BCS Question Bank

AHSAN
August 14, 2020
CONTENTS

Contents

Contents 2

1 Thoughts on PSolving, a note to thyself 2

1.1 Be DUMB, Keep it SIMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Combinatorics 3
2.1 Binomial Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Extremal Set Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Algorithmic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 Data Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.2 Minimal Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.3 Shortest Path Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.4 Other CP Tricks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.5 Fast Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.7 Algorithm Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.8 Covering Area with Squares . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4 Graph Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.1 Counting in Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.2 Algorithms in Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.5 Game Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.5.1 Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.5.2 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.6 Combinatorial Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.6.1 Chessboard Pieces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.7 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.7.1 Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.7.2 Generating Function Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.7.3 Sequence Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.7.4 Recurrent Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.8 Exploring Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
CONTENTS 3

2.8.1 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.8.1.1 Conway’s Soldiers . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.8.1.2 Triominos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.8.1.3 Dominos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.8.2 Coloring Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
2.9 Linear Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.10 Double Counting and Other Algebraic Methods . . . . . . . . . . . . . . . . . . . 95
2.10.1 Probabilistic Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.11 Bounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2.12 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
2.13 Local . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
2.14 CP Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
2.14.1 Cycle Finding Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
2.15 Permutations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

3 Algebra 123
3.1 Functional Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.1.1 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.2 Weird Ones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.3 FE cantonmathguy Seclected Problems . . . . . . . . . . . . . . . . . . . . . . . 132
3.4 Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.4.1 Techniques to remember . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.4.2 General Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.4.3 Root Hunting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3.4.4 NT Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3.4.5 Fourier Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.4.6 Irreducibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.5 Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
3.5.1 Basic Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
3.5.2 Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
3.5.3 Tricks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
3.5.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.5.4.1 Smoothing And Convexity . . . . . . . . . . . . . . . . . . . . . 160
3.6 Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
3.7 Ad-Hocs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
3.7.1 Factorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
3.7.2 Bounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
3.7.3 Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
3.8 Tricks and Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
3.8.1 Ad Hocs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

4 Geometry 171
4.1 First Portion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.2 Second Portion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
4.3 orthocenter–circumcircle–ninepoint circle . . . . . . . . . . . . . . . . . . . . . . . 181
CONTENTS 4

4.3.1 some figures, might be good for something, i dunno . . . . . . . . . . . . . 181

4.3.2 problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4.3.3 The line parallel to BC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
4.3.4 Simson Line and Stuffs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
4.3.5 Euler Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
4.4 Cevian and Circumcevian Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . 207
4.4.1 Circumcevian Triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
4.4.2 Cevian Triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
4.5 Incenter–Excenter Lemma stuff . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
4.6 Conjugates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
4.6.1 Isogonal Conjugate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
4.6.1.1 Symmedians . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
4.6.2 Isotonic Conjugate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
4.6.3 Reflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
4.7 Mixtilinear–Curvilinear–Normal In-Excricles . . . . . . . . . . . . . . . . . . . . . . 237
4.8 Circles and Radical Axises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
4.9 Complete Quadrilateral + Spiral Similarity . . . . . . . . . . . . . . . . . . . . . . 250
4.10 Projective Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
4.10.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
4.10.2 Cross Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
4.10.3 Involution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
4.10.4 Inversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
4.10.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
4.10.6 Projective Constructions . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
4.11 Parallelogram Stuff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
4.12 Length Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
4.13 Pedal Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
4.14 Pending Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
4.15 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
4.16 Research Stuffs for later . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
4.17 Big Pictures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
4.17.1 Parallel lines to the A-median . . . . . . . . . . . . . . . . . . . . . . . . . 306

5 Number Theory 309

5.1 Tricks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
5.1.1 Digit Sum or Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
5.1.2 Diophantine Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
5.1.3 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
5.1.4 NT Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
5.1.5 Construction Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
5.1.6 Sets satisfying certain properties . . . . . . . . . . . . . . . . . . . . . . . 311
5.1.7 Other Small Techniques to Remember . . . . . . . . . . . . . . . . . . . . 311
5.2 Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
5.2.1 Modular Arithmatic Theorems and Useful Results . . . . . . . . . . . . . . 315
5.3 Orders Modulo a Prime and Related Stuffs . . . . . . . . . . . . . . . . . . . . . . 316
CONTENTS 1

5.3.1 Cyclotomic Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316

5.3.2 Quadratic Residue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
5.3.2.1 Quadratic Residue . . . . . . . . . . . . . . . . . . . . . . . . . . 320
5.3.3 Zsigmondy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
5.3.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
5.4 Primes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
5.5 NT Functions and Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
5.6 Diophantine Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
5.7 Divisibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
5.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331

Problems 345

Theorems 352

Categories 354

Strategies 355

Misc 356
Chapter 1

Thoughts on PSolving, a note to thyself

1.1 Be DUMB, Keep it SIMPLE

Remember what Paul-lord said? Think wishfully, make dumb wishes. When first approaching the
problem, you can do whatever you want. You can loosen some constraints, you can add some new.
This works exceptionally well when you need to build an object from one given object, you can do
whatever you want. Putting additional constraints decreases the number of test cases. Sometimes
loosening some constraints help to give better observation of the problem.

Like in ISL 2016 N5, after deciding that we are going to build a pair (x2 , y2 ) from the previous pair
(x1 , y1 ), we should look for the most innocent looking relation between these four variable. Now it’s
time to play around, try dumb things. Rewriting the equation, we want to use the fact that x1 , x2
√
have to be on different sides of a. How can we insert this constraint in our equation in the most
simple and natural way? This is where we need to be dumb, and amature.

In Problem 5.8 the trick is to keep things simple. Making the most natural assumptions. In
construction problems, think of how the result can be achieved in the most natural way. Can we
make some extra assumptions that might result in the immediate proof the result’s existence?
Chapter 2

Combinatorics
2.1 Binomial Identities 4

2.1 Binomial Identities

Theorem 2.1.1 (Vandermonde’s identity) — For every positive integer m, n, k

  Xk   
m+n m n
=
k i k −i
i=0

   2  2  2
2n n n n
= + ··· +
n 0 1 n

Problem 2.1.1 (Somewhere) : Let n and m be positive integers with n < m. Prove that

n  
X
n−k 1 n 1
(−1) = m


m−k k (n + 1) n+1
k=0

Solution [MellowMelon]. The left side is the nth finite difference of the unique nth degree
polynomial satisfying P (x) = m−x 1
for x = 0, 1, . . . , n. This means that if an is the x n coefficient
of P , then the left side is equal to n! · an . Setting this equal to the right side, it suffices to show
that
1 1
an = m

= .
(n + 1)! n+1 m(m − 1) · · · (m − n)
Let Q(x) = (x − m)P (x) + 1. Then Q is degree n + 1, has roots 0, 1, . . . , n, and has leading
coefficient an . So
Q(x) = an x(x − 1)(x − 2) · · · (x − n).
We also know Q(m) = 1. Plugging x = m into the above equation and isolating an gives us
exactly what we need to show.

Remark. The idea to express the left hand side with finite differences is pretty simple. The crux
idea is to finding a Q that works. Since we have to show an (m) (m − 1) . . . (m − n) = 1, which
looks like a polynomial with leading coefficient an with roots 0, 1, 2, . . . n and Q(m) = 1. How can
we get that?
The takeaway is, if the problem requires you to find a polynomial that almost looks like the one
given in the problem, try to modify the given polynomial in some way first.
In this particular case was to modify P bu using the fact (x − m)P (x) = −1.
2.1 Binomial Identities 5

Solution [Hydroxide]. Let P be the unique polynomial of degree at most n such that P (0) =
P (1) = · · · = P (n) = 1. Obviously P (x) = 1, but we can also find P from Lagrange interpolation.
This gives us the equality of polynomials
n
X (x − 0) · · · (x − (k − 1))(x − (k + 1)) · · · (x − n)
= 1.
(k − 0) · · · (k − (k − 1))(k − (k + 1)) · · · (k − n)
k=0

Plugging in x = m and doing a ton of rearranging gives the desired result.

Remark. Both solutions can easily be extended to prove the more general result that if 0 ≤ p <
n < m are integers, then

n
mp
 
X
n−k 1 n p
(−1) k = m


m−k k (n + 1) n+1
k=0

Problem 2.1.2 (USA TST 2010 P8) : Let m, n be positive integers with m ≥ n, and let S be the
set of all n-term sequences of positive integers (a1 , a2 , . . . an ) such that a1 + a2 + · · · + an = m.
Show that

       
X
a1 a2 an n m n m n−2 n m n−1 n
1 2 ···n = n − (n − 1) + · · · + (−1) 2 + (−1)
n n−1 2 1
S

Solution [Combinatorial, MellowMelon]. Look at the right side, try to translate it to

combinatorial model. If we had used the inclusion-exclusion method, the right side would be the
number of ways to color m balls with n colors, with each color appearing at least once. Now our
remaining job is to prove the same for the left side.

Trying to interpret the left hand side, we first notice that we need to partition the balls in n parts
with sizes a1 , a2 , . . . an where the first partition will have 1 color to choose from, the second one
will have 2 and so on. We need to refine this idea for using it.

Solution [Generating Function, MellowMelon]. The first step of a gf solution is to decide on

a gf solution. Now we can think of the left hand side as a function of m, namely bm . So we can
make a generating function with it.
2.1 Binomial Identities 6

We let Fn (x) = b0 + b1 x + b2 x 2 + · · · and so we get:

Fn (x) = (x + x 2 + · · · )(2x + 4x 2 + · · · ) · · · (nx + n2 x 2 + · · · ),

x 2x nx
= ··· ,
1 − x 1 − 2x 1 − nx
n!x n
= .
(1 − x)(1 − 2x) · · · (1 − nx)

Now we need a generating function for the right hand side as well. The right hand side looks the
nth finite difference of the function m g0 (x) = x m , that is
       
n m n m n−2 n m n−1 n
m gn (0) = n − (n − 1) + · · · + (−1) 2 + (−1)
n n−1 2 1

Now we can make a generating function with this:

0 1
G(y ) = 0 gn (0)y + 1 gn (0)y ...

But this isn’t very useful as we are getting the nth finite differences on the right side, where we
want to work with the original polynomials.

So what we do is, we unravel the m gn (0) as original polynomials and get:

G(x, y ) = x 0 y 0 + x 1 y 1 + x 2 + y 2 . . .

Where our desired generating function for n is:

0 1
Gn (0, y ) = 0 gn (0)y + 1 gn (0)y + ...

And we want to show that

n!x n
Gn (0, y ) =
(1 − x)(1 − 2x) · · · (1 − nx)

Now we notice that we can actually use induction on n since we definded

Gn+1 (x, y ) = Gn (x + 1, y ) − Gn (x, y )

And so we are done.

Lemma 2.1.2 —
∞  
X n+k n 1
G(x) = x =
k (1 − x)k+1
n=0
2.1 Binomial Identities 7

Proof. We have the above relation because n+k can be interpret as the number of ways write


k
n as the sum of k + 1 numbers. And the generating function for that is:

G(x) = (1 + x + x 2 . . . )k+1

Problem 2.1.3 (Binom1) : Solve for positive integer n

       
2n 2n − 1 2n − 2 n n
− + − · · · + (−1) =?
0 1 2 n

P∞
Solution [Generating funtion, Tintarn]. Denote this sum an . Denote F (x) = n=0 an x
2n .

∞ ∞  
X X
k 2n − k 2n
F (x) = (−1) x
k
k=0 n=0
∞ ∞  
X
2 k
X 2n + k 2n
= (−x ) x .
k
k=0 n=0

Now, using the well-known power series

∞  
X n+k n 1
G(x) = x =
k (1 − x)k+1
n=0

we find
∞  
X 2n + k 2n G(x) + G(−x)
x =
k 2
n=0

and hence
∞ k ∞ k
x2 −x 2
 
1 X 1 X
F (x) = +
2(1 − x) x −1 2(1 + x) x +1
k=0 k=0

1 1
= +
2(x 2 − x + 1) 2(x 2 + x + 1)
x2 + 1
=
x4 + x2 + 1
1 − x4
=
1 − x6
= (1 − x 4 )(1 + x 6 + x 12 + x 18 + . . . )
= 1 − x 4 + x 6 − x 10 + x 12 − x 16 + . . . .
2.1 Binomial Identities 8

and hence a3k = 1, a3k+1 = 0, a3k+2 = −1 .

Solution [Combinaorial Model, MellowMelon]. Consider the number of ways to tile a 1 × 2n

rectangle with squares and dominoes. Let E be the set of ways to do it with an even number of
dominoes, and let O be the same for odd. We want to find |E| − |O|.

Trying some basic cases, we see that the only values we get are −1, 0, 1. So we want to find a
pairing between the two sets that is almost a bijection, but doesn’t work for only one element.
And that bijection is:

Consider the first consecutive pair of squares or domino which either comes at the very start or
comes right after another domino. Swap the pair of squares to a domino or vice versa.

Problem 2.1.4 (Binom 2) :

 2  2  2  2
n n n n n
− + − · · · + (−1) =?
0 1 2 n

Solution [Generating function, TinTarn]. Denote

∞   
X nk n
f (m, n) = (−1)
k m−k
k=0

We are looking for f (n, n). We have

∞ ∞  X ∞  
X
m
X n k n
f (m, n)x = (−1) xm
k m−k
m=0 k=0 m=0
"∞  #" ∞  #
X n X n
= (−x)k xm
k m
k=0 m=0
n n
= (1 − x) (1 + x)
= (1 − x 2 )n
∞  
t n
X
= (−1) x 2t .
t
t=0

Hence we see that for odd m we have f (m, n) = 0 and hence for odd n we have f (n, n) = 0 .
 
2k
Now, for n = 2k even, we find f (2k, 2k) = (−1) . k
k
2.1 Binomial Identities 9

Solution [Combinatorial, MellowMelon]. Consider the number of ways to paint the squares of
a 2 × n rectangle red and blue such that both rows have the same number of red squares. Let E
be the set of ways with an even number of red squares in each row, and let O be the same for
odd. We want to find |E| − |O|.

Pair the selections as follows: take the first column with both squares the same color, and flip
both colors. This maps elements in E to O and vice versa and is invertible, so except cases where
the pairing is undefined, E and O have the same number of elements.
2.2 Sets 10

2.2 Sets

2.2.1 Lemmas

Lemma 2.2.1 — Let S be a set with n elements, and let F be a family of subsets of S such that
for any pair A, B in F , A ∩ B 6= ∅. Then |F | ≤ 2n−1 .

Theorem 2.2.2 (Erdos Ko Rado theorem) — Suppose that A is a family of distinct subsets of
{1, 2 . . . n} such that each subset is of size r and each pair of subsets has a nonempty intersection,
and suppose that n ≥ 2r . Then the number of sets in A is less than or equal to the binomial
coefficient  
n−1
r −1

Lemma 2.2.3 — Let S be a set with n elements, and let F be a family of subsets of S such
that for any pair A, B in F , S is not contained by A ∪ B. Then |F | ≤ 2n−1 .

Lemma 2.2.4 ( Kleitman lemma) — A set family F is said to be downwards closed if the following
holds: if X is a set in F , then all subsets of X are also sets in F . Similarly, F is said to be upwards
closed if whenever X is a set in F , all sets containing X are also sets in F . Let F1 and F2 be
downwards closed families of subsets of S = {1, 2, ..., n}, and let F3 be an upwards closed family
of subsets of S. Then we have

|F1 | · |F2 |
|F1 ∩ F2 | ≥ (2.1)
2n
|F1 | · |F3 |
|F1 ∩ F3 | ≤ (2.2)
2n

Lemma 2.2.5 — Let S be a set with n elements, and let F be a family of subsets of S such
that for any pair A, B in F , A ∩ B 6= ∅ and A ∩ B 6= S. Then |F | ≤ 2n−2 .
2.2.1 Lemmas 11

Solution. Using the sets in lemma 1 and lemma 2, defining upwards and downwards sets like
in Kleitman’s Lemma.

Lemma 2.2.6 ( The Sunflower Lemma) — A sunflower with k petals and a core X is a family
of sets S1 , S2 , . . . , Sk such that Si ∩ Sj = X for each i 6= j. (The reason for the name is that
the Venn diagram representation for such a family resembles a sunflower.) The sets Si \ X are
known as petals and must be nonempty, though X can be empty. Show that if F is a family of
sets of cardinality s, and |F | > s!(k − 1)s , then F contains a sunflower with k petals.

Solution. Applying induction and considering the best case where |X| = 0
2.2.2 Extremal Set Theory 12

2.2.2 Extremal Set Theory

MIT 18.314 Lecture-8

Theorem 2.2.7 (Mirsky Theorem) — A set S with a chain of height h can’t be partitioned into
t anti-chains if t < h. In other words, the minimum number of sets in any anti-chain partition of
S is equal to the maximum height of the chains in S. (And Vice Versa)

Theorem 2.2.8 () — In any poset, the largest cardinality of an antichain is at most the smallest
cardinality of a chain-decomposition of that poset.

Theorem 2.2.9 (Dilworth’s Theorem) — Let P be a poset. Then there exist an antichain A and
a chain decomposition C of P such that |A| = |C|

Theorem 2.2.10 (Erdos-Szekeres Theorem) — Any sequence of ab + 1 real numbers contains

either a monotonically decreasing subsequence of length a + 1 or a monotonically increasing
subsequence of length b + 1. The more useful case is when a = b = n.

Problem 2.2.1 () : Let n ≥ 1 be an integer and let X be a set of n2 + 1 positive integers such
that in any subset of X with n + 1 elements there exist two elements x 6= y such that x|y . Prove
that there exists a subset {x1 , x2 . . . xn+1 ∈ X such that xi |xi+1 for all i = 1, 2, . . . n.
2.2.3 Problems 13

2.2.3 Problems

Problem 2.2.2 (USA TST 2005 P1) : Let n be an integer greater than 1. For a positive integer
m , let Sm = {1, 2, . . . , mn}. Suppose that there exists a 2n -element set T such that

1. each element of T is an m -element subset of Sm

2. each pair of elements of T shares at most one common element

3. each element of Sm is contained in exactly two elements of T

Determine the maximum possible value of m in terms of n.

Solution. We use double counting to find the ans, after that the rest is easy.

Problem 2.2.3 (Iran TST 2008 D3P1) : Let S be a set with n elements, and F be a family of
subsets of S with 2n−1 elements, such that for each A, B, C ∈ F , A ∩ B ∩ C is not empty. Prove
that the intersection of all of the elements of F is not empty.

Solution. Using Induction with this lemma.

Problem 2.2.4 (Romanian TST 2016 D1P2) : Let n be a positive integer, and let S1 , S2 , . . . Sn
be a collection of finite non-empty sets such that
X |Si ∩ Sj |
<1
|Si ||Sj |
1≤i<j≤n

Prove that there exist pairwise distinct elements x1 , x2 . . . xn such that xi is a member of Si for each
index i .

Solution. The Inductive proof reduces the problem to American Mathematical Monthly problem
E2309

Solution. The other approach is to focus on the given weird condition, and interpolate it to
something nice, like probabilistic condition.
2.2.3 Problems 14

Problem 2.2.5 (American Mathematical Monthly E2309) : If A1 , A2 ,. . . An are n nonempty subsets

of the set {1, 2, ..., n − 1}, then prove that

X Ai ∩ Aj
≥1
1≤i<j≤n
|Ai | · A j

Problem 2.2.6 (CGMO 2010 P1) : Let n be an integer greater than two, and let A1 , A2 , · · · , A2n
be pairwise distinct subsets of {1, 2, . . . n}. Determine the maximum value of
2n
X |Ai ∩ Ai+1 |
|Ai | · |Ai+1 |
i=1

Where A2n+1 = A1 and |X| denote the number of elements in X.

Problem 2.2.7 (ISL 2002 C5) : Let r ≥ 2 be a fixed positive integer, and let F be an infinite family
of sets, each of size r , no two of which are disjoint. Prove that there exists a set of size r − 1 that
meets each set in F .
HMMT 2016 Team Round: Fix positive integers r > s, and let F be an infinite family of sets, each
of size r , no two of which share fewer than s elements. Prove that there exists a set of size r − 1
that shares at least s elements with each set in F .

Solution [Focus on a set]. If we take an arbitrary set, we can say that there exists infinitely
many sets ∈ F which includes a fixed element from our test set. If we do this argument for r − 1
times, we get a set X of r − 1 elements, and an infinte family of sets that contains X completely.
At this point the problem is trivial.

Solution [Adding Elements]. Since it’s tricky to work with one family, why not introduce
another family, like the second monk. This solution generalizes the problem as such.

Problem 2.2.8 (ISL 1988 P10) : Let N = {1, 2 . . . , n}, n ≥ 2.

A collection F = {A1 , . . . , At } of subsets Ai ⊆ N, i = 1, . . . , t, is said to be separating, if for every
pair {x, y } ⊆ N, there is a set Ai ∈ F so that Ai ∩ {x, y } contains just one element.
F is said to be covering, if every element of N is contained in at least one set Ai ∈ F.
What is the smallest value f (n) of t, so there is a set F = {A1 , . . . , At } which is simultaneously
separating and covering.
2.2.3 Problems 15

Solution [Binary Representation]. Using Binary Representations for the elements as in or not
in, we get an easy bijection.

