Circumscribed quadrilaterals revisited / Darij Grinberg

[corrected and amended version, 7th of July 2021]

The aim of this note is to prove some new properties of circumscribed quadrilaterals
and give new proofs to classical ones.1
We start with some trivialities (Fig. 1).

a a

W X

ρ ρ b
O

d B
ρ
ρ b
D
d Y
Z c
c
C
Fig. 1

Let ABCD be a circumscribed quadrilateral, that is, a quadrilateral that has an

incircle. Let this incircle have the center O and the radius and touch its sides AB;
BC; CD; DA at the points X; Y; Z; W; respectively. Then, for obvious reasons, we
have OX ? AB; OY ? BC; OZ ? CD; OW ? DA and OX = OY = OZ = OW = :
Moreover, AW = AX; BX = BY; CY = CZ; DZ = DW; since the two tangents from
a point to a circle are equal in length. We denote
a = AW = AX; b = BX = BY ; c = CY = CZ; d = DZ = DW:
(Thus, we denote by a; b; c; d not, as usual, the sidelengths of the quadrilateral ABCD;
but the segments AW = AX; BX = BY; CY = CZ; DZ = DW:)
1
I am grateful to George Baloglou for correcting a mistake in Theorem 13.

1
 Then, the sidelengths of quadrilateral ABCD are

AB = AX + BX = a + b; BC = BY + CY = b + c;
CD = CZ + DZ = c + d; DA = DW + AW = d + a:

Hence,

AB + CD = (a + b) + (c + d) = (b + c) + (d + a)
= BC + DA (since b + c = BC and d + a = DA) :

Thus we have shown the maybe most famous fact about circumscribed quadrilaterals:

Theorem 1. If ABCD is a circumscribed quadrilateral2 , then AB +CD =

BC + DA:

In words: In a circumscribed quadrilateral, the sums of the lengths of opposite sides

are equal.
2
In the following, we assume in every theorem that ABCD is a circumscribed quadrilateral; and
we use all previously de…ned notations (for instance, O always stands for the center of the incircle of
ABCD).

2
 A

W X

D
Y
Z
C
Fig. 2

Now, let us get serious and turn to the …rst nontrivial result about circumscribed
quadrilaterals (Fig. 2):

Theorem 2. The four lines AC; BD; XZ; Y W concur at one point.3

This theorem is still rather well-known; it is problem 105 in [1] and also appears in
[6], [8] and [10]. Here we give two proofs of this theorem.
3
As already said, we are using all previously introduced notations. Thus, ABCD is a circumscribed
quadrilateral, and X; Y; Z and W are the points at which its incircle touches its sides AB; BC; CD
and DA.

3
 A

B
P

D
Y

C
Fig. 3

First proof of Theorem 2. (See Fig. 3.) Let P be the point of intersection of the
lines AC and Y W:
The lines BC and DA touch the incircle of the quadrilateral ABCD at the points Y
and W: Hence, by the tangent-chordal angle theorem, both angles ]CY W and ]DW Y
are equal to the chordal angle of the chord Y W in the incircle of the quadrilateral
ABCD: Thus, ]CY W = ]DW Y: In other words, ]CY P = 180 ]AW P: Thus,
sin ]CY P = sin ]AW P: But after the sine law in triangle AW P; we have AP = AW
sin ]AW P sin ]CY P
; and after the sine law in triangle CY P; we have CP = CY :
sin ]AP W sin ]CP Y
Thus,
sin ]AW P sin ]AW P
AW AW
AP
= sin ]AP W = sin ]AP W = AW = a :
CP sin ]CY P sin ]AW P CY c
CY CY
sin ]CP Y sin ]AP W
Now, let P 0 be the point of intersection of the lines AC and XZ: Then, we similarly
AP 0 a AP a AP AP 0
…nd = : Comparing this with = , we …nd = : This means that
CP 0 c CP c CP CP 0

4
the points P and P 0 divide the segment AC in the same ratio; hence, these points P
and P 0 coincide. Since the point P is the point of intersection of the lines AC and Y W;
and the point P 0 is the point of intersection of the lines AC and XZ; it thus follows
that the lines AC; XZ and Y W concur at one point. Similarly, we can verify that the
lines BD; XZ and Y W concur at one point. Hence, all four lines AC; BD; XZ and
Y W concur at one point, and Theorem 2 is proven.

a a

W X

d B
P

b
D
d Y
Z c
c
C
Fig. 4

This proof of Theorem 2 has a nice consequence (Fig. 4): The point of intersection of
the four lines AC; BD; XZ; Y W must obviously coincide with the point of intersection
P of the lines AC and Y W de…ned in the above proof of Theorem 2. However, we
AP a BP b
have shown that this point P satis…es = : Similarly, = : Thus, we get:
CP c DP d
Theorem 3. If P is the point of intersection of the lines AC; BD; XZ;
AP a BP b
Y W; then = and = :
CP c DP d
Note that this result appeared in [7] and [8].

5
 Second proof of Theorem 2. We will show that the lines AC; BD and XZ concur.
Then, analogously we can show that the lines AC; BD and Y W concur, and thus it
will follow that all four lines AC; BD; XZ and Y W concur, thus proving Theorem 2.

X'

Z
C
Fig. 5

(See Fig. 5.) Now, in order to show that the lines AC; BD and XZ concur, it
appears reasonable to apply the Brianchon theorem in a limiting case. However, one
has to be careful doing this. Here is how one should not proceed:
"Consider the degenerate hexagon AXBCZD (degenerate, since its adjacent sides
AX and XB lie on one line, and its adjacent sides CZ and ZD lie on one line). This
hexagon is obviously circumscribed, since all of its sides AX; XB; BC; CZ; ZD; DA
touch one circle (namely, the incircle of the quadrilateral ABCD). Hence, the main
diagonals AC; XZ and BD of this hexagon concur, and the proof is complete."
The mistake - to be more precise, the gap - in this argumentation becomes clear if
one applies it to the hexagon AX 0 BCZD; where X 0 is an arbitrary point on the line
AB: This hexagon, too, appears to be circumscribed, since all of its sides AX 0 ; X 0 B;
BC; CZ; ZD; DA touch one circle (namely, the incircle of the quadrilateral ABCD)

