic quadrilateral that is not a rectangle. Then
G
1
L = L(F, G
0
, G
2
) if and only if ABCD is an isosceles trapezoid or
1
x
+
1
z
=
1
y
+
1
w
. (4.2)
If ABCD is an isosceles trapezoid with y = w and x = z, then
G
1
=
A+B +C +D
L
, L = 2 +
2y
y +x +z
, 2 < L < 4 . (4.3)
If (4.2) holds, then
G
1
=
A+B +C +D
L
, L =
2(y +w)s
(y +w)s 2yw
, 2 < L
8
3
. (4.4)
Proof. Using (2.6), we see that
G
1
L
(w +x)A+ (x +y)B + (y +z)C + (z +w)D
2s
=
A+B +C +D
L
for some L. Using (2.1) then (2.3) then (4.1), the right hand side reduces to
_
L(w +x) 2s
_
wx
_
L(y +z) 2s
_
yz
=
_
L(x +y) 2s
_
xy
_
L(z +w) 2s
_
zw = 0 .
238 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
Adding and subtracting, we obtain
G
1
L (x z)
_
L 2
4yw
(x +z)(y +w) +y
2
+w
2
_
= 0 and (4.5)
(y w)
_
L 2
4xz
(x +z)(y +w) +x
2
+z
2
_
= 0 . (4.6)
Suppose rst that G
1
L. Then (4.5) and (4.6) hold, and we have two cases:
Case 1. Neither x z nor y w is zero. In this case,
L 2
1
=
4yw
(x +z)(y +w) +y
2
+w
2
=
4xz
(x +z)(y +w) +x
2
+z
2
. (4.7)
This is equivalent to each of the following, where the rst is obtained by adding
the numerator to twice the denominator in each fraction in (4.7):
L 2
L
=
2yw
(y +w)s
=
2xz
(x +z)s
(4.8)
L
L 2
2
s
=
1
x
+
1
z
=
1
y
+
1
w
. (4.9)
Then (4.2) follows from (4.9). The formula for L given in (4.4) follows from (4.8).
To show that 2 < L 8/3, we use (4.8) and the AM-GM inequality to see that
L 2
L
s
2
=
xz
x +z

(
x+z
2
)
2
x +z
=
x +z
4
.
The same holds if x+z is replaced by y+w. Adding the two statements, we obtain
L2
L
s
s
4
. This implies that L 8/3. That L > 2 follows from (4.8). Thus we
have proved that if G
1
L, and if x = z and y = w, then (4.2) and (4.4) hold.
Case 2. y = w and x = z. Then it follows from (4.5) that L = 2 +
2y
x+z+y
which immediately implies also that 2 < L < 4. This proves (4.3). The case x = z
and y = w is similar.
Thus we have proved the only if part of the theorem. For the converse,
suppose that (4.2) holds. Taking L as in (4.4) and using (4.2), we see that (4.8)
holds. Therefore (4.7) and hence (4.5) and (4.6) are satised. Therefore G
1
L.
Of course, if ABCD is an isosceles trapezoid, then G
1
L. This completes the
proof.
Corollary 4.2. In a cyclic quadrilateral, G
1
= G
2
if and only if it is a rectangle, or
an isosceles trapezoid each of whose equal legs is one third of the perimeter.
Proof. Take L = 3 in (4.3) of Theorem 4.1 to obtain y = x +z, i.e., y = s/3.
The next theorem can be given a proof along the line of that of Theorem 4.1.
However, the proof we give is geometrically very transparent.
Vol. 55 (2009) Centers of Quadrilaterals 239
A
B
C D
P
Q
F
C
G
0
Figure 7. Illustrating the proof of Theorem 4.3.
Theorem 4.3. Let ABCD be a cyclic quadrilateral that is not a rectangle. Then
C L = L(F, G
0
, G
2
) if and only if ABCD is a trapezoid or AC BD. If ABCD
is a trapezoid, say AB CD, then
FG
0
: FC = 1 + cos AFB : 2 . (4.10)
If AC BD, then FG
0
: FC = 1 : 2.
Proof. Let P, Q be the midpoints of AC, BD. Without loss of generality, we may
assume that c a and d b.
Suppose that C L and that AC is not perpendicular to BD. We are to
show that ABCD is a trapezoid. Since ABCD is not a rectangle, it is not possible
that P = Q = F. So we may assume that Q = F. If P = F, then C, being the
point where the perpendicular bisectors of AC and BD intersect, does not lie on
any of the diagonals, while L coincides with the diagonal BD. This contradicts the
assumption C L. Thus we may assume that P = F (and Q = F), and therefore
that c > a and d > b. Then FP = (c a)/2 and FQ = (db)/2. Since C L,
F, G
0
, C are collinear. Since G
0
is the midpoint of PQ, it follows that the diameter
FC in the cyclic quadrilateral FPCQ bisects PQ and hence FC PQ. Therefore
PF = QF; i.e., c a = d b. From this and ca = db we obtain c +a = d +b,
and hence c and a are the same as d and b in some order. From c > a and d > b
it follows that a = b and c = d, and ABCD is a trapezoid.
We have thus shown that if C L, then either AC BD or ABCD is a
trapezoid. The converse is obvious. In fact, if AC BD, then PCQF is a rectangle
and G
0
, being the midpoint of PQ, is the midpoint of FC. Thus C L. This also
holds trivially if ABCD is a trapezoid.
Suppose now that ABCD is a trapezoid with a = b < c = d, and let t =
cos AFB and = cos PFC. Since FC bisects PFQ, we have t = 2
2
1. It
is also clear that FG
0
: FC = FP :
FP

