Preface

This book showcases the synthetic problem-solving methods which fre-

quently appear in modern day Olympiad geometry, in the way we believe they
should be taught to someone with little familiarity in the subject. In some
sense, the text also represents an unofficial sequel to the recent problem col-
lection published by XYZ Press, 110 Geometry Problems for the International
Mathematical Olympiad, written by the first and third authors; but, the two
books can be studied completely independently of each other.
Lemmas in Olympiad Geometry is a project that started in the summer
of 2011, when the third author first taught the Geometric Proofs course at
the AwesomeMath Summer Camp. Some brief lecture notes were written
back then (with the intention of getting expanded), but nothing substantial
happened until last summer, when the second author came to the Cornell
camp as a teaching assistant for the same course. Ever since, we have all been
working together to make the current version of the manuscript possible, and
are excited to announce that it is ready.
The work is designed as a medley of the important Lemmas in classical
geometry in a relatively linear fashion: gradually starting from Power of a
Point and common results to more sophisticated topics, where knowing a lot
of techniques can prove to be tremendously useful. We treated each chapter as
a short story of its own and included numerous solved exercises with detailed
explanations and related insights that will hopefully make your journey very
enjoyable. Each chapter is also accompanied by a short list of problems that
we have carefully selected. These are problems that we have solved ourselves
on our own at some point, and so we are convinced that you are going to
appreciate them as well. The last chapter on three dimensional geometry is
the only chapter which is not followed by such a list of problems, since we
considered it as a bonus section, yet one that has beautiful problems which
are also relevant in other subdomains of geometry.
We wish you a pleasant reading and hope that you will enjoy Lemmas in
Olympiad Geometry as much as we enjoyed writing it.

The authors
Contents

Preface v

1 Power of a Point 1

2 Carnot and Radical Axes 15

3 Ceva, Trig Ceva, Quadrilateral Ceva 37

4 Menelaus’ Theorem 57

5 Desargues and Pascal 71

6 Jacobi’s Theorem 85

7 Isogonal Conjugates and Pedal Triangles 95

8 Simson and Steiner 111

9 Symmedians 129

10 Harmonic Divisions 149

11 Appendix A: Some Generalizations of Blanchet’s Theorem 165

12 Poles and Polars 173

13 Appendix B: An Incircle Related Perpendicularity 189

14 Homothety 199

15 Inversion 211

16 The Monge-D’Alembert Circle Theorem 223

viii

17 Mixtilinear and Curvilinear Incircles 233

18 Ptolemy and Casey 247

19 Complete Quadrilaterals 261

20 Apollonian Circles and Isodynamic Points 275

21 The Erdös-Mordell Inequality 287

22 Sondat’s Theorem and the Neuberg Cubic 301

23 Introduction to Complex Numbers 323

24 Complex Numbers in Olympiad Geometry 341

25 3D Geometry 361

Bibliography 369
Chapter 1

Power of a Point

One of the most important tools in Olympiad Geometry is the so-called

Power of Point Theorem, our first Lemma.

Theorem 1.1. Let Γ be a circle, and P a point. Let a line through P meet
Γ at points A and B, and let another line through P meet Γ at points C and
D. Then
P A · P B = P C · P D.

We announce the reader that we will be labeling our Lemmas as Theorems

not to follow any convention, but rather to emphasize their importance, since
after all they represent the main stars of our show.
C
A

A C
P
B
D
D

B
P

Proof. Of course, there are two configurations to consider here, depending

on whether P lies inside the circle or outside the circle. In the case when P
lies inside the circle, we have ∠P AD = ∠P CB and ∠AP D = ∠CP B, so that
triangles P AD and P CB are similar; hence
PA PC
= .
PD PB
2 Chapter 1

Rearranging then yields P A · P B = P C · P D.

When P lies outside the circle, we again have ∠P AD = ∠P CB and

∠AP D = ∠CP B, so again triangles P AD and P CB are similar. We get
the same result in this case. 