6
- if they are extended to lines (but this should not be a problem, since we are talking
about projective theorems, and thus arrangement shouldn’t matter). Thus, by the
Brianchon theorem, it seems to follow that the lines AC; X 0 Z and BD concur - but
this is nonsense for every point X 0 di¤erent from X:
So where is the mistake? The trick is: A geometrical theorem can be used in
a degenerate case if either its proof still functions in this case, or one can deduce
the degenerate case from the generic case by a limiting argument. Our application
of the Brianchon theorem to the hexagon AX 0 BCZD did not match any of these two
conditions; thus, it was not legitimate. Hence, there is no wonder the resulting assertion
was wrong.
However, one can rescue the above proof of Theorem 2. In order to do this, one
must …nd an argument that shows why the Brianchon theorem can be applied to the
degenerate hexagon AXBCZD; but not to the degenerate hexagon AX 0 BCZD with
X 0 6= X:
In order to …nd such an argument, let’s recall how the Brianchon theorem is derived
from the Pascal theorem using the polar transformation.
The Pascal theorem states: If six points A1 ; B1 ; C1 ; D1 ; E1 ; F1 lie on one circle, then
the points of intersection A1 B1 \ D1 E1 ; B1 C1 \ E1 F1 and C1 D1 \ F1 A1 are collinear;
here, if two "adjacent" points - i. e., for instance, the points A1 and B1 - coincide, then
the line A1 B1 has to be interpreted as the tangent to the circle at the point A1 ; and
not as an arbitrary line through the point A1 :
After the polar transformation, this becomes: If six lines a1 ; b1 ; c1 ; d1 ; e1 ; f1 touch
a circle, then the lines (a1 \ b1 ) (d1 \ e1 ) ; (b1 \ c1 ) (e1 \ f1 ) and (c1 \ d1 ) (f1 \ a1 )
are concurrent4 ; here, if two "adjacent" lines - i. e., for instance, the lines a1 and b1
- coincide, then the point of intersection a1 \ b1 has to be interpreted as the point of
tangency of the line a1 with the circle, and not as an arbitrary point on the line a1 :
In other words: The hexagon formed by the lines a1 ; b1 ; c1 ; d1 ; e1 ; f1 may be
degenerated, but if two adjacent sides lie on one line, then the vertex where these sides
meet must be the point of tangency of this line with the circle, and not just an arbitrary
point on this line.
This is ful…lled for the degenerate hexagon AXBCZD 5 , but not for the degenerate
hexagon AX 0 BCZD with X 0 6= X: Thus, the above argumentation for the hexagon
AXBCZD is correct - thus Theorem 2 is proven -, but the same argumentation for
the hexagon AX 0 BCZD is wrong.

Now, we head over to a less classical result, one noted by myself in 2003 (Fig. 6):

Theorem 4. Let the perpendicular to the line AB at the point A meet the
line BO at a point M: Let the perpendicular to the line AD at the point
A meet the line DO at a point N: Then, M N ? AC:
4
Hereby, we use the abbreviation G H for the line joining two points G and H:
5
The adjacent sides AX and XB of this hexagon lie on one line - and the vertex where they meet,
namely the vertex X; is indeed the point of tangency of this line with the circle. The same holds for
the adjacent sides CZ and ZD:

7
 A

M N

B
D

C
Fig. 6

In [4], this theorem appears as Theorem 1 and receives two proofs. Here is a di¤erent
proof of Theorem 4:
(See Fig. 7.) Let Lb and Ld be the orthogonal projections of the points B and D on
the line AC: Then, the lines BLb and DLd ; both being perpendicular to AC; must be
BLb BP
parallel to each other, and thus Thales yields = : But according to Theorem
DLd DP
BP b BLb b BLb DLd
3, we have = : Thus = ; or, equivalently, = :
DP d DLd d b d

8
 A

Lb
D B
P

Ld

C
Fig. 7

(See Fig. 8.) Let R be the orthogonal projection of the point M on the line AC:
Then, ]ARM = 90 : Compared with ]BLb A = 90 ; this yields ]ARM = ]BLb A: On
the other hand, ]M AB = 90 ; so that ]M AR = ]M AB ]Lb AB = 90 ]Lb AB:
But in the right-angled triangle ALb B; we have ]ABLb = 90 ]Lb AB: Comparing
these, we …nd ]M AR = ]ABLb : From ]ARM = ]BLb A and ]M AR = ]ABLb ; we
AR AM
see that the triangles ARM and BLb A are similar; thus, = :
BLb AB
On the other hand, the point M lies on the line BO; and from AM ? AB and
AM OX
OX ? AB it follows that AM k OX: Hence, by Thales, = : Thus, we obtain
AB BX
AR AM OX BLb
= = = ; so that AR = BLb = :
BLb AB BX b b b
Similarly, we can denote by R0 the orthogonal projection of the point N on the line
DLd BLb DLd
AC; and show that AR0 = : Since = ; we thus get AR = AR0 : Since
d b d
the points R and R0 both lie on the segment AC; this yields that these points R and
R0 coincide. Now, since the point R is the orthogonal projection of the point M on
the line AC; we have M R ? AC; so that the point M lies on the perpendicular to the
line AC at the point R: Similarly, the point N lies on the perpendicular to the line
AC at the point R0 : But since R = R0 ; these two perpendiculars coincide, and thus the

9
points M and N lie on one and the same perpendicular to the line AC: This means
M N ? AC; and Theorem 4 is proven.

M
R X

O
Lb
B
D

C
Fig. 8

In [2], Jean-Pierre Ehrmann showed an alternate approach to Theorem 4 with the

help of hyperbola properties. A corollary of this approach is the following fact:

Theorem 5. Denote the distances from the points B and D to the line
m n
M N by m and n; respectively. Then, = :
AB AD

10
 A

M
N

n O m

B
D

C
Fig. 9

Here is an elementary proof of Theorem 5. First, we focus on the points X; Y; Z;

W: We will use directed segments; in the following, the directed distance between two
points P1 and P2 will be denoted by P1 P2 (as opposed to the non-directed distance,
which we will continue to write as P1 P2 ). Also, we direct the lines AB; BC; CD; DA
in such a way that the directed segments AB; BC; CD; DA are positive (and thus the
segments BA; CB; DC; AD are negative). Then,

a = AW = AX; b = BX = BY ; c = CY = CZ; d = DZ = DW

becomes

a = W A = AX; b = XB = BY ; c = Y C = CZ; d = ZD = DW :

AT a
(See Fig. 10.) Now, let T be the point on the line AC satisfying = : Then,
TC c
TC c CT c
= ; what rewrites as = : Hence,
AT a TA a
AX BY CT a b c
= = 1:
XB Y C TA b c a

11
By the Menelaos theorem, applied to the triangle ABC and the points X; Y; T on its
sides AB; BC; CA; this yields that the points X; Y; T are collinear. In other words,
the point T lies on the line XY: As the de…nition of the point T is symmetric in B and
D; we can similarly show that this point T lies on the line ZW:
Note that we have thus shown an interesting side-result: Our point T lies on the lines
AT a
AC; XY and ZW and divides the segment AC in the ratio = : Comparing this
TC c
AP a AP a
with6 = (this is just the equation = from Theorem 3, rewritten using
PC c CP c
AT AP
directed segments), we see that = ; so that the point T is the harmonic
TC PC
conjugate of the point P with respect to the segment AC: Thus, we have shown:

Theorem 6. The lines AC; XY; ZW concur at one point T: This point
AT a
T divides the segment AC in the ratio = and is the harmonic
TC c
conjugate of the point P with respect to the segment AC:
6
Here, P denotes the point of intersection of the four lines AC; BD; XZ and Y W (as in Theorem
3).