=
2
: 1 = 1 +t : 2.
If AC BD, then G
0
is the midpoint of the diagonal PQ of rectangle FPCQ,
and therefore FG
0
: FC = 1 : 2, as desired. This completes the proof.
240 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
Remark 4.4. With reference to Theorem 4.3, cyclic quadrilaterals whose diagonals
are orthogonal to each other have attracted a good deal of attention in the liter-
ature. Section 10.4 (pp. 158159) of [41] is about such quadrilaterals, where they
are called orthodiagonal. Chapter 4 (pp. 3541) of [24], entitled On Quadrilaterals,
is devoted mainly to orthodiagonal quadrilaterals.
Corollary 4.5. If ABCD is a cyclic quadrilateral, then
C = F ABCD is a rectangle
C = G
0
ABCD is a rectangle
C = G
2
ABCD is a rectangle or an isosceles trapezoid whose diagonals
make equilateral triangles with its parallel sides .
Proof. If C = F, then a = b = c = d and ABCD is a rectangle. The second follows
from (4.10) and the fact that cos AFB = 1. The last follows from (4.10) and the
fact that FG
0
: FG
2
= 3 : 4, which follows from (2.5) and (2.7).
Theorem 4.6. Let ABCD be a cyclic quadrilateral that is not a rectangle. Then
C = G
1
if and only if ABCD is an isosceles trapezoid in which the equal legs y = w
and the parallel sides x and z are related by the relation
2y
2
y(x +z) (x
2
+z
2
) = 0 . (4.11)
Proof. In (2.6), replace C by cA/a and D by dB/b to obtain
G
1
= A+B, where (4.12)
2as = a(w +x) c(y +z) , 2bs = b(x +y) d(z +w) . (4.13)
Subtracting and adding and using (4.1), we obtain
2wxs 2xys = wx(w +x) yz(y +z) xy(x +y) +zw(z +w)
= (w y)
_
(w +y)(x +z) + (z
2
+x
2
)

(4.14)
2wxs + 2xys = (x z)
_
w
2
+y
2
+ (y +w)(x +z)

