More Characterizations of Tangential Quadrilaterals
More Characterizations of Tangential Quadrilaterals
More Characterizations of Tangential Quadrilaterals
Forum Geometricorum
Volume 11 (2011) 65–82. b b
FORUM GEOM
ISSN 1534-1178
Martin Josefsson
Abstract. In this paper we will prove several not so well known conditions for
a quadrilateral to have an incircle. Four of these are different excircle versions
of the characterizations due to Wu and Vaynshtejn.
1. Introduction
In the wonderful paper [13] Nicuşor Minculete gave a survey of some known
characterizations of tangential quadrilaterals and also proved a few new ones. This
paper can to some extent be considered a continuation to [13].
A tangential quadrilateral is a convex quadrilaterals with an incircle, that is a
circle tangent to all four sides. Other names for these quadrilaterals are1 tangent
quadrilateral, circumscribed quadrilateral, circumscribable quadrilateral, circum-
scribing quadrilateral, inscriptable quadrilateral and circumscriptible quadrilateral.
The names inscriptible quadrilateral and inscribable quadrilateral have also been
used, but sometimes they refer to a quadrilateral with a circumcircle (a cyclic
quadrilateral) and are not good choices because of this ambiguity. To avoid confu-
sion with so many names we suggest that only the names tangential quadrilateral
(or tangent quadrilateral) and circumscribed quadrilateral be used. This is sup-
ported from the number of hits on Google2 and the fact that both MathWorld and
Wikipedia uses the name tangential quadrilateral in their encyclopedias.
Not all quadrilaterals are tangential,3 hence they must satisfy some condition.
The most important and perhaps oldest such condition is the Pitot theorem, that a
quadrilateral ABCD with consecutive sides a, b, c and d is tangential if and only
if the sums of opposite sides are equal: AB + CD = BC + DA, that is
a + c = b + d. (1)
It is named after the French engineer Henri Pitot (1695-1771) who proved that this
is a necessary condition in 1725; that it is also a sufficient condition was proved by
the Swiss mathematician Jakob Steiner (1796-1863) in 1846 according to F. G.-M.
[7, p.319].
The proof of the direct theorem is an easy application of the two tangent theo-
rem, that two tangents to a circle from an external point are of equal length. We
know of four different proofs of the converse to this important theorem, all beauti-
ful in their own way. The first is a classic that uses a property of the perpendicular
bisectors to the sides of a triangle [2, pp.135-136], the second is a proof by con-
tradiction [10, pp.62-64], the third uses an excircle to a triangle [12, p.69] and the
fourth is an exquisite application of the Pythagorean theorem [1, pp.56-57]. The
first two of these can also be found in [3, pp.65-67].
Two similar characterizations are the following ones. If ABCD is a convex
quadrilateral where opposite sides AB and CD intersect at E, and the sides AD
and BC intersect at F (see Figure 1), then ABCD is a tangential quadrilateral if
and only if either of
BE + BF = DE + DF,
AE − AF = CE − CF.
These are given as problems in [3] and [14], where the first condition is proved
using contradiction in [14, p.147]; the second is proved in the same way4.
F
b
D b
c
b
C
d
b
A
b b b
E
a B
4In [3, pp.186-187] only the direct theorems (not the converses) are proved.
More characterizations of tangential quadrilaterals 67
Proof. In a convex quadrilateral ABCD, let the incircles in triangles ABC, CDA,
BCD and DAB be tangent to the diagonals AC and BD at the points X, Y , Z
and W respectively (see Figure 2). First we prove that
ZW = 1 a − b + c − d = XY.
2
Using the two tangent theorem, we have BW = a − w and BZ = b − z, so
ZW = BW − BZ = a − w − b + z.
In the same way DW = d − w and DZ = c − z, so
ZW = DZ − DW = c − z − d + w.
Adding these yields
2ZW = a − w − b + z + c − z − d + w = a − b + c − d.
Hence
ZW = 12 a − b + c − d
where we put an absolute value since Z and W can “change places” in some quadri-
laterals; that is, it is possible for W to be closer to B than Z is. Then we would
have ZW = 12 (−a + b − c + d).
C
b
z
z
b
c−z
b
D b
W b−z
d−w b
Z
b
w
b b b
A w a−w B
Theorem 2. The incircles in the four overlapping triangles formed by the diago-
nals of a convex quadrilateral are tangent to the sides in eight points, two per side,
making one distance between tangency points per side. It is a tangential quadri-
lateral if and only if the sums of those distances at opposite sides are equal.