Problem 2.2.9 (Iran TST 2013 D1P2) : Find the maximum number of subsets from {1, ..., n}
such that for any two of them like A, B if A ⊂ B then |B − A| ≥ 3. (Here |X| is the number of
elements of the set X.)

Solution. By partitioning the maximum set of subsets into groups which contain the number
n and which don’t and Induction on n we can show that the maximum number of subset is
2n − (−1)n
3
.

Problem 2.2.10 (Putnam 2005 B4) : For positive integers m and n, let f (m, n) denote the
number of n-tuples (x1 , x2 , . . . , xn ) of integers such that |x1 | + |x2 | + · · · + |xn | ≤ m. Show that
f (m, n) = f (n, m).

Solution. Try to show Bijection between the result and choosing m or n objects from m + n
m+n
objects or show that the result is n .


2.3 Algorithmic 16

2.3 Algorithmic

• Handout by Cody Johnson

2.3.1 Data Structures

Data Structure (Binary Heap) — Segment tree where, node n has children 2n+1, 2n+2.

1 timus 1862, Sum of operations

Data Structure (Disjoint Set) — This data structure keeps track of connectivity by assigning a
representative to a connected subset. And for any two nodes, u, v they are connected iff they have
the same representative.

2.3.2 Minimal Spanning Tree

Definition (Minimum Spanning Tree) — A minimum spanning tree or minimum weight spanning
tree is a subset of the edges of a connected, edge-weighted (un)directed graph that connects all
the vertices’s together, without any cycles and with the minimum possible total edge weight.

Figure 2.1: Minimal Spanning Tree

2.3.3 Shortest Path Problem 17

Lemma 2.3.1 ( MST Cut) — For any cut of the graph, the lightest edge in that cut-set is in
evey MST of the graph.

Algorithm (Kruskal’s Algorithm) — Kruskal’s algorithm is a ‘minimum-spanning-tree algorithm’

which finds an edge of the least possible weight that connects any two trees in the forest. It is
a greedy algorithm in graph theory as it finds a minimum spanning tree for a connected weighted
graph adding increasing cost arcs at each step.

Solution. To optimize this algorithm, Disjoint Set DS is used.

Algorithm (Prim’s Algorithm) — Greedily build the tree by adding edges one by one. At one step
we add the minimal cost edge that connects the tree to the vertices’s that are not in the tree.

2.3.3 Shortest Path Problem

Definition (Shortest Path Problem) — Finding the shortest path between two nodes in a weighted
or unweighted graph.

Algorithm (Breadth-First Search) — This algo runs from a node and “levelizes” the other nodes.

Algorithm (Dijkstra’s Algorithm) — It picks the unvisited vertex with the lowest distance, cal-
culates the distance through it to each unvisited neighbor, and updates the neighbor’s distance if
smaller.

2.3.4 Other CP Tricks

Theorem 2.3.2 (Swap Sort) — In any swap sorting algorithm, the number of swaps needed has
the same parity.
2.3.5 Fast Fourier Transform 18

Algorithm (Convex Hull Trick) — Given a lot of lines on the plane, and a lot of queries each
asking for the smallest value for y among the lines for a given x, the optimal strategy is to sort
the lines according to their slopes, and adding them to a stack, checking if they are relevant to the
‘minimal’ convex hull of those lines.

2.3.5 Fast Fourier Transform

Let A(x) = a0 + a1 x + . . . an−1 x n−1 be a polinomial and let ω n = 1 be a nth root of unity. We use
FFT to multiply two polynomials in n log n time.

Definition (DFT Matrix) — Let ω be a nth root of unity. The DFT Matrix is a n × n matrix
given by:  
1 1 1 ... 1
1
 ω ω2 . . . ω n−1  
1 ω 2 ω4 . . . ω 2(n−1) 
W =
 .. .. .. .. .. 

. . . . . 
n−1 2(n−1) 2
1 ω ω . . . ω (n−1)
And its inverse is given by:
 
1 1 1 ... 1
1 ω −1 ω −2 . . . ω −(n−1) 
1 ω −2 ω −4

. . . ω −2(n−1) 
W −1 = 1

n  .. .. .. .. ..

. . . . .


2
1 ω −(n−1) ω −2(n−1) . . . ω −(n−1)

Definition (Discrete Fourier Transform) — The Discrete Fourier transform (DFT) of the poly-
nomial A(x) (or equivalently the vector of coefficients (a0 , a1 , . . . , an−1 ) is defined as the values of
the polynomial at the points x = ωnk , i.e. it is the vector:
DFT (a0 , a1 , . . . , an−1 ) = (y0 , y1 , . . . , yn−1 ) = A ωn0 , A ωn1 , . . . , A ωnn−1

In other words, we can write DFT(A) as:

    
y0 1 1 1 ... 1 a0
y1  1
   ω ω2 . . . ω n−1  a1 
 
DFT(A) = y2  = 1
   ω2 ω4 . . . ω 2(n−1)  a2 
 
 ..   .. .. .. .. .. .
 .  . . . . .   .. 

2
yn 1 ω n−1 ω 2(n−1) . . . ω (n−1) an
2.3.5 Fast Fourier Transform 19

Definition (Inverse Discrete Fourier Transform) — The Inverse Discrete Fourier Transform of
values of the polynomial (y0 , y1 , . . . , yn−1 ) are the coefficients of the polynomial (a0 , a1 , . . . , an−1 )

InverseDFT (y0 , y1 , . . . , yn−1 ) = (a0 , a1 , . . . , an−1 )

Thus, if a direct DFT computes the values of the polynomial at the points at the n -th roots, the
inverse DFT can restore the coefficients of the polynomial using those values.

Algorithm (Multiplication of two polynomials) — Say we have A, B two polynomials, and we want
to compute (A · B)(x). Then we first component-wise multiply DFT(A) and DFT(B), and then
retrieve the coefficients of A · B by applying the Inverse DFT.

Algorithm (Fast Fourier Transform) — We want to compute the DFT(A) in n log n time. Suppose
n is a power of 2, ω n = 1, and A(x) = a0 x 0 + a1 x 1 + · · · + an−1 x n−1 . We use divide and conquer
by considering:
n
A0 (x) = a0 x 0 + a2 x 1 + · · · + an−2 x 2 −1
n
A1 (x) = a1 x 0 + a3 x 1 + · · · + an−1 x 2 −1

and A(x) = A0 (x 2 ) + xA1 (x 2 )

n n
−1 −1
After we have computed DFT(A0 ) = (yi0 )i=0
2
and DFT(A0 ) = (yi1 )i=0
2
, we can compute (yi )n−1
i=0
by:  n
 yk + ω yk , for k <
 0 k 1
2
yk =
 y 0 − ω k y 1 , for k ≥ n

k k
2

Algorithm (Inverse FFT) — Since by definition we know DFT(A) = W × A, we have:

2.3.6 Problems 20

2.3.6 Problems

Problem 2.3.1 (Iran TST 2018 P1) : Let A1 , A2 , ..., Ak be the subsets of {1, 2, 3, ..., n} such that
for all 1 ≤ i, j ≤ k:Ai ∩ Aj 6= ∅. Prove that there are n distinct positive integers x1 , x2 , ..., xn such
that for each 1 ≤ j ≤ k:
lcmi∈Aj {xi } > lcmi ∈A
/ j {xi }

Solution. Apply induction on either k or n.

Problem 2.3.2 (ISL 2016 C1) : The leader of an IMO team chooses positive integers n and k
with n > k , and announces them to the deputy leader and a contestant. The leader then secretly
tells the deputy leader an n -digit binary string, and the deputy leader writes down all n -digit binary
strings which differ from the leader’s in exactly k positions. (For example, if n = 3 and k = 1 , and
if the leader chooses 101 , the deputy leader would write down 001, 111 and 100.) The contestant
is allowed to look at the strings written by the deputy leader and guess the leader’s string. What is
the minimum number of guesses (in terms of n and k ) needed to guarantee the correct answer?

Solution. Small cases check.

Problem 2.3.3 (ISL 2005 C2) : Let a1 , a2 , . . . be a sequence of integers with infinitely many positive
and negative terms. Suppose that for every positive integer n the numbers a1 , a2 , . . . , an leave n
different remainders upon division by n. Prove that every integer occurs exactly once in the sequence
a1 , a2 , . . ..

Solution. Constructing for the beginning.

Problem 2.3.4 (IOI Practice 2017) : C plays a game with A and B. There’s a room with a table.
First C goes in the room and puts 64 coins on the table in a row. Each coin is facing either heads
or tails. Coins are identical to one another, but one of them is cursed. C decides to put that coin
in position c. Then he calls in A and shows him the position of the cursed coin. Now he allows A to
flip some coins if he wants (he can’t move any coin to other positions). After that A and C leave
the room and sends in B. If B can identify the cursed coin then C loses, otherwise C wins.
The rules of the game are explained to A and B beforehand, so they can discuss their strategy
before entering the room. Find the minimum number k of coin flips required by A so that no matter
what configuration of 64 coins C gives them and where he puts the cursed coin, A and B can win
with A flipping at most k coins.
2.3.6 Problems 21

Find constructions for k = 32, 8, 6, 3, 2, 1

Solution. XOR XOR XOR binary representation

Problem 2.3.5 (Codeforces 987E) : Petr likes to come up with problems about randomly gener-
ated data. This time problem is about random permutation. He decided to generate a random
permutation this way: he takes identity permutation of numbers from 1 to n and then 3n times
takes a random pair of different elements and swaps them. Alex envies Petr and tries to imitate
him in all kind of things. Alex has also come up with a problem about random permutation. He
generates a random permutation just like Petr but swaps elements 7n + 1 times instead of 3n times.
Because it is more random, OK?!

You somehow get a test from one of these problems and now you want to know from which one.

Solution. This theorem kills this problem instantly.

Problem 2.3.6 (USAMO 2013 P6) : At the vertices’s of a regular hexagon are written six non-
negative integers whose sum is 20032003 . Bert is allowed to make moves of the following form:
he may pick a vertex and replace the number written there by the absolute value of the difference
between the numbers written at the two neighboring vertices. Prove that Bert can make a sequence
of moves, after which the number 0 appears at all six vertices.

Solution. Firstly what comes into mind is to decrease the maximum value, but since this is a
P6, there must be some mistakes. Surely, we can’t follow this algo in the case (k, k, 0, k, k, 0).
But this time, the sum becomes even. So we have to slowly minimize the maximum, keeping the
sum odd. And since only odd number on the board is the easiest to handle, we solve that case
first, and the other cases can be easly handled with an additional algo.

Problem 2.3.7 (ISL 2012 C1) : Several positive integers are written in a row. Iteratively, Alice
chooses two adjacent numbers x and y such that x > y and x is to the left of y , and replaces the
pair (x, y ) by either (y + 1, x) or (x − 1, x). Prove that she can perform only finitely many such
iterations.

Solution. Easy invariant.

2.3.6 Problems 22

Problem 2.3.8 (AoPS) : There is a number from the set {1, −1} written in each of the vertices’s
of a regular do-decagon ( 12 -gon). In a single turn we select 3 numbers going in the row and
change their signs. In the beginning all numbers, except one are equal to 1. Can we transfer the
only −1 into adjacent vertex after a finite number of turns?

Solution. Algo+Proof =⇒ Invariant.

Problem 2.3.9 (ISL 1994 C3) : Peter has three accounts in a bank, each with an integral number
of dollars. He is only allowed to transfer money from one account to another so that the amount
of money in the latter is doubled. Prove that Peter can always transfer all his money into two
accounts. Can Peter always transfer all his money into one account?

Solution. Since we want to decrease the minimum, and one of the most simple way is to
consider Euclidean algorithm. So we sort the accounts, A < B < C, and write B = qA + r , and
do some experiment to turn B into r .

Problem 2.3.10 (MEMO 2008, Team, P6) : On a blackboard there are n ≥ 2, n ∈ Z+ numbers.
In each step we select two numbers from the blackboard and replace both of them by their sum.
Determine all numbers n for which it is possible to yield n identical number after a finite number of
steps.

Solution. The pair thing rules out the case of odds. For evens, we make two identical sets,
and focus on only one of the sets, with an additional move x → 2x available to use. Since we can
now change the powers of 2 at out will at any time, we only focus on the greatest odd divisors.
Our aim is to slowly decrease the largest odd divisor.

Problem 2.3.11 (USA Dec TST 2016, P1) : Let S = {1, . . . , n}. Given a bijection f : S → S
an orbit of f is a set of the form {x, f (x), f (f (x)), . . . } for some x ∈ S. We denote by c(f ) the
number of distinct orbits of f . For example, if n = 3 and f (1) = 2, f (2) = 1, f (3) = 3, the two
orbits are {1, 2} and {3}, hence c(f ) = 2.

Given k bijections f1 , . . ., fk from S to itself, prove that

c(f1 ) + · · · + c(fk ) ≤ n(k − 1) + c(f )

where f : S → S is the composed function f1 ◦ · · · ◦ fk .

2.3.6 Problems 23

Solution. Induction reduces the problem to the case of k = 2. Then another inuction on
c(f1 ) solves the problem. The later induction works on the basis of the fact that a “swap” in the
bijection changes the number of cycles by 1 (either adds +1 or −1).

Problem 2.3.12 (Cody Johnson) : Consider a set of 6 integers S = {a1 . . . a6 }. At on step, you
can add +1 or −1 to all of the 6 integers. Prove that you can make a finite number of moves so
that after the moves, you have a1 a5 a6 = a2 a4 a6 = a3 a4 a5

Problem 2.3.13 (ISL 2014 A1) : Let a0 < a1 < a2 . . . be an infinite sequence of positive integers.
Prove that there exists a unique integer n ≥ 1 such that
a0 + a1 + a2 + · · · + an
an < ≤ an+1 .
n

Solution. My idea was to construct the sequence with the assumtion that the condition is
false. It leads to either all of the right ineq false or the condition being true.

Solution. The magical solution: defining bn = (an − an−1 ) + · · · + (an − a1 ) which eases the
inequlity.

Solution. The beautiful solution: defining δi as an = a0 + ∆1 + ∆2 + ... + ∆n for all n, i .

Solution. Another idea is to first prove the existence and then to prove the uniqueness.

Problem 2.3.14 (ISL 2014 N3) : For each positive integer n, the Bank of Cape Town issues coins of
denomination n1 . Given a finite collection of such coins (of not necessarily different denominations)
with total value at most most 99 + 21 , prove that it is possible to split this collection into 100 or
fewer groups, such that each group has total value at most 1.

1 1 1
Solution. Notice that the sum of the geometric series S = + 2 + 3 . . . is 1. And in
2 2 2
another problem we partitioned the set of integers into subsets with each subset starting with
an odd number k and every other elements of the subset being 2i ∗ k. We do similarly in this
problem, and partition the set of the coins in a similar way. Then we take the first 100 sets whose
sum is less than 1 and insert the other left coins in these sets, with the condition that the sum
2.3.6 Problems 24

1
of all of the coins is 99 + . Solu
2

Solution. Replacing 100 by n, we show that for all n the codition is valid. Assume otherwise.
Take the minimal n for which the condition does not work. Ta-Da! We can show that if n does
not work, so doesn’t n − 1. Solu

k
Solution. Or just be an EChen and prove the result for at most k − with k groups.
2k + 1

Solution. Very similar to this problem

Problem 2.3.15 (China TST 2006) : Given positive integer n, find the biggest real number C
which satisfy the condition that if the sum of the reciprocals ( n1 is the reciprocal of n) of a set of
integers (They can be the same.) that are greater than 1 is less than C, then we can divide the set
of numbers into no more than n groups so that the sum of reciprocals of every group is less than 1.

Problem 2.3.16 () : In a n ∗ n grid, every cell is either black or white. A ‘command’ is a pair of
integers, i, j ≤ n, after which all of the cells in the i th row and the j th column (meaning a total of
2n − 1 cells) will switch the state. Our goal is to make every cell of the same state.

n2
1. Prove that if it can be done, it can be done in less than 2 commands.

2. Prove that it can always be done if n is even.

3. Prove or disprove for odd n.

Solution. ( a ) is really easy, just take into account that flipping all cells result in the switch of
all of the cells. And the question did not ask for an algorithm.

Solution. ( b ) is also easy, notice that we can pair the columns and then make them look like
the same, with a compound command. A better algo is to take the original algo and to modify
it like, take one cell, then do the original move on all cells in the row and column of this cell.
2.3.6 Problems 25

Solution. ( c ) uses Linear Algebra, which I dont know yet, or... use double counting to build
the criteria of the fucntion f : states → subset of moves being bijective.

Problem 2.3.17 (OIM 1994, PSMiC) : In every square of an n × n board there is a lamp. Initially
all the lamps are turned off. Touching a lamp changes the state of all the lamps in its row and its
column (including the lamp that was touched). Prove that we can always turn on all the lamps and
find the minimum number of lamps we have to touch to do this.

Problem 2.3.18 (AtCoder GC043 B) : Given is a sequence of N digits a1 , a2 . . . aN , where each

element is 1, 2, or 3. Let xi,j defined as follows:

• x1,j := aj (1 ≤ j ≤ N)

• xi,j := |xi−1,j − xi−1,j+1 | (2 ≤ i ≤ N and 1 ≤ j ≤ N + 1 − i )

Find xN,1 .

Solution. Since |x − y | ≡ x + y (mod 2), we can determine the parity of XN,1 using binomial
coefficient. Which in turn we can get in O(n) with bitwise operator. Now we have to distinguish
between 0, 2. For 2, all of the rows starting with the second one should have only 2 and 0. Where
we can apply the same algorithm as before and find whether the final digit is 2 or 0.

Problem 2.3.19 (Timus 1578) : The very last mammoth runs away from a group of primeval
hunters. The hunters are fierce, hungry and are armed with bludgeons and stone axes. In order to
escape from his pursuers, the mammoth tries to foul the trail. Its path is a polyline (not necessarily
simple). Besides, all the pairs of adjacent segments of the polyline form acute angles (an angle of
0 degrees is also considered acute).

After the mammoth vanished, it turned out that it had made exactly N turns while running away.
The points where the mammoth turned, as well as the points where the pursuit started and where
the pursuit ended, are known. You are to determine one if the possible paths of the mammoth.

Problem 2.3.20 (CodeForces 744B) : Given a hidden matrix of n × n, n ≤ 1000 where for every
i , M(i,i)=0 , Luffy’s task is to find the minimum value in the n rows, formarly spoken, he has to
find values minj=1...n, j6=i M(i,j) . To do this he can ask the computer questions of following types:
In one question, Luffy picks up a set, a1 , a2 . . . ak with ai , k ≤ n. And gives the computer this set.
The computer will respond with n integers. The i -th integer will contain the minimum value of
2.3.6 Problems 26

minj=1...k M(i,aj ) . And on top of this, he can only ask 20 questions. Luffy being the stupid he is,
doesn’t even have any clue how to do this, you have to help him solving this problem.

Solution. If we draw the diagonal in the matrix, we see that we can fit boxes of 2i × 2i in
there depending on the i ’s value. Now after we have decomposed the matrix into such boxes, we
can choose several from them to ask a question. The trick is that for every row, there must be
questions asked from each of the boxes this row covers and no question from here must contain
the (i, i ) cell.

Solution. The magical solution goes as following: For i ≤ 10 , for every k = 1 . . . n , include
k in the question if the i th bit of k ’s binary form is 0. And then for the second round include k
in the question if the i th bit of k ’s binary form is 1.

Problem 2.3.21 () : Alice wants to add an edge (u, v ) in a graph. You want to know what this
edge is. So, you can ask some questions to Alice. For each question, you will give Alice 2 non-empty
disjoint sets S and T to Alice, and Alice will answer "true" iff u and v belongs to different sets.
You can ask atmost 3 ∗ dlog2 |V |e questions to Alice. Describe a strategy to find the edge (u, v ).

Solution. First find one true answer in dlog2 |V |e questions, and then get the result out of
these two sets in 2 ∗ dlog2 |V |e questions.

Solution. The magical solution goes as following: In the i th question, S =

x : i bit of x is 0, T = x : i th bit of x is 1
th

Problem 2.3.22 (USAMO 2015 P4) : Steve is piling m ≥ 1 indistinguishable stones on the squares
of an n × n grid. Each square can have an arbitrarily high pile of stones. After he finished piling
his stones in some manner, he can then perform stone moves, defined as follows. Consider any
four grid squares, which are corners of a rectangle, i.e. in positions (i , k), (i , l), (j, k), (j, l) for some
1 ≤ i, j, k, l ≤ n, such that i < j and k < l. A stone move consists of either removing one stone
from each of (i , k) and (j, l) and moving them to (i , l) and (j, k) respectively,j or removing one
stone from each of (i , l) and (j, k) and moving them to (i , k) and (j, l) respectively.