. (4.15)
Letting t = cos AFB, we obtain from (4.12) that
G
1
A
2
G
1

2
= a
2
2A (A+B) = a
2
2a
2
2abt . (4.16)
Using Lemma 2.2, we obtain
G
1
= C G
1
A
2
G
1
C
2
= G
1
B
2
G
1
D
2
= 0
2a + 2bt +c a = 2at + 2b +d b = 0 . (4.17)
The last equations are obtained from (4.16) and its analogues.
Suppose now that G
1
= C. Then the right hand side of (4.17) holds. Elimi-
nating t, we obtain (a c 2a)(2a) = (b d 2b)(2b), or
(as cs 2as)(2as) = (bs ds 2bs)(2bs) .
Using (4.13), we obtain
_
as cs a(w +x) +c(y +z)
_
(2as) =
_
bs ds b(x +y) +d(z +w)
_
(2bs) .
Vol. 55 (2009) Centers of Quadrilaterals 241
Replacing s by x +y +z +w and a, b, c, d, respectively, by wx, xy, yz, zw as in
(4.1)), and simplifying, we obtain
wy(x z)(2wxs 2xys) +xz(w y)(2wxs + 2xys) = 0 .
Using (4.14) and (4.15), we obtain (xz)(wy)P = 0, where P = wy[(w+y)(x+
z) +(z
2
+x
2
)] + xz[w
2
+y
2
+(y +w)(x+z)] > 0. Therefore x = z or y = w. Since
ABCD is not a rectangle, we may assume that y = w and x = z. Then ABCD is
an isosceles trapezoid with a = b, c = d, and x : z = a : c.
Also, it follows from (4.13) that = . From (4.13) and (4.17), we obtain
2xs = x(y +x) z(y +z) = (x z)(x +y +z) and 2x +2xt = x z. Dividing,
we obtain
1 +t =
s
x +y +z
, t =
y
x +y +z
. (4.18)
On the other hand, applying the law of cosines in triangles ABF, CDF, and ADF,
we obtain x
2
= 2a
2
(1 t), z
2
= 2c
2
(1 t), xz = 2ac(1 t), y
2
= a
2
+c
2
+ 2act.
Multiplying the last one by 2(1t) and using the rst three, we obtain 2y
2
(1t) =
x
2
+z
2
+ 2txz, and therefore
t =
2y
2
x
2
z
2
2y
2
+ 2xz
. (4.19)
Eliminating t in (4.18) and (4.19), we obtain (4.11).
Thus we have proved the only if part. To prove the converse, suppose
ABCD is an isosceles trapezoid with y = w and such that (4.11) holds. Using
(4.19), we obtain (4.18). Also, it follows from (4.14) and (4.15) that = and
2xs = (x z)(y + x + z). From this and (4.18), we see that x z = 2x(1 + t).
From x : z = a : c, we obtain a c = 2a(1 + t), thus proving (4.17), and hence
G
1
= C. This completes the proof.
Theorem 4.7 combines Theorem 3.5, Corollaries 4.2 and 4.5, and Theorem 4.6.
Theorem 4.7. For a cyclic quadrilateral ABCD that is not a rectangle, consider
the 10 possible coincidences between pairs of its 5 centers F, G
0
, G
1
, G
2
, C. Then
after relabeling, we have
G
1
= G
2
w = y = x +z , (4.20)
C = G
2
w = y , AFB = 60