Proof. According to the two tangent theorem, AZ = AY , BS = BT , CU = CV
and DW = DX, see Figure 3. Using the Pitot theorem, we get
AB + CD = BC + DA
⇔ AZ + ZS + BS + CV + V W + DW = BT + T U + CU + DX + XY + AY
⇔ ZS + V W = T U + XY
after cancelling eight terms. This is what we wanted to prove.
C
b
V b
W
D b
b b
U
X b
b
T
Y b
b b b b
A B
Z S
The configuration with the four incircles in the last two theorems has other in-
teresting properties. If the quadrilateral ABCD is cyclic, then the four incenters
are the vertices of a rectangle, see [2, p.133] or [3, pp.44-46].
A third example where the Pitot theorem is used to prove another characteri-
zation of tangential quadrilaterals is the following one, which is more or less the
same as one given as a part of a Russian solution (see [18]) to a problem we will
discuss in more detail in Section 4.
Theorem 3. A convex quadrilateral is subdivided into four nonoverlapping trian-
gles by its diagonals. Consider the four tangency points of the incircles in these
triangles on one of the diagonals. It is a tangential quadrilateral if and only if the
distance between two tangency points on one side of the second diagonal is equal
to the distance between the two tangency points on the other side of that diagonal.
Proof. Here we cite the Russian proof given in [18]. We use notations as in Fig-
′ ′
ure 4 and shall prove that the quadrilateral has an incircle if and only if T1 T2 =
′ ′
T3 T4 .
More characterizations of tangential quadrilaterals 69
D b
b
T3
T4 b b b C
′ ′
T4 b
T3
b
T4
′′
′
b
T2
b
P b
T2
b
′′
T1 ′
b
T1
b b b
A B
T1
D
b
b
b
b
b
r3 b C
b
h4 h3
r4 b
h2 b
b
b
b
P r2
b
h1
r1
b b b b
A B
In [13] Nicuşor Minculete proved in two different ways that another characteri-
zation of tangential quadrilaterals is5
1 1 1 1
+ = + , (2)
h1 h3 h2 h4
where h1 , h2 , h3 and h4 are the altitudes in triangles ABP , BCP , CDP and
DAP from P to the sides AB, BC, CD and DA respectively, see Figure 5. These
two characterizations are closely related to the following one.
Theorem 4. A convex quadrilateral ABCD is tangential if and only if
1 1 1 1
+ = +
R1 R3 R2 R4
where R1 , R2 , R3 and R4 are the exradii to triangles ABP , BCP , CDP and
DAP opposite the vertex P , the intersection of the diagonals AC and BD.
Proof. In a triangle, an exradius Ra is related to the altitudes by the well known
relation
1 1 1 1
=− + + . (3)
Ra ha hb hc
5
Although he used different notations.
More characterizations of tangential quadrilaterals 71
R3
b
b
C
b
D P
b
R2
hA
hB
b
h1 b
R4
b b b b B
A
R1
C
b
D b
b
b b
A B
This problem appeared at the CTK Exchange6 on September 17, 2003 [17],
where it was debated for a month. On Januari 2, 2004, it migrated to the Hy-
acinthos problem solving group at Yahoo [15], and after a week a full synthetic
solution with many extra properties of the configuration was given by Darij Grin-
berg [8] with the help of many others.
So why was this problem called Christopher Bradley’s conjecture? In November
2004 a paper about cyclic quadrilaterals by the British mathematician Christopher
Bradley was published, where the above problem was stated as a conjecture (see
[4, p.430]). Our guess is that the conjecture was also published elsewhere more
than a year earlier, which explains how it appeared at the CTK Exchange.
A similar problem, that is almost the converse, was given in 1998 by Toshio
Seimiya in the Canadian problem solving journal Crux Mathematicorum [16]:
Suppose ABCD is a convex cyclic quadrilateral and P is the intersection of
the diagonals AC and BD. Let I1 , I2 , I3 and I4 be the incenters of triangles
P AB, P BC, P CD and P DA respectively. Suppose that I1 , I2 , I3 and I4 are con-
cyclic. Prove that ABCD has an incircle.
The next year a beautiful solution by Peter Y. Woo was published in [16]. He
generalized the problem to the following very nice characterization of tangential
quadrilaterals:
When a convex quadrilateral is subdivided into four nonoverlapping triangles
by its two diagonals, then the incenters of the four triangles are concyclic if and
only if the quadrilateral has an incircle.
6
It was formulated slightly different, where the use of the word inscriptable led to a
misunderstanding.