As a very important special case, when P lies outside the circle and P C is
tangent to the circle, we have that
P A · P B = P C 2.

Conversely, the above represents a very useful criterion for proving concy-
clities.

Theorem 1.2. Let A, B, C, D be four distinct points. Let the lines AB and
CD intersect at P . Assume that either P lies on both line segments AB and
CD, or P lies on neither line segment. Then A, B, C, D are concyclic if and
only if P A · P B = P C · P D.

Proof. Going backwards, the relation P A · P B = P C · P D is equivalent to

PA PC
= ,
PD PB
which combined with ∠AP D = ∠CP B (which holds in both configurations
described above) yields that triangles AP D and CP B are similar. Thus, we
get that ∠P AD = ∠P CB, which in both cases implies that A, B, C, D are
concyclic. 

This tells us that no matter what chord XY we take through P (with X,

Y on the circle), the value P X · P Y is constant. This constant is called the
power of P with respect to the circle considered. In particular, if Γ(O, R) is
the circle with center O and radius R, then if we consider the chord XY that
passes through the center O (i.e. we choose the diameter of the circle passing
through P ), we get that
P X · P Y = kOP 2 − R2 k

//We say that the points lying on the circle Γ have zero power with respect
to Γ!

We emphasize this interplay between products and differences of squares

with the following exercise.
Power of a Point 3

Delta 1.1. (IMO 2011 Shortlist) Let A1 A2 A3 A4 be a non-cyclic quadrilateral.

Let O1 and r1 be the circumcenter and the circumradius of triangle A2 A3 A4 .
Define O2 , O3 , O4 and r2 , r3 , r4 in a similar way. Prove that
1 1 1 1
+ + + = 0.
O1 A21 − r12 O2 A22 − r22 O3 A23 − r32 O4 A24 − r42

A2 A3

B1

A1

A4

Proof. Let M be the point of intersection of the diagonals A1 A3 and

A2 A4 . On each diagonal choose a direction and let x, y, z, and w be the
signed distances from M to the points A1 , A2 , A3 , A4 , respectively. Let ω1 be
the circumcircle of triangle A2 A3 A4 and let B1 be the second intersection of
ω1 and A1 A3 (thus, B1 = A3 if and only if A1 A3 is tangent to ω1 ). Since the
expression O1 A21 − r12 is the power of the point A1 with respect to ω1 , we get

O1 A21 − r12 = A1 B1 · A1 A3 .

On the other hand, from the equality M B1 · M A3 = M A2 · M A4 , we obtain

yw
M B1 = .
z
Hence, it follows that
 yw  z−x
O1 A21 − r12 = − x (z − x) = (yw − xz).
z z
Doing the same thing for the other three expressions, we then get that
4  
X 1 1 z w x y
= − + − = 0,
i=1
Oi A2i − ri2 yw − xz z−x w−y x−z y−w
4 Chapter 1

as claimed. This completes the proof. 

By the way, this will not be the only time we will make use of signed
distances in this material. Usually, we can assume without loss of generality a
certain position of the points in our diagram - however, in problems involving
lots of circles, the computations involving the Power of Point Theorem are not
the same for all configurations; hence, we often need to take extra care when
dealing with signs.

Some warm-up problems now! We begin with another simple interplay

between the two formulas for the power of a point.

Delta 1.2. (Euler’s Theorem) In a triangle ABC with circumcenter O, incenter

I, circumradius R, and inradius r, prove that

OI 2 = R(R − 2r).

B C

Proof. Let AI meet the circumcircle again at D. In this case, the Power
of Point Theorem applied for I yields

IA · ID = R2 − OI 2 .
r
Thus, we would like to show that IA · ID = 2Rr. First, note that IA = sin A
2
(draw the perpendicular from I to AB and apply the Law of Sines in the right
Power of a Point 5

triangle that you obtain). Next, note that

∠BID = ∠BAD + ∠ABI = ∠DAC + ∠IBC = ∠DBC + ∠IBC = ∠IBD;

hence ID = BD = 2R sin A2 , where the last equality comes from the (extended)
Law of Sines in triangle ABD. Hence, we get that
r A
IA · ID = A
· 2R sin = 2Rr,
sin 2 2

as desired. This completes the proof. 