, (4.21)
C = G
1
w = y , 2y
2
y(x +z) (x
2
+z
2
) = 0 . (4.22)
No other coincidence can occur.
5. Coincidence of centers of a circumscriptible quadrilateral
In this section, we specialize to the case when ABCD is circumscriptible, and we
denote the incenter by I. We again adhere to the notations in Section 2.
242 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
Circumscriptible parallelograms are clearly rhombi. As for kites, it is useful
to mention the obvious facts that if ABCD is circumscriptible, then the conditions
x = w, y = z, b = d, = are all equivalent.
We start with two fundamental theorems. The rst is commonly referred
to as Pithots theorem and has appeared quite frequently; see [13, pp. 2728],
[2, pp. 6567], [41, Theorem 10.3.1, p. 156], [25, pp. 6264], [40, Exercise 10.19,
p. 260], and [6]. It is remarked in [40] that Pithots theorem is a theorem in Neutral
Geometry.
Theorem 5.1. ABCD is circumscriptible if and only if x +z = y +w.
In view of Theorem 5.1, no matter how one exes a circumscriptible quadri-
lateral, it remains circumscriptible. Thus, unlike a cyclic quadrilateral, a circum-
scriptible one is not determined by its side lengths x, y, z, w. On the other hand,
a circumscriptible quadrilateral is completely determined by , , , , which in
turn can be any positive numbers. In fact, this is true for circumscriptible n-gons
for all n, as shown in [18].
Theorem 5.2. The circumscriptible quadrilateral ABCD is a kite if and only if
AC BD.
Proof. Suppose that AC BD. It follows from Pythagoras theorem that x
2
+z
2
=
y
2
+w
2
and it follows from Theorem 5.1 that x +z = y +w. Therefore xz = yw.
From this and x + z = y + w, we conclude that x and z are the same as y and w
in some order. Hence ABCD is a kite. The converse is trivial.
Note that if ABCD is a kite, say = , and if it is not a rhombus, then AC
is the only axis of symmetry and consists of the points xed by all isometries of
the kite. Thus F, G
0
, G
1
, G
2
, I all lie on AC. This will be freely used.
Theorem 5.3 is due to Newton who discovered it in the context of determining
the locus of the centers of all ellipses that can be inscribed in a given convex
quadrilateral; see [12, Problem 49, pp. 217219]. Here and elsewhere, the line
joining two points is understood to be a singleton if the two points coincide.
Theorem 5.3. In a circumscriptible quadrilateral, the incenter I lies on the line
segment joining the midpoints of the diagonals.
Theorem 5.4 is taken from [37, Theorem 1], where it is attributed, together
with generalizations to polygons of any number of sides and to polyhedra in space,
to the 1943 paper [8]. The analogue for tetrahedra appears in [10, p. 69]. The
analogue for triangles appears in [7, pp. 137138], [32], [26, pp. 225227], and the
rst row in [28, Table 5.5, p. 143]. It was recently rediscovered in [3], where the
authors generalized it to circumscriptible gures much more general than polygons.
The same authors generalized their results to higher dimensional spaces in [4].
Theorem 5.4. In a circumscriptible quadrilateral, I, G
2
, and G
1
are collinear and
IG
2
: G
2
G
1
= 2 : 1.
Vol. 55 (2009) Centers of Quadrilaterals 243
It follows that the coincidences I = G
2
, G
2
= G
1
, G
1
= I are equivalent, and
each is equivalent to I = G
2
= G
1
. They are treated in the next theorem.
Theorem 5.5. Let ABCD be a circumscriptible quadrilateral that is not a rhombus.
Then two, and hence all, of the centers I, G
2
, G
1
coincide if and only if = =
+ or = = +.
Proof. Suppose I = G
2
= G
1
.
If I = G
0
, then G
0
= G
2
and hence ABCD is a rhombus, by Theorem 3.5.
If I = G
0
and ABCD is not a kite, we will reach the contradiction I = G
0
.
Since ABCD is not a kite, none of its diagonals bisects the other. Letting P, Q be
the midpoints of the diagonals, as in Figure 7, we see that G
0
, being the midpoint
of PQ, is the only point common to the lines FG
0
and PQ. Since I lies on PQ,
by Theorem 5.3, and G
2
lies on FG
0
, and since I = G
2
, it follows that I = G
0
, a
contradiction. Thus ABCD is a kite, and we may assume that = and = .
Thus it remains to show that if = , then = = + G
2
= I.
From (2.7) and B +D = O, it follows that G
2
= (A+C)/3. Therefore
9
_
AG
2

2
CG
2

2
_
= 9
_
_
_
_
_
A
A+C
3
_
_
_
_
2

_
_
_
_
C
A+C
3
_
_
_
_
2
_
= 2AC
2
2C A
2
= 3(a
2
c
2
)
= 3
_
AD
2
CD
2
_
, (by Pythagoras Theorem)
= 3
_
( +)
2
( +)
2
_
= 3( )( + + 2) .
Also, AI
2
CI
2
=
2

