Application of Algebra and Calculus

in Euclidean Geometry

Huy Nguyen, Trung Le

Le Hong Phong High School - Ho Chi Minh City

Contents
1 Introduction 5

Problem 0 5
Proof 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Proof 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Proof 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Proof 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Proof 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Proof 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Problem 1 13

2 Background 14

3 Example problems 16

Problem 2 17

Problem 3 (USAMO 2001) 18

Problem 4 (USAMO 2008) 19

Problem 5 (VMO 2006, Board A) 20

Problem 6 (Vietnam TST 2014) 21

1
Problem 7 (VMO 2008) 24

4 Practice Exercises 25

Exercise 1 (China Girls MO 2002) 25

Exercise 2 (Italy TST 2000) 25

Exercise 3 (Mixtinilear Incircle) 25

Exercise 4 (USA TST 2004) 26

Exercise 5 (Ho Chi Minh City Regional Team Selection 2012) 26

Exercise 6 (IMO 2013) 26

Exercise 7 (IMO Shortlist 2005) 26

Exercise 8 (IMO Shortlist 2011) 26

Exercise 9 (All-Russian MO 2009) 26

Exercise 10 (Balkan MO 2007) 26

Exercise 11 (China TST 2008) 27

Exercise 12 (Balkan MO 2013) 27

Exercise 13 (IMO Shortlist 2005 *) 27

5 Solution to Practice Exercises 27

Solution to Exercise 1 (China Girls MO 2002) 27

Solution to Exercise 2 (Italy TST 2000) 28

Solution to Exercise 3 (Mixtilinear Incircle) 29

Solution to Exercise 4 (USA TST 2004) 30

Solution to Exercise 5 (Ho Chi Minh City Regional Team Selection 2012) 32

Solution to Exercise 6 (IMO 2013) 33

2
Solution to Exercise 7 (IMO Shortlist 2005) 36

Solution to Exercise 8 (IMO Shortlist 2011) 37

Solution to Exercise 9 (All-Russian MO 2009) 39

Solution to Exercise 10 (Balkan MO 2007) 41

Solution to Exercise 11 (China TST 2008) 42

Solution to Exercise 12 (Balkan MO 2013) 44

Solution to Exercise 13 (IMO Shortlist 2005) 45

3
Glossary
Abbreviations

MO Abbrev. for Mathematical Olympiad, the highest-level math contest for

high school students in a region - for example, VMO (Vietnam), USAMO
(America), IMO (International).
TST Abbrev. for Team Selection Test, a test given to top performers in a na-
tional olympiad to select the nation’s representative team for an international
olympiad.
Mathematical notions

a, b, c, r, R Unless otherwise specified, in 4ABC, a, b, c denote the lengths of segments

BC, CA, AB respectively. p is its semiperimeter, i.e., p = 12 (a + b + c). r and
R denote its inradius and circumradius respectively.
S4XY Z Denote the area of 4XY Z.
(XY Z) For any three non-collinear points X, Y, Z, denote the circumcircle of 4XY Z.
(I; r) and Denote the circle with center I and radius r. When it is clear from the context
(I) that there is only one circle with center I, the parameter r can be omitted.
−→
AB The algebraic length of segment AB. If AB points in the positive direction,
−→ −→
AB = |AB|; otherwise, AB = −|AB|.
(ABCD) For any four collinear points (A, B, C, D), denote the cross-ratio

AC BC
(ABCD) = : .
AD BD

When (ABCD) = −1, A and B are harmonic conjugates of each other with
respect to C and D.
(AB, CD) The oriented angle starting at AB and ending at CD.

4
1 Introduction
We begin with an analysis of a simple geometry problem, proposed by Dr. Trinh Le, deputy
delegate of the Vietnam IMO team.

Problem 0

In 4ABC consider a point D on segment BC. Let I and J be the midpoints of DB and
DC respectively. The perpendicular bisectors of DB and DC intersect AB at M and AC
at N respectively. Let O be the circumcenter of 4ABC . Prove that AM ON is a cyclic
quadrilateral.

O
B D C

Figure 1: Prove that AM ON is a cyclic quadrilateral.

The following six solutions were devised by six students from the Ho Chi Minh City VMO
team. Let us first take a look at their approaches and analyze the pros and cons of each.

5
Proof 1

A (b,c)

U K V
N

B (-1,0) D (d,0) T (0,0) C (1,0)

Let T be the midpoint of BC. Consider the Cartesian coordinate system T xy where T
is the origin, T x is BC and T y is the perpendicular bisector of BC.
Denote the following coordinate points: B(−1, 0), C(1, 0), D(d, 0) and A(a, b).
Let U and V be the midpoints of AB and AC respectively, and let K be the projection of
O on M N . Furthermore, for any line l, denote n~l as its normal vector.
We can then compute the followings:
b(x + 1) b(x − 1)
1. (AB) : y = , (AC) : y = .
a+1 a−1
       
a−1 b a+1 b d − 1 b(d + 1) d + 1 b(d − 1)
2. U , , V , ,M , , N , .
2 2 2 2 2 2(a + 1) 2 2(a − 1)
a2 + b 2 − 1
 
−→ −−→
3. xO = xT = 0, OU k nAB ⇒ O 0; .
2ab
 
b(a − d) d−1 b(d + 1)
4. (M N ) : 2 x− +y− = 0.
a −1 2 2(a + 1)
−−→ xK b(a − d)
5. OK k −
n−→
MN ⇒ a2 +b2 −1
= 2 .
yK − 2b a −1

Observe that yU = yv = 2b , so it turns out the line U V is (U V ) : y = 2b . Furthermore, the

point K is uniquely determined from (4) and (5), and we notice that xK = d−a b


2
, yK = 2
d−a b


satisfies both. Hence K 2 , 2 and therefore K ∈ U V . By the converse Simson’s theorem
AM ON is cyclic. (QED)

6
Proof 2

Let N 0 6= A be the intersection point of (AM O) and AC. We prove AN 0 = AN , which

would imply N 0 ≡ N . Denote DB = x, then

x x a−x
BM = , AM = c − , AN = b − . (1)
2 cos B 2 cos B 2 cos C

N’ = N

O
B D C

Since AM ON 0 is a cyclic quadrilateral, by Ptolemy’s theorem:

AO · M N 0 = AM · ON 0 + AN 0 · OM. (2)

Further observe that ∠OM N 0 = ∠OAC = 900 − B and ∠ON 0 M = ∠OAB = 900 − C.
Hence, in 4OM N 0 , applying the sine Rule, we have

ON 0 cos B
= .
OM cos C
cos C
Multiplying both sides of (2) by OM
we have

MN0 ON 0
(AO cos C) · = (AM cos C) · + AN 0 cos C.
OM OM

which yields

0 sin ∠M ON 0 cos B
AN cos C = (AO cos C) · 0
− (AM cos C) ·
 sin ∠ON M   cosC
a sin A x 
= · cos C · − c− cos B
2 sin A cos C 2 cos B
a+x
= − c cos B.
2

7
From (1) we also have
a−x
AN cos C = b cos C − ,
2
hence AN 0 = AN would be equivalent to

a+x a−x
− c cos B = b cos C − ,
2 2

which can be rewritten as a = b cos C + c cos B, which is true. Thus AN 0 = AN and

therefore N 0 ≡ N . (QED)