Two ways of piling the stones are equivalent if they can be obtained from one another by a sequence
of stone moves.
2.3.6 Problems 27

How many different non-equivalent ways can Steve pile the stones on the grid?

Solution. Building an invariant, we see that only the sum of the columns is not sufficient. So
to get more control, we take the row sums into account as well.

Problem 2.3.23 (ISL 2003 C4) : Let x1 , . . . , xn and y1 , . . . , yn be real numbers. Let A = (aij )1≤i,j≤n
be the matrix with entries (
1, if xi + yj ≥ 0;
aij =
0, if xi + yj < 0.
Suppose that B is an n × n matrix with entries 0, 1 such that the sum of the elements in each row
and each column of B is equal to the corresponding sum for the matrix A. Prove that A = B.

Solution. If done after this problem, this problem seems straigthforward.

Problem 2.3.24 (India TST 2017 D1 P3) : Let n ≥ 1 be a positive integer. An n × n matrix is
called good if each entry is a non-negative integer, the sum of entries in each row and each column
is equal. A permutation matrix is an n × n matrix consisting of n ones and n(n − 1) zeroes such
that each row and each column has exactly one non-zero entry.

Prove that any good matrix is a sum of finitely many permutation matrices.

Solution. Same algo as above. Either distributing uniformly or gathering all in a diagonal

Problem 2.3.25 (Tournament of Towns 2015F S7) : N children no two of the same height stand
in a line. The following two-step procedure is applied: first, the line is split into the least possible
number of groups so that in each group all children are arranged from the left to the right in
ascending order of their heights (a group may consist of a single child). Second, the order of
children in each group is reversed, so now in each group the children stand in descending order of
their heights. Prove that in result of applying this procedure N − 1 times the children in the line
would stand from the left to the right in descending order of their heights.

Solution. It’s obvious that we need to find some invariant or mono-variant. Now, an idea, we
need to show that for any i , for it to be on it’s rightful place, it doesn’t need more than N − 1
moves. How do we show that? Another idea, think about the bad bois on either of its sides.
Now, observation, ‘junctions’ decrease with each move. Find the ‘junctions’.
2.3.6 Problems 28

Problem 2.3.26 (Polish OI) : Given n jobs, indexed from 1, 2 . . . n. Given two sequences of reals,
{ai }ni=1 , {bi }ni=1 where, 0 ≤ ai , bi ≤ 1. If job i starts at time t , then the job takes hi (t) = ai t + bi
time to finish. Order the jobs in a way such that the total time taken by all of the jobs is the
minimum.

Solution. Example of a problem which is solved by investigating two adjacent objects in the
optimal arrangement.

Problem 2.3.27 (CodeForces 960/C) : Pikachu had an array with him. He wrote down all the
non-empty subsequences of the array on paper. Note that an array of size n has 2n − 1 non-empty
subsequences in it.

Pikachu being mischievous as he always is, removed all the subsequences in which
Maximum element of the subsequence − Minimum element of subsequence ≥ d

Pikachu was finally left with X subsequences.

However, he lost the initial array he had, and now is in serious trouble. He still remembers the
numbers X and d. He now wants you to construct any such array which will satisfy the above
conditions. All the numbers in the final array should be positive integers less than 1018 .

Note the number of elements in the output array should not be more than 104 . If no answer is
possible, print −1.

Problem 2.3.28 (ARO 2005 P10.3, P11.2) : Given 2005 distinct numbers a1 , a2 , . . . , a2005 . By
one question, we may take three different indices 1 ≤ i < j < k ≤ 2005 and find out the set
of numbers {ai , aj , ak } (unordered, of course). Find the minimal number of questions, which are
necessary to find out all numbers ai .

Solution. The key idea is to ask questions such that it is connected to multiple other questions,
and each question uniquely finds out multiple elements together. One by itself immediately after
the question has been asked, and one after the next question which is related to this one has
been asked. As we find out three elements’ values after one question, first, second, third, so, let
us find first from the previous question, second from the current question, third from the next
question.
2.3.6 Problems 29

Problem 2.3.29 (IOI 2007 P3) : You are given two sets of integers A = {a1 , a2 . . . an } and
B = {b1 , b2 . . . bn } such that ai ≥ bi . At move i you have to pick bi distinct integers from the
set Ai = {1, 2, . . . ai }. In total, (b1 + b2 + · · · + bn ) integers are selected, but not all of these are
distinct. Suppose k distinct integers have been selected, with multiplicities c1 , c2 , c3 . . . ck . Your
score is defined as
Xk
ci (ci − 1)
i=1
Give an efficient algorithm to select numbers in order to “minimize” your score.

Solution. Some investigation shows that if ci > cj + 1 and i > j , then we can always minimize
the score. and if i < j , then we can minimize the score only when i , j ∈ Ak but i has been taken
at move k , but j hasn’t. So in the minimal state, either both i , j has been taken at move k ,
or ak < j. So the idea is to take elements from Ai as large as possible, and then taking smaller
values after wards if the ci value of a big element gets more than that of a small element. In
this algorithm, we see that we greedily manipulate ci . So it is a good idea to greedily choose ci
’s from the very beginning.

Solution. Solution Algo: at step i , take the set {c1 , c2 . . . cai } and take the smallest bi from
this set, and add 1 to each of them (in other words, take their index numbers as the numbers to
take).

Problem 2.3.30 () : Given n numbers {a1 , a2 , ..., an } in arbitrary order, you have to select k of
them such that no two consecutive numbers are selected and their sum is maximized.

Solution. Notice that if ai is the maximum value, and if ai is not counted in the optimal
solution, then both of ai−1 , ai+1 must be in the optimal solution, and ai−1 + ai+1 > ai . And if ai
is counted in the optimal solution, then none of ai−1 , ai+1 can be counted in the optimal solution.
So either way, we can remove these three and replace them by a single element to use induction.
So remove ai−1 , ai+1 and replace ai by ai−1 + ai+1 − ai .

Problem 2.3.31 (USAMO 2010 P2) : There are n students standing in a circle, one behind the
other. The students have heights h1 < h2 < · · · < hn . If a student with height hk is standing
directly behind a student with height hk−2 or less, the
two students are permitted to switch places.
Prove that it is not possible to make more than n3 such switches before reaching a position in
which no further switches are possible.

Problem 2.3.32 (Serbia TST 2015 P3) : We have 2015 prisinoers. The king gives everyone a hat
2.3.6 Problems 30

coloured in one of 5 colors. Everyone sees all hats expect his own. Now,the King orders them in a
line(a prisioner can see all guys behind and in front of him). The king asks the prisinoers one by one
does he know the color of his hat.If he answers NO,then he is killed.If he answers YES,then answers
which color is his hat,if his answers is true,he goes to freedom,if not,he is killed.All the prisinors can
hear did he answer YES or NO,but if he answered YES,they don’t know what did he answered(he
is killed in public).They can think of a strategy before the King comes,but after that they can’t
comunicate.What is the largest number of prisinors we can guarentee that can survive?

Problem 2.3.33 (Taiwan TST 2015 R3D1P1) : A plane has several seats on it, each with its own
price, as shown below. 2n − 2 passengers wish to take this plane, but none of them wants to sit
with any other passenger in the same column or row. The captain realize that, no matter how he
arranges the passengers, the total money he can collect is the same. Proof this fact, and compute
how much money the captain can collect.

Figure 2.2

Problem 2.3.34 (ISL 2014 N1) : Let n ≥ 2 be an integer, and let An be the set

An = {2n − 2k | k ∈ Z, 0 ≤ k < n}.

Determine the largest positive integer that cannot be written as the sum of one or more (not
necessarily distinct) elements of An .

Solution. Inductive approach

Problem 2.3.35 (Codeforces 330D) : Biridian Forest

Solution. Generalize the condition for a meet-up.

2.3.6 Problems 31

Problem 2.3.36 (Codeforces 1270F) : Let’s call a binary string s awesome, if it has at least 1
symbol 1 and length of the string is divisible by the number of 1 in it. In particular, 1, 1010, 111
are awesome, but 0, 110, 01010 aren’t.

You are given a binary string s of size ≤ 2 × 105 . Count the number of its awesome substrings.

√
Solution. The constraint tells us the algorithm should be of complexity O(n n). Playing with
the funtion f (i ) and the divisibility condition gives us some ground to work with. Since we need
√
≤ c × n queries around the full string, we remember the prime sieve trick.

Problem 2.3.37 (IMO 1986 P3) : To each vertex of a regular pentagon an integer is assigned, so
that the sum of all five numbers is positive. If three consecutive vertices are assigned the numbers
x, y , z respectively, and y < 0, then the following operation is allowed: x, y , z are replaced by
x + y , −y , z + y respectively. Such an operation is performed repeatedly as long as at least one of
the five numbers is negative. Determine whether this procedure necessarily comes to an end after
a finite number of steps.

Solution. Notice how starting from one negative and moving to consective number on one side
moves the number by one vertex nicely.

Problem 2.3.38 (CodeForces 1379E) : Let G be a tree with n vertices that has a root, and every
vertex has either 2 children or no child. A vertex is said to be “good” if the two subtrees of it, say
X, Y, satisfy either V (X) ≥ 2V (Y ) or V (Y ) ≥ 2V (X) (here V (X) is the number of vertices of tree
X). Find all k such that there is a tree with n vertices with k good vertices, and give a construction
to build the tree.

Solution. Try recursion, the pattern is quite easy to find.

2.3.7 Algorithm Analysis 32

2.3.7 Algorithm Analysis

Problem 2.3.39 (GQMO 2020 P4) : Prove that, for all sufficiently large integers n, there exists n
numbers a1 , a2 , . . . , an satisfying the following three conditions:
2n
Each number ai is equal to either −1, 0 or 1. At least of the numbers a1 , a2 , . . . , an are non-zero.
5
a1 a2 an
The sum + + ··· + is 0.
1 2 n
Note: Results with 2/5 replaced by a constant c will be awarded points depending on the value of c
2.3.8 Covering Area with Squares 33

2.3.8 Covering Area with Squares

• A nice blog post by ankogonit

Problem 2.3.40 (Brazilian MO 2002, ARO 1979) : Given a finite collection of squares with to-
tal area at least 4, prove that you can cover a unit square completely with these squares (with
overlapping allowed, of course).

Solution. Maybe motivated by the number 4 and how nice it would be if all the squares had
2’s power side lengths, the idea is to shrink every square to a side with side of 2’s power.

Problem 2.3.41 (ARO 1979’s Sharper Version) : Given a finite collection of squares with total area
at least 3, prove that you can cover a unit square completely with these squares (with overlapping
allowed).

Solution. The idea is to greedily cover the unit square by covering the lowest row uncovered.
And then using boundings to prove that it is possible.

Problem 2.3.42 () : Prove that a finite collection of squares of total area 1

2 can be placed inside
a unit square without overlap.

Solution. The same idea as before.

Problem 2.3.43 (Tournament of Towns Spring 2012 S7) : We attempt to cover the plane with
an infinite sequence of rectangles, overlapping allowed.

1. Is the task always possible if the area of the n−th rectangle is n2 for each n?

2. Is the task always possible if each rectangle is a square, and for any number N, there exist
squares with total area greater than N?

Solution. Identical algo and proving technique as above.

Solution. Using the first problem in this subsection to find a better algo.
2.3.8 Covering Area with Squares 34

Problem 2.3.44 (ISL 2006 C6) : A holey triangle is an upward equilateral triangle of side length n
with n upward unit triangular holes cut out. A diamond is a 60◦ − 120◦ unit rhombus.

Prove that a holey triangle T can be tiled with diamonds if and only if the following condition holds:
Every upward equilateral triangle of side length k in T contains at most k holes, for 1 ≤ k ≤ n.

Solution. Think of induction and how you can deal with that.

Problem 2.3.45 (Putnam 2002 A3) : Let N be an integer greater than 1 and let Tn be the number
of non empty subsets S of {1, 2, ....., n} with the property that the average of the elements of S is
an integer. Prove that Tn − n is always even.

Solution. Try to show an bijection between the sets with average n which has k elements
(Here k is an even integer) and the sets with average n but with number of elements k + 1. This
implies that the number of such sets is even.

Problem 2.3.46 (USAMO 1998) : Prove that for each n ≥ 2, there is a set S of n integders such
that (a − b)2 divides ab for every distinct a, b ∈ S.

Solution. Induction comes to the rescue. Trying to find a way to get from n to n + 1, we see
that we can shift the integers by any integer k. So after shifting, what stays the same, and what
changes?
2.4 Graph Theory 35

2.4 Graph Theory

Stuck? Try These: Turning grids into graphs

• One common way to turn a grid into graphs is to create a bipartite graph between the
columns and rows such that ci and rj are connected iff (i , j) is marked. This way we
can find cycles alternating row and column.

• Creating a bipartite graph between all rows and columns and particular objects. This
helps to prove matching.

Lemma 2.4.1 ( Bipartite Graph Criteria) — Any graph having only even cycles are Bipartite.

1 AoPS 3 Problem
2 ISL 2004 C3 4 Problem

Theorem 2.4.2 (Euler’s Polyhedron Formula) — • For any polyhedron with E, V, F edges, ver-
tices’s and faces resp. the following relation holds

V +F =E+2

• In a planar graph with V vertices, E edges and C cycles, the following condition is always
satisfied:
V −E+C =1

Lemma 2.4.3 ( Criteria of partitioning a graph into disconnected sub-graphs) — If there exist
no three vertices, u, v , w that uv ∈ E(G) also uw , v w ∈ E(G), the graph can be partitioned into
equivalence classes based on their non-neighbors.

Problem 2.4.1 (China TST 2015 T1 D2 P3) : There are some players in a Ping Pong tournament,
where every 2 players play with each other at most once. Given:
1. Each player wins against at least a players, and loses to at least b players. (a, b ≥ 1)
2.4 Graph Theory 36

2. For any two players A, B, there exist some players P1 , ..., Pk (k ≥ 2) (where P1 = A,Pk = B),
such that Pi wins against Pi+1 (i = 1, 2..., k − 1)
Prove that there exist a + b + 1 distinct players Q1 , ...Qa+b+1 , such that Qi wins against Qi+1
(i = 1, ..., a + b).

Remark. Typical largest path, some workaround with given constraints problem.

Solution. Take the largest path starting from a1 to an .

Definition— Assume that n ≤ a + b. Since this is the largest path, a edges coming out of
an are all in S = {a1 , a2 , . . . an }, and b edges going in a1 are all in S. Let S1 be the set of
vertices that an wins against, and S2 be the set of vertices that a1 loses against. Moreover,
let al be the smallest element of S1 , and an−k be the largest element of S2 (smallest means
leftmost in the part, and largest means rightmost).

Let S 0 = {al , al+1 , . . . an−k−1 , an−k }. Since n ≤ a + b, we have S 0 6= ∅.

We also define:
S10 = {ai | ai defeats ai+1 where ai+1 ∈ S1 ∩ S 0 }
S20 = {ai | ai−1 defeats ai where ai−1 ∈ S2 ∩ S 0 }

Figure 2.3

Now, note that for any x ∈ S20 , y ∈ S10 , there is a path between x, y with n vertices. So for all
x ∈ S20 , there does not exist a vertex outside of S that defeats x. And for all y ∈ S10 , there
doesn’t exist a vertex outside of S that loses to y , because of the maximality of n.

We show that, S10 ∩ S20 6= ∅. Then there would exist a vertex that doesn’t have any edge outside
of S, meaning it has at least a + b games inside S, proving the result.
2.4 Graph Theory 37

We have, S10 , S20 ⊂ S 0 ∪ {al−1 , an−k+1 }. We have,

|S10 | ≥ a − k + 2 [∵ an has at most k − 2 vertices in {an−k+1 , . . . an }]

|S20 | ≥b−l +3

But |S 0 | = n − (l − 1) − (k) + 2 ≤ a + b − l − k + 3 < |S10 | + |S20 |. So S10 ∩ S20 6= ∅, and we are

done.

Problem 2.4.2 (ARO 2005 P10.8) : A white plane is partitioned in to cells (in a usual way). A
finite number of cells are coloured black. Each black cell has an even (0, 2 or 4) number of adjacent
(by the side) white cells. Prove that one may colour each white cell in green or red such that every
black cell will have equal number of red and green adjacent cells.

Solution. First we join the white cells like this:

Figure 2.4

Now, notice that the plane have been divided into some cycles (a black cell that has no adjacent
black cells is a cycle itself). So we can color the plane blue and yellow in a way that no region
has the same color as its neighbors. We can do this because at any junction, there are an even
number of regions connected because of the problem condition.

Now we focus on our graph that we created connected the white cells. Take any cycle on it. If
we have a “slanted” edge, then both of the nodes are inside a region of either blue or yellow. But
in a “straight” edge, the two nodes are in different colored region.

We know that there are an even number of slanted edges, which is trivial to prove (using the
fact that any cycle on a grid system has even number of nodes, and on these cycles, most of
the edges (the straight ones) have even lenght, but only the slanted one has odd lenghts on the
sides). It is also easy to see that there are an even number of straight edges, because of going
in and out of the regions of a fixed color.

So our cycle has an even number of nodes and thus bipartite. We can color the graph with two
colors, so that along each edge, the two nodes are of different color.
2.4 Graph Theory 38

Remark. There is a simple coloring using this solution. After we color the regions of the plane
with blue and yellow, we number each column with integers. Then on the odd numbered columns,
we color all the white cells that are in yellow region green and blue region red. And on the even
numbered columns, we do the opposite. It is easy to check that this coloring works using the graph
we created before.

Problem 2.4.3 (AtCoder GC033 C) : Takahashi and Aoki will play a game on a tree. The tree has
N vertices numbered 1 to N, and the i -th of the N − 1 edges connects Vertex ai and Vertex bi
At the beginning of the game, each vertex contains a coin. Starting from Takahashi, he and Aoki
will alternately perform the following operation:

• Choose a vertex v that contains one or more coins, and remove all the coins from v .

• Then, move each coin remaining on the tree to the vertex that is nearest to v among the
adjacent vertices of the coin’s current vertex.

The player who becomes unable to play, loses the game. That is, the player who takes his turn
when there is no coin remaining on the tree, loses the game. Determine the winner of the game
when both players play optimally.

Solution. First transform the game by removing the idea of coins, and replacing it with deleting
vertices. Now, notice that the longest path in this tree (i.e. the diameter) strictly decreases by
1 or 2 each turn depending on the move. So it’s just a basic predetermined game.

Problem 2.4.4 (ARO 1999 P9.8) : There are 2000 components in a circuit, every two of which
were initially joined by a wire. The hooligans Vasya and Petya cut the wires one after another.
Vasya, who starts, cuts one wire on his turn, while Petya cuts one or three. The hooligan who cuts
the last wire from some component loses. Who has the winning strategy?

Solution [Copycat]. The P-Hooligan Petya has a winning strategy, for he can be follow the old
cunning trick of never losing. How does he do it?
He starts by secretly partitioning the vertices in two 1000 degree subsets. He calls them A =
{a1 , a2 . . . a1000 } and B = {b1 , b2 . . . b1000 }. He then connects ai with bi with an edge with an
invisible marker that only he can see.
Now the game begins. Petya copies Vasyas moves following these rules:

1. If Vasya removes an edge ai – aj , where i 6= j, then Patya removes the edges ai – bj , aj –

bi and bi – bj .

2. If Vasya removes aj – bi , where i 6= j, then Patya removes the other three edges from the
above rule.
2.4 Graph Theory 39

3. If Vasya removes ai –bi , then Patya looks for another bj , such that ai –bj exists. Then by
the symmetry so far maintanined, aj –bi and aj –bj exist too. And Patya can remove aj –bj ,
and swap the names of bj and bi .
But if he can’t, then that would mean after Vasya’s move ai would become isolated, and
Patya would win.

It is easy to see that the above moves are possible since Patya is always maintaining symmetry
between A, B. So he can’t move means Vasya has already disconnected one of the vertices.

Remark. The case with 4 vertices and 6 vertices give an idea to copy the opponent’s moves.

Problem 2.4.5 (ISL 2001 C3) : Define a k-clique to be a set of k people such that every pair of
them are acquainted with each other. At a certain party, every pair of 3-cliques has at least one
person in common, and there are no 5-cliques. Prove that there are two or fewer people at the
party whose departure leaves no 3-clique remaining.

Solution. Casework with the point where most of the triangles are joined.

Problem 2.4.6 (ARO 2017 P9.1) : In a country some cities are connected by one-way flights (there
is at most one flight between two cities). City A is called “available" from city B, if there is a flight
from B to A, maybe with some transits. It is known, that for every 2 cities P and Q, there exists a
city R, such that P and Q are both available from R. Prove, that exist city A, such that every city
is available from A.

Solution. Basic induction exercise.