2
= ( )( +). Therefore
AG
2

2
CG
2

2
= AI
2
CI
2
= = + .
Since BG
2

2
DG
2

2
= BI
2
DI
2
= 0, it follows from Lemma 2.2 that
G
2
= I if and only if = = +, as desired.
The next theorem treats the coincidence I = G
0
. A proof similar to that of
Theorem 5.5 can be given, but we chose to give a more elementary argument.
Theorem 5.6. Let ABCD be a circumscriptible quadrilateral that is not a rhombus.
Then I = G
0
if and only if + = + . In this case, F, G
0
= I, G
2
, G
1
are
distinct and collinear, and FG
0
: G
0
G
2
= IG
2
: G
2
G
1
= 6 : 2 : 1.
Proof. Let P, Q be the midpoints of AC,BD, as in Figure 8. Then G
0
is the
midpoint of PQ. Applying Apollonius theorem to medians AG
0
, CG
0
in triangles
APQ, CPQ, and then to medians AQ, CQ in triangles ABD, CBD, we obtain
4
_
AG
0

2
CG
0

2
_
= 2
_
AQ
2
CQ
2
_
= AB
2
+AD
2
CB
2
CD
2
= ( +)
2
+ ( +)
2
( +)
2
( +)
2
= 2( )( + + +) .
244 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
P
Q
G0
A B
C
D


Figure 8. Illustrating the proof of Theorem 5.6.
Also, AI
2
CI
2
=
2

2
= ( )( +). Therefore
AG
0

2
CG
0

2
= AI
2
CI
2
( )( + ) = 0 .
By symmetry, we obtain a similar statement with A, C replaced by B, D. By
Lemma 2.2, we have G
0
= I ()(+) = ()(+) = 0.
Since ABCD is not a rhombus, and cannot both be 0, and therefore
G
0
= I + = +.
We nally remark that if F = I, then the diagonals of ABCD bisect the
vertex angles, and ABCD is a rhombus. This, together with the results in Theo-
rems 3.5, 5.5, and Theorem 5.6, are summarized in the following theorem.
Theorem 5.7. For a circumscriptible quadrilateral that is not a rhombus, consider
the 10 possible coincidences between pairs of its 5 centers F, G
0
, G
1
, G
2
, I. Then
G
1
= G
2
I = G
1
I = G
2
G
1
= G
2
= I
= = + , or = = + (5.1)
I = G
0
+ = + . (5.2)
No other coincidence can occur.
Remark 5.8. Note that quadrilaterals in the family (5.2) are not kites and have no
geometrical symmetry, i.e., have a trivial group of isometries. In fact, it is easy to
see that parallelograms, kites, and isosceles trapezoids are the only quadrilaterals
with nontrivial groups of isometries; see [36].
6. Coincidence of centers of a bicentric quadrilateral
Here again, we take a convex quadrilateral ABCD, and we adhere to the notation
in Section 2.
We start with a useful characterization of bicentric quadrilaterals. For a proof,
see [38] or [21].
Theorem 6.1. If ABCD is circumscriptible, then it is cyclic if and only if = .
Vol. 55 (2009) Centers of Quadrilaterals 245
Theorem 6.2. For a bicentric quadrilateral, I = C if and only if it is a square.
Proof. Let R and r be the circumradius and the inradius, respectively. By Pythago-
ras theorem,
2
= R
2
r
2
. Similarly,
2
=
2
=
2
= R
2
r
2
. Therefore
= = = . Thus the quadrilateral has equal sides and is therefore a square.
Theorem 6.3. For a bicentric quadrilateral that is not a square, I = G
0
if and only
if ABCD is a circumscriptible isosceles trapezoid, i.e., an isosceles trapezoid each
of whose equal legs is one fourth of the perimeter.
Proof. Let ABCD be a bicentric quadrilateral. It follows from Theorem 5.6 that
I = G
0
if and only if + = +. By Theorem 6.1, = . Therefore I = G
0
if and only if and are the same as and in some order. Thus either (, )
equals (, ) or (, ), as desired.
Apart from the pair (I, C), treated already in Theorem 6.2, the remaining
pairs of centers have all been considered in the previous sections, and the results
are summarized in Theorems 4.7 and 5.7, where one nds the ve families dened
by (w = y = x+z), (w = y, AFB = 60