Proof 3

O’ = O

D
B C

(AM N ) intersects the diameter AK of (ABC) at O0 6= A. We will prove that AK =

2AO0 , which implies O0 ≡ O.
Observe that 4O0 M N ∼ 4KBC so

OM MN O0 N
= = = k,
KB BC KC

which means that

M N = kBC, O0 M = kKB, O0 N = kKC. (3)

8
Applying Ptolemy’s theorem to cyclic quadrilaterals AM O0 N and ABKC we have

AO0 · M N = AM · O0 N + AN · O0 M. (4)

AK · BC = AB · KC + AC · KB. (5)

Dividing both sides of (4) by k we have

AO · BC = AM · KC + AN · KB. (6)

From Solution 2, with DB = x, we also have

x a−x
AM = c − , AN = b − .
2 cos B 2 cos C

Further note that that KB = 2R cos C and KC = 2R cos B. Looking at (5) and (6), to
show AK = 2AO0 we need to prove that

AB · KC + AC · KB = 2(AM · KC + AN · KB),

which is equivalent to
 
 x  a−x
c cos B + b cos C = 2 cos B c − + 2 cos C b − ,
2 cos B 2 cos C

which reduces to a = b cos C +c cos B, which is true. Hence O ≡ O0 ∈ (AM N ) and therefore
AM ON is cyclic. (QED)

9
Proof 4

U V

B R I D S J C

Let U, V be the midpoints of AB, AC respectively and R, S be the projections of U, V

onto BC respectively. It is easy to see that IR = JS. Indeed, on axis BC, denote the
−−→
direction of BC as the positive direction. We then have

x c cos B − x
IR = BR − BI = BU cos(BR, BU ) − = ,
2 2
a−x −b cos C + a − x
JS = CS − CJ = CV cos(CS, CA) + = .
2 2

Since a = b cos C + c cos B, IR = JS so IR = JS.

Applying Thales’s theorem for U R k M I and V S k N J we have

MU BU 1 NV CN 1
= = , = = .
IR BR cos B SJ CJ cos C

Hence
MU IR cos C cos C OU
= · = = .
NV JS cos B cos B OV
Observe that 4OM U and 4ON V are directionally similar. Let f be the spiral similarity
OU
with center O, angle (OU, OV ) and ratio k = OV
, then f (U ) = V, f (M ) = N .
Hence (M O, M U ) ≡ (N O, N V ) (mod π), so AM ON is a cyclic quadrilateral. (QED)

10
Proof 5

E
A

X
N

B D C

Let E be the reflection of D across M N . We then have

∠M EN = ∠M DN = 1800 − ∠M DB − ∠N DC = 1800 − B − C = A.

Hence, AEN M is a cyclic quadrilateral, which implies ∠AM E = ∠AN E. Note that
M E = M D = M B and N E = N D = N C, so 4M BE and 4N CE are isosceles. In
other words, ∠AM E = 2∠M BE and ∠AN E = 2∠N CE. Therefore ∠M BE = ∠N CE, so
AECB is cylic and E ∈ (ABC).
Observe that OM and ON are the perpendicular bisectors of EB and EC respectively. Let
X = OM ∩ BE, Y = ON ∩ CE, then EXOY is a cyclic quadrilateral.
Hence
∠M ON = 1800 − ∠XEY = 1800 − A,

which means AM ON is cyclic. (QED)

11
 Proof 6

O’ = O

B D E C

Let E 6= D be the other intersection point of (DM N ) and BC.

We have ∠EM N = ∠EDN = C, A = ∠M DN = ∠M EN so 4EM N ∼ 4ACB.
Let O0 be the orthocenter of 4EM N . We observe that BM O0 E, AM O0 N, CN O0 E are
cyclic quadrilaterals, hence

∠O0 BE = ∠O0 M E = ∠O0 N E = ∠O0 CE,

so O0 C = O0 B. Similarly we have O0 A = O0 B (= O0 C) and thus o0 is the orthocenter of

4ABC . In other words, O0 ≡ O.
We also have ∠M ON = 1800 − ∠M DN = 1800 − A, which means AM ON is cyclic.

Remark 1. We can observe the relation among the six solutions given above. While all solutions
involve the additional construction of some geometric objects (new point, new line, or new circle),
the more sophisticated this construction is, the simpler the subsequent proof becomes.

• Solution 1 is the most simple approach: by considering the problem in the context of
Cartesian coordinates. It is then straightforward to calculate the coordinates of every
point. The proof does apply a small twist, however, by invoking Simson’s theorem to turn
the problem of cyclicity into that of collinearity, thereby avoiding equations of circles which
are usually fairly complicated.

• Solution 2 and 3 “reverse” the problem by constructing the point N 0 6= A as the intersection
of AC and (AM O). The motivation here is that, instead of having to deduce cyclicity,

12
 we assume cyclicity is already given. What remains is to show that both N 0 and N
share a common unique property, which in turn implies that N 0 is indeed N . This is
a recommended approach when one has not figured out how to effectively use all of the
information provided by the problem.

• Solution 4 builds upon the realization that all properties of M and N depend only on the
location of D on BC, i.e., the length of the segment DB. Hence we try to convert all
computations to those involving only the segments on BC. Thanks to this approach, there
are much less algebraic work to do.

• Solution 5 and 6 are two purely geometric solutions that one needs rich experience in
geometry to come up with. In particular, the idea of solution 5 is that since ∠M DN = A,
we create a reflection of D in order to obtain a cyclic quadrilateral. In solution 6, we use
the following lemma:

Lemma 1. Given 4ABC with three points X, Y, Z on BC, CA, AB respectively such that
4XY Z ∼ 4ABC. It follows that the orthocenter of 4XY Z is also the circumcenter of
4ABC.

While solutions such as 5 and 6 are always desirable, it can be difficult to identify the
underlying geometrical insights that in turn lead to such concise solutions, especially for students
not strong in geometry1 . The goal of this article, then, is to provide a systematic approach
to tackling geometry problems using the tools from algebra and calculus. We will show how
algebraic techniques along the line of Solution 1 and 2 can be utilized to solve a variety of
Olympiad-level problems.

Remark 2. The given problem is in fact one in a series of similar geometric properties.

Problem 1

Let O, I and H be the circumcenter, incenter and orthocenter of 4ABC respectively. Con-
sider three points D ∈ BC, M ∈ CA, N ∈ AB.

• Prove that if BD = BM and CD = CN then I ∈ (AM N ).

• Prove that if M D = M B and N D = N C then O ∈ (AM N ).

• Prove that if DM = DB and DN = DC then H ∈ (AM N ).

1
including us, the authors.

13
2 Background
We first highlight a number of theorems and solutions that are highly applicable. Note that
there are a number of textbook formulas - Pythagorean’s Theorem, Thales’ theorem, Law of
sines and Law of cosines - that are not introduced but also relevant here.

Theme 1. Useful equatilies.

• (Ptolemy) If ABCD is a cyclic quadrilateral then

AC · BD = AB · CD + BC · AD.

• (Stewart) For any collinear points A, B, C and arbitrary point M ,

M A2 · BC + M B 2 · CA + M C 2 · AB = −AB · BC · CA.