Problem 2.4.7 (Turkey National MO 2002 P3) : Graph Airlines (GA) operates flights between
some of the cities of the Republic of Graphia. There are at least three GA flights from each city,
and it is possible to travel from any city in Graphia to any city in Graphia using GA flights. GA
decides to discontinue some of its flights. Show that this can be done in such a way that it is still
possible to travel between any two cities using GA flights, yet at least 2/9 of the cities have only
one flight.

Problem 2.4.8 (Tournament of Towns 2009 S6) : Anna and Ben decided to visit Archipelago with
2009 islands. Some pairs of islands are connected by boats which run both ways. Anna and Ben
are playing during the trip:
2.4.1 Counting in Graph 40

Anna chooses the first island on which they arrive by plane. Then Ben chooses the next island which
they could visit. Thereafter, the two take turns choosing an island which they have not yet visited.
When they arrive at an island which is connected only to islands they had already visited, whoever’s
turn to choose next would be the loser. Prove that Anna could always win, regardless of the way
Ben played and regardless of the way the islands were connected.

Solution. The copycat idea, as always in games. How can Anna make sure she gets a move
after Ben? Maybe she can “pair” vertices and follow them along. What happens if she tries this
strategy?

2.4.1 Counting in Graph

Lemma 2.4.4 ( Average of Degrees) — In a graph G with n vertexes, let E be the set of all
edges. Assign an integer fi to every vertex vi such that fi equals to the everage degree of the
neighbors of vi . We have,
Xn
fi ≥ 2|E|
i=1

Lemma 2.4.5 — In a graph G with n vertexes, let E be the set of all edges. Assign an integer
gi to every vertex vi such that gi equals to the maximum degree among its neighbors. We have,
n
X
gi ≥ 2|E|
i=1

Problem 2.4.9 (USA TST 2014 P3) : Let n be an even positive integer, and let G be an n-vertex
2
graph with exactly n4 edges, where there are no loops or multiple edges (each unordered pair of
distinct vertices is joined by either 0 or 1 edge). An unordered pair of distinct vertices {x, y } is said
to be amicable if they have a common neighbor (there is a vertex z such that xz and y z are both
n/2
edges). Prove that G has at least 2 2 pairs of vertices which are amicable.

Solution. Define friendship in a different way, bounding below, keeping in mind the equality
case. Then using the previous lemma.
2.4.1 Counting in Graph 41

Theorem 2.4.6 (Turan’s theorem) — Let G be any graph with n vertices, such that G is Kr +1
-free. Then G is the “Turán’s Graph” and is a complete r partite graph. And the number of edges
in G is at most

r − 1 n2
  2
1 n
· = 1− ·
r 2 r 2
A special case of Turán’s theorem for n = 2 is the Mantel’s Theorem.
 2  It states that the maximal
n
triangle free graph is a complete bipartite graph with at most edges.
4

Figure 2.5: Turán’s Graph

Proof. We need to prove that the maximal graph is the r partite one, and the rest will follow.
We can directly try to prove that this graph is r colorable, but that is quite troublesome. Instead,
we try to show that, we can partition the vertices of G into equivalence classes based on their
non-neighbors. Since this is imply the former. So we need to prove that this holds for this graph.

The way it is done is quite interesting. We need to show that if the criteria doesn’t hold in this
graph, then this graph is not the maximal graph. How are we going to do that? We compare
the degrees of u, w , and replace either u by w or w by u to get a graph with more edges and
without the nasty situation.

Problem 2.4.10 () : 155 birds P1 , P2 , . . . , P155 are sitting down no the boundary of a circle C.
Two birds Pi , Pj are mutually visible if the angle at the center of their cord, m(Pi Pj ) ≤ 10◦ . Find
the smallest number of mutually visible pairs of birds.

Problem 2.4.11 () : For a pair A = (x1 , y1 ) and B = (x2 , y2 ) of points on the coordinate plane, let
d(A, B) = |x1 −x2 |+|y1 −y2 |. We call a pair (A, B) of unordered points harmonic if 1 < d(A, B) ≤ 2.
Determine the maximum number of harminc pairs among 100 points in the plane.
2.4.1 Counting in Graph 42

Problem 2.4.12 (Swell coloring) : Let Kn denote the complete graph on n vertices, that is, the
graph with n vertices’s such that every pair of vertices’s is connected by an edge. A swell coloring
of Kn is an assignment of a color to each of the edges such that the edges of any triangle are
either all of distinct colors or all the same color. Further, more than one color must be used in total
(otherwise trivially if all edges are the same color we would have a swell coloring). Show that if Kn
√
can be swell colored with k colors, then k ≥ n + 1.

Solution. Concentrate on only one vertex.

Problem 2.4.13 (Belarus 2001) : Given n people, any two are either friends or enemies, and
friendship and enmity are mutual. I want to distribute hats to them, in such a way that any two
friends possess a hat of the same color but no two enemies possess a hat of the same color.
Each person can receive multiple hats. What is the minimum number of colors required to always
guarantee that I can do this?

Solution. In this problem, finding the worst case is a big help, because once the answer is
guessed, the things become really clear.

Problem 2.4.14 (ELMO 2017 P5) : The edges of K2017 are each labeled with 1, 2 or 3 such that
any triangle has sum of labels at least 5. Determine the minimum possible average of all labels.
(Here K2017 is defined as the complete graph on 2017 vertices’s, with an edge between every pair
of vertices’s.)

Solution. A starting idea to get the ans: if we discard of all the 2-edges, we see that in any
triangle, one edge has to be a 3-edge. So... Turan-kinda...

Solution. After getting the ans, and thinking about approaching inductively, if we remove only
one vertex, there will be pairs to consider. But if we remove two vertices, we will only need to
consider single vertices after the removal of these two vertices.

Now which pair of vertices are the best choice to remove? Before doing that, lets first think how
much change will we get in the sum after we remove two vertices. Since we have the ans, we do
quick maffs:

m(4m + 1) − (m − 1)(4m − 3) = 8m − 3 = 4 × (2m − 1) + 1

Doesn’t this indicate that we remove a 1-edge, so the other edges coming out of the two vertices
will sum up to be at least 4 ∗ (2m − 1).
2.4.1 Counting in Graph 43

Solution. The solution by bern is very pretty. What he probably had thought was:

If we pick a vertex, say u, and take an 1-edge from this vertex to another vertex v , we see
that there are at least as many 3-edges in u than there are 1-edges in v . Now if to get a more
accurate value of d3 (u) (defined naturally), we need to take the maximum of the values d1 (v )
for all v ’s connected to u.

Now we need to evaluate the number of 3 edges from the d1 values. Can we put a bound on
this sum? We have this lemma, does this help? Turns out that it does.

What left is to sum it all up to see if we can get the ans.

Problem 2.4.15 (ARO 2005 P9.4) : 100 people from 50 countries, two from each countries, stay
on a circle. Prove that one may partition them onto 2 groups in such way that neither no two
countrymen, nor three consecutive people on a circle, are in the same group.

Variant: There are 100 people from 25 countries sitting around a circular table. Prove that they
can be separated into four classes, so that no two countrymen are in the same class, nor any two
people sitting adjacent in the circle.

Solution. Thinking of the most natural way of eliminating the consecutive condition – pair two
consecutive verices.

Problem 2.4.16 (Romanian TST 2012 P4) : Prove that a finite simple planar graph has an orien-
tation so that every vertex has out-degree at most 3.

Problem 2.4.17 (USA TST 2006 P1) : A communications network consisting of some terminals
is called a 3-connector if among any three terminals, some two of them can directly communicate
with each other. A communications network contains a windmill with n blades if there exist n pairs
of terminals {x1 , y1 }, {x2 , y2 }, . . . , {xn , yn } such that each xi can directly communicate with the
corresponding yi and there is a hub terminal that can directly communicate with each of the 2n
terminals x1 , y1 , . . . , xn , yn . Determine the minimum value of f (n), in terms of n, such that a 3
-connector with f (n) terminals always contains a windmill with n blades.

Solution. Windmills won’t be there if among any 2n + 1 vertices, there were one vertex that
were not connected to any of the other 2n vertices. So that means that we are dealing Turan-
kinda config here. So we can make several ‘compact’ graphs that are mutually disconnected, and
each have at most 2n verices. Guessing from this, the ans is probably of some form k ∗ 2n + 1.
Now we have another condition to consider, 3-connector. Lets see, if we had 3 disconnected
2.4.2 Algorithms in Graph 44

componets, the resulting graph wouldn’t be a 3-connector. Done...

Problem 2.4.18 () : Graph G on n vertices has the property that the degree of every vertex is
greater than 2. Prove that for every 0 < k < n, there is a simple path with lenght at least n/k or,
k cycles, such that every cycle has at least one node which none of the other cycles has, and its
lenght is not divisible by 3.

Problem 2.4.19 (ISL 2005 C4) : Let n ≥ 3 be a fixed integer. Each side and each diagonal of a
regular n-gon is labelled with a number from the set {1; 2; ...; r } in a way such that the following
two conditions are fulfilled:
• Each number from the set {1, 2, . . . r } occurs at least once as a label.
• In each triangle formed by three vertices of the n-gon, two of the sides are labelled with the same
number, and this number is greater than the label of the third side.

1. Find the maximal r for which such a labelling is possible.

2. For this maximal value of r , how many such labellings are there?

Solution [Extremal]. Take the edges labeled with r , and delete them. Study what is left. For
the second part, formulate a recursive function, and try out small cases to find pattern.

2.4.2 Algorithms in Graph

Definition (Cut) — A cut is a partition of the vertices of a graph into two disjoint subsets. Any
cut determines a cut-set, the set of edges that have one endpoint in each subset of the partition.
These edges are said to cross the cut. In a connected graph, each cut-set determines a unique cut,
and in some cases cuts are identified with their cut-sets rather than with their vertex partitions.

(a) Minimum Cut (b) Maximum Cut

2.4.2 Algorithms in Graph 45

Theorem 2.4.7 (Prufer sequence) — Consider a labeled tree T with vertices’s {1, 2, ..., n}. At
step i , remove the leaf with the smallest label and set the i th element of the Prüfer sequence to
be the label of this leaf’s neighbour. Prove that a Prüfer sequence of length n − 2 defines a Tree
with length n.

Figure 2.6: A labeled tree with Prüfer sequence 4, 4, 4, 5.

Problem 2.4.20 (German TST 2004 E7P3) : We consider graphs with vertices colored black or
white. “Switching" a vertex means: coloring it black if it was formerly white, and coloring it white
if it was formerly black.

Consider a finite graph with all vertices colored white. Now, we can do the following operation:
Switch a vertex and simultaneously switch all of its neighbours (i. e. all vertices connected to this
vertex by an edge). Can we, just by performing this operation several times, obtain a graph with all
vertices colored black?

Solution. A classical example of creating complex moves from counter cases.

Problem 2.4.21 (ARO 2014 P9.8) : In a country of n cities, an express train runs both ways
between any two cities. For any train, ticket prices either direction are equal, but for any different
routes these prices are different. Prove that the traveler can select the starting city, leave it and
go on, successively, n − 1 trains, such that each fare is smaller than that of the previous fare. (A
traveler can enter the same city several times.)

Problem 2.4.22 (Generalization) : Let A be a set of n points in the space. From the family of all
segments with endpoints in A , q segments have been selected and colored yellow. Suppose that
all yellow segments are of different length. Prove that there exists a polygonal line composed of m
yellow segments, where m ≥ 2q n , arranged in order of increasing length.
2.4.2 Algorithms in Graph 46

Solution. Make one person go to every node. Then let the two people on the two sides of the
most expensive edge swap their position. This ensures that every edge was used exactly 2 times.
Using PHP, we have the desired result. Another solution is by this

Problem 2.4.23 () : Given a bipartite graph, prove that the minimum number of colors required to
color the edges of the graph such that no node is adjacent to 2 edges of same color is the maximum
degree of the graph.

Problem 2.4.24 () : For every bipartite graph prove that it’s edges can be bicolored so that each
deg
node is adjacent to atmost d e edges of any color.
2

Solution. Using the main property of a bipartite graph.

Solution. After finding the cycle solution, to optimize it, we recall that we can find a Eulerian
Path (if it exists) in O(V + E). Now we want to make the graph have a Eulerian path, so we add
a vertice to both sides of the graph, and join them with odd vertices from the other side.

Problem 2.4.25 (ISL 2005 C1) : A house has an even number of lamps distributed among its
rooms in such a way that there are at least three lamps in every room. Each lamp shares a switch
with exactly one other lamp, not necessarily from the same room. Each change in the switch shared
by two lamps changes their states simultaneously. Prove that for every initial state of the lamps
there exists a sequence of changes in some of the switches at the end of which each room contains
lamps which are on as well as lamps which are off.

Problem 2.4.26 (ISL 2013 C3) : A crazy physicist discovered a new kind of particle which he called
an i -mon, after some of them mysteriously appeared in his lab. Some pairs of i -mons in the lab
can be entangled, and each i -mon can participate in many entanglement relations. The physicist
has found a way to perform the following two kinds of operations with these particles, one operation
at a time.

1. If some i -mon is entangled with an odd number of other i -mons in the lab, then the physicist
can destroy it.

2. At any moment, he may double the whole family of i -mons in the lab by creating a copy I 0
of each i -mon I. During this procedure, the two copies I 0 and J 0 become entangled if and
2.4.2 Algorithms in Graph 47

only if the original i -mons I and J are entangled, and each copy I 0 becomes entangled with
its original i -mon I ; no other entanglements occur or disappear at this moment.

Prove that the physicist may apply a sequence of much operations resulting in a family of i -mons,
no two of which are entangled.

Solution. As there are an integer number of i -mons, it is quite natural to use induction. We
try to find an algorithm to reduce the number of particles.
Another way to do this is to consider the chromatic number of the graph. If we can show that
this number reduces after some move, then we are done by induction.

Problem 2.4.27 (ISL 2005 C2) : A forest consists of rooted (i. e. oriented) trees. Each vertex
of the forest is either a leaf or has two successors. A vertex v is called an extended successor of
a vertex u if there is a chain of vertices’s u0 = u, u1 , u2 . . . ut−1 , ut = v with t > 0 such that the
vertex ui+1 is a successor of the vertex ui for every integer i with 0 ≤ i ≤ t − 1.

Let k be a nonnegative integer. A vertex is called dynastic if it has two successors and each of
these successors has at least k extended successors.

Prove that if the forest has n vertices, then there are at most n
k+2 dynastic vertices.

Solution. Trying to apply induction, we realize the bound is very loosy. That’s why when we
try to add in the inductive step, the value becomes larger than the bound. To stop that overflow,
we tighten the bound.

Solution. The second and dummy approach is to first doing some smaller cases, finding small
infos, taking the root, seeing that the bound doesnt work, but it would work if one of the
successors of the root would have exactly or less than 2k + 3 successors. As we can’t always
guarantee that, we look for such a vertex with 2k + 3 successors. We do some work with it and
by induction its done.

Problem 2.4.28 (All Russia 2017 9.1) : In a country some cities are connected by oneway flights
(There are no more then one flight between two cities). City A called "available" for city B , if
there is flight from B to A , maybe with some transfers. It is known, that for every 2 cities P and
Q exist city R , such that P and Q are available from R. Prove, that exist city A , such that every
city is available for A.
2.4.2 Algorithms in Graph 48

Problem 2.4.29 (Jacob Tsimerman Induction) : There are 2010 ninjas in the village of Konoha
(what? Ninjas are cool.) Certain ninjas are friends, but it is known that there do not exist 3 ninjas
such that they are all pairwise friends. Find the maximum possible number of pairs of friends.(If
ninja A is friends with ninja B , then ninja B is also friends with ninja A.)

Problem 2.4.30 (USA TST 2011 D3P2) : Let n ≥ 1 be an integer, and let S be a set of integer
pairs (a, b) with 1 ≤ a < b ≤ 2n . Assume |S| > n · 2n+1 . Prove that there exists four integers
a < b < c < d such that S contains all three pairs (a, c), (b, d) and (a, d).

Solution. Using Induction to the first and last half of the set S shows us the hardest part of
the problem. Then ordering the left and right elements with some sort of hierarchy is all the work
left to do.

Problem 2.4.31 (ISL 2016 C6) : There are n ≥ 3 islands in a city. Initially, the ferry company
offers some routes between some pairs of islands so that it is impossible to divide the islands into
two groups such that no two islands in different groups are connected by a ferry route.

After each year, the ferry company will close a ferry route between some two islands X and Y . At
the same time, in order to maintain its service, the company will open new routes according to the
following rule: for any island which is connected to a ferry route to exactly one of X and Y , a new
route between this island and the other of X and Y is added.

Suppose at any moment, if we partition all islands into two nonempty groups in any way, then it is
known that the ferry company will close a certain route connecting two islands from the two groups
after some years. Prove that after some years there will be an island which is connected to all other
islands by ferry routes.

Solution. It is only natural to use induction on this kinda problems. After some trying, we
see that if we remove 1 node, We get to nowhere, but if we remove 2 nodes, we get something
interesting. So now focus on those two nodes and the rest of the nodes separately. Its not hard
from there.

Solution. As it seems, the separation of the graph was the main observation. We can call this
trick Bringing Order in the Chaos.
2.4.2 Algorithms in Graph 49

Problem 2.4.32 (ARO 2013 P9.5) : 2n real numbers with a positive sum are aligned in a circle.
For each of the numbers, we can see there are two sets of n numbers such that this number is on
the end. Prove that at least one of the numbers has a positive sum for both of these two sets.

Solution. Since there is nothing specfic about the sum, we may safely assume that it is 0,
because (1) probably it works, and (2) it makes things more convenient. How we do that? we
decrease every number by the average.

Now, Consider every block of n consecutive blockes of numbers. When are two blocks connected?
When they share the same end. What if we consider them as vertices, and this “connectivity” as
edges? We see that cycles pop out.

And we make use of the fact that our sum is 0. So signs are sure to bet flipped at the opposite
side, and there are odd and even -ness in cycles that we can use.

Problem 2.4.33 (USA TST 2011 P2) : In the nation of Onewaynia, certain pairs of cities are
connected by roads. Every road connects exactly two cities (roads are allowed to cross each other,
e.g., via bridges). Some roads have a traffic capacity of 1 unit and other roads have a traffic
capacity of 2 units. However, on every road, traffic is only allowed to travel in one direction. It is
known that for every city, the sum of the capacities of the roads connected to it is always odd. The
transportation minister needs to assign a direction to every road. Prove that he can do it in such a
way that for every city, the difference between the sum of the capacities of roads entering the city
and the sum of the capacities of roads leaving the city is always exactly one.

Solution. As there are two types of subgraph, 1 -type and 2 -type. By some work-arounds, we
see that we have to work distinctly in both types of graphs. Firstly, if we work in type- 1 , we
see after making a path from node x, y , the degrees of x, y will be {1, −1} and the degrees of
other nodes on the path will be the same. After that, we make every nodes have degree either
{1, −1}. So after this operation we remove the 1 -edges. Now, when dealing with the type- 2
sub-graph. Start over from zero, we see that when making a path between nodes x, y the degree
of those two changes parity, and other nodes on the path stays the same. So select two odd
nodes....

Solution. Dealing with two different kind of edges simultaneously is messy, so we work with
graph 1 and graph 2 differently. Now on both graphs, we can remove cycles. And in graph 2 , we
see that we can remove any big paths if there is a edge 1 joining the two endpoints. Since if the
new graph works then the previous graph works too. [Several cases to show here] And if there is
no edge joining the two endpoints, replace the path by joining the two endpoints by a edge 2.
2.4.2 Algorithms in Graph 50

Now there are only edge 1 s, and lone edge 2 s. Now dividing the graph 1 into paths of edge 1 ,
and dealing with several small cases, we are done.

Problem 2.4.34 (Iran TST 2009 P6) : We have a closed path that goes from one vertex to another
neighboring vertex, on the vertices of a n × n square which pass throgugh each vertex exactly once.
Prove that we have two adjacent vertices such that if we cut the path at these two points then the
length of each open paths is at least n2 /4.

Solution. Draw a path, doesn’t it look like a snake? Now can we relate the area of the tiled
path with its perimeter? If we could do that, we would be able to replace two neighboring vertices
by an edge inside the path, which seems to make the problem simpler.

Problem 2.4.35 (OC Chap2 P2) : Arutyun and Amayak perform a magic trick as follows. A
spectator writes down on a board a sequence of N (decimal) digits. Amayak covers two adjacent
digits by a black disc. Then Arutyun comes and says both closed digits (and their order). For which
minimal N can this trick always work? NOTE: Arutyun and Amayak have a strategy determined
beforehand.

Solution. We have to actually find a bijection between all of the combinations the spectator
can create, and all of the combinations that Arutyun might see when he comes back. Which tells
us to use “Perfect Matching" tricks.

Solution. Existential proof: for this trick to always work, they have to make a bijection from
a set of N digits with two covered, to an unique set of N digits. Consider a bijection from the
set of 0 − 9 strings with length N to the set of 0 − 9 strings with length N with 2 adjacent digits
unknown. There exist a bijection iff the two sets satisfy Hall’s Marriage Theorem. By double
counting we get the value of N from here.