Note that for any given point P in plane, the above method can be extended
to generate an identity for OP 2 .

Delta 1.3. Let ABC be an acute-angled triangle and let D be the foot of
the A-altitude. Let H be a point on the segment AD. Prove that H is the
orthocenter of triangle ABC if and only if DB · DC = AD · HD.
A

B C
D

A0

Proof. Let A0 be the second intersection of the line AD with the circum-
circle of triangle ABC. We know that A0 is the reflection of the orthocenter
across BC (if not, try angle chasing). Thus, if H is the orthocenter of ABC,
then the computing power of D with respect to the circumcircle gives us

DB · DC = AD · DA0 = AD · HD,

as desired. Conversely, we have that DB ·DC = AD ·HD and also DB ·DC =

AD · HA0 (the power of D with respect to the circumcircle); thus HD = HA0 ,
and so H needs to be the orthocenter of ABC, as claimed. 
6 Chapter 1

Although very simple, this proves to be a very useful criterion for showing
that a point lying on an altitude of a triangle is the orthocenter. Let’s see a
couple of problems where this may come in handy.

Delta 1.4. (USA TSTST 2012) In scalene triangle ABC, let the feet of the
perpendiculars from A to BC, B to CA, C to AB be A1 , B1 , C1 , respectively.
Denote by A2 the intersection of lines BC and B1 C1 . Define B2 and C2
analogously. Let D, E, F be the respective midpoints of sides BC, CA, AB.
Show that the perpendiculars from D to AA2 , E to BB2 and F to CC2 are
concurrent.
A

C1
A3

B1

B D A1 C A2

Proof. Let H be the orthocenter of triangle ABC. We claim that H is the

desired point of concurrency. Let A3 be the foot of perpendicular from D to
line AA2 . Since AA1 ⊥ BC and DA3 ⊥ AA2 , quadrilateral A3 A1 DA is cyclic.
By Power of a Point, we have A2 C1 ·A2 B1 = A2 A3 ·A2 A. Again, by Power of a
Point (this time with respect to the nine point circle of triangle ABC) A2 A1 ·
A2 D = A2 C1 ·A2 B1 , so combining these equations, A2 C1 ·A2 B1 = A2 A3 ·A2 A,
implying quadrilateral A3 C1 B1 A is cyclic by Theorem 1.2. But H lies on
the circumcircle of this quadrilateral, since HC1 ⊥ AB and HB1 ⊥ AC. It
follows that ∠HA3 A = 180◦ − ∠HB1 A = 90◦ , so points D, H, A3 are collinear.
Defining B3 and C3 analogously, similar arguments show that points E, H, B3
and F, H, C3 are also collinear, so the lines in the problem are concurrent at
H as claimed. 

Delta 1.5. (IMO Shortlist 1998) Let I be the incenter of triangle ABC. Let
K, L and M be the points of tangency of the incircle of triangle ABC with
Power of a Point 7

sides AB, BC, and CA, respectively. The line ` passes through B and is
parallel to KL. The lines M K and M L intersect ` at the points R and S
respectively. Prove that ∠RIS is acute.
A

R
K

M
I
X

B L C

Proof. First note that

∠KRB = ∠M KL = ∠M LC = ∠SLB

and
∠RKB = ∠AKM = ∠KLM = ∠LSB
Thus, triangle BKS is similar to triangle BRL. This means that BS · BR =
BL2 . Now let X be the midpoint of segment KL. We have that X lies on the
altitude from I to RS and also that BX = BL cos B2 and BI = cosBL
B which
2
means that BX · BI = BR · BS. Hence, by Delta 1.3, X is the orthocenter
of triangle RIS. But since X is the projection of I onto line KL it’s clear
that X lies inside of triangle RIS which implies that this triangle is acute as
desired. 