• (Carnot) For any four arbitrary points A, B, C, D,

AD⊥BC ⇔ AB 2 − AC 2 = DB 2 − DC 2 .

a+b+c
• (Heron) The area of 4ABC with side lengths a, b and c and semiperimeter p = 2
is
p
S4ABC = p(p − a)(p − b)(p − c).

Theme 2. Triangle lengths.

Given 4ABC (AB < AC) with altitude AD and bisector AF . Let E be the tangent point
of 4ABC ’s incircle and BC. We then see that
a2 + c 2 − b 2 a+c−b ac
BD = , BE = , BF = ,
2a 2 b+c
which in turn yields
(b − c)(b + c − a)
DE = BE − BD = .
2a
Furthermore, from Stewart’s theorem, we have
√
p
2 bcp(p − a) 2bc A
AD = AB · AC − DB · DC = = · cos .
b+c b+c 2
Theme 3. Trigonometric expressions.

In any 4ABC, we have

r s
A p(p − a) A (p − b)(p − c)
cos = , tan = ,
2 bc 2 p(p − a)
b 2 + c 2 − a2 A b+c−a
cot A = , cot = .
4S 2 2r
14
Theme 4. Concurrency and collinearity.

In 4ABC consider points D ∈ BC, E ∈ CA, F ∈ AB. Let L be another point on BC.

• (Ceva) AD, BE, CF are concurrent if and only if

DB EA F C
· · = −1.
DC EB F A

• (Menelaus) L, E, F are collinear if and only if

LB EA F C
· · = 1.
LC EB F A

L B D C

Combining the two theorems also shows that

DB LB
: = −1.
DC LC

In other words, if AD, BE, CF are concurrent and L, E, F are collinear then (DLBC) = −1 =
(LDBC).

Theme 5. Oriented angles.

Oriented angles are considered in mod π and have the following properties:

• (AB, CD) ≡ (AB, DC) ≡ (BA, CD).

• a k b if and only if (a, b) ≡ 0.

• a⊥b if and only if (a, b) ≡ π2 .

• (a, b) ≡ −(b, a).

• (a, b) ≡ (a, c) + (c, b).

15
 • A, B, C, D are cyclic if and only if (AB, AD) ≡ (CB, CD).

• AT is tangent to (ABC) if and only if (AT, AB) ≡ (CA, CB).

Theme 6. Cartesian coordinates.

Besides textbook formulas on vectors and line equations, we introduce the expressions of
several useful geometric constructs in the context of Cartesian coordinates.
Given 4ABC with centroid G, incenter I, orthocenter H, circumcenter O and excenter J
opposite to A, we have:
−
→ −→ −→ → −
• aIA + bIB + cIC = 0 , so
 
axA + bxB + cxC ayA + byB + cyC
I , .
a+b+c a+b+c

−→ −−→ −→ → −
• GA + GB + GC = 0 , so
 
xA + xB + xC y A + y B + y C
G , .
3 3

−−→ −−→ −−→ −→

• AH · BC = 0 and BH · AC = 0, so

xH (xC − xB ) + yH (yC − yB ) = xA (xC − xB ) + yA (yC − yB ),

xH (xC − xA ) + yH (yC − yA ) = xB (xC − xA ) + yB (yC − yA ).

• OA2 = OB 2 = OC 2 , so

− 2xO xA + x2A − 2yO yA + yA2

= − 2xO xB + x2B − 2yO yB + yB2

= − 2xO xC + x2C − 2yO yC + yC2 .

−→ −→ −→ −→ −→ −−→ −→ −→
• (AJ, AB) = (AJ, AC) and (BJ, BC) = (BJ, AB), so
−→ −→ −→ −→ −→ −−→ −→ −→
AJ · AB AJ · AC BJ · BC BJ · AB
= , = ,
c b a c

which yields  
−axA + bxB + cxC −ayA + byB + cyC
J , .
−a + b + c −a + b + c

3 Example problems

16
 Problem 2

Consider 4ABC with incircle (I) touching BC at D. Let DE be a diameter of (I). The
excircle to vertex A touches BC at L. Prove that A, E, L are collinear.

a+c−b
Proof. We know from the definition of L that CL = BD = 2
. To prove that A, E, L are
collinear, let L0 be the intersection of AE and BC. It remains to show that CL0 = CL, so L0 ≡ L.
To make use of DE⊥BC let’s create an altitude AH. In this case, 4AHL has DE k AH - since
we already know DH, DE and AH, Thales would give us CL0 , as desired.
H1

B H D C
L’ = L

J Figure 2: Prove that A, E, L are collinear.

Let L0 be the intersection of AE and BC. Draw an altitude AH of 4ABC . WLOG assume
AB ≤ AC, so H lies between B and D.
Let BL0 = x. As DE k AH we have
DL0 ED 2r a 2a
0
= = = = ,
HL AH AH p a+b+c
which leads to
x − BD DL0 2a
= 0
= .
x − BH HL a+b+c
Having
a2 + c 2 − b 2 a+c−b
BH = , BD =
2a 2
17
then gives us
−2a · BH + (a + b + c) · BD b+a−c
x= = ,
b+c−a 2
so
a+c−b
CL0 = a − x = = CL.
2
In other words, L0 ≡ L and therefore A, E, L are collinear.

Remark 3. This example serves as a strong lemma for many Olympiad problems. Let’s take a
look at one of them next.

Problem 3: USAMO 2001

The incircle (I) of 4ABC touches CA and CB at E1 and D1 respectively. Consider

E2 ∈ CA and D2 ∈ CB such that CE2 = AE1 and CD2 = BD1 . AD2 intersects BE2 at
P . Let Q be the intersection of (I) and AD2 that is closer to A. Prove that AQ = D2 P .

a+b−c
Proof. Construct an altitude AH of 4ABC . As BD2 = CD1 = 2
, D1 Q is in fact a diameter
of (I), so D1 Q k AH, from which we can calculate AQ.
To make use of the intersecion P , we can apply Menelaus to the transversal line BP E2 of
4CAD2 , which would give us P D2 .
A

Q
E1

I E2

B H D1 D2 C

Figure 3: Prove that AQ = DP2 .

Construct an altitude AH of 4ABC . According to Problem 2, D1 Q is the diameter of (I),

so AH k D1 Q. This would imply that
QD2 QD1 AD2 − QD2 AH − QD1
= ⇒ = .
AD2 AH AD2 AH

18
In other words,
2S 2S
AQ AH − QD1 a
− p p−a b+c−a
= = 2S
= = . (7)
AD2 AH a
p b+c+a
According to Menelaus, since BP E2 is a transversal line of 4CAD2 ,

P D2 BC E2 A
· · =1
PA BD2 E2 C
P D2 2a a+b−c
· · =1
PA a+b−c b+c−a
P D2 b+c−a
=
PA 2a
P D2 b+c−a b+c−a
= = . (8)
AD2 b + c − a + 2a b+c+a

From (7) and (8) we have AQ = D2 P .

Remark 4. Many more applications of the lemma introduced in Problem 2 can be found in [1].
For now we will move on to a different application of Menelaus’ theorem.

Problem 4: USAMO 2008

In an acute, scalene triangle 4ABC, let M, N, P be the midpoints of BC, CA, AB re-
spectively. The perpendicular bisectors of AB and AC intersect ray AM at D and E
respectively. BD intersects CE at F . Prove that AP F N is a cyclic quadrilateral.