Problem 2.4.36 (Simurgh 2019 P3) : Call a graph symmetric, if one can put its vertices on the
plane such that it becomes symmetric wrt a line (which doesn’t pass through any vertex). Find the
minimum value of k such that (the edges of) every graph on 100 vertices, can be decomposed into
k symmetric subgraph.

Problem 2.4.37 (RMM 2020 P3) : Let n ≥ 3 be an integer. In a country there are n airports and
n airlines operating two-way flights. For each airline, there is an odd integer m ≥ 3, and m distinct
airports c1 , . . . , cm , where the flights offered by the airline are exactly those between the following
pairs of airports: c1 and c2 ; c2 and c3 ; . . . ; cm−1 and cm ; cm and c1 .
2.4.2 Algorithms in Graph 51

Prove that there is a closed route consisting of an odd number of flights where no two flights are
operated by the same airline.

Solution [Weird Induction]. Fix one vertice, merge all neighbors with it that has a unique airline
between them.

Solution [Element of Time]. Add one edge from each cycle one at a time, without creating a
cycle. Our objective is to show that when we reach the maximum stage where one edge creates
a cycle, that cycle is of odd length.
2.5 Game Theory 52

2.5 Game Theory

• Zawad’s Game Theory Pset

2.5.1 Games

Definition (Nimbers) — Nimbers are simply ‘Nim values’ which are assigned to a game configu-
ration - these values are written as 0, ∗1, ∗2, ∗3 . . . We shall first describe how to obtain the Nim
values for the game Squaring the Number. First, the Nim value of n = 0 is assigned 0, since it is
a state in which neither player has a valid move. We then recursively adopt the following rule for
each n : find all the possible moves from n and pick the smallest Nim value which
does not occur among all these possible moves.

Theorem 2.5.1 (Sprague-Grund Theorem) — The Sprague–Grundy theorem states that every
impartial game under the normal play convention is equivalent to a nimber.

Game (Chip Firing Game) — Let G = (V, E) be a graph without any loops or multiedges. Let a
number of si chips be stacked on vertex i . The game follows with the player choosing a vertex i ,
taking di chips from it (si − di > 0), and sending one chip to each of the neighbors of the vertex
where di is the degree of i . The Problem of this game is to determine when the game will be infinte.

• If N is the total number of edges in G, and S is the total number of chips, then

Game (Cutting a stack in half) — Given a number of stacks, at his/her move, a player can choose
a stack with even number stones, and divide it in two stacks with the same number of stones.

Game (Cutting a stack in several) — Given a number of stacks, at his/her move, a player can
choose a stack, and divide it in several stacks with the same number of stones.
2.5.2 Problems 53

2.5.2 Problems

Problem 2.5.1 (USAMO 2008 P5) : Three non-negative real numbers r1 , r2 , r3 are written on
a blackboard. These numbers have the property that there exist integers a1 , a2 , a3 , not all zero,
satisfying a1 r1 + a2 r2 + a3 r3 = 0. We are permitted to perform the following operation: find two
numbers x , y on the blackboard with x ≤ y , then erase y and write y − x in its place. Prove that
after a finite number of such operations, we can end up with at least one 0 on the blackboard.

Solution. When can’t get info out of the reals, try the integers. Observe the integers, and
check if they have any invariant. Rule of thumb of finding an invariant.

Problem 2.5.2 (USAMO 2014 P1) : Let k be a positive integer. Two players A and B play a
game on an infinite grid of regular hexagons. Initially all the grid cells are empty. Then the players
alternately take turns with A moving first. In his move, A may choose two adjacent hexagons in the
grid which are empty and place a counter in both of them. In his move, B may choose any counter
on the board and remove it. If at any time there are k consecutive grid cells in a line all of which
contain a counter, A wins. Find the minimum value of k for which A cannot win in a finite number
of moves, or prove that no such minimum value exists.

Solution. Trying to block A. We see that if we could alternately color the points black and
white, we could’ve found some strategy for B. But the triangle grid doesn’t seem very friendly.
How can we color the triangles? And don’t forget the details idiot.

Problem 2.5.3 (Indian TST 2004) : The game of pebbles is played as followed: Initially there is
one pebble at (0, 0). In a move one can remove the pebble at (i , j) and put one pebble each at
(i + 1, j) and (i , j + 1) , given that both (i + 1, j) and (i , j + 1) were empty. Prove that at any point
in the game, there will be a pebble at some lattice point (a, b) with a + b ≤ 3.

Solution. Two from one, means if the weight is reduced by half in the second level, then the
sum would be the same.

Problem 2.5.4 (ISL 1998 C7) : A solitaire game is played on an m × n rectangular board, using
mn markers which are white on one side and black on the other. Initially, each square of the board
contains a marker with its white side up, except for one corner square, which contains a marker
with its black side up. In each move, one may take away one marker with its black side up, but
must then turn over all markers which are in squares having an edge in common with the square of
2.5.2 Problems 54

the removed marker. Determine all pairs (m, n) of positive integers such that all markers can be
removed from the board.

Solution. If we remove one marker, then this cell becomes useless. So the neighbors to this
cell will act like they are not connected to this cell. Now if a cell is connected to w white cells,
and b black cells, then the resulting board state will have b − w more cells. Now only this info
doesn’t build up an invariant. Notice that as we are doing moves, we are reducing neighborhood
relations as well, in other words, neighborhood relations decrease by b + w . So if we consider the
sum W + E where W is the number of all white cells, and E is the number of all neighborhood
relations, we get an invariant on this value.

Problem 2.5.5 (ARO 1999 P10.1) : There are three empty jugs on a table. Winnie the pooh,
Rabbit, and Piglet put walnuts in the jugs one by one. They play successively, with the initial
determined by a draw. Thereby Winnie the pooh plays either in the first or second jug, Rabbit in
the second or third, and Piglet in the first or third. The player after whose move there are exactly
1999 walnuts loses the games. Show that Winnie the pooh and Piglet can cooperate so as to make
Rabbit lose.

Problem 2.5.6 (USAMO 2004, P4) : Alice and Bob play a game on a 6 by 6 grid. On his or her
turn, a player chooses a rational number not yet appearing in the grid and writes it in an empty
square of the grid. Alice goes first and then the players alternate. When all squares have numbers
written in them, in each row, the square with the greatest number in that row is colored black.
Alice wins if she can then draw a line from the top of the grid to the bottom of the grid that stays
in black squares, and Bob wins if she can’t. (If two squares share a vertex, Alice can draw a line
from one to the other that stays in those two squares.) Find, with proof, a winning strategy for
one of the players.

Problem 2.5.7 (RMM 2019 P1) : Amy and Bob play the game. At the beginning, Amy writes
down a positive integer on the board. Then the players take moves in turn, Bob moves first. On
any move of his, Bob replaces the number n on the blackboard with a number of the form n − a2 ,
where a is a positive integer. On any move of hers, Amy replaces the number n on the blackboard
with a number of the form nk , where k is a positive integer. Bob wins if the number on the board
becomes zero. Can Amy prevent Bob’s win?

Solution. Decent.

Problem 2.5.8 (ISL 2015 C4) : Let n be a positive integer. Two players A and B play a game in
which they take turns choosing positive integers k ≤ n. The rules of the game are:
2.5.2 Problems 55

1. A player cannot choose a number that has been chosen by either player on any previous turn.

2. A player cannot choose a number consecutive to any of those the player has already chosen
on any previous turn.

3. The game is a draw if all numbers have been chosen; otherwise the player who cannot choose
a number anymore loses the game.

The player A takes the first turn. Determine the outcome of the game, assuming that both players
use optimal strategies.

Solution. Look at the simplest things, first produce data like a good boy, and then see what A
has to do to win, or at least draw that he can’t because B is an asshole.

Problem 2.5.9 (ISL 2012 C4) : Players A and B play a game with N ≥ 2012 coins and 2012 boxes
arranged around a circle. Initially A distributes the coins among the boxes so that there is at least
1 coin in each box. Then the two of them make moves in the order B, A, B, A, . . . by the following
rules:

1. On every move of his B passes 1 coin from every box to an adjacent box.

2. On every move of hers A chooses several coins that were not involved in B’s previous move
and are in different boxes. She passes every coin to and adjacent box.

Player A’s goal is to ensure at least 1 coin in each box after every move of hers, regardless of how
B plays and how many moves are made. Find the least N that enables her to succeed.

Solution. Investigate B’s move, see how and where he can make 0’s

Problem 2.5.10 (ISL 2009 C1) : Consider 2009 cards, each having one gold side and one black
side, lying on parallel on a long table. Initially all cards show their gold sides. Two player, standing
by the same long side of the table, play a game with alternating moves. Each move consists of
choosing a block of 50 consecutive cards, the leftmost of which is showing gold, and turning them
all over, so those which showed gold now show black and vice versa. The last player who can make
a legal move wins.
1. Does the game necessarily end?
2. Does there exist a winning strategy for the starting player?

Remark. Trying out small cases doesn’t help in this problem. Rather exploring what inevitably has
to happen helps to notice patterns.
2.5.2 Problems 56

Solution. The first part of the problem is trivial induction usage.

For the second half, notice that card 1 has to chosen only once. And cards 2, . . . 50 have to
chosen an even number of times each. Again, card 51 has to be taken an odd number of times.
Inductively, cards 1, 51 . . . 1951 were chosen odd number of times each, and all other cards were
chosen even number of times each. Since that means the parity of total number of moves in this
game is even.
2.6 Combinatorial Geometry 57

2.6 Combinatorial Geometry

• Combinatorial Geometry - Maria Monk (MOP 2010)

Stuck? Try These

• Consider the convex hull made up of the points.

• Consider the extreme points: smallest or highest x or y coordinate.

• Find the triangle (quadrilateral, pentagon, etc.) with the vertices being the points from
your set S, so that the area of the triangle is minimal/maximal.

Theorem 2.6.1 (Helly’s Theorem) — Let X1 , ..., Xn be a finite collection of convex subsets of
Rd , with n > d. If the intersection of every d + 1 of these sets is nonempty, then the whole
collection has a nonempty intersection; that is,
n
\
Xj 6= ∅
j=1
2.6 Combinatorial Geometry 58

Problem 2.6.1 (ARO 2013 P9.4) : N lines lie on a plane, no two of which are parallel and no three
of which are concurrent. Prove that there exists a non-self-intersecting broken line A1 A2 A3 . . . AN
with N parts, such that on each of the N lines lies exactly one of the N segments of the line.

Problem 2.6.2 (EGMO 2017 P3) : There are 2017 lines in the plane such that no three of them go
through the same point. Turbo the snail sits on a point on exactly one of the lines and starts sliding
along the lines in the following fashion: she moves on a given line until she reaches an intersection
of two lines. At the intersection, she follows her journey on the other line turning left or right,
alternating her choice at each intersection point she reaches. She can only change direction at
an intersection point. Can there exist a line segment through which she passes in both directions
during her journey?

Solution. The condition that tells us to go either right or left, seems very non-rigorous. So to
rigorize this condition, instead of using right or left condition in the direction, we consider what’s
on our right and left. (INTUITION) After some experiment we see (not all of us) that if we color
the plane with two colors in a way where every neighboring regions have different colors, we find
some interesting stuff. (CREATIVITY) With this we are done. Color the Plane

Problem 2.6.3 (ISL 2006 C2) : Let P be a regular 2006 -gon. A diagonal is called good if its
endpoints divide the boundary of P into two parts, each composed of an odd number of sides of P .
The sides of P are also called good.
Suppose P has been dissected into triangles by 2003 diagonals, no two of which have a common
point in the interior of P . Find the maximum number of isosceles triangles having two good sides
that could appear in such a configuration.

Solution. The straight way, induction.

Solution. The intuitive way, bijection. There are at most n good triangles, there are 2n edges,
so a mapping that takes a two edges to a single good triangle must exist. Finding it is not that
hard.

Problem 2.6.4 (ARO 2014 P9.3) : In a convex n -gon, several diagonals are drawn. Among these
diagonals, a diagonal is called good if it intersects exactly one other diagonal drawn (in the interior
of the n -gon). Find the maximum number of good diagonals.
2.6 Combinatorial Geometry 59

Solution. There can be two cases, two good diagonals intersecting each other, and no two
good diagonals intersecting each other. In the first case, we just use induction, and in the later,
all of the good diagonals create a “triangulation” of the polygon, which gives us the numbers.

Problem 2.6.5 (ISL 2013 C2, IMO 2013 P2) : A configuration of 4027 points in the plane is
called Colombian if it consists of 2013 red points and 2014 blue points, and no three of the points
of the configuration are collinear. By drawing some lines, the plane is divided into several regions.
An arrangement of lines is good for a Colombian configuration if the following two conditions are
satisfied:

1. No line passes through any point of the configuration.

2. No region contains points of both colors.

Find the least value of k such that for any Colombian configuration of 4027 points, there is a good
arrangement of k lines.

Solution. Obviously a n00b would think about induction. The only problem occurs when the
convex hull completely consists of red points. In this case, after some investigation, we should
get the sandwiching two points idea.

Solution. Another way of inductive approach is like this, as the problem condition says that no
region contains points of both colors, which means if we connect any two red and blue points,
some line must bisect this segment. Now it is known that there is non intersecting partition of
the points in to red-blue segments. So suppose in such a partition, we draw bisectors of each
segments. Now there will be some holes in this proof. We see that to fill these holes, we have to
focus on two red points with their respective blue partners, and draw the two bisectors in a way
that separates the two red points form the blue points. So to remove further holes, we get the
sandwiching idea.

Problem 2.6.6 (ILL 1985) : Let A and B be two finite disjoint sets of points in the plane such that
no three distinct points in A ∪ B are collinear. Assume that at least one of the sets A, B contains
at least five points. Show that there exists a triangle all of whose vertices’s are contained in A or
in B that does not contain in its interior any point from the other set.

Solution. Concentrating on one of the sets five points such that there is no other points of
the same set inside the hull of those five points.
2.6 Combinatorial Geometry 60

Problem 2.6.7 (APMO 1999 P5) : Let S be a set of 2n + 1 points in the plane such that no
three are collinear and no four concyclic. A circle will be called “Good” if it has 3 points of S on its
circumference, n − 1 points in its interior and n − 1 points in its exterior. Prove that the number of
good circles has the same parity as n.

Solution. When thinking about induction, got a feeling that double counting with the number
of good circles going through pairs of points might be useful, because a good circle will be counted
three times, if we can show that every pair has odd number of good circles, we are done. So, take
a pair. Now we need to ‘sort’ the points somehow. See that, we can’t sort the points in a trivial
way with numbers, so moving to angles. Now setting conditions for a point inside of a circle in
terms of angles, we see amazing patter, and an easy way to calculate the number of good circle
of that pair of points.

Problem 2.6.8 (ISL 2014 C1) : Let n points be given inside a rectangle R such that no two of
them lie on a line parallel to one of the sides of R. The rectangle R is to be dissected into smaller
rectangles with sides parallel to the sides of R in such a way that none of these rectangles contains
any of the given points in its interior. Prove that we have to dissect R into at least n + 1 smaller
rectangles.

Solution. Work with the largest continuous segments, and their endpoints.

Problem 2.6.9 (ISL 2007 C2) : A rectangle D is partitioned in several (≥ 2) rectangles with sides
parallel to those of D. Given that any line parallel to one of the sides of D, and having common
points with the interior of D, also has common interior points with the interior of at least one
rectangle of the partition; prove that there is at least one rectangle of the partition having no
common points with D’s boundary.

Solution. There existing such a rectangle means that there is a rectangular region inside of the
original rectangle. So what if we walked along the segments, and cut a smaller rectangle from
the inside of the rectangle? Like the way in the game.
2.6 Combinatorial Geometry 61

Figure 2.7

Solution. Starting from one corner, and taking the opposite corner of the rectangle containing
that corner, we use infinite decent to reach a contradiction.

Solution. Using ISL 2014 C1 as a lemma.

Solution. Take one side of the square. Take a “sandwiched” rectangle touching that side. If
no such rectangle exists, then it’s just a special case that can be dealt with ease.

Problem 2.6.10 (ISL 2003 C2) : Let D1 , D2 . . . , Dn be closed discs in the plane. (A closed disc is
the region limited by a circle, taken jointly with this circle.) Suppose that every point in the plane
is contained in at most 2003 discs Di . Prove that there exists a disc Dk which intersects at most
7 · 2003 − 1 = 14020 other discs Di .

Solution. Just go with the natural idea.

2.6 Combinatorial Geometry 62

Problem 2.6.11 (ISL 2003 C3) : Let n ≥ 5 be a given integer. Determine the greatest integer k
for which there exists a polygon with n vertices (convex or not, with non-selfintersecting boundary)
having k internal right angles.

Solution. double count

Problem 2.6.12 (Tournament of Towns 2015S S4) :

Solution. Just use what’s the most natural, POP, on one vertex point.

Problem 2.6.13 (USAMO 2007 P2) : A square grid on the Euclidean plane consists of all points
(m, n), where m and n are integers. Is it possible to cover all grid points by an infinite family of
discs with non-overlapping interiors if each disc in the family has radius at least 5?

Problem 2.6.14 (MEMO 2015 T4) : Let N be a positive integer. In each of the N 2 unit squares
of an N × N board, one of the two diagonals is drawn. The drawn diagonals divide the N × N board
into K regions. For each N, determine the smallest and the largest possible values of K.

Figure 2.8
2.6 Combinatorial Geometry 63

Solution. An Algorithmic Approach: Consider each diagonal as 0 or 1, prove that the maximum
configuration is the one with alternating 0, 1s and the minimum one is the one with all 0s.

Solution. A Counting Approach: Just count and bound with the minimum areas of the regions.

Problem 2.6.15 () : In every cells of a m × n grid, one of the two diagonals are drawn. Prove that
there exist a path on these diagonals from left to right or from up to bottom of the grid.

Solution. First remove the cycles, then take the largest path from left to right, and use
induction.

Problem 2.6.16 (Math Price for Girls 2017 P4) : A lattice point is a point in the plane whose two
coordinates are both integers. A lattice line is a line in the plane that contains at least two lattice
points. Is it possible to color every lattice point red or blue such that every lattice line contains
exactly 2017 red lattice points? Prove that your answer is correct.

Solution. Transfinite induction.

Problem 2.6.17 (China TST 2016 T3P2) : In the coordinate plane the points with both coordinates
being rational numbers are called rational points. For any positive integer n, is there a way to use n
colours to colour all rational points, every point is coloured one colour, such that any line segment
with both endpoints being rational points contains the rational points of every colour?

Solution. Transfinite induction

Problem 2.6.18 (IGO 2018 A3) : Find all possible values of integer n > 3 such that there is a
convex n-gon in which, each diagonal is the perpendicular bisector of at least one other diagonal.

Solution. Taking maximum terminal triangle.

Problem 2.6.19 (Lithuania ??) : Prove that in every polygon there is a diagonal that cuts off a
triangle and lies completely within the polygon.

Problem 2.6.20 (Romanian TST 2008 T1P4) : Prove that there exists a set S of n − 2 points
inside a convex polygon P with n sides, such that any triangle determined by 3 vertices of P contains
2.6 Combinatorial Geometry 64

exactly one point from S inside or on the boundaries.

Solution. Checking small cases inductively quickly shows a construction.

Problem 2.6.21 (Iran TST ??) : In an isosceles right-angled triangle shaped billiards table, a ball
starts moving from one of the vertices adjacent to hypotenuse. When it reaches to one side then
it will reflect its path. Prove that if we reach to a vertex then it is not the vertex at initial position

Problem 2.6.22 (APMO 2018 P4) : Let ABC be an equilateral triangle. From the vertex A we
draw a ray towards the interior of the triangle such that the ray reaches one of the sides of the
triangle. When the ray reaches a side, it then bounces off following the law of reflection, that is, if
it arrives with a directed angle α, it leaves with a directed angle 180◦ − α. After n bounces, the ray
returns to A without ever landing on any of the other two vertices. Find all possible values of n.

Solution. Reflect the whole board when just reflecting the ball doesn’t seem to be helping.
GLOBAL

Problem 2.6.23 (ISL 2007 C5) : In the Cartesian coordinate plane define the strips Sn = {(x, y )|n ≤
x < n + 1}, n ∈ Z and color each strip black or white. Prove that any rectangle which is not a
square can be placed in the plane so that its vertices have the same color.

Solution. Proceed step by step. See what happens if the parity of a, b are different. Then the
case with two coprimes. In this case, we want to tilt the rectangle to some extent where the
desired result is achieved. We just need to show that this is possible. A bit of wishful thinking
and a bit of algebra does the rest.