12
 A

D
Fig. 10

(See Fig. 11.) Now, let M 0 be the orthogonal projection of the point B on the line
M N: Then, the distance m from the point B to the line M N equals to the segment
BM 0 ; so we have m = BM 0 :
On the other hand, BM 0 ? M N; combined with M N ? AC; yields BM 0 k AC; so
that ]M 0 BA = ]XAT:

13
 A

M M' N

X
O m

Y
D C

Fig. 11

Since ]M M 0 B = 90 and ]M AB = 90 ; the points M 0 and A lie on the circle with

diameter M B: Thus, the quadrilateral AM 0 BM is cyclic, so that ]BM 0 A = 180
]AM B: On the other hand, in the right-angled triangle AM B; we have ]AM B =
90 ]ABM: But since the point M lies on the line BO; i. e. on the angle bisector
of the angle ABC (since the point O is the incenter of the quadrilateral ABCD), we
]ABC
have ]ABM = : Finally, since BX = BY; the triangle XBY is isosceles, so
2
that its base angle ]BXY equals
180 ]XBY ]XBY ]ABC
]BXY = = 90 = 90 :
2 2 2

14
Thus,
]ABC
]BM 0 A = 180 ]AM B = 180 (90 ]ABM ) = 90 + ]ABM = 90 +
2
]ABC
= 180 90 = 180 ]BXY = ]AXT:
2

Since ]M 0 BA = ]XAT and ]BM 0 A = ]AXT; the triangles BM 0 A and AXT

BM 0 AX
are similar. Thus, = : Since m = BM 0 and a = AX; we can rewrite this as
AB TA
m a n a m n
= : Similarly, = : Comparing these, we …nd = ; which proves
AB TA AD TA AB AD
Theorem 5.

In the remainder of the article, we will study some metric identities for the circum-
scribed quadrilateral (Fig. 12).

W A' X

ρ
O
B' B

D
Y

C
Fig. 12

The points X and Y; being the points of tangency of the incircle of the quadrilateral
ABCD with its sides AB and BC; are symmetric to each other with respect to the

15
angle bisector BO of the angle ABC: Hence, the segment XY is perpendicular to the
line BO and is bisected by this line. So the midpoint B 0 of the segment XY lies on
the line BO: Similarly, the midpoint A0 of the segment W X lies on the line AO:
Now, from XY ? BO we see that ]XB 0 O = 90 ; while from OX ? AB we have
]BXO = 90 : Thus, ]XB 0 O = ]BXO: Also, trivially, ]XOB 0 = ]BOX: Thus, the
OB 0 OX
triangles XB 0 O and BXO are similar, so that = ; and thus OB OB 0 =
OX OB
OX 2 = 2 :
OB OA0
Similarly, OA OA0 = 2 : Hence, OB OB 0 = OA OA0 ; so that = 0
:
OA OB
Together with ]BOA = ]A0 OB 0 ; this yields the similarity of triangles BOA and
A0 OB 0 : Consequently,

A0 B 0 OA0 OA0 OA OA0 2

= ; thus A0 B 0 = AB = AB = AB :
AB OB OB OA OB OA OB
Now, the points A0 and B 0 are the midpoints of the sides W X and XY of triangle
2
YW YW
W XY ; thus, A0 B 0 = : Hence, AB = : Consequently,
2 OA OB 2
YW OA OB
AB = 2
:
2
Similar relations must obviously hold for BC; CD and DA: We summarize:

Theorem 7. We have
YW OA OB XZ OB OC
AB = 2
; BC = 2
;
2 2
YW OC OD XZ OD OA
CD = 2
; DA = 2
:
2 2

(See Fig. 13.)

16
 A

W X

ρ ρ
O

B
ρ
ρ
D
Y
Z
C
Fig. 13

These equations can be used for deriving some other formulas. For instance, AB =
Y W OA OB
2
transforms into
2

2 YW 2 2 AB
OA OB = AB : = :
2 YW
Similarly,
2
2 CD
OC OD = :
YW
Thus,
2
2 AB
OA OB YW AB
= = :
OC OD 2 2 CD CD
YW
OB OC BC
Similarly, = : So we have shown:
OD OA DA

17
 Theorem 8. We have
AB OA OB BC OB OC
= ; = :
CD OC OD DA OD OA

Proving these equations was a 10th grade problem in the 4th round of the 14th
DeMO (East German mathematical olympiad) 1974/75. We furthermore have

AB BC AB BC OA OB OB OC
= = (by Theorem 8)
CD DA CD DA OC OD OD OA
OB 2
= ;
OD2
or, equivalently,
OB 2 OD2
= :
AB BC CD DA
OA2 OC 2
Similarly, = : Thus we arrive at the following:
DA AB BC CD
Theorem 9. We have
OB 2 OD2 OA2 OC 2
= ; = :
AB BC CD DA DA AB BC CD

This also appears with proof in [5].

Now we show a harder identity given in the China IMO TST 2003 ([8]):

Theorem 10. We have

p
OA OC + OB OD = AB BC CD DA:

Proof of Theorem 10. (See Fig. 14.) Let X 0 and Z 0 be the antipodes of the
7
points X and Z on the incircle of the quadrilateral ABCD , or, in other words,
the re‡ections of the points X and Z with respect to the center O of this incircle.
Then, the segment XX 0 is a diameter of the incircle of the quadrilateral ABCD; and
thus ]XY X 0 = 90 ; so that Y X 0 ? XY: On the other hand, XY ? BO: Hence,
Y X 0 k BO; so that ]XX 0 Y = ]BOX: Together with ]XY X 0 = ]BXO (since
]XY X 0 = 90 and ]BXO = 90 ) this entails that the triangles XX 0 Y and BOX are
X 0Y OX OX
similar; consequently, 0 = ; so that X 0 Y = X 0 X : Now, X 0 X = 2 OX
X X OB OB
(since the point X 0 is the re‡ection of X in O), and thus

OX 2 OX 2 2 2
X 0 Y = 2 OX = = :
OB OB OB
7
The antipode of a point P on a circle k is de…ned as the point P 0 on the circle k such that the
segment P P 0 is a diameter of k:

18
 A

X
ρ

X'
D
Y

C
Fig. 14

Similarly,

2 2 2 2 2 2
Z 0Y = ; Z 0W = ; X 0W = :
OC OD OA
Finally, X 0 Z 0 = XZ; since the points X 0 and Z 0 are the re‡ections of the points X and
Z in the point O; and re‡ections preserve distances.

19
 A

Z'

W X

X'
D
Y
Z
C
Fig. 15

(See Fig. 15.) Now, the points X 0 ; Y; Z 0 ; W all lie on the incircle of the quadrilateral
ABCD; thus, the quadrilateral X 0 Y Z 0 W is cyclic, so that, after the Ptolemy theorem,

X 0 Y Z 0 W + X 0 W Z 0 Y = X 0 Z 0 Y W:

According to the above formulas, this becomes

2 2 2 2 2 2 2 2
+ = XZ Y W; i. e.
OB OD OA OC
4 1 1
4 + = XZ Y W; i. e.
OB OD OA OC
OA OC + OB OD
4 4 = XZ Y W:
OA OB OC OD
Hence,
XZ Y W OA OB OC OD
OA OC + OB OD = : (1)
4 4

20
But Theorem 7 yields

AB BC CD DA
Y W OA OB XZ OB OC YW OC OD XZ OD OA
= 2 2 2 2
2 2 2 2
2
XZ Y W OA OB OC OD
= ;
4 4

so that
XZ Y W OA OB OC OD p
= AB BC CD DA:
4 4
Hence, (1) becomes
p
OA OC + OB OD = AB BC CD DA;

and Theorem 10 is proven.