//Another way to prove that X is the orthocenter of triangle RIS is to

prove that triangle RXS is self-polar with respect to the incircle of triangle
ABC.
8 Chapter 1

We continue with a computational problem from the USA Mathematical

Olympiad from 1998.

Delta 1.6. (USAMO 1998) Let C1 and C2 be concentric circles, with C2 in the
interior of C1 . From a point A on C1 one draws the tangent AB to C2 (B ∈ C2 ).
Let C be the second point of intersection of AB with C1 , and let D be the
midpoint of AB. A line passing through A intersects C2 at E and F in such
a way that the perpendicular bisectors of DE and CF intersect at a point M
AM
on AB. Find, with proof, the ratio M C.

M B D
C A
E

Proof. Let O be the common center of the concentric circles C1 , C2 . The

tangency point B is the midpoint of the chord AC, because AC is perpendic-
ular to the radius OB of the circle C2 , and O is also the center of the circle
C1 . The power of the point A with respect to circle C2 is AE · AF = AB 2 .
But since B is the midpoint of AC and D the midpoint of AB, we have
that AD · AC = AB 2
2 · 2AB = AB as well. Hence, by Theorem 1.2,
quadrilateral CDEF is cyclic. The intersection M of the perpendicular
bisectors of its diagonals CE, DF is its circumcenter. If this circumcen-
ter is to be on its side CD, it must be the midpoint of this side, hence
DM = M C = DC 3
2 . Since DC = 2 AB, we now have DM = M C = 4 AB
3

and AM = AD + DM = AB 3 5 AM
2 + 4 AB = 4 AB and so M C = 3 .
5


We continue with a beautiful IMO problem, where Power of Point can be

used in a surprising way.

Delta 1.7. (IMO 2009) Let ABC be a triangle with circumcenter O. The
points P and Q are interior points of the sides CA and AB respectively. Let
Power of a Point 9

K, L and M be the midpoints of the segments BP, CQ and P Q. respectively,

and let Γ be the circle passing through K, L and M . Suppose that the line
P Q is tangent to the circle Γ. Prove that OP = OQ.
A

L P

B C

Proof. Since line P Q is tangent to Γ, we have that ∠QM K = ∠M LK.

Since M K is the P -midline of triangle P QB we have that M K k AB so
∠QM K = ∠AQM . Hence, ∠AQP = ∠M LK. Similarly we get that
∠M KL = ∠AP Q, so triangles M KL and AP Q are similar. Therefore
AQ AP AP AQ
= =⇒ = =⇒ AP · P C = AQ · BQ.
ML MK BQ PC
Thus, P and Q have the same power with respect to the circumcircle of triangle
ABC, so OP = OQ as desired. 

We end this section with a cute result due to Hiroshi Haruki (according to
[18]).

Delta 1.8. (Haruki’s Lemma) Given two non-intersecting chords AB and CD

in a circle and a variable point P on the arc AB remote from points C and
D, let E and F be the intersections of chords P C, AB, and of P D, AB,
respectively. Prove that the value of
AE · BF
EF
does not depend on the position of P .
10 Chapter 1

A E F B
G

C D

Proof. The proof relies on the fact that the angle ∠CP D is constant. We
begin by constructing the circumcircle of triangle P ED. Define point G to be
the intersection of this circle with the line AB. Note that ∠EGD = ∠EP D
as they are subtended by the same chord ED of the circumcircle of triangle
P ED; these angles remain constant as P varies on the arc AB. Hence, for all
positions of P , ∠EGD remains fixed and, therefore, point G remains fixed on
the line AB. It follows that BG is constant. On the other hand, by Power of
Point, we have that AF · F B = P F · F D and EF · F G = P F · F D. Hence,

(AE + EF ) · F B = EF · (F B + BG),

and AE · F B = EF · BG. Therefore, we conclude that

AE · BF
= BG,
EF

a constant. 