Problem 2.6.24 (APMO 2018 P3) : A collection of n squares on the plane is called tri-connected
if the following criteria are satisfied:

1. All the squares are congruent.

2. If two squares have a point P in common, then P is a vertex of each of the squares.

3. Each square touches exactly three other squares.

How many positive integers n are there with 2018 ≤ n ≤ 3018, such that there exists a collection
of n squares that is tri-connected?
2.6 Combinatorial Geometry 65

Solution. Play around to find that 6k for k > 4 is good. Then play around a little bit more for
a different construction. Another construction for 6k gives rise to a construction for 10k. Which
integers can be written as a sum of 6k and 10k?

Problem 2.6.25 (Iran 2005) : A simple polygon is one where the perimeter of the polygon does
not intersect itself (but is not necessarily convex). Prove that a simple polygon P contains a
diagonal which is completely inside P such that the diagonal divides the perimeter into two parts
both containing at least n3 − 1 vertices. (Do not count the vertices which are endpoints of the
diagonal.)

Solution. Triangulate.

Problem 2.6.26 (ISL 2008 C3) : In the coordinate plane consider the set S of all points with
integer coordinates. For a positive integer k, two distinct points a, B ∈ S will be called k-friends
if there is a point C ∈ S such that the area of the triangle ABC is equal to k. A set T ⊂ S will
be called k-clique if every two points in T are k-friends. Find the least positive integer k for which
there exits a k-clique with more than 200 elements.

Solution. When does ax + by = c have integer solution? Fix one point as origin and check
other points friendliness with other points.

Problem 2.6.27 (ISL 2015 C2) : We say that a finite set S of points in the plane is balanced if,
for any two different points A and B in S, there is a point C in S such that AC = BC. We say that
S is centre-free if for any three different points A, B and C in S, there is no points P in S such
that P A = P B = P C.

1. Show that for all integers n ≥ 3, there exists a balanced set consisting of n points.

2. Determine all integers n ≥ 3 for which there exists a balanced centre-free set consisting of n
points.

Solution. Simple, think about circles, then think about “center-free” in a graph theoritic manner.
2.6.1 Chessboard Pieces 66

2.6.1 Chessboard Pieces

Lemma 2.6.2 — What is the maximum number of knights that can be placed on a chessboard
such that no two knights attack each other?

Solution. A knight’s move always changes the color of the cell.

Problem 2.6.28 (IMO 2018 P4) : A site is any point (x, y ) in the plane such that x and y are
both positive integers less than or equal to 20.

Initially, each of the 400 sites is unoccupied. Amy and Ben take turns placing stones with Amy
going first. On her turn, Amy places a new red stone on an unoccupied
√ site such that the distance
between any two sites occupied by red stones is not equal to 5. On his turn, Ben places a new
blue stone on any unoccupied site. (A site occupied by a blue stone is allowed to be at any distance
from any other occupied site.) They stop as soon as a player cannot place a stone.

Find the greatest K such that Amy can ensure that she places at least K red stones, no matter
how Ben places his blue stones.

Solution. Using the maximum knight problem as a lemma.

Problem 2.6.29 () : How many rooks can be placed on an n × n board such that each rook attacks
at most one other rook?

Solution. Use graphs with one set of degrees being rows, and the other set of degrees being
columns.

Problem 2.6.30 (Eight queens puzzle) : How many queens can be placed on an n × n board such
that no queen attacks another queen?
2.6.1 Chessboard Pieces 67

Figure 2.9

Figure 2.10
Figure 2.11

Problem 2.6.31 (Serbia National D2P2) : How many queens can be placed on an n × n board such
that each queen attacks at most one other queen?

Problem 2.6.32 (BdMO 2019 P10) : Define a new chess piece named warrior. it can either go
three steps forward and one step to the side, or t2wo steps forward and two steps to the side in
some orientation. In a 2020 × 2020 chessboard, prove that the mazimum number of warriors so
2
that none of them attack each other is leass htan or equal to of the number of cells.
5

Solution. Color and partition

Problem 2.6.33 (RMM 2019 P4) : Prove that for every positive integer n there exists a (not
necessarily convex) polygon with no three collinear vertices, which admits exactly n diffferent trian-
gulations.
(A triangulation is a dissection of the polygon into triangles by interior diagonals which have no
common interior points with each other nor with the sides of the polygon)
2.6.1 Chessboard Pieces 68

Figure 2.12: Fixes

Solution.

Problem 2.6.34 (China TST 2015 T1D2P1) : Prove that : For each integer n ≥ 3, there exists
the positive integers a1 < a2 < · · · < an , such that for i = 1, 2, · · · , n − 2 , With ai , ai+1 , ai+2 may
be formed as a triangle side length , and the area of the triangle is a positive integer.

Solution. First of all we dont need to limit us to integers, we can work with rationals. We
want to build a4 from a1 , a2 , a3 . with a4 > a3 while keeping the area rational i.e. keeping the
height and base rational.

Problem 2.6.35 (Codeforces 1158D) : You are given n points on the plane, and a sequence S of
length n − 2 consisting of L and R. You need to generate a sequence of the points a1 , a2 . . . an
such that

• the polyline a1 a2 . . . an is not self intersecting.

• the directed segment ai+1 ai+2 is on the left side of the the directed segment ai ai+1 if Si = L,
and on the right side if Si = R.
2.7 Sequences 69

2.7 Sequences

2.7.1 Lemmas

Theorem 2.7.1 (Van der Waerden’s Theorem) — For any given positive integersr and k, there
is some number N such that if the integers {1, 2, ..., N} are colored, each with one of r different
colors, then there are at least k integers in arithmetic progression all of the same color.

2.7.2 Generating Function Lemmas

Theorem 2.7.2 (Catalan Recursion) — The infinite series defined as following:

n
Y
a0 = a1 = 1, an = ai an−i+1 = a0 an−1 + a1 an−2 · · · + an−1 a0
i=0

1 2n
has the general term an = Cn =


n+1 n

1 Problem to do with generating function

2.7.3 Sequence Problems

Sequences - Alexander Remorov

Problem 2.7.1 (ISL 1990) : Assume that the set of all positive integers is decomposed into r
(disjoint) subsets A1 ∪ A2 ∪ · · · ∪ Ar = N. Prove that one of them, say Ai , has the following
property: There exists a positive m such that for any k one can find numbers a1 , a2 , . . . , ak in Ai
with 0 < aj+1 − aj ≤ m, (1 ≤ j ≤ k − 1).
2.7.3 Sequence Problems 70

Problem 2.7.2 (Result by Erdos, Dividing the integers into arithmetic progressions) : Let d1 , d2 , . . . , dk
be differences of k arithmetic progressions that partition N. Show that di = dj for some i , j.

Problem 2.7.3 (APMO 1999 P1) : Find the smallest positive integer n with the following property:
there does not exist an arithmetic progression of 1999 real numbers containing exactly n integers.

Problem 2.7.4 (APMO 1999 P2) : Let a1 , a2 , . . . be a sequence of real numbers satisfying ai+j ≤
ai + aj for all i , j = 1, 2, . . . . Prove that
a2 a3 an
a1 + + + ··· + ≥ an
2 3 n
for each positive integer n.

a2
Problem 2.7.5 (ISL 1994 A1) : Let a0 = 1994 and an+1 = an +1 n
for each nonnegative integer n.
Prove that 1994 − n is the greatest integer less than or equal to an , 0 ≤ n ≤ 998

Solution. Take the differences.

Problem 2.7.6 (ISL 2007 C4) : Let A0 = (a1 , . . . , an ) be a finite sequence of real numbers. For
each k ≥ 0, from the sequence Ak = (x1 , . . . , xk ) we construct a new sequence Ak+1 in the following
way.

1. We choose a partition {1, . . . , n} = I ∪ J, where I and J are two disjoint sets, such that the
expression
X X
xi − xj
i∈I j∈J

attains the smallest value. (We allow I or J to be empty; in this case the corresponding sum
is 0.) If there are several such partitions, one is chosen arbitrarily.

2. We set Ak+1 = (y1 , . . . , yn ) where yi = xi + 1 if i ∈ I, and yi = xi − 1 if i ∈ J.

Prove that for some k, the sequence Ak contains an element x such that |x| ≥ n2 .

Solution. Suppose the contrary. Now, since Ai can only attain finite values, So Ai = Aj for
some i, j. Now, we are taking about changes here, so we need to think of some invariants. Firstly
2.7.3 Sequence Problems 71

the sum, it’s not much of an help, because it doesn’t give us much control. So kinda sum-ish
invariant with a bit more control is the sum of squares. We combine these two ideas.

Problem 2.7.7 (ISL 2009 A6) : Suppose that s1 , s2 , s3 , . . . is a strictly increasing sequence of
positive integers such that the sub-sequences
ss1 , ss2 , ss3 , . . . and ss1 +1 , ss2 +1 , ss3 +1 , . . .
are both arithmetic progressions. Prove that the sequence s1 , s2 , s3 , . . . is itself an arithmetic pro-
gression.

Solution. First notice that the two arithmetic sequences has the same common difference.
Then notice that the diffrences of the original sequence is bounded. Another advice, give every-
thing names. After naming the smallest difference and the largest differnce, we get two different
inequalities, from where we deduce that the difference is constant.

Problem 2.7.8 (ISL 2013 N2) : Assume that k and n are two positive integers. Prove that there
exist positive integers m1 , . . . , mk such that
2k − 1
   
1 1
1+ = 1+ ··· 1 + .
n m1 mk

Solution. Just induct, and think wishfully.

Problem 2.7.9 (USAMO 2007 P1) : n be a positive integer. Define a sequence by setting a1 = n
and, for each k > 1, letting ak be the unique integer in the range 0 ≤ ak ≤ k − 1 for which
a0 + a1 · · · + ak is divisible by k. Prove that for any n the sequence ai eventually becomes constant.

Solution. Investigate and done.

Problem 2.7.10 (APMO 2015 P3) : A sequence of real numbers a0 , a1 , ... is said to be good if the
following three conditions hold.

1. The value of a0 is a positive integer.

ai
2. For each non-negative integer i we have ai+1 = 2ai + 1 or ai+1 =
ai + 2
3. There exists a positive integer k such that ak = 2014.

Find the smallest positive integer n such that there exists a good sequence a0 , a1 , ... of real numbers
with the property that an = 2014.
2.7.3 Sequence Problems 72

Solution. We will rename the sequence and call it {si }, with s0 = x ∈ Z. Now let,

ai x + bi
si =
ci x + di
At the beginning we have a0 = d0 = 1, b0 = c0 = 0. It is easy to prove that

ai+1 + ci+1 = 2(ai + ci )

bi+1 + di+1 = 2(bi + di )

So it follows that
ai + ci = 2i = bi + di
Also, by induction (which is easy to prove), we have

ai − bi = di − ci = 1

Suppose for some k, sk = 2014. So,

ak x + bk
= 2014 (2.3)
ck x + dk

2014dk − bk 2015dk − 2k
=⇒ x = = k (2.4)
ak − 2014ck 2 − 2015ck

2015(dk − ck )
= −1 (2.5)
2k − 2015ck

But gcd(2015, 2k − 2015ck ) = 1, which implies 2k − 2015 − ck = 1. Solving for k with CRT
gives us 60|k.
Now we have to prove that there is a sequence with s60 = 2014. Solving (1), s0 = 2014, and,

2014 · 260 + 1 2014 · 260 − 2014

a60 = b60 =
2015 2015

260 − 1 260 + 2014

c60 = d60 =
2015 2015
We show that we can make (a60 , c60 ) from (a0 , c0 ) = (1, 0). We prove it by induction, that
(ak , ck ) can take any form (2k − i , i ) with i ∈ {0, 1, . . . 2k − 1}.
2.7.4 Recurrent Sequences 73

2.7.4 Recurrent Sequences

• WOOT 2010-11 Recursion

Theorem 2.7.1
Sum of Geometric Sequences Every recurrent sequence can be written as a sum of some
geometric sequences. Given a recurrent sequence,

xn = a1 xn−1 + a2 xn−2 + · · · + ak xn−k

Then xn can be written as

xn = c1 r1n + c2 r2n + · · · + cl rln
For all ci if ri are the roots of the characteristic polynomial of the recursion. Which is:

x k − a1 x k−1 − a2 x k−2 · · · − ak = 0

If there are double roots, say r1 = r2 = r3 , then we instead have,

xn = c1 r1n + c2 n r2n + c3 n2 r n · · · + cl rln

Reversely, we can say that a sequence defined by a sum of geometric recurrent series is a
recursion.

Lemma 2.7.3 — Let Fn be the nth Fibonacci number. Then the following holds:

Fn2 + Fn+1
2
= F2n+1

Proof. Expanding the general form of the terms, and showing that an = Fn2 + Fn+1
2 − F2n+1 is
a recursion by Theorem 2.7.1.

Theorem 2.7.2 (title=Repertoire Method)

Given a recurrent function defined by

f (n) = A(n)α + B(n)β + C(n)γ

2.7.4 Recurrent Sequences 74

We plug in different values for f (n), for example, f (n) = 1, n, 2n etc. for which the values
are known from the recursion, and then solve for A, B, C.
2.8 Exploring Configurations 75

2.8 Exploring Configurations

Problems where there is some kind of a configuration is given, the question usually asks to proof or
find some specific properties of the configuration.

2.8.1 Problems

Problem 2.8.1 (ARO 2018 P11.5) : On the table, there’re 1000 cards arranged on a circle. On
each card, a positive integer was written so that all 1000 numbers are distinct. First, Vasya selects
one of the card, remove it from the circle, and do the following operation: If on the last card taken
out was written positive integer k, count the k th clockwise card not removed, from that position,
then remove it and repeat the operation. This continues until only one card left on the table. Is it
possible that, initially, there’s a card A such that, no matter what other card Vasya selects as first
card, the one that left is always card A?

Remark. We want to find a configuration, where one card, let’s call it a, gets skipped over all the
time. Now, controlling the skipping for every move is kinda hard. Instead of doing that, we want
to control only one move that skips a, and all other moves will go to the next card in the clockwise
rotation. And the card before a will skip over a, and land of the next card. That being said, the
construction is rather trivial.

Solution. Consider the numbering,

1, 1001! + 1, 1002! + 1, . . . 1998! + 1, 1999! + 2

It’s easy to check that it works.

Problem 2.8.2 (APMO 2017 P3) P : Let A(n) denote the number of sequences a1 ≥ a2 ≥ · · · ≥ ak
of positive integers for which ki=1 ak = n and each ai + 1 is a power of two. Let B(n) denote
the number of sequences b1 ≥ b2 ≥ · · · ≥ bk of positive integers for which ki=1 bk = n and
P

each inequality bj ≥ 2bj+1 holds (j = 1, 2 . . . m − 1). Prove that |A(n)| = |B(n)| for every positive
integer.

Solution. A sequence of the first type can be rewritten as:

n = x1 + 3x2 + 7x3 · · · + (2i − 1)xi + . . . (2k − 1)xk
Where xi are non-negative integers. This motivates us to find a way to represent bi as sums
2.8.1 Problems 76

of (2i − 1)xi . Then since bj ≥ 2bj+1 , we write: bi = 2bi−1 + xi with xi being non-negative
integers.

Problem 2.8.3 (ISL 2008 C4) : Let n and k be positive integers with k ≥ n and k − n an even
number. Let 2n lamps labeled 1, 2 . . . 2n be given, each of which can be either on or off. Initially,
all the lamps are off. We consider sequences of steps: at each step one of the lamps is switched
(from on to off or from off to on).

Let N be the number of such sequences consisting of k steps and resulting in the state where lamps
1 through n are all on, and lamps n + 1 through 2n are all off.

Let M be number of such sequences consisting of k steps, resulting in the state where lamps 1
through n are all on, and lamps n + 1 through 2n are all off, but where none of the lamps n + 1
through 2n is ever switched on.

Determine M.
N

Solution. These type of problems most of the time have bijection or algo solutions. Think of
a way to perform bijection from the set S{M} → S{N}. Find an algorithm to get a sequence of
the first type from a sequence of the second type.

Problem 2.8.4 (USAMO 1996 P4) : An n -term sequence (x1 , x2 , . . . , xn ) in which each term is
either 0 or 1 is called a binary sequence of length n. Let an be the number of binary sequences of
length n containing no three consecutive terms equal to 0, 1, 0 in that order. Let bn be the number
of binary sequences of length n that contain no four consecutive terms equal to 0, 0, 1, 1 or 1, 1,
0, 0 in that order. Prove that bn+1 = 2an for all positive integers n.

Solution. These type of problems cries for a nice bijection. That is a way to get from a → b
and vice versa. What if there is no 0, 0, 1, 1 ? Or what if there is no 0, 1, 0 ? What is an one
way bijection?

Problem 2.8.5 (APMO 2017 P1) : We call a 5 -tuple of integers arrangeable if its elements can
be labeled a, b, c, d, e in some order so that a − b + c − d + e = 29. Determine all 2017 -tuples
of integers n1 , n2 , n3 . . . n2017 such that if we place them in a circle in clockwise order, then any 5
-tuple of numbers in consecutive positions on the circle is arrangeable.
2.8.1 Problems 77

Solution. EChen trick.

Problem 2.8.6 (ISL 2004 C1) : There are 10001 students at an university. Some students join
together to form several clubs (a student may belong to different clubs). Some clubs join together
to form several societies (a club may belong to different societies). There are a total of k societies.
Find all possible values of k so that the following conditions are satisfied:

1 Each pair of students are in exactly one club.

2 For each student and each society, the student is in exactly one club of the society.
3 Each club has an odd number of students. In addition, a club with 2m + 1 students ( m is a
positive integer) is in exactly m societies.

Solution. Just Double-Counting.

Problem 2.8.7 (ISL 2002 C1) : Let n be a positive integer. Each point (x, y ) in the plane, where
x and y are non-negative integers with x + y < n , is coloured red or blue, subject to the following
condition: if a point (x, y ) is red, then so are all points (x 0 , y 0 ) with x 0 ≤ x and y 0 ≤ y . Let A be
the number of ways to choose n blue points with distinct x -coordinates, and let B be the number
of ways to choose n blue points with distinct y -coordinates. Prove that A = B.

Problem 2.8.8 (USAMO 2012 P2) : A circle is divided into 432 congruent arcs by 432 points. The
points are colored in four colors such that some 108 points are colored Red, some 108 points are
colored Green, some 108 points are colored Blue, and the remaining 108 points are colored Yellow.
Prove that one can choose three points of each color in such a way that the four triangles formed
by the chosen points of the same color are congruent.

Solution. Double counting saves the day :) The trick is to rotate ;)

Problem 2.8.9 (European Mathematics Cup 2018 P1) : Call a partition of n a set a1 , . . . ak with
a1 ≤ a2 · · · ≤ ak and a1 + a2 · · · + ak =. A partition of a positive integer is ‘even’ if all of its
elements are even numbers. Similarly, a partition is ‘odd’ if all of its elements are odd. Determine
all positive integers n such that the number of even partitions of n is equal to the number of odd
partitions of n.

Solution. Bijection.
2.8.1 Problems 78

Problem 2.8.10 (APMO 2008 P2) : Students in a class form groups each of which contains exactly
three members such that any two distinct groups have at most one member in common. Prove
that, when the class size is 46 , there is a set of 10 students in which no group is properly contained.

Solution. Taking the maximum set that follows the “in which no group is properly contained”
rule. Now the elements that are not in this set, we can connect this element to only one of the
pairs from the set. Now defining a bijection, and counting the elements, we are done.

Problem 2.8.11 (IMO SL 1985) : A set of 1985 points is distributed around the circumference of
a circle and each of the points is marked with 1 or −1. A point is called “good” if the partial sums
that can be formed by starting at that point and proceeding around the circle for any distance in
either direction are all strictly positive. Show that if the number of points marked with −1 is less
than 662 , there must be at least one good point.

Solution. First thing to notice, the number 3 ∗ 661 + 2 = 1985. And these numbers are
completely random. So what if we try to replace 1985 by n ? Will the condition still hold?

Problem 2.8.12 (IMO 2011 P4) : Let n > 0 be an integer. We are given a balance and n weights
of weight 20 , 21 , · · · , 2n−1 . We are to place each of the n weights on the balance, one after another,
in such a way that the right pan is never heavier than the left pan. At each step we choose one of
the weights that has not yet been placed on the balance, and place it on either the left pan or the
right pan, until all of the weights have been placed. Determine the number of ways in which this
can be done.

Solution. Writing the whole process as a sum, we see that only 20 is the odd term here, if we
remove that we can divide by 2 to get a recursive formula.

Solution. Calculating wrt to the last placed weight.

Solution. Getting recursive formula considering the position of 2n−1 .

Problem 2.8.13 (USAMO 2008 P3) : Let n be a positive integer. Denote by Sn the set of points
2.8.1 Problems 79

(x, y ) with integer coordinates such that

1
|x| + y + < n.
2

A path is a sequence of distinct points (x1 , y1 ), (x2 , y2 ), . . . , (x` , y` ) in Sn such that, for i = 2, . . . , ` ,
the distance between (xi , yi ) and (xi−1 , yi−1 ) is 1 (in other words, the points (xi , yi ) and (xi−1 , yi−1 )
are neighbors in the lattice of points with integer coordinates). Prove that the points in Sn cannot
be partitioned into fewer than n paths (a partition of Sn into m paths is a set P of m nonempty
paths such that each point in Sn appears in exactly one of the m paths in P ).