In the following, we shall denote by jP1 P2 :::Pn j the (non-directed) area of an arbi-
trary polygon P1 P2 :::Pn :
Furthermore, we denote the interior angles of the quadrilateral ABCD by

= ]DAB; = ]ABC; = ]BCD; = ]CDA:

Now, we are going to show the following:

Theorem 11. We have

(a + c) (b + d)
OA OC = + ; OB OD = ;
+
sin sin
2 2
OA OC a+c (a + b + c + d)
= ; OA OC + OB OD = + :
OB OD b+d sin
2

21
 A

X
ρ c
U
O

D ρ B

Z c
C
Fig. 16

Proof of Theorem 11. (See Fig. 16.) Let U be the point on the ray XB satisfying
U X = c: Comparing this with c = CZ; we get U X = CZ: Furthermore, ]OXU =
90 = ]OZC and OX = OZ: Thus, the triangles OXU and OZC are congruent, so
that OU = OC and ]XOU = ]ZOC:
Since the point O; being the incenter of the quadrilateral ABCD; lies on the angle
]DAB
bisector of its angle DAB; we have ]XAO = = ; in the right-angled triangle
2 2
AXO; we thus obtain ]XOA = 90 ]XAO = 90 : Similarly, ]ZOC = 90 ;
2 2
since ]XOU = ]ZOC; this becomes ]XOU = 90 : Hence, ]AOU = ]XOA +
2
+ +
]XOU = 90 + 90 = 180 ; so that sin ]AOU = sin :
2 2 2 2
From AX = a and U X = c; we conclude that AU = AX + U X = a + c:
Now, the area of a triangle equals half of the product of two of its sides and the
sine of the angle between them; applying this to triangle AOU; we get jAOU j =
1 +
OA OU sin ]AOU ; since OU = OC and sin ]AOU = sin ; this becomes
2 2
1 +
jAOU j = OA OC sin :
2 2

22
 On the other hand, the area of a triangle equals half of the product of a side with
the respective altitude; applied to the triangle AOU (in which OX is the altitude to
1
the side AU ), this yields jAOU j = AU OX; since AU = a + c and OX = ; this
2
1
rewrites as jAOU j = (a + c) :
2
1 + 1
Comparing the equations jAOU j = OA OC sin and jAOU j = (a + c) ;
2 2 2
+
we see that OA OC sin = (a + c) ; and thus
2
(a + c)
OA OC = + :
sin
2
Similarly,
(b + d)
OB OD = :
+
sin
2
Now, by the sum of angles in the quadrilateral ABCD; we have + + + = 360 ; so
+ + + + + 360 + +
that + = = = 180 ; and thus sin = sin :
2 2 2 2 2 2
Hence, the equation

(b + d) (b + d)
OB OD = becomes OB OD = + :
+
sin sin
2 2
Thus, 0 1
B (a + c) C
@ + A
OA OC sin a+c
=0 2 1=
OB OD b+d
B (b + d) C
@ + A
sin
2
and
(a + c) (b + d) (a + b + c + d)
OA OC + OB OD = + + + = + :
sin sin sin
2 2 2
Therefore, Theorem 11 is proven.
Now, Theorem 11 asserts

(a + b + c + d)
OA OC + OB OD = + ;
sin
2
while Theorem 10 states that
p
OA OC + OB OD = AB BC CD DA:

23
Hence,
(a + b + c + d) p
+ = AB BC CD DA:
sin
2
+
Comparing these two equalities, and multiplying by sin , we …nd
2
p +
(a + b + c + d) = AB BC CD DA sin :
2
(See Fig. 13.) Now, the area of a right-angled triangle equals half of the product of
1
its two catets; for the right-angled triangle AW O; this yields jAW Oj = AW OW =
2
1 1
a : Similarly, jAXOj = a ; and thus jAW OXj = jAW Oj + jAXOj =
2 2
1 1
a + a = a : Similarly, jBXOY j = b ; jCY OZj = c and jDZOW j = d :
2 2
Hence,

jABCDj = jAW OXj + jBXOY j + jCY OZj + jDZOW j = a + b +c +d

p +
= (a + b + c + d) = AB BC CD DA sin :
2
Thus, we conclude:

Theorem 12. The area jABCDj of a circumscribed quadrilateral ABCD

equals
p +
jABCDj = AB BC CD DA sin :
2
This is not an unknown formula; however it is usually derived from the generalized
Brahmagupta formula for the area of an arbitrary quadrilateral ([9]), which, in turn, is
proven by a long trigonometric calculation. Instead, we have given a rather long, yet
synthetic proof of Theorem 12.

Next, we are going to prove a result due to A. Zaslavsky, M. Isaev and D. Tsvetov
which was given in the …nal (…fth) round of the Allrussian Mathematical Olympiad
2005 as problem 7 for class 11 ([11]):

Theorem 13. The incenter O of a circumscribed quadrilateral ABCD

coincides with the centroid of the quadrilateral ABCD if and only if either8
the quadrilateral ABCD is a rhombus or OA OC = OB OD: (See Fig.
17.)

Here, the centroid of the quadrilateral ABCD is de…ned as follows:

Let E; F; G; H be the midpoints of the sides AB; BC; CD; DA of the quadri-
lateral ABCD: Then, according to the Varignon theorem, the quadrilateral EF GH
is a parallelogram, so that its two diagonals EG and F H bisect each other. In other
8
The words “either/or”are being used in a non-exclusive meaning here (i.e., the statement “either
A or B” allows for the possibility that both A and B hold).

24
words, the segments EG and F H have a common midpoint. This midpoint is called
the centroid of the quadrilateral ABCD:

F
D
G
C
Fig. 17

Now, let’s prove Theorem 13. In order to do this, we have to verify two assertions:
Assertion 1: If the point O is the centroid of the quadrilateral ABCD; then either
the quadrilateral ABCD is a rhombus or OA OC = OB OD.
Assertion 2: If either the quadrilateral ABCD is a rhombus or OA OC = OB OD,
then the point O is the centroid of the quadrilateral ABCD:
Before we establish any of these assertions, we start with a few observations holding
for every circumscribed quadrilateral ABCD (Fig. 18):
AB a+b
Since the point E is the midpoint of the segment AB; we have AE = = ;
2 2

25
and thus
a+b a+b
EX = jAX AEj = a since AX = a and AE =
2 2
a b ja bj
= = :
2 2

jc dj
Similarly, GZ = :
2
Also, note that the triangles EOX and GOZ are right-angled at their vertices X
and Z; since ]OXE = 90 and ]OZG = 90 :

B
D

G
Z
C
Fig. 18

Now, we are going to establish Assertions 1 and 2.

Proof of Assertion 1. We distinguish between two cases:
Case 1: We have a + c 6= b + d.
Case 2: We have a + c = b + d.