), (w = y, 2y
2
y(x+z)(x
2
+z
2
) = 0),
( = = + ), ( + = + ). The last family, corresponding to I = G
0
, is
treated in Theorem 6.3. We now show that none of the four remaining families
contains bicentric quadrilaterals.
For the rst, use Theorem 5.1 to obtain x + z = y + w, contradicting y =
w = x + z. For the second, use the law of cosines to obtain the contradiction
(x + z)
2
= (2y)
2
= 4(x
2
+ xz + z
2
) > (x + z)
2
. The third one, together with
x+z = 2y yield the contradiction 0 = 2y
2
y(x+z) (x
2
+z
2
) = (x
2
+z
2
). For
the fourth, use = to obtain + = , =
2
. Thus , are the zeros of
the quadratic T
2
T +
2
. This is impossible since the discriminant is negative.
We now summarize these results in the following theorem.
Theorem 6.4. For a bicentric quadrilateral ABCD that is not a square, consider
the 15 possible coincidences between pairs of its 6 centers F, G
0
, G
1
, G
2
, C, I. Then
(I, G
0
) is the only pair of centers that may coincide, and this happens precisely
when ABCD is a circumscriptible trapezoid. This is equivalent, after relabeling,
to w = y = (x +z)/2.
References
[1] S. Abu-Saymeh and M. Hajja, Coincidence of centers for scalene triangles. Forum
Geom. 7 (2007), 137146.
[2] T. Andreescu and B. Enescu, Mathematical Olympiad Treasures. Birkh auser, Boston,
2004.
[3] T. M. Apostol and M. A. Mnatsakanian, Figures circumscribing circles. Amer. Math.
Monthly 111 (2004), 853863.
[4] T. M. Apostol and M. A. Mnatsakanian, Solids circumscribing spheres. Amer. Math.
Monthly 113 (2006), 521540.
246 A. Al-Sharif, M. Hajja and P. T. Krasopoulos Result.Math.
[5] M. Berman, Problem 148. Two Year College Math. J. 10 (1979), 294; Solution, ibid
12 (1981) 6466.
[6] C. A. Bishop, Problem 78.E. Math. Gaz. 79 (1995), 129.
[7] C. J. Bradley, Challenges in Geometry. Oxford Univ. Press, N. Y., 2005.
[8] E. Brassine, Sur quelques proprietes des centres de gravite. J. Mathematiques (1943),
4648.
[9] B. Carter, The i-centroid of an n-simplex. Amer. Math. Monthly 68 (1961), 914917.
[10] P. Couderc and A. Balliccioni, Premier Livre du Tetra`edre. Gauthier-Villars, Paris,
1935.
[11] R. Courant and H. Robbins, What is Mathematics?. Oxford U. Press, N. Y., 1978.
[12] H. D orrie, 100 Great Problems of Elementary Mathematics. Dover, N. Y., 1965.
[13] C. V. Durrell and A. Robson, Advanced Trigonometry. G. Bell and Sons, Ltd., Lon-
don, 1961.
[14] A. L. Edmonds, M. Hajja, and H. Martini, Coincidences of simplex centers and re-
lated facial structures. Beitr. Algebra Geom. 46 (2005), 491512.
[15] A. L. Edmonds, M. Hajja, and H. Martini, Orthocentric simplices and their centers.
Results Math. 47 (2005), 266295.
[16] Euclid, The Elements. Sir Thomas L. Heath, editor, Dover, N. Y., 1956.
[17] Euclids Elements. aleph0.clarku.edu/
~
djoyce/mathhist/alexandria.html
[18] D. E. Gurarie, Problem 10303. Amer. Math. Monthly 100 (1993), 401; solution ibid
101 (1994), 10191020.