Solution. Graph + Partition, coloring is just natural. Again, the edges join two neighbor lattice
points, so checkerboard coloring. But checkerboard doesn’t do much good. So the next thing we
try is to apply some derivations of it, pseudo!!! Well, overkill.

Solution. For all n, induction is very natural. The optimal partition (the most beautiful one)
and the longest path in it, say P , gives us a way to perform induction. As always, we suppose a
partition with n − 1 paths. As there are a lot of partitions, we need to choose a certain partition,
say M. Again as our goal is to include P in M. So suppose that the set with all the points in P is
A. And further more, suppose that in M there is a path Q with |Q ∩ A| being maximal among all
other partitions of the points. Some some easy case work shows that we must have P ∈ M.

Problem 2.8.14 (USAMO 2013 P2) : For a positive integer n ≥ 3 plot n equally spaced points
around a circle. Label one of them A , and place a marker at A. One may move the marker forward
in a clockwise direction to either the next point or the point after that. Hence there are a total
of 2n distinct moves available; two from each point. Let an count the number of ways to advance
around the circle exactly twice, beginning and ending at A , without repeating a move. Prove that
an−1 + an = 2n for all n ≥ 4.

Solution. Problems where there are multiple possible value of a function regardless of the
current position, one of dealing with these is to assigning labels of these possible values to
each points of the function, and this will give a combinatorial model and a way to deal it with
bijection.

Solution. First investigate the problem condition, an + an−1 = 2n , now, 2n means the number
of differently coloring every point black or white, and the left side is the number of such paths
for n and n − 1. Which means we should try to color the points and see what happens.
2.8.1 Problems 80

Proof. EChen’s solution: In this problem, the main obstacle seems to be the circle condition.
And on top of that, on can land on the starting point. So things are pretty messed up here.
What we want to do is to make things a little bit more easy to deal with. So our best option is
to change the problem so that we get the similar problem with a different explanation. So we
change the condition circle with matrix, 2 round with 2 rows. n points with n entries in each
rows. What we get now is the same problem, just a bit easier to deal with. We call this Tweak
The Problem strategy.

Problem 2.8.15 () : 10 persons went to a bookstore. It is known that: Every person has bought
3 kinds on books and for every 2 persons, there is at least one kind of books which they both have
bought. Let mi be the number of the persons who bought the i th kind of books and M = max{mi }
Find the smallest possible value of M.

Problem 2.8.16 (ARO 2006 11.3) : On a 49 × 69 rectangle formed by a grid of lattice squares, all
50 · 70 lattice points are colored blue. Two persons play the following game: In each step, a player
colors two blue points red, and draws a segment between these two points. (Different segments can
intersect in their interior.) Segments are drawn this way until all formerly blue points are colored
red. At this moment, the first player directs all segments drawn - i. e., he takes every segment AB,
−→ −→
and replaces it either by the vector AB, or by the vector BA. If the first player succeeds to direct
→
−
all the segments drawn in such a way that the sum of the resulting vectors is 0 , then he wins; else,
the second player wins.
Which player has a winning strategy?

Proof. The basic idea comes from wishing that the first player might be able to copy the second
player to “nullify” his moves. But since this isn’t always possible, because no nice symmetry exists
on the board, the idea of coloring the board with dominoes and copying moves wrt the dominoes
comes.

Problem 2.8.17 (ISL 2002 C3) : Let n be a positive integer. A sequence of n positive integers
(not necessarily distinct) is called full if it satisfies the following condition: for each positive integer
k ≥ 2, if the number k appears in the sequence then so does the number k − 1, and moreover
the first occurrence of k − 1 comes before the last occurrence of k. For each n, how many full
sequences are there?

Proof. After guessing the ans, the first thing that I did was to draw a level based graph.
Suppose that a full sequence has k different entries. Then the top level contains the positions of
k in the sequence sorted from left to right. The next level contains the positions of k − 1 in the
sequence sorted so, and so on till the last level. What I noticed is that if we draw arrows pointing
2.8.1 Problems 81

from a larger integer to a smaller integer, the only arrows (or more like relations between entries
of the sequence) we need to worry about are the arrows pointing left to right in each levels, and
the arrows from the last entry of level i to the first entry of level i + 1. After this, if we try with
a smaller case, we see that this leads to a bijection from the set of sequences of length n with n
different integers to the set of full-sequences of length n.

Solution. Another bijection approach is as followed, in a full-sequence, on first run, go from

right to left, placing integers starting with 1 onwards on the 1’s in the sequence. on the second
run continue counting and placing integers on the 2’s and so on.

Solution. Another idea is to prove an = nan−1 . To do this, remove the rightmost 1 and do
some casework.

Problem 2.8.18 (ISL 1994 C2) : In a certain city, age is reckoned in terms of real numbers rather
than integers. Every two citizens x and x 0 either know each other or do not know each other.
Moreover, if they do not, then there exists a chain of citizens x = x0 , x1 , . . . , xn = x 0 for some
integer n ≥ 2 such that xi−1 and xi know each other. In a census, all male citizens declare their
ages, and there is at least one male citizen. Each female citizen provides only the information
that her age is the average of the ages of all the citizens she knows. Prove that this is enough to
determine uniquely the ages of all the female citizens.

Solution. Describing the problem using matrix and vector spaces, the problem reduces to well
known theorems of linear algebra.

Problem 2.8.19 (ISL 2003 C1) : Let A be a 101-element subset of the set S = {1, 2, . . . , 1000000}.
Prove that there exist numbers t1 , t2 , . . . , t100 in S such that the sets

Aj = {x + tj | x ∈ A}, j = 1, 2, . . . , 100

are pairwise disjoint.

Solution. just count...

Problem 2.8.20 (EGMO 2017 P5) : Let n ≥ 2 be an integer. An n-tuple (a1 , a2 , . . . , an ) of not
necessarily different positive integers is expensive if there exists a positive integer k such that

(a1 + a2 )(a2 + a3 ) . . . (an−1 + an )(an + a1 ) = 22k−1

a) Find all integers n ≥ 2 for which there exists an expensive n-tuple.

2.8.1 Problems 82

b) Prove that for every odd positive integer m there exists an integer n ≥ 2 such that m belongs to
an expensive n-tuple.

Remark. gutaguti solution

Solution. All odd n works, you can prove for even n by +1, −1 addition. For the second part,
start with the odd number, move on both side, you will eventually reach 1.

Problem 2.8.21 (USAMO 2006 P2) : For a given positive integer k find, in terms of k, the
minimum value of N for which there is a set of 2k + 1 distinct positive integers that has sum greater
than N but every subset of size k has sum at most N2 .

Solution. Compactness is the optimal decision.

Problem 2.8.22 (MOP Problem) : Prove that for any positive integer c, there exists an integer n
such that n has more 1’s in its binary expansion than n2 + c does.

Solution. For x = 2a − 1, x and x 2 have the same number of 10 s. So does x = 2a − 2b . But

what if increase the number of 10 s in this x by substracting 1? Let x = 2a − 2b − 1. This might
work if we can choose nice a, b’s

Problem 2.8.23 (EGMO 2015 P2) : A domino is a 2 × 1 or 1 × 2 tile. Determine in how many
ways exactly n2 dominoes can be placed without overlapping on a 2n × 2n chessboard so that every
2 × 2 square contains at least two uncovered unit squares which lie in the same row or column.

Solution. Notice how each of the four kind of dominoes needs to be in a group. So if we
separated them into blocks, inverstigation shows that there can only be 4 blocks and each strictly
attached to the sides. The reason why this is happening is pretty obvious. Now those blocks create
two paths between the two opposite vertices of the square. This gives our desired bijection.

Problem 2.8.24 (USA TST 2006 P5) : Let n be a given integer with n greater than 7 , and let
P be a convex polygon with n sides. Any set of n − 3 diagonals of P that do not intersect in
the interior of the polygon determine a triangulation of P into n − 2 triangles. A triangle in the
triangulation of P is an interior triangle if all of its sides are diagonals of P. Express, in terms of n,
the number of triangulations of P with exactly two interior triangles, in closed form.

Solution. Just mindless calculation...

2.8.1 Problems 83

Problem 2.8.25 (ISL 2010 C3) : 2500 chess kings have to be placed on a 100 × 100 chessboard
so that

1. no king can capture any other one (i.e. no two kings are placed in two squares sharing a
common vertex);

2. each row and each column contains exactly 25 kings.

Find the number of such arrangements. (Two arrangements differing by rotation or symmetry are
supposed to be different.)

Solution. In a 2 × 2 box, one can place only one king. So we divide the board in that way, and
explore...

Problem 2.8.26 (USA Winter TST 2018 P3) : At a university dinner, there are 2017 mathemati-
cians who each order two distinct entrées, with no two mathematicians ordering the same pair of
entrées. The cost of each entrée is equal to the number of mathematicians who ordered it, and
the university pays for each mathematician’s less expensive entrée (ties broken arbitrarily). Over all
possible sets of orders, what is the maximum total amount the university could have paid?

Solution [Grid Representation]. We put the information in a grid in the usual manner. We
take the configuration for which the score (which we define to be the total amount the university
has to pay) is maximum.

Definition— Let Ci be the price of the dish i . Let Pj be the pair of dishes j ordered.

We now “sort” the grid in the following way:

• Sort the columns first, from the least expensive to the most.

• Now we have 2017 binary strings as rows. We sort them decreasingly from top to bottom.

We end up with something like this:

2.8.1 Problems 84

Figure 2.13

Definition— We make some sets:

Ai = {x|x ordered dish i but didn’t order any of the dishes j < i }

Let ai = |Ai |. We also call Ai block, the xth rows with x ∈ Ai . It is easy to see that the
committee has to pay for dish i ai times.

Take a j such that aj = 0.

If j < k, let

Sj = {x | In the j th column, there is a 1 in the set of the rows from Ax }

2.8.1 Problems 85

Figure 2.14

In the example of the previous diagram, S4 = {1, 2, 3}

Take the largest element of Sj , namely t. We want to move the 1 in the At block of column j
to another column to the right. Notice that this won’t decrease the score, because for each of
the mathematician after the block At , the score would be non decreasing. If we could do this,
we could do this inductively, moving the jth column to the left each move, eventually making it
disappear.
So suppose we can’t do this. So there is no column on the right of t, that does not have a 1 in
the block At . It is straightforward to deduce that N = t + at .
From there, we can say:
at+m ≤ at − m and at−m ≤ at + m
So,
X
2017 = ai ≤ (at + t − 1) + . . . 1
N(N − 1)
=⇒ 2017 ≤
2
=⇒ N ≥ 65

We know that the score, S = Ck ak . where ai is strictly decreasing and Ci is non decreasing.
P

So the maximum sum would have Ci = K for some constant (by rearrangement inequality). And
to have K maximum, we need N minimum or N = 64.
And if j > k, the same reason holds. So N = 65, a1 = 63, a2 = 62 . . . a6 3 = 1. But we need
another ai = 1, that has to be on its own.

Remark. Easy to think in grids, but it was quite difficult to formulate the solution rigorously, I
still am not 100% convinced myself. Next time when I feel like it, I will reconstruct the solution.
Roughly it is: take the final 0 value columns, prove that there is only one such, and prove the
2.8.1 Problems 86

bounds on the solution before using that instead. Then use general bounding to find the maximum.
It’s not hard, it’s just I don’t have time now. GAH!! HSC!!

Remark. Didn’t read other solutions too, gonna give a todo

2.8.1.1 Conway’s Soldiers

Problem 2.8.27 (ISL 1993 C5) : On an infinite chessboard, a solitaire game is played as follows:
at the start, we have n2 pieces occupying a square of side n. The only allowed move is to jump over
an occupied square to an unoccupied one, and the piece which has been jumped over is removed.
For which n can the game end with only one piece remaining on the board?

Solution. We want to find an invariant. So we need to find a weight for each of the cells
such that any two consecutive cells’ values equals to the values of the two cells on the two sides.
Some mind bashing gives the idea of mod 3. And a construction for the other n’s can be easily
generated after some casework.

Problem 2.8.28 (ARO 1999 P4) : A frog is placed on each cell of a n × n square inside an infinite
chessboard (so initially there are a total of n × n frogs). Each move consists of a frog A jumping
over a frog B adjacent to it with A landingh inithe next cell and B disappearing (adjacent means two
n2
cells sharing a side). Prove that at least 3 moves are needed to reach a configuration where no
more moves are possible.

Solution. In the final stage, no two neighboring cells are occupied. Could we double count the
number of frogs with this information? What about the number of frogs in the original n × n
board? Another small information needed for this is that we need 2 moves to “empty” a 2 × 2
board.

2.8.1.2 Triominos

Problem 2.8.29 (ARO 2011 P10.8) : A 2010 × 2010 board is divided into corner-shaped figures
of three cells. Prove that it is possible to mark one cell in each figure such that each row and each
column will have the same number of marked cells.

Solution. First we will mark the corner pieces of the triominos. Then shift the mark to either
of the legs. Our objective is to show that we can always do this. First if we only focus on the
rows, we can easily show that this can be done using some counting. To show that we can do the
same for columns as well, we create a graph from columns and rows to triominos which should
be operated on, and using Hall’s Marriage we pove the result.
2.8.1 Problems 87

Problem 2.8.30 (St. Petersburg 2000) : On an infinite checkerboard are placed 111 non-overlapping
corners, L-shaped figures made of 3 unit squares. Suppose that for any corner, the 2 × 2 square
containing it is entirely covered by the corners. Prove that one can remove some number between
1 and 110 of the corners so that the property will be preserved.

2.8.1.3 Dominos

Problem 2.8.31 () : An m × n rectangular grid is covered by dominoes. Prove that the vertices of
the grid can be coloured using three colours so that any two vertices a distance 1 apart are colored
with different colours if and only if their segment lies on the boundary of a domino.

Solution. Create a graph with the midpoints of the dominos.

2.8.2 Coloring Problems 88

2.8.2 Coloring Problems

Problem 2.8.32 (EGMO 2017 P2) : Find the smallest positive integer k for which there exists
a colouring of the positive integers Z>0 with k colours and a function f : Z>0 → Z>0 with the
following two properties:

1. For all positive integers m, n of the same colour, f (m + n) = f (m) + f (n).

2. There are positive integers m, n such that f (m + n) 6= f (m) + f (n).

In a colouring of Z>0 with k colours, every integer is coloured in exactly one of the k colours. In
both (i ) and (i i ) the positive integers m, n are not necessarily distinct.

Solution. Firstly a modular coloring shows that 1 < k ≤ 2. For k = 2 we do some trivial case
works.

Problem 2.8.33 (ISL 2002 C2) : For n an odd positive integer, the unit squares of an n × n
chessboard are coloured alternately black and white, with the four corners coloured black. A it
tromino is an L-shape formed by three connected unit squares. For which values of n is it possible
to cover all the black squares with non-overlapping trominos? When it is possible, what is the
minimum number of trominos needed?

Solution. First find the first ans and a configuration that works. Then guess the second ans,
and see from where that might come from, usually these anses come from some special set of
problems, where bijection is applicable.

Problem 2.8.34 (Codeforces 101954/G) : Two Knights are given on a chessboard, one black one
white. Which player has a winning possibility?

Solution. A knight’s move always changes the color of the cell.

Problem 2.8.35 (ARO 1993 P10.4) : Thirty people sit at a round table. Each of them is either
smart or dumb. Each of them is asked: "Is your neighbor to the right smart or dumb?" A smart
person always answers correctly, while a dumb person can answer both correctly and incorrectly. It
is known that the number of dumb people does not exceed F . What is the largest possible value
of F such that knowing what the answers of the people are, you can point at at least one person,
knowing he is smart?
2.8.2 Coloring Problems 89

Solution. We see that the strings of truth only exist either when all people are dumb or the
last one is the truthful one. Now we take the longest such string, and this sting has to be of the
second kind. To prove this, we use bounding with the given constraint.

Problem 2.8.36 (Tournament of Towns 2015S S6) : An Emperor invited 2015 wizards to a festival.
Each of the wizards knows who of them is good and who is evil, however the Emperor doesn’t know
this. A good wizard always tells the truth, while an evil wizard can tell the truth or lie at any
moment. The Emperor gives each wizard a card with a single question, maybe different for different
wizards, and after that listens to the answers of all wizards which are either “yes” or “no”. Having
listened to all the answers, the Emperor expels a single wizard through a magic door which shows
if this wizard is good or evil. Then the Emperor makes new cards with questions and repeats the
procedure with the remaining wizards, and so on. The Emperor may stop at any moment, and after
this the Emperor may expel or not expel a wizard. Prove that the Emperor can expel all the evil
wizards having expelled at most one good wizard.

Solution. There is only one problem with the cyclic arrangement, that is what if all the answers
are ‘yes’ ? We get rid of this problem by trying small case with n = 3 and trying the most simple
way to connect this strategy to any n. Simplicity is the key.

Problem 2.8.37 (ISL 2007 C1) : Let n > 1 be an integer. Find all sequences a1 , a2 , . . . an2 +n
satisfying the following conditions:

1. ai ∈ {0, 1} for all 1 ≤ i ≤ n2 + n

2. for all 0 ≤ i ≤ n2 − n

ai+1 + ai+2 + . . . + ai+n < ai+n+1 + ai+n+2 + . . . + ai+2n

Solution. n + 1 blocks of n, each strictly greater than the previous one. means the sums of
the blocks have to be 0, 1, . . . n. construction’s easy from examples of 2, 3.

Problem 2.8.38 (ISL 2016 C2) : Find all positive integers n for which all positive divisors of n can
be put into the cells of a rectangular table under the following constraints: each cell contains a
distinct divisor;
the sums of all rows are equal; and
the sums of all columns are equal.

Solution. Check the sizes.

2.8.2 Coloring Problems 90

Problem 2.8.39 (ISL 2007 C3) : Find all positive integers n for which the numbers in the set
S = {1, 2, . . . , n} can be colored red and blue, with the following condition being satisfied: The set
S × S × S contains exactly 2007 ordered triples (x, y , z) such that:

1. the numbers x, y , z are of the same color

2. the number x + y + z is divisible by n.

Solution. Trying out small cases, noticing patter. It doesn’t matter ‘which’ numbers are red,
but ‘how’ many numbers are red.

Problem 2.8.40 (ISL 2014 C4) : Construct a tetromino by attaching two 2 × 1 dominoes along
their longer sides such that the midpoint of the longer side of one domino is a corner of the other
domino. This construction yields two kinds of tetrominoes with opposite orientations. Let us call
them S- and Z-tetrominoes, respectively.

Assume that a lattice polygon P can be tiled with S-tetrominoes. Prove that no matter how we
tile P using only S- and Z-tetrominoes, we always use an even number of Z-tetrominoes.

Solution. So after we are determined to do coloring, it is not very hard to come up with a
coloring. Start from stracth type coloring. Color one square at a time, this might take several
tries.

Figure 2.15
2.9 Linear Algebra 91

2.9 Linear Algebra

• Algebraic Techniques - Yufei Zhao

Lemma 2.9.1 ( Multiplication Order) —

(AB)T = B T AT

ABCD = A(BC)D

Theorem 2.9.2 (Linear map that reverses the order of multiplication) — Let F be a field, and
m a positive integer. The only F -linear maps s : Mm (F ) → Mm (F ) which satisfy

s (Xk Xk−1 ...X1 ) = s (X1 ) s (X2 ) ...s (Xk ) ∀k ∈ N, Xi ∈ Mm (F )

are maps of the form s (P ) = UP T U −1 ∀P ∈ Mm (F ) for U an invertible m × m matrix over F .

Proof. Let s be a map that satisfies the problem condition. Let t be another linear map defined
by t(P ) = s(P )T ∀P ∈ M(F ).
This map t also satisfies

t (Xk Xk−1 ...X1 ) = t (X1 ) t (X2 ) ...t (Xk ) ∀k ∈ N, Xi ∈ Mm (F )

The rest uses advanced algebra :cold_sweat: something called Noether-Skolem theorem...
maybe later...

Problem 2.9.1 (ISL 2009 C3) : Let n be a positive integer. Given a sequence ε1 , . . . , εn−1 with
εi = 0 or εi = 1 for each i = 1 , . . . , n − 1 , the sequences a0 , . . . , an and b0 , . . . , bn are
constructed by the following rules:

a0 = b0 = 1, a1 = b1 = 7,
(
2ai−1 + 3ai , if εi = 0,
ai+1 = for each i = 1, . . . , n − 1,
3ai−1 + ai , if εi = 1,
(
2bi−1 + 3bi , if εn−i = 0,
bi+1 = for each i = 1, . . . , n − 1.
3bi−1 + bi , if εn−i = 1,

Prove that an = bn .
2.9 Linear Algebra 92

Solution [Algebraic, darij grinberg]. We’ll first translate the problem in matrix language, then
find a linear transformation to prove it.