26
 Let us …rst consider Case 1. The point O is the centroid of the quadrilateral
ABCD; that is, the midpoint of the segment EG: Thus, OE = OG: Also, OX = OZ:
Hence, the two right-angled triangles EOX and GOZ have the hypotenuse and one
catet in common; thus, they are congruent, and we conclude that EX = GZ: Since
ja bj jc dj
EX = and GZ = ; this yields ja bj = jc dj : Thus, either a b = c d;
2 2
or a b = d c: Now, a b = d c would lead to a + c = b + d; what is impossible
since we have a + c 6= b + d (because we are in Case 1). Hence, it remains only the
possibility a b = c d; that is, a + d = b + c: Similarly to a b = c d; we can
prove that a d = c b; and thus 2a = (a + d) + (a d) = (b + c) + (c b) = 2c: In
other words, a = c: Similarly, b = d: Hence, opposite sides of the quadrilateral ABCD
are equal; this means that the quadrilateral ABCD is a parallelogram, and since it is
circumscribed, it must be a rhombus (in fact, among all parallelograms, only rhombi
are circumscribed). Thus, we have shown that the quadrilateral ABCD is a rhombus
in Case 1.
OA OC a+c
Now, let us consider Case 2. In this case, a+c = b+d. As we have =
OB OD b+d
from Theorem 11, this yields OA OC = OB OD. Thus, OA OC = OB OD holds
in Case 2.
Hence, we have shown that the quadrilateral ABCD is a rhombus in Case 1, and
that OA OC = OB OD in Case 2. Since these cases cover all possibilities, we conclude
that either the quadrilateral ABCD is a rhombus or OA OC = OB OD. Assertion
1 is proven.
Proof of Assertion 2. Assume that either the quadrilateral ABCD is a rhombus or
OA OC = OB OD. We can WLOG assume that OA OC = OB OD (because the
case when the quadrilateral ABCD is a rhombus is trivial for symmetry reasons).
OA OC a+c
From Theorem 11, we have = ; so that OA OC = OB OD
OB OD b+d
ja bj
immediately yields a + c = b + d: Hence, a b = d c; and thus EX = =
2
jd cj jc dj
= = GZ: Furthermore, OX = OZ: Thus, the two right-angled triangles
2 2
EOX and GOZ have the same catets; hence, they are congruent, and it follows that
OE = OG: So the point O lies on the perpendicular bisector of the segment EG:
Similarly, the point O lies on the perpendicular bisector of the segment F H:
Since the circumscribed quadrilateral ABCD is convex, and E; F; G; H are the
midpoints of its sides, the lines EG and F H cannot be parallel. Thus, the perpendicular
bisectors of the segments EG and F H are not parallel as well; therefore, they have
one and only one common point. This common point is obviously the centroid of the
quadrilateral ABCD (since this centroid is the common midpoint of the segments EG
and F H and thus lies on their perpendicular bisectors).
But as we have shown that the point O lies on the perpendicular bisectors of the
segments EG and F H; the point O must be this common point. Hence, the point
O is the centroid of the quadrilateral ABCD: Assertion 2 is shown, and the proof of
Theorem 13 is complete.

Now we return to the case of an arbitrary circumscribed quadrilateral ABCD: We

prove an identity formulated by Pengshi in [12]:

27
 Theorem 14. The radius of the incircle of the circumscribed quadrilat-
eral ABCD satis…es

2 bcd + cda + dab + abc

= :
a+b+c+d

Our proof of this theorem will only slightly di¤er from Anipoh’s in [12]; the key is
the following lemma:

Theorem 15. Let x; y; z; w be four angles such that x + y + z + w = 180 :

Then,

tan x + tan y + tan z + tan w

= tan y tan z tan w + tan z tan w tan x + tan w tan x tan y + tan x tan y tan z:

Proof of Theorem 15. From x + y + z + w = 180 it follows that x + y =

180 (z + w) ; so that tan (x + y) = tan (180 (z + w)) = tan (z + w) and thus
tan (x + y) + tan (z + w) = 0: But the addition formulas for the tan function yield
tan x + tan y tan z + tan w
tan (x + y) = and tan (z + w) = ; hence, tan (x + y) +
1 tan x tan y 1 tan z tan w
tan x + tan y tan z + tan w
tan (z + w) = 0 becomes + = 0: Multiplication by
1 tan x tan y 1 tan z tan w
(1 tan x tan y) (1 tan z tan w) yields

(tan x + tan y) (1 tan z tan w) + (tan z + tan w) (1 tan x tan y) = 0;

thus

(tan x + tan y tan z tan w tan x tan y tan z tan w)

+ (tan z + tan w tan x tan y tan z tan w tan x tan y) = 0;

thus

tan x + tan y + tan z + tan w

= tan y tan z tan w + tan z tan w tan x + tan w tan x tan y + tan x tan y tan z:

This proves Theorem 15.

Now we come to the proof of Theorem 14: With the notations ; ; ; for the
angles of the quadrilateral ABCD; we have

+ + + = ]DAB + ]ABC + ]BCD + ]CDA = 360

(by the sum of angles in the quadrilateral ABCD). Now set x = ;y= ;z = ;
2 2 2
w = : Then,
2
+ + + 360
x+y+z+w = + + + = = = 180 :
2 2 2 2 2 2

28
Thus, Theorem 15 yields

tan x + tan y + tan z + tan w

= tan y tan z tan w + tan z tan w tan x + tan w tan x tan y + tan x tan y tan z:

(See Fig. 16.) During the proof of Theorem 11, we have shown that ]XAO = :
2
Since OX ? AB; the triangle AXO is right-angled at X: Hence, OX = AX tan ]XAO;
so that = a tan x (since OX = ; AX = a and ]XAO = = x). Thus, tan x = ;
2 a
similarly, tan y = ; tan z = ; and tan w = : Hence, the equality
b c d
tan x + tan y + tan z + tan w
= tan y tan z tan w + tan z tan w tan x + tan w tan x tan y + tan x tan y tan z

(which was just proved) becomes

+ + + = + + + :
a b c d b c d c d a d a b a b c
Multiplication by abcd yields
3 3 3 3
bcd + cda + dab + abc = a+ b+ c+ d:

In other words,
3
(bcd + cda + dab + abc) = (a + b + c + d) ; so that
2 bcd + cda + dab + abc
= ;
a+b+c+d
which proves Theorem 14.

We now introduce another notation: If P is a point, and g is a line, then we denote

by dist (P ; g) the (undirected) distance from the point P to the line g: We will often
use the following fact:

Area-distance relation: For any three points U; V; W we have

1
jU V W j = V W dist (U ; V W ) : (2)
2

This fact is just a restatement of the fact that the area of a triangle equals
1
sidelength corresponding altitude
2
(because in triangle U V W; the altitude from U to V W is dist (U ; V W )).