P
N

F
O

B M C

Figure 4: Prove that AP F N is a cylic quadrilateral.

19
Proof. To make use of the intersection F , we can apply Menelaus to the transversal line DEM
MB
or F EC. Note that using DEM would involve the ratio MC
= 1, since M is the midpoint of
BC; this would make the calculations simpler.
It’s easy to see that DA = DB and F A = F C. Let O be the circumcenter of 4ABC, then O
lies on both EN and DP .
4BF C has transversal line DEM so by Menelaus:

DF EC M B DF EC
1= · · = · ,
DB EF M C DB EF

so
DF DB AD
= = .
EF EC AE
Hence F A is the external bisector at F of 4DEF . Observe that ∠AP O = ∠AN O = 90◦ so
AP ON is cyclic. To show that AP F N is also cyclic, we will prove that ∠AF O = 90◦ as well,
or, in other words, F O is the internal bisector at F of 4DEF .
Indeed, 4ADB has DA = DB and median DP so DP is also the angular bisector of ∠ADB. In
other words, DO is the external bisector at D of 4DEF . Similarly, EO is the external bisector
at F of 4DEF . Hence O is the excenter relative to F of 4DEF . It then follows that F O is
the internal bisector at F of 4DEF .

Problem 5: VMO 2006, Board A

Let ABCD be a convex quadrilateral. Consider a moving point M on AB such that M 6= A

and M 6= B. Let N 6= M be the other intersection of (M AC) and (M BD). Prove that N
is always on a fixed circle and M N always goes through a fixed point.

Proof. Let I be the intersection of AC and BD, then

(CI, CN ) ≡ (CA, CN ) ≡ (M A, M N )

≡ (M B, M N ) ≡ (DB, DN )

≡ (DI, DN ).

So C, I, D, N are cyclic or, in other words, N is always on the fixed circle (CDI).
Let t be the line that goes through I and is parallel to AB. t intersects M N at K.
Observe (M A, M N ) ≡ (KI, KN ) because AM k KI, while (M A, M N ) ≡ (CI, CN ) as well
because A, M, C, N are cyclic. Hence (KI, KN ) ≡ (CI, CN ) so C, I, K, N are cyclic. It follows
that K is the other intersection of t and (CDI) besides I.

20
 A
M
B

D
C

Figure 5: Prove that AP F N is a cylic quadrilateral.

As both t and (CDI) are fixed, K is a fixed point, and this is the point that M N always goes
through.

Remark 5. This problem is an example of when oriented angles work best. Due to the posi-
tioning of M , it’s very hard to prove that C, I, D, N are cyclic with normal angles; both cases
of ∠ICN = ∠IDN or ∠ICN = 180◦ − ∠IDN can happen under different circumstances. With
oriented angle, we can simply express this condition as (CI, CN ) ≡ (DI, DN ).

Problem 6: Vietnam TST 2014

Prove that:

(a) In 4ABC with altitude AD, consider a point P on AD. P B intersects AC at E and
P C intersects AB at F . If AEDF is a cyclic quadrilateral, prove that

PA A
= (tan B + tan C) · cot .
PD 2

(b) Consider 4ABC with orthocenter H. Let P be a moving point on AH. The line
perpendicular to AC at C intersects BP at M . The line perpendicular to AB at B
intersects CP at N . Let K be the projection of A on M N . Prove that ∠BKC +∠M AN
is constant.

Proof. We first state, without proof, a common application of the harmonic series:

21
Lemma 2. Consider 4ABC with altitude AD and a point P on AD. BP intersects AC at E
and CP intersects AB at F . Then DA is the angular bisector of ∠DHF .

Using this lemma, we see that ∠AEF = ∠ADF = ∠ADE = ∠AF E, so 4AEF is isosceles
at A. Now consider two cases:

F P

B D C

PA
Figure 6: Prove that PD
= (tan B + tan C) · cot A2 .

(a) • If AB = AC then AD is the angular bisector of A. It then follows that DF ⊥AB, DE⊥AC.
Hence
DC 2 a2 4b2 − a2
EC = = , EA = AC − EC = .
AC 4b 4b
Applying Menelaus to the transversal line EP B of 4ADC yields
PA EA BC 4b2 − a2
= · = .
PD EC BD 2a2
Note that ∠B = ∠C = 90◦ − ∠A
2
, so we also have
A A (b + c − a)2 (2b − a)2
(tan B + tan C) · cot = 2 cot2 = = ,
2 2 2r2 r2
Using algebraic manipulations and noting that 4ABC has b = c, r = Sp , we see that
(2b − a)2 2 p
2
= (2b − a) ·
r2 S2
p2
= (2b − a)2 ·
p(p − a)(p − b)(p − c)
4(2b + a)2
= (2b − a)2 · 2
a (2b + a)(2b − a)
2 2
4b − a
= .
2a2
PA
Hence PD
= (tan B + tan C) cot A2 .

22
 • If AB 6= AC then DB 6= DC. Applying Lemma 2, we see that DA is the angular
bisector of ∠EDF . Furthermore, AF DE is cyclic, so AE = AF = x.
We now see that

a2 + c 2 − b 2 a2 + b 2 − c 2 2S A b+c−a
DB = , DC = , AD = , cot = ,
2a 2a a 2 2r

which implies
 
A 1 1 A
(tan B + tan C) · cot = AD + · cot
2 DB DC 2
 
2S 2a 2a b+c−a
= 2 2 2
+ 2 2 2
·
a a +c −b a +b −c 2r
2
2a (b + c − a)(b + c + a)
= 2 . (9)
(a + c2 − b2 )(a2 + b2 − c2 )

As AD, BE, CF are collinear, according to Ceva,

DB EC F A DB x b−x
1= · · = · · .
¯ DC EA F B DC c − x x

which yields
b · DB − c · DC DC(c − b)
x= , b−x= .
DB − DC DB − DC
Applying Menelaus to the transversal line EP B of 4ADC:

PA EA BC x a b · DB − c · DC a
= · = · = · . (10)
PD EC BD b − x BD DC(c − b) DB
a2 +c2 −b2 a2 +b2 −c2
Substituting DB = 2a
and DC = 2a
into (10) and simplifying the expres-
sion, we get

PA 2a2 (b + c − a)(b + c + a) A
= 2 2 2 2 2 2
= (tan B + tan C) · cot . (11)
PD (a + c − b )(a + b − c ) 2

(b) Let I be the intersection of N B and M C, then I lies on (O).

Note that AKBN and AKCM are cyclic, so

∠N AM + ∠BKC = ∠KBI + ∠KCI + ∠BKC = 360◦ − ∠BIC,

which is constant.

23
 Problem 7: VMO 2008

Consider 4ABC with median AD and a line d perpendicular to AD. Let M be a mobile
point on d and let E, F be the midpoints of M B, M C respectively. The line passing through
E and perpendicular to d intersects AB at P . The line passing through F and perpendicular
to d intersects AC at Q. Let d0 be the line passing through M and perpendicular to P Q.
Prove that d0 always goes through a fixed point.

Proof. Since this problem only concerns perpendicular and parallel lines, a natural solution
would be to consider it the context of a two-dimensional coordinate Oxy. To simplify subsequent
calculations we will pick D as the origin.