Haruki’s Lemma can be used to give a very short proof of the so-called
Butterfly Theorem, a very popular result in projective geometry.

Delta 1.9. (Butterfly Theorem). Let M be the midpoint of chord P Q of a

given circle, through which two other chords AB and CD are drawn; AD cuts
P Q at X and BC cuts P Q at Y . Then, M is also the midpoint of XY .
Power of a Point 11

A C

M Y
X Q
P

D
B

Proof. We think of A and C as being two positions of the variable point

traversing the circle. Then, Haruki’s lemma tells us that
XP · M Q MP · Y Q
= ,
XM YM
which, because of M P = M Q, is simplified to
XP YQ
= .
XM YM
Adding 1 to both sides gives
XP + XM YQ+YM
= .
XM YM
Applying M P = M Q again, we obtain the required XM = Y M . This com-
pletes the proof. 
12 Chapter 1

Assigned Problems

Epsilon 1.1. Let ABC be an acute triangle. Let the line through B perpen-
dicular to AC meet the circle with diameter AC at points P and Q, and let
the line through C perpendicular to AB meet the circle with diameter AB at
points R and S. Prove that P , Q, R, S are concyclic.

Epsilon 1.2. Let ABC be an acute-angled triangle with circumcenter O and

orthocenter H. Prove that

OH 2 = R2 (1 − 8 cos A cos B cos C).

Epsilon 1.3. Let ABC be a triangle and let D, E, F be the feet of the altitudes,
with D on BC, E on CA, and F on AB. Let the parallel through D to EF
meet AB at X and AC at Y . Let T be the intersection of EF with BC and
let M be the midpoint of side BC. Prove that the points T , M , X, Y are
concyclic.

Epsilon 1.4. (Kazakhstan MO 2008) Suppose that B1 is the midpoint of the arc
AC, containing B, of the circumcircle of triangle ABC, and let Ib be the B-
excircle’s center. Assume that the external angle bisector of ∠ABC intersects
AC at B2 . Prove that B2 I is perpendicular to B1 IB , where I is the incenter
of ABC.

Epsilon 1.5. (IMO 2000) Two circles Γ1 and Γ2 intersect at M and N . Let `
be the common tangent to Γ1 and Γ2 so that M is closer to ` than N is. Let
` touch Γ1 at A and Γ2 at B. Let the line through M parallel to ` meet the
circle Γ1 again at C and the circle Γ2 again at D. Lines CA and DB meet
at E; lines AN and CD meet at P ; lines BN and CD meet at Q. Show that
EP = EQ.

Epsilon 1.6. Let C be a point on a semicircle Γ of diameter AB and let D

be the midpoint of the arc AC. Let E be the projection of D onto the line
BC and F the intersection of the line AE with the semicircle. Prove that BF
bisects the line segment DE.

Epsilon 1.7. Let A, B, C be three points on a circle Γ with AB = BC. Let

the tangents at A and B meet at D. Let DC meet Γ again at E. Prove that
the line AE bisects the segment BD.

Epsilon 1.8. (EGMO 2012) Let ABC be a triangle with circumcenter O. The
points D, E, F lie in the interiors of the sides BC, CA, AB respectively, such
that DE is perpendicular to CO and DF is perpendicular to BO. (By interior
Power of a Point 13

we mean, for example, that the point D lies on the line BC and D is between
B and C on that line.) Let K be the circumcenter of triangle AF E. Prove
that the lines DK and BC are perpendicular.

Epsilon 1.9. (IMO Shortlist 2013) Let ABC be a triangle with ∠B > ∠C. Let
P and Q be two different points on line AC such that ∠P BA = ∠QBA =
∠ACB and A is located between P and C. Suppose that there exists an
interior point D of segment BQ for which P D = P B. Let the ray AD intersect
the circumcircle of triangle ABC at R 6= A. Prove that QB = QR.