From now on, we let P be the point of intersection of the four lines AC; BD; XZ
and Y W (as in Theorem 3).
Now, we record an easy corollary of Theorem 3 (Fig. 4):

29
 Theorem 16. We have
jAP Bj jBP Cj jCP Dj jDP Aj
= = = : (3)
ab bc cd da

1
Proof of Theorem 16. By the area-distance relation, jBAP j = AP dist (B; AP )
2
1
and jBCP j = CP dist (B; CP ) ; so that
2
1
jAP Bj jBAP j AP dist (B; AP ) AP dist (B; AP )
= = 2 = : (4)
jBP Cj jBCP j 1 CP dist (B; CP )
CP dist (B; CP )
2
dist (B; AP )
Now, = 1 (since dist (B; AP ) = dist (B; CP ) ; because AP and CP are
dist (B; CP )
AP a jAP Bj a
the same line), and = by Theorem 3. Hence, (4) simpli…es to = 1=
CP c jBP Cj c
a ab jAP Bj jBP Cj jBP Cj jCP Dj jCP Dj jDP Aj
= ; so that = : Similarly, = and = :
c bc ab bc bc cd cd da
This proves Theorem 16.

Now we shall show a result by A. Zaslavsky from [13] (see also [14]) (Fig. 19):

30
 A

B
P

C
Fig. 19

Theorem 17. We have

1 1 1 1
+ = + :
dist (P ; AB) dist (P ; CD) dist (P ; BC) dist (P ; DA)

Proof of Theorem 17. Due to the equation (3), we can de…ne a number

jAP Bj jBP Cj jCP Dj jDP Aj

= = = = :
ab bc cd da
Then, jAP Bj = ab:
1
By the area-distance relation, jP ABj = AB dist (P ; AB) ; so that
2
2 jP ABj 2 jAP Bj 2 ab
dist (P ; AB) = = = (as jAP Bj = ab and AB = a + b) ;
AB AB a+b

31
and thus
1 2 ab a+b 1 a+b 1 1 1
=1 = = = + :
dist (P ; AB) a+b 2 ab 2 ab 2 a b

1 1 1 1
Similarly, = + ; so that
dist (P ; CD) 2 c d

1 1 1 1 1 1 1 1 1 1 1 1 1
+ = + + + = + + + :
dist (P ; AB) dist (P ; CD) 2 a b 2 c d 2 a b c d

Similarly,
1 1 1 1 1 1 1
+ = + + + :
dist (P ; BC) dist (P ; DA) 2 a b c d
Thus,
1 1 1 1
+ = + ;
dist (P ; AB) dist (P ; CD) dist (P ; BC) dist (P ; DA)
and Theorem 17 is proven.

Next comes a result whose part a) appeared in [15] (with a di¤erent proof) (Fig.
20):

32
 A

a a

W X

b
O
d HZ B
HX P
b
HY
D
HW
d Y
Z c
c
C
Fig. 20

Theorem 18. Let HX ; HY ; HZ ; HW be the orthocenters of triangles AOB;

BOC; COD; DOA:
a) The points P; HX ; HY ; HZ ; HW are collinear.
b) Using directed segments, we have

P HX P HY P HZ P HW
= = = :
ab bc cd da

33
 A

HZ B

b
BY HY

D Y
d
Z C
Fig. 21

Proof of Theorem 18. (See Fig. 21.) Let BY be the foot of the altitude of triangle
BOC issuing from B: Then, the lines BBY and OY are two altitudes of triangle BOC
(for BBY ; this is clear, and for OY it follows from OY ? BC), and thus intersect at
the orthocenter HY of this triangle. Hence, ]BY HY = 90 and

]Y BHY = ]CBBY = 90 ]BCBY (in the right-angled triangle BBY C)

= 90 ]BCO:

Thus we have shown that ]BY HY = 90 and ]Y BHY = 90 ]BCO: Similarly,

]DZHZ = 90 and ]ZDHZ = 90 ]DCO:
The point O; being the incenter of the quadrilateral ABCD; lies on the angle
bisector of the angle BCD: Thus, ]BCO = ]DCO:
From ]BY HY = 90 = ]DZHZ and ]Y BHY = 90 ]BCO = 90 ]DCO =
BHY
]ZDHZ ; it follows that triangles BY HY and DZHZ are similar. Therefore, =
DHZ
BY BHY b
: Since BY = b and DZ = d; this becomes = :
DZ DHZ d

34
 The line BHY is the line BBY ; thus, BBY ? CO yields BHY ? CO: Similarly,
DHZ ? CO: Consequently, BHY k DHZ :

HZ B
P0
b
HY

D Y
d
Z C
Fig. 22

(See Fig. 22.) Now, denote by P0 the point of intersection of the lines HY HZ and
BP0 BHY BHY b
BD: Since BHY k DHZ ; the Thales theorem yields = : Since = ;
DP0 DHZ DHZ d
BP0 b BP b BP0 BP
this becomes = : But Theorem 3 asserts = : Thus, = : Hence,
DP0 d DP d DP0 DP
the points P0 and P divide the segment BD in the same ratio (both internally, as
one can see by arrangement considerations9 ). Hence, these points P0 and P must
coincide. Thus, P0 2 HY HZ yields P 2 HY HZ : Hence, the lines P HY and P HZ
coincide. Similarly, the lines P HZ and P HW coincide, and the lines P HW and P HX
coincide. Thus, all four lines P HX ; P HY ; P HZ ; P HW coincide, i. e., the points P;
HX ; HY ; HZ ; HW are collinear. Theorem 18 a) is proven.
P0 HY BHY
Because of BHY k DHZ ; the Thales theorem implies = : As we saw
P0 HZ DHZ
9
One could also avoid arrangement considerations by working consequently with directed segments,
but this would require more theory.

35
 P HY BHY BHY b
above, P0 = P; so this becomes = : Together with = ; this yields
P HZ DHZ DHZ d
P HY b P HY b
= : With directed segments, this transforms into = (as arrange-
P HZ d P HZ d
P HY
ment considerations show that the directed ratio is negative). Thus, d P HY =
P HZ
P HY P HZ P HY P HZ
b P HZ ; so that = : Dividing by c yields = : Similarly,
b d bc cd
P HW P HZ P HW P HX P HX P HY P HZ P HW
= and = : Thus, = = = ;
da cd da ab ab bc cd da
and Theorem 18 b) is proven. This completes the proof of Theorem 18.

Now, we come to some properties of the incircles of triangles AP B; BP C; CP D

and DP A:

OW
OX

P B

OY
D
OZ

C
Fig. 23

36
 (See Fig. 23.) Let OX ; OY ; OZ and OW be the incenters10 of triangles AP B; BP C;
CP D and DP A: Let X ; Y ; Z and W be the inradii11 of triangles AP B; BP C; CP D
and DP A:
Since OY is the incenter of triangle BP C, the line P OY is the internal angle bisector
of angle BP C; thus the external angle bisector of angle AP B:
Since OX is the incenter of triangle AP B, the line P OX is the internal angle bisector
of angle AP B.
Since the internal and external angle bisectors of an angle are always mutually
orthogonal, we thus conclude that P OX ? P OY . Hence, ]OX P OY = 90 : Similarly,
]OY P OZ = 90 ; ]OZ P OW = 90 and ]OW P OX = 90 : Because of ]OX P OZ =
]OX P OY + ]OY P OZ = 90 + 90 = 180 , the points OX ; P and OZ lie on one line.
Furthermore, the points OX and OZ lie on di¤erent sides of the point P (since OX ,
being the incenter of triangle AP B, lies inside the angle AP B, while OZ , being the
incenter of triangle CP D, lies inside the angle CP D; but the angles AP B and CP D
are opposite angles). Hence, the point P lies on the segment OX OZ . Similarly, the
point P lies on the segment OY OW . These two segments OX OZ and OY OW thus meet
at P . They furthermore meet at a right angle (since P OX ? P OY ).
Now, we state two rather surprising results:
10
The incenter of a triangle means the center of its incircle.
11
The inradius of a triangle means the radius of its incircle.