A d0

P K

F
H
E
d Q

H0 K0
B C
D

Fixed

Figure 7: Prove that d0 always goes through a fixed point.

Consider the Cartesian coordinates Dxy with origin D and Dy ≡ DA. As Dx⊥Dy and
d⊥DA, we have Dx k d.
Now consider the following point coordinates: A(0, a), B(b, c), C(−b, −c), where a, b 6= 0. We
can then calculate
−→ −→
AB = (b, c − a), AC = (−b, −c − a),

which yields
(AB) : (a − c)x + by − ab = 0,

(AC) : (a + c)x − by + ab = 0.

24
Note that d is fixed and d k Dx so yM is fixed. Let yM = h, then
   
b + xM h + c xM − b h − c
E , , F , .
2 2 2 2
Let d1 , d2 be the lines passing through E, F respectively and perpendicular to d. It follows that
xM + b xM − b
(d1 ) : x = , (d2 ) : x = ,
2 2
so  
xM + b (a − c)(xM + b)
P = d1 ∩ AB = ,a − ,
2 2b
 
xM − b (a + c)(xM − b)
Q = d2 ∩ AC = ,a + .
2 2b
 
−→ axM − bc
As a result, P Q = −b, . We can now calculate
b
axM − bc
(d0 ) : −b(x − xM ) + · (y − h) = 0,
b
which can be rewritten as
b2
   
2 bc
b x−
− (axM − bc) y − h + = 0.
a a
 
b2
Hence d0 always passes through a fixed point with coordinates bca , h − a
.

Remark 6. This was a very difficult problem in VMO 2008, with 81.2% participants getting
a 0/7 score on it. A number of purely geometrical solutions to this problem can be viewed
here. The fixed point in question is the intersection of HH 0 and KK 0 , where H, K are the
projections of B, C on d respectively and H 0 , K 0 are the reflections of H about AB and K about
AC respectively. An analytical solution, in our case, turns out to be much simpler.

4 Practice Exercises
Exercise 1 (China Girls MO 2002). An acute 4ABC has three heights AD, BE and CF
respectively. Prove that the perimeter of triangle DEF is not over half of the perimeter of
triangle ABC.

Exercise 2 (Italy TST 2000). Let ABC be an isosceles right triangle and M be the midpoint
of its hypotenuse AB. Points D and E are taken on the legs AC and BC respectively such that
AD = 2DC and BE = 2EC. Lines AE and DM intersect at F . Show that F C bisects ∠DF E.

Exercise 3 (Mixtinilear Incircle). Consider 4ABC with circumcircle (O). A circle ω touches
(O) and also touches AB, AC at D, E respectively. Let J be the incenter of 4ABC . Prove that
D, J, E are collinear.

25
Exercise 4 (USA TST 2004). Let ABC be a triangle. Choose a point D in its interior. Let ω1
be a circle passing through B and D and ω2 be a circle passing through C and D so that the other
point of intersection of the two circles lies on AD. Let ω1 and ω2 intersect side BC at E and
F , respectively. Denote by X the intersection of DF , AB and Y the intersection of DE, AC.
Show that XY k BC.

Exercise 5 (Ho Chi Minh City Regional Team Selection 2012). 4ABC has circumcircle (O)
ˆ that does not contain A (D 6= B, D 6= C).
and heights AM, BN . Let D be a point on the arc BC
DA intersects BN at Q; DB intersects AM at P . Let I be the midpoint of P Q. Prove that
M, N, I are collinear.

Exercise 6 (IMO 2013). Let the excircle of 4ABC opposite the vertex A be tangent to the
side BC at the point A1 . Define the points B1 on CA and C1 on AB analogously, using the
excircles opposite B and C, respectively. Suppose that the circumcentre of (A1 B1 C1 ) lies on the
circumcircle (ABC). Prove that 4ABC is right-angled.

Exercise 7 (IMO Shortlist 2005). Let 4ABC be an acute-angled triangle with AB 6= AC. Let
H be the orthocenter of 4ABC, and let M be the midpoint of the side BC. Let D be a point
on the side AB and E a point on the side AC such that AE = AD and the points D, H, E
are on the same line. Prove that the line HM is perpendicular to the common chord of the
circumscribed circles of triangle 4ABC and triangle 4ADE.

Exercise 8 (IMO Shortlist 2011). Let ABC be a triangle with incentre I and circumcircle ω.
Let D and E be the second intersection points of ω with AI and BI, respectively. The chord
DE meets AC at a point F , and BC at a point G. Let P be the intersection point of the line
through F parallel to AD and the line through G parallel to BE. Suppose that the tangents to ω
at A and B meet at a point K. Prove that the three lines AE, BD and KP are either parallel
or concurrent.

Exercise 9 (All-Russian MO 2009). Let be given a parallelogram ABCD and two points A1 , C1
on its sides AB, BC, respectively. Lines AC1 and CA1 meet at P . Assume that the circumcircles
of triangles AA1 P and CC1 P intersect at the second point Q inside triangle ACD. Prove that
∠P DA = ∠QBA.

Exercise 10 (Balkan MO 2007). Let ABCD a convex quadrilateral with AB = BC = CD, with
AC not equal to BD and E be the intersection point of it’s diagonals. Prove that AE = DE if
and only if ∠BAD + ∠ADC = 120.

26
Exercise 11 (China TST 2008). Let ABC be a triangle, let AB > AC. Its incircle touches side
BC at point E. Point D is the second intersection of the incircle with segment AE (different
from E). Point F (different from E) is taken on segment AE such that CE = CF . The ray
CF meets BD at point G. Show that CF = F G.

Exercise 12 (Balkan MO 2013). In 4ABC, the excircle ωa opposite A touches AB at P and

AC at Q, while the excircle ωb opposite B touches BA at M and BC at N . Let K be the
projection of C onto M N and let L be the projection of C onto P Q. Show that the quadrilateral
M KLP is cyclic.

Exercise 13 (IMO Shortlist 2005 *). Given 4ABC satisfying AC + BC = 3 · AB. The incircle
of 4ABC has center I and touches the sides BC and CA at the points D and E, respectively.
Let K and L be the reflections of the points D and E with respect to I. Prove that the points A,
B, K, L lie on one circle.

5 Solution to Practice Exercises

Solution to Exercise 1: China Girls MO 2002

An acute triangle ABC has three heights AD, BE and CF respectively. Prove that the
perimeter of triangle DEF is not over half of the perimeter of triangle ABC.

B D C

Figure 8: Prove that DE + EF + F D ≤ 21 (AB + BC + CA).

27
Proof. From 4AEF ∼ 4ABC ∼ 4DBF ∼ 4DEC we get

EF = BC · cos A = a · cos A

DE = AB · cos C = c · cos C

F D = AC · cos C = b · cos B

We need to prove that

1
EF + DE + F D ≤ (AB + BC + CA),
2
which is equivalent to
1
a · cos A + b · cos B + c · cos C ≤ (a + b + c)
2
2 2 2
X b +c −a 1
⇔ a· ≤ (a + b + c)
2bc 2
X
2 2 2 2
⇔ a (b + c − a ) ≤ abc(a + b + c)

⇔ 2(a2 b2 + b2 c2 + c2 a2 ) ≤ a4 + b4 + c4 + abc(a + b + c).