More characterizations of tangential quadrilaterals 73
J3
b
C J2
b b
I3
D b
b
b
I2
I4 P
J4 b
b
I1
b b B
A
J1
b
Figure 8 suggests that J1 J3 ⊥J2 J4 and I1 I3 ⊥I2 I4 . These are true in all convex
quadrilaterals, and the proof is very simple. The incenters and excenters lies on the
angle bisectors to the angles between the diagonals. Hence we have ∠J4 P J1 =
∠I4 P I1 = 12 ∠DP B = π2 .
Another characterization related to the configuration of Christopher Bradley’s
conjecture is the following one. This is perhaps not one of the nicest characteriza-
tions, but the connection between opposite sides is present here as well as in many
others. That the equality in the theorem is true in a tangential quadrilateral was
established at [5].
Theorem 6. A convex quadrilateral ABCD with diagonals intersecting at P is
tangential if and only if
(AP + BP − AB)(CP + DP − CD) (BP + CP − BC)(DP + AP − DA)
= .
(AP + BP + AB)(CP + DP + CD) (BP + CP + BC)(DP + AP + DA)
Proof. In a triangle ABC with sides a, b and c, the distance from vertex A to the
incenter I is given by
q
(s−a)(s−b)(s−c) r r
r s bc(s − a) bc(−a + b + c)
AI = A
= q = = . (5)
sin 2 (s−b)(s−c) s a+b+c
bc
More characterizations of tangential quadrilaterals 75
In a quadrilateral ABCD, let the incenters in triangles ABP , BCP , CDP and
DAP be I1 , I2 , I3 and I4 respectively. Using (5), we get
r
P A · P B(AP + BP − AB)
P I1 = ,
AP + BP + AB
r
P C · P D(CP + DP − CD)
P I3 = .
CP + DP + CD
Thus in all convex quadrilaterals
AP · BP · CP · DP (AP + BP − AB)(CP + DP − CD)
(P I1 · P I3 )2 =
(AP + BP + AB)(CP + DP + CD)
and in the same way we have
AP · BP · CP · DP (BP + CP − BC)(DP + AP − DA)
(P I2 · P I4 )2 = .
(BP + CP + BC)(DP + AP + DA)
In [16], Woo proved that P I1 · P I3 = P I2 · P I4 if and only if ABCD has an
incircle. Hence it is a tangential quadrilateral if and only if
AP · BP · CP · DP (AP + BP − AB)(CP + DP − CD)
(AP + BP + AB)(CP + DP + CD)
AP · BP · CP · DP (BP + CP − BC)(DP + AP − DA)
=
(BP + CP + BC)(DP + AP + DA)
from which the theorem follows.
5. Iosifescu’s characterization
In [13] Nicuşor Minculete cites a trigonometric characterization of tangential
quadrilaterals due to Marius Iosifescu from the old Romanian journal [11]. We had
never seen this nice characterization before and suspect no proof has been given in
English, so here we give one. Since we have had no access to the Romanian journal
we don’t know if this is the same proof as the original one.
Theorem 7 (Iosifescu). A convex quadrilateral ABCD is tangential if and only if
x z y w
tan · tan = tan · tan
2 2 2 2
where x = ∠ABD, y = ∠ADB, z = ∠BDC and w = ∠DBC.
Proof. Using the trigonometric formula
u 1 − cos u
tan2 = ,
2 1 + cos u
we get that the equality in the theorem is equivalent to
1 − cos x 1 − cos z 1 − cos y 1 − cos w
· = · .
1 + cos x 1 + cos z 1 + cos y 1 + cos w
This in turn is equivalent to
(1 − cos x)(1 − cos z)(1 + cos y)(1 + cos w)
= (1 − cos y)(1 − cos w)(1 + cos x)(1 + cos z). (6)
76 M. Josefsson
D b
c
z b
C
y
d
q b
w
x
b b
A a B
C
b
J1
b
b
a
Ra
b b
A c B
8This is a characterization of tangential quadrilaterals, but that’s not important for the proof of
this theorem.
9Here it is important that Theorem 6 is a characterization of tangential quadrilaterals.
More characterizations of tangential quadrilaterals 79
JCP |B C
b b
D
b
b
JCP |D
P
b
JAP |B b
b b
A B
JAP |D
Proof. We use the notation JAP |B for the excenter in the excircle tangent to side
AP opposite B in triangle ABP . Using the Lemma in triangles ABP , BCP ,
CDP and DAP yields (see Figure 11)
(P JAP |B )2 AB + AP − BP
= ,
AP · BP AB − AP + BP
(P JCP |D )2 CD + CP − DP
= ,
CP · DP CD − CP + DP
(P JCP |B )2 BC + CP − BP
= ,
CP · BP BC − CP + BP
(P JAP |D )2 DA + AP − DP
= .