37
 A

ρW ρX
OW
OX

P B

OY
D ρY
OZ
ρZ

C
Fig. 24

1 1 1 1
Theorem 19. (See Fig. 24.) We have + = + :
X Z Y W

38
 A

OW
OX

B
P

D OY
OZ

C
Fig. 25

Theorem 20. (See Fig. 25.) The points OX ; OY ; OZ and OW lie on one
circle.

Theorem 19 comes from [17], while Theorem 20 comes from [18]. In order to prove
both theorems, we need a lemma from triangle geometry:

Lemma 21. (See Fig. 26.) Let ABC be a triangle12 . Let be the inradius
of triangle ABC. Let jABCj be the area of triangle ABC:
(a) We have
2 jABCj
= :
BC + CA + AB
12
In this lemma (and its proof), we are working with an “empty slate”; i.e., we forget all notations
that we have previously introduced. Thus, in particular, no longer means the radius of the incircle
of a quadrilateral ABCD.

39
 (b) Let I be the incenter of triangle ABC. Then,

AI = (5)
sin ]IAC
and
AB + AC BC = 2 cot ]IAC: (6)

ρ
C

A
Fig. 26

Proof of Lemma 21. (See Fig. 27.) Let I be the incenter of triangle ABC. Let X,
Y and Z be the points at which the incircle of triangle ABC touches its sides BC; CA
and AB. 13 Then, clearly IX ? BC; IY ? CA; IZ ? AB and IX = IY = IZ = .
Furthermore, we have AY = AZ; BZ = BX and CX = CY; since the two tangents
from a point to a circle are equal in length.
13
Of course, these points X; Y and Z have nothing to do with the points X; Y and Z that were
introduced at the beginning of this article.

40
 B

ρ
C

A Y

Fig. 27

Since IY is the perpendicular from I onto the line CA, we have14 dist (I; CA) =
IY = . Thus, the area-distance relation (2) (applied to the points I, C and A instead
of U , V and W ) yields
1 1
jICAj = CA dist (I; CA) = CA
2 2
1 1
(since dist (I; CA) = ). Similarly, jIBCj = BC and jIABj = AB . Since
2 2
the point I lies inside of triangle ABC, we now have

jABCj = jIBCj + jICAj + jIABj

1 1 1
= BC + CA + AB
2 2 2
1
= (BC + CA + AB) :
2
Solving this for , we …nd
2 jABCj
= :
BC + CA + AB
This proves Lemma 21 (a).
On to part (b). Triangle AY I is right-angled at Y (since IY ? CA). Hence,
IY = AI sin ]IAY , so that
IY
AI = = (since IY = and ]IAY = ]IAC) :
sin ]IAY sin ]IAC
This proves (5).
14
Again, we are using the notation dist (P ; g) for the distance from a point P to a line g.

41
 In the right-angled triangle AY I, we also have AY = IY cot ]IAY = cot ]IAC
(since IY = and ]IAY = ]IAC).
Furthermore, BC = BX + CX = BZ + CY (since BX = BZ and CX = CY ) and
AB = AZ + BZ and AC = AY + CY . These three equalities lead to

AB + AC BC = (AZ + BZ) + (AY + CY ) (BZ + CY )

= AZ + AY = AY + AY (since AZ = AY )
= 2 AY = 2 cot ]IAC (since AY = cot ]IAC) :

This proves (6), and thus completes the proof of Lemma 21 (b).
Now, we return to our circumscribed quadrilateral ABCD that we have been study-
ing (before Lemma 21). In particular, we shall again use the notations introduced
throughout this article (before Lemma 21). We shall now prove Theorem 19 and The-
orem 20:
Proof of Theorem 19. Because of (3), we can de…ne a number

jAP Bj jBP Cj jCP Dj jDP Aj

= = = = :
ab bc cd da
Hence, jAP Bj = ab: Thus, ab = jAP Bj =
6 0 (since P does not lie on the line AB),
so that 6= 0. Therefore, 2 6= 0.
AP a BP b
Theorem 3 yields = and = .
CP c DP d
AP a AP CP
From = , we obtain AP c = CP a, so that = . Hence, we can
CP c a c
de…ne a number
AP CP
= = :
a c
Thus, AP = a and CP = c.
BP b BP DP
From = , we obtain BP d = DP b, so that = . Hence, we can
DP d b d
de…ne a number
BP DP
= = :
b d
Thus, BP = b and DP = d.
Applying Lemma 21 (a) to the triangle AP B and its inradius X (instead of the
triangle ABC and its inradius ), we …nd

2 jAP Bj 2 ab
X = =
P B + BA + AP BP + AB + AP
(since jAP Bj = ab and P B = BP and BA = AB). Consequently,

2 2 BP + AB + AP b + (a + b) + a
= = =
X 2 ab ab ab
BP + AB + AP
(since BP = b and AB = a + b and AP = a)
+1 +1
= + (by simple computation) :
a b

42
Similarly,
2 +1 +1
= + :
Z c d
Adding these two equalities, we …nd
2 2 +1 +1 +1 +1
+ = + + + :
X Z a b c d

Similarly,
2 2 +1 +1 +1 +1
+ = + + +
Y W b c d a
+1 +1 +1 +1
= + + + :
a b c d
Comparing the last two equalities, we …nd
2 2 2 2
+ = + :
X Z Y W

Dividing this equality by 2 (this is allowed, since 2 6= 0), we obtain

1 1 1 1
+ = + :
X Z Y W

This proves Theorem 19.

Proof of Theorem 20. (See Fig. 25.) We know that the point P lies on the segment
OX OZ . Hence, ]OZ P D = ]OX P B (as opposite angles). Furthermore, ]OX P B +
]OY P B = ]OX P OY = 90 (since P OX ? P OY ), so that ]OY P B = 90 ]OX P B.
However, the line P OY is the internal angle bisector of angle BP C (since OY is the
incircle of triangle BP C); thus, we have ]OY P C = ]OY P B = 90 ]OX P B. Hence,

cot ]OY P C = cot (90 ]OX P B) = tan ]OX P B

and
sin ]OY P C = sin (90 ]OX P B) = cos ]OX P B:
The triangle AP B has incenter OX and inradius X . In other words, the triangle
P AB has incenter OX and inradius X . Hence, we can apply Lemma 21 (b) to the
triangle P AB, its incenter OX and its inradius X (instead of the triangle ABC, its
incenter I and its inradius ). Thus, we obtain
X
P OX = (7)
sin ]OX P B
and
PA + PB AB = 2 X cot ]OX P B: (8)
Similarly to (8), we obtain

PC + PD CD = 2 Z cot ]OZ P D

43
(by applying Lemma 21 (b) to the triangle P CD, its incenter OZ and its inradius Z ).
Adding this equality to (8), we …nd

(P A + P B AB) + (P C + P D CD)
= 2 X cot ]OX P B + 2 Z cot ]OZ P D
= 2 X cot ]OX P B + 2 Z cot ]OX P B (since ]OZ P D = ]OX P B)
= 2 ( X + Z ) cot ]OX P B: (9)

Similarly (or by cyclic permutation of the vertices A, B, C and D), we …nd

(P B + P C BC) + (P D + P A DA)
= 2 ( Y + W ) cot ]OY P C
= 2 ( Y + W ) tan ]OX P B (10)

(since cot ]OY P C = tan ]OX P B).