According to Schur’s Inequality, we have

X
a4 + b4 + c4 + abc(a + b + c) ≥ (a3 b + ab3 ).

while based on AM-GM we have

X X √ X
(a3 b + ab3 ) ≥ 2 a3 b · ab3 = 2 a2 b 2 .

Hence 2(a2 b2 + b2 c2 + c2 a2 ) ≤ a4 + b4 + c4 + abc(a + b + c).

Solution to Exercise 2: Italy TST 2000

Let ABC be an isosceles right triangle and M be the midpoint of its hypotenuse AB.
Points D and E are taken on the legs AC and BC respectively such that AD = 2DC and
BE = 2EC. Lines AE and DM intersect at F . Show that F C bisects ∠DF E.

Proof. Let AC = BC = x we have

√
x 2 2x x 2x2
AM = , AD = BE = , CD = CE = , DE 2 = ,
2 3 3 9
which yields
2 2 2 π 5x2
M E = M B + BE − 2M B · BE · cos = .
4 18
Hence AM 2 − AD2 = M E 2 − DE 2 . According to Carnot’s Theorem, AE⊥DM .
Hence DCEF is cyclic, and we also have CD = CE, so ∠DF C = ∠DEC = ∠CDE = ∠CEF .
In other words, F E bisects ∠DF E.

28
 A

F
D

C
E B

Figure 9: Prove that F C bisects ∠DF E.

Solution to Exercise 3: Mixtilinear Incircle

Consider 4ABC with circumcircle (O). A circle ω touches (O) and also touches AB, AC
at D, E respectively. Let J be the incenter of 4ABC . Prove that D, J, E are collinear.

Proof. Denote I as the center of ω and let ID = IE = x, we have

x r √
IA = , JA = , OI = R − x, OJ = R2 − 2Rr.
sin A2 sin A2
According to Stewart’s Theorem,

OA2 · IJ + OI 2 · JA = OJ 2 · AI + IA · AJ · JI,

which is equivalent to
rx(x − r)
R2 (x − r) + (R − x)2 r = (R2 − 2Rr)x + ,
sin2 A2
which implies that sin2 A
2
= x−r
x
, so

IJ · IA = (AI − AJ) · AI = AI 2 − AI · AJ
1 x
= (x2 − rx) = · (x2 − rx)
sin2 A
2
x−r

= x2 = IE 2 .

Hence JE⊥AI. By similar arguments, JD⊥AI. Thus D, J, E are collinear.

29
 A

B C

Figure 10: Prove that E, D, F are collinear.

Solution to Exercise 4: USA TST 2004

Let ABC be a triangle. Choose a point D in its interior. Let ω1 be a circle passing through
B and D and ω2 be a circle passing through C and D so that the other point of intersection
of the two circles lies on AD. Let ω1 and ω2 intersect side BC at E and F , respectively.
Denote by X the intersection of DF , AB and Y the intersection of DE, AC. Show that
XY k BC.

Proof. Let S be the intersection of AD and BC and R be the other intersection of ω1 and ω2
besides D, then R lies on AD.
Applying Menelaus to a transversal line XDF 4ABS and line ACS of 4EDY we have

AX AD SF AY AD SE
= · and = · .
XB SD BF CY SD CE

30
 Y X

B F S E

Figure 11: Prove that XY k BC.

AX AY SF SE
To prove that XY k BC we need BX
= CY
, or, in other words, BF
= CE
.
Observe that SF · SC = SR · SD = SE · SB, so

SF · CE = SF (SE − SC) and SE · BF = SE(SF − SB).

We note that 4SDC ∼ 4SF R and 4SDB ∼ 4SER so

SF · SC = SD · SR = SE · SB,

which in turn means SF · CE = SE · BF . Thus XY k BC.

31
 Solution to Exercise 5: Ho Chi Minh City Regional Team Selection 2012

ˆ that
4ABC has circumcircle (O) and heights AM, BN . Let D be a point on the arc BC
does not contain A (D 6= B, D 6= C). DA intersects BN at Q; DB intersects AM at P .
Let I be the midpoint of P Q. Prove that M, N, I are collinear.

B M C

Figure 12: Prove that M, N, I are collinear.

Proof. Let ∠CAD = ∠CDB = x. Consider the following ratios:

NQ SAN Q AN · AQ · sin x AQ sin x

= = = , (12)
NH SAN H AN · AH · sin ∠HAC AH cos C
MP SBM P BM · BP · sin HBC
\ BP sin x
= = = . (13)
MH SBM H BM.BH. sin x BH cos C

Observe that ∠AQH + ∠BP H = 180◦ , so DP HQ is cyclic. Therefore

sin ∠AHQ sin ∠BHP

= .
sin ∠AQH sin ∠BP H

32
 AQ BP NQ MP
It then follows that AH
= BH
, and, according to (12) and (13), NH
= MH
.
Now consider three points M, I, N and 4HP Q, we see that

IP N Q M H NQ MH
· · = · = 1,
IQ N H M P NH MP

According to the converse Menelaus’ theorem, M, N, I are collinear.

Solution to Exercise 6: IMO 2013

Let the excircle of 4ABC opposite the vertex A be tangent to the side BC at the point
A1 . Define the points B1 on CA and C1 on AB analogously, using the excircles opposite
B and C, respectively. Suppose that the circumcentre of (A1 B1 C1 ) lies on the circumcircle
(ABC). Prove that 4ABC is right-angled.

Proof. We will extend the problem and prove both directions: 4ABC is right-angled if and only
if the circumcentre of (A1 B1 C1 ) lies on the circumcircle of 4ABC.

• If 4ABC is right-angled, WLOG assume ∠BAC = 90◦ . Let O be the midpoint of BC

and M be the midpoint of the arc BC
d that contains A in the circumcircle (ABC). As

M C = M B, CB1 = BC1 and ∠M CB1 = ∠M BC1 ,

we get 4M CB1 = 4M BC1 and therefore M B1 = M C1 . We now show that M A1 = M C1 .

Note that

a+b−c a c−b a2 + (b − c)2

BA1 = , OM = , OA1 = , M A21 = .
2 2 2 4

Let H be the projection of M on AB. 4M BC is a right isosceles triangle so M B = M C =

√a . By Carnot’s Theorem,
2

M B 2 − M A2 = BH 2 − AH 2 = (BH + AH)(BH − AH),

which would give us M A:

a2 √
− M A2 = c(c − 2 · M A)
2
b + c2
2 √
− M A 2 = c2 − 2 · c · M A
2 √
2
MA = (c − b).
2

33
 C

B1

A1
Q

H C1

A
B

K M

Figure 13: Prove that 4ABC is right-angled.

Applying the law of cosine on 4M AC1 :

M C12 = M A2 + AC12 − 2 · M A · AC1 · cos 45◦

√
(b − c)2 (a + c − b)2 2 a+c−b c−b
= + −2· · · √
2 4 2 2 2
2 2
a + (b − c)
= .
4
Hence M A21 = M C12 , so M A1 = M C1 = M B1 , which means M is the circumcenter of
(A1 B1 C1 ).

• If the circumcenter of (A1 B1 C1 ) lies on the circumcircle (ABC), WLOG assume

M , the circumcenter of (A1 B1 C1 ), lies on the arc BC
d that contains A of (ABC).