[19] M. Hajja, Triangle centres : some questions in euclidean geometry. Internat. J. Math.
Ed. Sci. Tech. 32 (2001), 2137.
[20] M. Hajja, Coincidences of centers of edge-incentric, or balloon, simplices. Results
Math. 49 (2006), 237263.
[21] M. Hajja, A condition that a circumscriptible quadrilateral be cyclic. Forum Geom.
8 (2008), 103106.
[22] M. Hajja and P. Walker, Equifacial Tetrahedra. Internat. J. Math. Ed. Sci. Tech. 32
(2001), 501508.
[23] M. Hajja and P. Walker, Equifaciality of tetrahedra whose incenter and Fermat
Torricelli center coincide. J. Geometry Graphics 9 (2005), 3741.
[24] R. Honsberger, Episodes in the Nineteenth and Twentieth Century Euclidean Geom-
etry. New Mathematical Library No. 37, MAA, Washington, DC, 1995.
[25] R. Honsberger, In Polyas Footsteps, Miscellaneous Problems and Essays. Dolciani
Mathematical Expositions No. 19, MAA, Washington, DC, 1997.
[26] R. A. Johnson, Advanced Euclidean Geometry, Dover, N. Y., 1960.
[27] L. S. Johnston, Problem E 206. Amer. Math. Monthly 43 (1936), 241; solution, ibid
43 (1936) 640641.
[28] C. Kimberling, Triangle Centers and Central Triangles. Congr. Numer. 129 (1998),
1285.
[29] J. D. E. Konhauser and S. Wagon, Problem 10662. Amer. Math. Monthly 105 (1998),
464; solution, ibid 106 (1999), 871.
Vol. 55 (2009) Centers of Quadrilaterals 247
[30] Y. S. Kupitz and H. Martini, Geometric aspects of the generalized FermatTorricelli
problem. Intuitive Geometry, Bolyai Society Math. Studies, Vol. 6, pp. 55127, 1997.
[31] A. Liu, Hungarian Problem Book III. New Mathematical Library No. 42, MAA,
Washington, DC, 2001.
[32] M. Longuet-Higgins, A fourfold point of concurrence lying on the Euler line of a
triangle. Math. Intelligencer 22 (2000), 5459.
[33] Z. A. Melzak, Invitation to Geometry. John Wiley, N. Y., 1983.
[34] A. Myakishev, On two remarkable lines related to a quadrilateral. Forum Geom. 6
(2006), 289295.
[35] I. Pinelis, Cyclic polygons with given edge lengths: Existence and uniqueness.
J. Geom. 82 (2005), 156171.
[36] S. A. Robertson, Classifying triangles and quadrilaterals. Math. Gaz. 61 (1977), 38
49.
[37] G. C. Shephard, Centroids of polygons and polyhedra. Math. Gaz. 74 (1990), 4243.
[38] A. Sinefakopoulos, Problem 10804. Amer. Math. Monthly 107 (2000), 462; solution,
ibid 108 (2001), 378.
[39] E. H. Smart, C. G. of a quadrilateral lamina. Math. Gaz. 10 (1920), 144.
[40] G. A. Venema, Foundations of Geometry. Prentice Hall, New Jersey, 2006.
[41] P. Yiu, Notes on Euclidean Geometry (1998). http://www.math.fau.edu/Yiu/
Geometry.html
Abdullah Al-Sharif
Physics Department
Yarmouk University
Irbid
Jordan
e-mail: alsharif@yu.edu.jo
Mowaaq Hajja
Mathematics Department
Yarmouk University
Irbid
Jordan
e-mail: mowhajja@yahoo.com
mhajja@yu.edu.jo
Panagiotis T. Krasopoulos
Skra 59
176 73 Kallithea
Athens
Greece
e-mail: pankras@in.gr
Received: July 6, 2008.
Revised: January 31, 2009.
Accepted: February 2, 2009.

Das könnte Ihnen auch gefallen