From (9), we obtain

2 ( X + Z ) cot ]OX P B
= (P A + P B AB) + (P C + P D CD)
= P A + P B + P C + P D (AB + CD)
= P A + P B + P C + P D (BC + DA) (by Theorem 1)
= (P B + P C BC) + (P D + P A DA)
= 2 ( Y + W ) tan ]OX P B (by (10)) :

Hence, ( X + Z ) cot ]OX P B =( Y + W ) tan ]OX P B, so that

sin ]OX P B
X + Z tan ]OX P B cos ]OX P B
= =
Y + W cot ]OX P B cos ]OX P B
sin ]OX P B
(sin ]OX P B)2
= : (11)
(cos ]OX P B)2

Now, we recall the equality (7); this equality says that

X
P OX = :
sin ]OX P B
Similarly, we …nd
Z
P OZ =
sin ]OZ P D
(by applying Lemma 21 (b) to the triangle P CD, its incenter OZ and its inradius Z ).
Since ]OZ P D = ]OX P B, we can rewrite this as
Z
P OZ = : (12)
sin ]OX P B

44
Multiplying the equalities (7) and (12), we …nd
X Z
P OX P OZ =
sin ]OX P B sin ]OX P B
X Z
= : (13)
(sin ]OX P B)2
Similarly (or by cyclic permutation of the vertices A, B, C and D), we …nd
Y W
P OY P O W = :
(sin ]OY P C)2
Since sin ]OY P C = cos ]OX P B, we can rewrite this as
Y W
P OY P OW = :
(cos ]OX P B)2
Dividing this equality by the equality (13), we obtain
P OY P OW
P OX P O Z
Y W
(cos ]OX P B)2 Y W (sin ]OX P B)2 Y W X + Z
= = =
X Z X Z (cos ]OX P B)2 X Z Y + W
(sin ]OX P B)2
!
(sin ]OX P B)2 X + Z
since (11) entails =
(cos ]OX P B)2 Y + W

X + Z Y + W 1 1 1 1
= = + +
X Z Y W X Z Y W

X + Z 1 1 Y + W 1 1
since = + and = +
X Z X Z Y W Y W
= 1
1 1 1 1
(since Theorem 19 yields + = + ). Hence, P OY P OW = P OX P OZ ,
X Z Y W
so that
P OX P OZ = P OY P OW : (14)
Now, we shall use directed segments; in the following, the directed distance between
two points P1 and P2 will be denoted by P1 P2 (as opposed to the non-directed distance,
which we will continue to write as P1 P2 ). We direct the lines OX OZ and OY OW
arbitrarily. Then, P OX P OZ = P OX P OZ (since the point P lies on the segment
OX OZ ) and P OY P OW = P OY P OW (similarly). Hence,

P OX P OZ = P O X P OZ = P O Y P OW (by (14))
= P OY P O W :

By the converse of the intersecting chords theorem, we can conclude from this that the
points OX ; OY ; OZ and OW lie on one circle (since the point P lies on the segments
OX OZ and OY OW ). Theorem 20 is thus proved.

45
 References

[1] D. O. Shkljarskij, N. N. Chenzov, I. M. Jaglom, Izbrannye zadachi i teoremy

elementarnoj matematiki: Chastj 2 (Planimetrija), Moscow 1952.
[2] Jean-Pierre Ehrmann, Hyacinthos message #6740.
http://groups.yahoo.com/group/Hyacinthos/messages/6740?viscount=1&xm=1&m=e
[3] Mecrazywong, MathLinks topic #52681 (circumscribed quadrilateral).
http://www.mathlinks.ro/Forum/viewtopic.php?t=52681
[4] Darij Grinberg, A theorem on circumscribed quadrilaterals.
http://www.cip.ifi.lmu.de/~grinberg/geometry2.html
[5] Darij Grinberg, MathLinks topic #4748 (Circumscribed quadrilaterals).
http://www.mathlinks.ro/Forum/viewtopic.php?t=4748
[6] Alexander Bogomolny, A Property of Circumscribed Quadrilaterals.
http://cut-the-knot.com/Curriculum/Geometry/CircumQuadri.shtml
[7] Alexander Bogomolny, Ceva in Circumscribed Quadrilateral.
http://www.cut-the-knot.org/Generalization/BrianchonCeva.shtml
[8] Darij Grinberg, CTK Exchange College Math topic #419 (Ceva in Circumscribed
Quadrilateral).
http://www.cut-the-knot.org/htdocs/dcforum/DCForumID6/419.shtml
[9] Darij Grinberg, MathLinks post #112006 (The generalized Brahmagupta for-
mula).
http://www.mathlinks.ro/Forum/viewtopic.php?t=15973
[10] Sprmnt21, MathLinks topic #19549 (Corollary of Brianchon’s theorem).
http://www.mathlinks.ro/Forum/viewtopic.php?t=19549
[11] Orl, MathLinks post #220212 (quadrilateral ABCD without parallel sides is
circumscribed).
http://www.mathlinks.ro/Forum/viewtopic.php?t=35309
[12] Pengshi, Anipoh, Yetti, MathLinks topic #28864 (Looks so easy […nd inradius
of circumscribed quadrilateral]).
http://www.mathlinks.ro/Forum/viewtopic.php?t=28864
[13] A. Zaslavsky, Problem M1887 and solution, Kvant 3/2004, page 19 (in Russian).
http://kvant.mirror1.mccme.ru/djvu/2004_03.djvu
[14] MathLinks topic #173533 (Problem M1887 in russian journal "Kvant" no.6/2003).
http://www.mathlinks.ro/viewtopic.php?t=173533
[15] Yetti, MathLinks topic #121204 (Five points collinear).
http://www.mathlinks.ro/viewtopic.php?t=121204
[16] Pig‡y, Yetti, Juancarlos: MathLinks topic #21758.
http://www.mathlinks.ro/Forum/viewtopic.php?t=21758
[17] I. Vaynshtejn, N. Vasiljev, V. Senderov: Problem M1495 and solution, Kvant
6/1995, pages 27-28 (in Russian).
http://kvant.mccme.ru/1995/06/resheniya_zadachnika_kvanta_ma.htm
[18] I. Vaynshtejn: Problem M1524 and solution, Kvant 3/1996, pages 25-26 (in
Russian).
http://kvant.mccme.ru/1996/03/resheniya_zadachnika_kvanta_ma.htm

46