34
 We then have M B1 = M C1 , and also note that from the property of excircle, BC1 = CB1 .
It follows that 4M B1 C = 4M C1 B, so M B = M C, or, in other words, M is the midpoint
of arc BC.
d

Let H and K be the projections of M on AB and AC respectively. As AB > AC, note

that H lies inside the segment AB and K lies outside the segment AC. It’s easy to see
that 4M HB = 4M KC, so M H = M K and therefore AH = AK. In other words,

180◦ − A 180◦ − A
   
AB − AM · sin = AC + AM · sin ,
2 2
A
AB − AC = 2AM cos ,
 2 
C −B A
sin C − sin B = 2 sin cos ,
2 2

which means tan A2 = 1 and therefore ∠BAC = 90◦ . Thus, 4ABC is right-angled.

Solution to Exercise 7: IMO Shortlist 2005

Let 4ABC be an acute-angled triangle with AB 6= AC. Let H be the orthocenter of

4ABC, and let M be the midpoint of the side BC. Let D be a point on the side AB and
E a point on the side AC such that AE = AD and the points D, H, E are on the same
line. Prove that the line HM is perpendicular to the common chord of the circumscribed
circles of triangle 4ABC and triangle 4ADE.

Proof. Let O and O0 be the center of the circumcircles (ABC) and (ADE) respectively. Let I
and P be the midpoints of AH and the arc BC
d not containing A of (ABC) respectively. AP

intersects HM at N and (ADE) at N 0 . We will show that N ≡ N 0 .

Draw the diameter AA0 of (ABC). As BHCA0 is a parallelogram, H, M, A0 are collinear and
OM is the midsegment of 4AHA0 . Now observe that

AH = 2OM = 2R cos A and M P = OP − OM = R(1 − cos A).

Furthermore, since AH k M P ,

AN AH 2 cos A
= = . (14)
AP AH + M P 1 + cos A

Note that  
A A
sin ∠AHD = sin B + , sin ∠ADE = cos .
2 2

35
 A

K
I

O0

D
H O

N’ = N
B

M
C

A0
P

Figure 14: Prove that HM ⊥AK.

Applying the sine rule in 4ADH we get

 
1 A
AD = · AH sin B + . (15)
cos A2 2
As 4ADE is isosceles and AP is the angular bisector of A, AN 0 is the diameter of (ADE).
Hence AD = AN 0 cos A2 . Combining this with (15) we get
   
0 1 A 4R cos A A
AN = · AH sin B + = · sin B + . (16)
cos2 A2 2 1 + cos A 2
On the other hand,  
A
AP = 2R sin ∠ABP = 2R sin B + ,
2
so (14) yields  
2 cos A 4R cos A A
AN = · AP = · sin B + .
1 + cos A 1 + cos A 2
Combining this with (16) we get AN = AN 0 ; in other words, N ≡ N 0 . Hence H, N, A0 are
collinear. Also, O, O0 , I 0 are collinear and lie on the midsegment of 4AHA0 .
As a result, HM k OI, while we also have OI⊥AN , so HM ⊥AN .

Remark 7. A similar problem has appreared in Vietnam TST 2006: Consider 4ABC with
orthocenter H. The exterior angular bisector of ∠BHC intersects AB and AC at D and E

36
respectively. The angular bisector of ∠BAC interesects (ADE) at K 6= A. Prove that HK goes
through the midpoint of BC.

Solution to Exercise 8: IMO Shortlist 2011

Let ABC be a triangle with incentre I and circumcircle ω. Let D and E be the second
intersection points of ω with AI and BI, respectively. The chord DE meets AC at a point
F , and BC at a point G. Let P be the intersection point of the line through F parallel
to AD and the line through G parallel to BE. Suppose that the tangents to ω at A and
B meet at a point K. Prove that the three lines AE, BD and KP are either parallel or
concurrent.

A E
H

F
L

B
P

G C

Figure 15: Prove that AE, BD and KP are either parallel or concurrent.

Proof. As ∠IAF = ∠DAC = ∠BAD = ∠DEB = ∠F EI, we see that AIF E is cyclic. Sim-
ilarly, BDGI is cyclic. Denote ω1 as the circle passing through A, I, F, E and ω2 as the circle
passing through B, D, G, I.
If ω1 touches ω2 then denote d as the two circles’ common tangent line. Otherwise, let L be the
other intersection point of ω1 and ω2 besides I and denote d as the line IL.

37
d intersects (ABL) and (F GL) at K 0 , P 0 6= L respectively. Note that

(AB, BK 0 ) ≡ (AL, LK 0 ) ≡ (AL, LI) ≡ (AE, EI) ≡ (AE, EB) ≡ (AB, AB).

Therefore BK 0 ≡ BK. Similarly, AK 0 ≡ AK. It follows that K ≡ K 0 . On the other hand,

(P 0 F, F G) ≡ (P 0 L, LG) ≡ (IL, LG) ≡ (ID, DG) ≡ (AD, DE) ≡ (P F, F G).

Hence P 0 F ≡ P F and similarly P 0 G ≡ P G. As a result, P 0 ≡ P . Thus d passes through K

and P . Finally, as KP, AE, BD are [...] of ω1 and ω2 , ω and ω1 , ω and ω2 respectively, they are
either parallel or concurrent.

Solution to Exercise 9: All-Russian MO 2009

Consider a parallelogram ABCD and two points A1 , C1 on its sides AB, BC, respectively.
Lines AC1 and CA1 meet at P . Assume that the circumcircles of triangles AA1 P and CC1 P
intersect at the second point Q inside triangle ACD. Prove that ∠P DA = ∠QBA.

A1 B
A

C1

D
C

Figure 16: Prove that ∠P DA = ∠QBA.

Proof. Let E be the intersection of AD and CP , then

sin ∠ABP AA1 sin ∠ADP EP CD
= , · = .
sin ∠CBQ CC1 sin ∠CDP P C ED
In addition,
EP EA AD EA AA1 CD − AA1
= , =1− =1− = ,
PC CC1 ED ED CD CD

38
so

CD · AD CD · AD CD · AD
ED = = =
CD − AA1 AB − AA1 BA1
sin ∠ADP EA CD EA BA1
⇒ = · CD·AD = ·
sin ∠CDP CC1 BA1
BC CC1
sin ∠ADP AA1 BA1 AA1 sin ∠ABQ
⇒ = · = = . (17)
sin ∠CDP BA1 CC1 CC1 sin ∠CBQ

Let ∠ABC = ∠ADC = α and consider

sin(α − x)
f : (0, π) → R, f (x) = .
sin x

Note that f is a decreasing function on (0, π), and from (17) we see that f (∠P DA) = f (∠QBA),
so ∠P DA = ∠QBA.

Solution to Exercise 10: Balkan MO 2007

Let ABCD a convex quadrilateral with AB = BC = CD, with AC not equal to BD

and E be the intersection point of it’s diagonals. Prove that AE = DE if and only if
∠BAD + ∠ADC = 120.

B C

A D

Figure 17: Prove that AE = DE if and only if ∠BAD + ∠ADC = 120.

Proof. We will prove both directions:

39
• If AE = ED, then 4EAD is isoceles at E so ∠EAD = ∠EDA = β. Applying the sine
rule in 4ABD and 4ACD we get

AB AD CD AD
= , = .
sin β sin ∠ABD sin β sin ∠ACD

Also, AB = CD so sin ∠ABD = sin ∠ACD, or ∠ABD + ∠ACD = 180◦ .