AP · DP DA − AP + DP
From Theorem 8 we get that ABCD is a tangential quadrilateral if and only if
(P JAP |B )2 (P JCP |D )2 (P JCP |B )2 (P JAP |D )2
· = · ,
AP · BP CP · DP CP · BP AP · DP
80 M. Josefsson
which is equivalent to
Now JAP |B JCP |D and JCP |B JAP |D are straight lines through P since they are
angle bisectors to the angles between the diagonals in ABCD. According to the
intersecting chords theorem and its converse, (9) is a condition for the excenters to
be concyclic.
There is of course a similar characterization where the excircles are opposite the
vertices A and C.
We conclude with a theorem that resembles Theorem 4, but with the excircles
in Theorem 10.
Proof. We use notations on the altitudes as in Figure 12, which are the same as in
the proof of Theorem 4. From (3) we have
1 1 1 1
= − + + ,
Ra hB hA h1
1 1 1 1
= − + + ,
Rb hB hC h2
1 1 1 1
= − + + ,
Rc hD hC h3
1 1 1 1
= − + + .
Rd hD hA h4
These yield
1 1 1 1 1 1 1 1
+ − − = + − − .
Ra Rc Rb Rd h1 h3 h2 h4
Hence
1 1 1 1 1 1 1 1
+ = + ⇔ + = + .
Ra Rc Rb Rd h1 h3 h2 h4
Since the equality to the right is a characterization of tangential quadrilaterals ac-
cording to (2), so is the equality to the left.
Even here there is a similar characterization where the excircles are opposite the
vertices A and C.
More characterizations of tangential quadrilaterals 81
b
Rb C
b
hC Rc
D
b
h3 h2 b
P
h4 b
hD
b
hB
hA
h1
Ra
b b
A B
Rd b
References
[1] I. Agricola and T. Friedrich, Elementary geometry, American Mathematical Society, 2008.
[2] N. Altshiller-Court, College Geometry, Dover reprint, 2007.
[3] T. Andreescu and B. Enescu, Mathematical Olympiad Treasures, Birkhäuser, Boston, 2004.
[4] C. J. Bradley, Cyclic quadrilaterals, Math. Gazette, 88 (2004) 417–431.
[5] crazyman (username), tangential quadrilateral, Art of Problem Solving, 2010,
http://www.artofproblemsolving.com/Forum/viewtopic.php?t=334860
[6] N. Dergiades, Hyacinthos message 8966, January 6, 2004.
[7] F. G.-M, Exercices de Géométrie, Cinquième édition (in French), Éditions Jaques Gabay, 1912.
[8] D. Grinberg, Hyacinthos message 9022, January 10, 2004.
[9] E. W. Hobson, A Treatise on Plane and Advanced Trigonometry, Seventh Edition, Dover Pub-
lications, 1957.
[10] R. Honsberger, In Pólya’s Footsteps, Math. Assoc. Amer., 1997.
[11] M. Iosifescu, Problem 1421, The Mathematical Gazette (in Romanian), no. 11, 1954.
[12] K. S. Kedlaya, Geometry Unbound, 2006, available at
http://math.mit.edu/˜kedlaya/geometryunbound/
[13] N. Minculete, Characterizations of a tangential quadrilateral, Forum Geom., 9 (2009) 113–118.
[14] V. Prasolov, Problems in Plane and Solid Geometry, translated and edited by D. Leites, 2006,
available at http://students.imsa.edu/˜tliu/Math/planegeo.pdf
[15] Rafi (username), Hyacinthos message 8910, January 2, 2004.
[16] T. Seimiya and P. Y. Woo, Problem 2338, Crux Math., 24 (1998) 234; solution, ibid., 25 (1999)
243–245.
[17] sfwc (username), A conjecture of Christopher Bradley, CTK Exchange, 2003,
http://www.cut-the-knot.org/cgi-bin/dcforum/forumctk.cgi?az=
read_count&om=383&forum=DCForumID6
[18] I. Vaynshtejn, Problem M1524, Kvant (in Russian) no. 3, 1996 pp. 25-26, available at http:
//kvant.mccme.ru/1996/03/resheniya_zadachnika_kvanta_ma.htm
82 M. Josefsson