Let ∠ACD = x and denote M as the intersection of AB and CD. We now have

∠M BD = ∠BAD + ∠BDC = ∠BAC + 2β,

which implies x − 2β = ∠BAC = ∠CBA. By similar arguments,

∠CDB = 180◦ − x − 2β.

On the other hand, ∠BEA = ∠EBC + ∠BCE, so ∠DBC = 4β − x. Combining this with
∠CBD = ∠CDB, we get the equation

180◦ − x − 2β = 4β − x,

which yields x = 30◦ , from which it’s easy to deduce that ∠BAD + ∠ADC = 120.

• If ∠BAD + ∠ADC = 120◦ , let ∠EAD = α and ∠EDA = β. It follows that α + β = 60◦ .
Applying the sine rule in 4ABD and 4ACD we get

sin α sin ∠ACD

= .
sin β sin ∠ABD

Let ∠CDB = ∠CBD = x, we can calculate

∠ACD = 180◦ − α − β − x = 120◦ − x.

Now note that ∠AEB = ∠EBC + ∠ECB, so

α + β − x = ∠ACB = ∠CAB,

which implies
∠ABD = 180◦ − 2α − 2β + x = 60◦ + x.

Finally, we see that

sin α sin ∠ACD 60◦ + x
= = = 1,
sin β sin ∠ABD 120◦ − x
which means sin α = sin β, so α = β. Consequently, AE = DE.

40
 Solution to Exercise 11: China TST 2008

Let ABC be a triangle, let AB > AC. Its incircle touches side BC at point E. Point
D is the second intersection of the incircle with segment AE (different from E). Point F
(different from E) is taken on segment AE such that CE = CF . The ray CF meets BD
at point G. Show that CF = F G.

M E H C
B

Figure 18: Prove that CF = F G.

Proof. Let I be the incenter of 4ABC and M be the midpoint of BC. Construct the heights
AH of 4ABC and CT of 4CEF .
EM EF
To prove that CF = F G, we will show that M F k BD or, equivalently, EB
= ED
. Note that
a+c−b
EM BE − BM 2
− a2 c−b
= = a+c−b
= .
EB EB 2
a+c−b
On the other hand,
2
2
a + b − c a2 + b 2 − c 2

2 2 2 4SABC
AE = AH + HE = + −
a2 2 2a
(a + b + c)(b + c − a)(c + a − b)(a + b − c) (c − b)2 (b + c − a)2
= + . (18)
4a2 4a2
(I) touches AB at K, then
 2
2 b+c−a
AD · AE = AK = . (19)
2

41
As ED · EA = AE 2 − AD · AE, from (18) and (19) we get
2
4S 2 (c − b)2 (b + c − a)2

b+c−a
ED · EA = 2 + −
a 4a2 2
2 2
4S (b + c − a) (c − b − a)(c − b + a)
= 2 +
a 4a2
(a + b − c)(b + c − a)(c + a − b)
= . (20)
2a
As AT HC is cyclic,

EF · EA = 2EI · EA = 2 · EH · EC
 2
a + c2 − b 2 a + c − b

a+b−c
=2· − ·
2a 2 2
(c − b)(b + c − a)(a + b − c)
= . (21)
2a
From (20) and (21), we get
EF c−b EM
= = ,
ED a+c−b EB
which yields CF = F G.

Solution to Exercise 12: Balkan MO 2013

In 4ABC, the excircle ωa opposite A touches AB at P and AC at Q, while the excircle ωb

opposite B touches BA at M and BC at N . Let K be the projection of C onto M N and
let L be the projection of C onto P Q. Show that the quadrilateral M KLP is cyclic.

Proof. Let Ib , Ic be the excenters opposite B and C in 4ABC respectively. Let ωc be the excircle
opposite to C. ωc touches BC at T . S is the intersection of M N and P Q.
Let X, Y be the projections of M on P Q and P on M N respectively. AX intersects P Y at H.
HC intersects M N at S1 and P Q at S2 .
Observe that
b+c−a B a+c−b A
CK = · cos , CL = · cos .
2 2 2 2
On the other hand, HA · HX = HP · HY so
YH MH sin B2
= = .
XH PH sin A2
In addition,
A b+c−a B c+a−b
tan = , tan = ,
2 2T Ic 2 2T Ic
which yields
B A
(a + c − b) · tan = (b + c − a) · tan ,
2 2
42
 M

Ic
A
Ib

K
H

B C N

P Q
S = S1 = S2
X L

Figure 19: Prove that M KLP is cyclic.

which implies that

CK (b + c − a) · cos B2 sin B2 YH
= A
= A
= .
CL (a + c − b) cos 2 sin 2 XH
Now, as CK k Y H and CL k XH,

S1 C CK CL S2 C
= = = ,
S1 H YH XH S2 H

so S1 ≡ S2 and therefore H, C, S are collinear.

As ∠SKL = 900 − ∠CKL and ∠SY X = 900 − ∠HY X, ∠SKL = ∠HY X, so KL k XY . It
follows that ∠Y XP = ∠KLP , while we also have ∠Y XP = 180◦ − ∠P M Y . Thus ∠KLP =
180◦ − ∠P M Y and so M P LK is cyclic.

43
 Solution to Exercise 13: IMO Shortlist 2005

Given 4ABC satisfying AB + BC = 3CA. The incircle of 4ABC has center I and touches
the sides AB and BC at the points D and E, respectively. Let K and L be the reflections
of the points D and E with respect to I. Prove that the points A, C, K, L lie on one circle.

L
D

C E F B

Figure 20: Prove that A, C, K, L lie on one circle.

Proof. We have c + a = 3b. Note that if b = c then a = 2b = b + c, which contradicts the triangle
inequality. Hence b 6= c, and WLOG assume b < c.
−−→
AL intersects BC at F . Assume the positive direction is that of vector BC, then

a+c−b a+b−c
EF = CF − CE = −BE − CE = − = c − b,
2 2

so EF = |EF | = c − b. Now 4LF E and 4CLE have

LE 2r LE 4r
tan ∠LF C = = , tan ∠LCE = = ,
EF c−b CE a+c−b

so
tan ∠LF C + tan ∠LCE
tan ∠ALC = tan(∠LF C + ∠LCE) =
  1 − tan ∠LF C · tan ∠LCE
8r2
 
2r 4r
= + : 1−
c−b a+b−c (c − b)(a + b − c)
2r(a + c − b)
= .
(c − b)(a + b − c) − 8r2

44
As a + c = 3b, we can compute the inradius r of 4ABC from

S2 (p − a)(p − b)(p − c) (b + c − a)(a + b − c)

r2 = = = .
p2 p 8

It then follows that

2r(a + c − b)
tan ∠ALC =
(c − b)(a + b − c) − (b + c − a)(a + b − c)
2r(a + c − b) 2br
= = .
(a + b − c)(a − 2b) (2b − c)(b − c)

By similar arguments,
2br
tan ∠AKC = .
(2b − a)(b − a)
From a + c = 3b it is easy to see that (2b − c)(b − c) = (2b − a)(b − a). Thus tan ∠ALC =
tan ∠AKC. In other words, ∠ALC = ∠AKC, so ALKC is cyclic.

45