Definition— Let M2 (Q) be the ring of 2 ×  2 matrices

 over the rational numbers.
3 2 1 3
Definte two matrices A, B ∈ M2 (Q) by A = and B = .
1 0 1 0
For every i ∈ {1, 2, ..., n − 1}, define a matrix Ki ∈ M2 (Q) by

Ki = εi B + (1 − εi ) A

This clearly yields that Ki = A if εi = 0,and that

 Ki =B if εi =
 1.
ai+1 ai
For every i ∈ {1, 2, ..., n − 1}, we have = Ki . By induction we have
ai ai−1
   
an 7
= Kn−1 Kn−2 ...K1
an−1 1
 

7
an = 10 Kn−1 Kn−2 ...K1
1
We need to prove that
   

7
7
10 Kn−1 Kn−2 ...K1 = 10 K1 K2 ...Kn−1
1 1

In order to do this, it is clearly enough to define some map s : M2 (Q) → M2 (Q) which satisfies
1. s (Kn−1 Kn−2 ...K1 ) = K1 K2 ...Kn−1    
7 7
2. Every matrix P ∈ M2 (Q) satisfies 1 0 P



= 1 0 s (P )
1 1
 
7 1
Let U be the invertible matrix ∈ M2 (Q). Let s : M2 (Q) → M2 (Q) be the map
1 2
defined by
s (P ) = UP T U −1 , ∀P ∈ M2 (Q)
It’s easy to check with computation that this mapping satisfies (2). We need to show it satisfies
(1) too.
First let’s list some properties of s. We have,

1. s(I2 ) = I2 where I2 is the identity matrix.

2. s(XY ) = s(Y )s(X), trivial by Lemma 2.9.1
3. s(A) = A and s(B) = B, easy to show with some computation.
4. It follows that s(Ki ) = Ki .

(1) trivially follows from these properties. So s satisfies the problem.

2.9 Linear Algebra 93

Remark. The above solution followed a rather standard procedure (translating linear recurrences
into matrix multiplication - this is the same trick that solves many problems about Fibonacci num-
bers) until the point where we "guessed" the matrix U and the map s. How did we do that?
The motivation is the following: We need a map s which satisfies (1) and (2). We forget about
(2) for a moment, and try to satisfy (1) only.
The easiest way to ensure that (1) holds for every choice of n and ε1 , ε2 , ..., εn−1 is to choose s
as a linear map satisfying s (A) = A and s (B) = B (this immediately guarantees that s (Ki ) = Ki
for every i , because Ki = εi B + (1 − εi ) A is a linear combination of A and B) and satisfying
s (Xk Xk−1 ...X1 ) = s (X1 ) s (X2 ) ...s (Xk ) for any k ∈ N and any 2 × 2 matrices X1 , X2 , ..., Xk .
This condition s (Xk Xk−1 ...X1 ) = s (X1 ) s (X2 ) ...s (Xk ) is fulfilled, for example, when the map s
has the form s (P ) = UP T U −1 for every P ∈ M2 (Q) for U an invertible 2 × 2 matrix. Actually
it is fulfilled only in this case, as I explain further below, but as for now let us at least agree that
s (P ) = UP T U −1 for every P ∈ M2 (Q) is a good point to start.
So now we are searching for a 2 × 2 matrix U such that the map s defined by s (P ) = UP T U −1 for
every P ∈ M2 (Q) satisfies s (A) = A, s (B) = B and (2). These conditions give linear equations
on
 the entries
 of this matrix U, and the only matrix U which solves all of them is (up to scaling)
7 1
∈ M2 (Q). It is now clear how to proceed from here.
1 2

Solution [Bijection, evan chen]. Let Ai = 2i ai . So the recursive relations now are:

Ai+1 = 8Ai−1 + 6Ai or 12Ai−1 + 2Ai

Now we build a bijective relation. Imagine n rooms in a row with n − 1 doors, numbered i ,
between them. We have 14 colors, {?, 1, 2, . . . 13} to paint the rooms. So the two sides of the
doors will get different coloring, let these two colors be i , j. We need to follow some certain rules:

1. If i is ?, then j can be any color.

2. If the door is labeled with 0, and i ∈ {?, j − 2, j − 1, j, j + 1, j + 2} (mod 13)
3. If the door is labeled with 1, i ∈ {?, j}.

We prove that Ai is the number of ways the first i rooms can be painted (and Bi the number of
ways the last i rooms).
First, the door between the rooms i , i + 1 was labeled 0. If Ai 6= ?, then we have 6 ways to paint
i + 1. From there comes 6Ai ways of painting the first i + 1 rooms. If Ai = ?, then there are 14
ways of painting i + 1. But 6 of those ways have already been counted. And the rest of the 8
new colors will come from Ai−1 . So 8Ai−1 ways from there. In total:

Ai+1 = 8Ai−1 + 6Ai

And if the door was labeled 1, then 6, 8 becomes 2, 12. And our result follows.
2.9 Linear Algebra 94

Remark. We first have to settle to search for a bijective solution. So we ask the question, “ai ’s
are the number of ways to do something. How can we define that?”
Now the coefficients of the recurresions doesn’t add up. It would be better if we had them being
equal. So we do that, nice and easy.
So the coefficients now add up to 14. So for the i + 1-th object, we have 14 choices, and those
choices get divided depending on the choice for the i -th object. From here, it is natural to think
about those rules.
2.10 Double Counting and Other Algebraic Methods 95

2.10 Double Counting and Other Algebraic Methods

Problem 2.10.1 (ISL 2012 C3) : In a 999×999 square table some cells are white and the remaining
ones are red. Let T be the number of triples (C1 , C2 , C3 ) of cells, the first two in the same row and
the last two in the same column, with C1 , C3 white and C2 red. Find the maximum value T can
attain.

Solution. Expicitly count the value, and bound it using ineq...

2.10.1 Probabilistic Methods 96

2.10.1 Probabilistic Methods

Problem 2.10.2 (ISL 2006 C3) : Let S be a finite set of points in the plane such that no three
of them are on a line. For each convex polygon P whose vertices are in S, let a(P ) be the number
of vertices of P , and let b(P ) be the number of points of S which are outside P . A line segment,
a point, and the empty set are considered as convex polygons of 2, 1, and 0 vertices respectively.
Prove that for every real number x
X
x a(P ) (1 − x)b(P ) = 1,
P

where the sum is taken over all convex polygons with vertices in S.

Solution. This can be done by strong induction and double counting. Counting for every
possibile subsets with every points inside of the convex hull and at least one point on the convex
hull outside of the set.

Solution. The beautiful solution on the other hand uses probability. Color each point black or
white, then translate the condition in terms of probability. :D
2.11 Bounding 97

2.11 Bounding

Problem 2.11.1 (RMM 2019 P3) : Given any positive real number ε, prove that, for all but finitely
many positive integers v , any graph on v vertices with at least (1 + ε)v edges has two distinct
simple cycles of equal lengths. (Recall that the notion of a simple cycle does not allow repetition
of vertices in a cycle.)

Solution. If suppose there are x cycles at some point, each distinct in size. Double-counting
x2
the number of edges in those cycles: there are at least edges. Since there are at most 2n
2
x2
edges in the graph, there is one edge that is contained in at least cycles.
4n
Now, if we delete that edge, and keep doing that until there is no cycle left in the graph, how
many steps might we need to make?
Notice that x < 4n. If we let 4n = c, we can rephrase our question:

Lemma 2.11.1 — Given m < c, let ai ≥0 be a sequence defined by:

x2
a0 = m, ai+1 = ai − d e
c
It is clear that eventually for some t, at will become 0. What is the upper bound for t?

x2
Suppose we are at x now. The decrement we need to make now is k := . The decrement we
c
x k x
need to make when we are at is √ . AND at each step in going from x → changes the
2 2 2
decrement very little. So we can take the average decrement of all the steps and calculate the
number of steps with that fixed.
So, essentially Greedy Algorithm From this intuition,
r and some calculation, we get the idea of
c
thinking about steps needed to cross an interval of
4

Solution. It is easier to control cycles in a graph where there is no cycle! And the best kind of
spanning tree is a BFS tree.
Now, every edge that is not in this tree creates a cycle. We know that the lengths of these cycles
are all distinct. So we can lower bound the sum of the lengths. Whenever we want to bound
something, it is always a good idea to look for a way to represent the variables differently. This
is where the “jumping over vertices” idea comes from.
2.11 Bounding 98

Solution. Think of the cycles as binary strings. Addition of these cycles are XOR operation.
Now, let M be the set of all cycle lenghts. If we take some cycles and add them, they correspond
to some subset of M. This will give us an upper bound for the sum of some lengths from M.
Now, we want these sums to be distinct. How to do that? We are adding cycles together right?
So we need to bring order in there. We will only take cycles from a certain set, so that it holds.
These cycles need to be linearly independent of themeselvs. That gives us the construction for
the set.
After this, the rest is just bounding.

Problem 2.11.2 (ISL 2018 C5) : Let k be a positive integer. The organising commitee of a tennis
tournament is to schedule the matches for 2k players so that every two players play once, each
day exactly one match is played, and each player arrives to the tournament site the day of his first
match, and departs the day of his last match. For every day a player is present on the tournament,
the committee has to pay 1 coin to the hotel. The organisers want to design the schedule so as to
minimise the total cost of all players’ stays. Determine this minimum cost.
2.12 Problems 99

2.12 Problems

Problem 2.12.1 (ARO 2018 10.3) : A positive integer k is given. Initially, N cells are marked on
an infinite checkered plane. We say that the cross of a cell A is the set of all cells lying in the same
row or in the same column as A. By a turn, it is allowed to mark an unmarked cell A if the cross of
A contains at least k marked cells. It appears that every cell can be marked in a sequence of such
turns. Determine the smallest possible value of N.

Solution. First find the construction.

Problem 2.12.2 (ARO 2018 P9.5) : On the circle, 99 points are marked, dividing this circle into
99 equal arcs. Petya and Vasya play the game, taking turns. Petya goes first; on his first move,
he paints in red or blue any marked point. Then each player can paint on his own turn, in red or
blue, any uncolored marked point adjacent to the already painted one. Vasya wins, if after painting
all points there is an equilateral triangle, all three vertices’s of which are colored in the same color.
Could Petya prevent him?

Solution. Think of what Petya must do to prevent immediate losing.

Problem 2.12.3 (ISL 2004 C2) : Let n and k be positive integers. There are given n circles in
the plane. Every two of them intersect at two distinct points, and all points of intersection they
determine are pairwise distinct (i. e. no three circles have a common point). No three circles have
a point in common. Each intersection point must be colored with one of n distinct colors so that
each color is used at least once and exactly k distinct colors occur on each circle. Find all values of
n ≥ 2 and k for which such a coloring is possible.

Problem 2.12.4 (ISL 2004 C3) : The following operation is allowed on a finite graph: Choose an
arbitrary cycle of length 4 (if there is any), choose an arbitrary edge in that cycle, and delete it
from the graph. For a fixed integer n ≥ 4 , find the least number of edges of a graph that can be
obtained by repeated applications of this operation from the complete graph on n vertices’s (where
each pair of vertices’s are joined by an edge).

Solution. Walk backwards. or the same thing with Bipartite Graphs.

2.12 Problems 100

Problem 2.12.5 (Iran TST 2012 P4) : Consider m + 1 horizontal and n + 1 vertical lines ( m, n ≥ 4
) in the plane forming an m × n table. Cosider a closed path on the segments of this table such
that it does not intersect itself and also it passes through all (m − 1)(n − 1) interior vertices’s
(each vertex is an intersection point of two lines) and it doesn’t pass through any of outer vertices.
Suppose A is the number of vertices’s such that the path passes through them straight forward, B
number of the table squares that only their two opposite sides are used in the path, and C number
of the table squares that none of their sides is used in the path. Prove that A = B − C + m + n − 1.

Problem 2.12.6 (AoPS) : Given 2n + 1 irrational numbers, prove that one can pick n from them
s.t. no two of the choosen n sum up to a rational number.

Solution. Use a graph theory representation.

Problem 2.12.7 (Bulgarian IMO TST 2004, Day 3, Problem 3) : Prove that among any 2n + 1
irrational numbers there are n + 1 numbers such that the sum of any k of them is irrational, for all
k ∈ {1, 2, 3, . . . , n + 1}.

Solution. We first create a set B such that any linear combination of the elements in it are
irrational. Then for convenience, we add 1 to it, so that now the sum equals to 0 of any linear
combinations. An algorithm for building it comes into our mind, which leaves some other original
elements, which we then later add to the final solution set A along with the elements in the set
B except 1.

Problem 2.12.8 (ISL 1997 P4) : An n×n matrix whose entries come from the set S = {1, 2, ..., 2n−
1} is called a “silver matrix” if, for each i = 1, 2, ..., n , the i -th row and the i -th column together
contain all elements of S. Show that:

1 there is no silver matrix for n = 1997 ;

2 silver matrices exist for infinitely many values of n.

Solution. Proving that for odd n ’s isn’t hard. Then A small try-around with n = 2, 4 , we see
a pattern that leads to a construction for 2n

Problem 2.12.9 () : A rectangle is completely partitioned into smaller rectangles such that each
smaller rectangles has at least one integral side. Prove that the original rectangle also has at least
2.12 Problems 101

one integral side.

Solution. Try a special grid system with .5 × .5 boxes.

Solution. Consider the number of corners in the rectangle.

Problem 2.12.10 (ISL 2004 C5) : A and B play a game, given an integer N, A writes down 1 first,
then every player sees the last number written and if it is n then in his turn he writes n + 1 or 2n
, but his number cannot be bigger than N. The player who writes N wins. For which values of N
does B win?

Solution. Trying with smaller cases, it’s easy. Using most important game theory trick.

Problem 2.12.11 (ISL 2006 C1) : We have n ≥ 2 lamps L1 , L2 . . . Ln in a row, each of them being
either on or off. Every second we simultaneously modify the state of each lamp as follows: if the
lamp Li and its neighbors (only one neighbor for i = 1 or i = n , two neighbors for other i ) are in
the same state, then Li is switched off; otherwise, Li is switched on. Initially all the lamps are off
except the leftmost one which is on.

1 Prove that there are infinitely many integers n for which all the lamps will eventually be off.
2 Prove that there are infinitely many integers n for which the lamps will never be all off

Problem 2.12.12 (ISL 2006 C4) : A cake has the form of an n × n square composed of n2 unit
squares. Strawberries lie on some of the unit squares so that each row or column contains exactly
one strawberry; call this arrangement A.
Let B be another such arrangement. Suppose that every grid rectangle with one vertex at the top
left corner of the cake contains no fewer strawberries of arrangement B than of arrangement A.
Prove that arrangement B can be obtained from A by performing a number of switches, defined as
follows:
A switch consists in selecting a grid rectangle with only two strawberries, situated at its top right
corner and bottom left corner, and moving these two strawberries to the other two corners of that
rectangle.
2.12 Problems 102

Solution. When the first approach fails, don’t throw that idea yet. Stick to it, as it is most
probably the closest to a correct solution. Taking the smallest rectangle with 0 ’s equal to 1 ’s,
we see that we can ’shrink’ the rectangle. Which leads to a solution instantly.

Problem 2.12.13 (ISL 2014 C2) : We have 2m sheets of paper, with the number 1 written on
each of them. We perform the following operation. In every step we choose two distinct sheets; if
the numbers on the two sheets are a and b , then we erase these numbers and write the number
a + b on both sheets. Prove that after m2m−1 steps, the sum of the numbers on all the sheets is
at least 4m .

Solution. When you know that the problem can be solved using invariants, go through all of
the possible invariants (from the rules of thumb). Don’t give up on one so quickly. And product
and sum are actually more close than you think. Because if you are told to prove some bound
on the sum, then product can come very handy. After all there is AM-GM to connect sum and
product.

Problem 2.12.14 (ISL 2016 C3) : Let n be a positive integer relatively prime to 6. We paint the
vertices’s of a regular n -gon with three colours so that there is an odd number of vertices’s of each
colour. Show that there exists an isosceles triangle whose three vertices’s are of different colours.

Solution. Double Count with the number of points of each colors.

Problem 2.12.15 (Iran TST 2002 P3) : A “2-line” is the area between two parallel lines. Length
of “2-line” is distance of two parallel lines. We have covered unit circle with some “2-lines”. Prove
sum of lengths of “2-lines” is at least 2.

Solution. Consider the “2-line” of the largest length.

Problem 2.12.16 (ARO 2008 P9.5) : The distance between two cells of an infinite chessboard is
defined as the minimum number to moves needed for a king to move from one to the other. On
the board are chosen three cells on pairwise distances equal to 100. How many cells are there that
are at the distance 50 from each of the three cells?

Problem 2.12.17 (USAMO 1986 P2) : During a certain lecture, each of five mathematicians fell
asleep exactly twice. For each pair of mathematicians, there was some moment when both were
2.12 Problems 103

asleep simultaneously. Prove that, at some moment, three of them were sleeping simultaneously.

Problem 2.12.18 (Mexican Regional 2014 P6) : Let A = n × n be a {0, 1} matrix, where each
row is different. Prove that you can remove a column such that the resulting n × (n − 1) matrix
has n different rows.

Solution. Try to represent the sets in a nicer way, with graph. or. Induction on the number of
columns deleted and the number or different rows being there.

Problem 2.12.19 (IMO 2017 P5) : An integer N ≥ 2 is given. A collection of N(N + 1) soccer
players, no two of whom are of the same height, stand in a row. Show that Sir Alex can always
remove N(N − 1) players from this row leaving a new row of 2N players in which the following N
conditions hold:
( 1 ) no one stands between the two tallest players,
( 2 ) no one stands between the third and fourth tallest players,
..
.
( N ) no one stands between the two shortest players.

Solution. N(N + 1) , rows, removing . . . these things just begs for to be arranged in a
systematic order. As arranging thing in a matrix is the simplest way, we arrange the bad-bois in
a N · (N + 1) matrix. Now finding the algorithm is not very hard.

Problem 2.12.20 (ISL 1990 P3) : Let n ≥ 3 and consider a set E of 2n − 1 distinct points on
a circle. Suppose that exactly k of these points are to be colored black. Such a coloring is good
if there is at least one pair of black points such that the interior of one of the arcs between them
contains exactly n points from E. Find the smallest value of k so that every such coloring of k
points of E is good.

Solution. Creating a graph and using Alternating Chains Technique

Problem 2.12.21 (USAMO 1999 P1) : Some checkers placed on an n × n checkerboard satisfy
the following conditions:
2.12 Problems 104

1 every square that does not contain a checker shares a side with one that does;
2 given any pair of squares that contain checkers, there is a sequence of squares containing checkers,
starting and ending with the given squares, such that every two consecutive squares of the sequence
share a side.

Prove that at least (n2 − 2)/3 checkers have been placed on the board.

Solution. As the problem simply seems to exist, we can’t count how much contribution a
checker cntaining square contributes to the whole board. So we place one at a time and see the
changes.

Generalization 2.12.21.1 (USAMO 1999 P1 generalization) : Find the smallest positive integer m
such that if m squares of an n × n board are colored, then there will exist 3 colored squares whose
centers form a right triangle with sides parallel to the edges of the board.

Problem 2.12.22 (ISL 2013 C1) : Let n be an positive integer. Find the smallest integer k with
the following property; Given any real numbers a1 , · · · , ad such that a1 + a2 + · · · + ad = n and
0 ≤ ai ≤ 1 for i = 1, 2, · · · , d , it is possible to partition these numbers into k groups (some of
which may be empty) such that the sum of the numbers in each group is at most 1.

Solution. Think about the worst case where d is the minimum and the ans is d , it would only
be possible if each ai > 12 but this can’t be true, so, the ans is 2n − 1. Now the ques should
become obvious.

Problem 2.12.23 (Brazilian Olympic Revenge 2014) : Let n a positive integer. In a 2n × 2n board,
1 × n and n × 1 pieces are arranged without overlap. Call an arrangement maximal if it is impossible
to put a new piece in the board without overlapping the previous ones. Find the least k such that
there is a maximal arrangement that uses k pieces.

Solution. Intuition gives that there is at least one n -mino in each row. But we can easily
guess that there is no maximal arrangement with 2n minos. Suppose in a maximal arrangement,
there are no vertical n -mino, that means there are more than 2n + 1 n-minos. So suppose that
there is at least one vertical suppose that it lies in a column i between 1 and n. Then we have
that there is at least one n -mino in each column in between 1 and i . If there is one in between 1
and 2n , say j , then there is one in each of the columns on the right side of it. Then we count
horizontal n -minos, we show that 2n + 1 is the answer.
2.12 Problems 105

Problem 2.12.24 (ISL 2008 C1) : In the plane we consider rectangles whose sides are parallel to the
coordinate axes and have positive length. Such a rectangle will be called a box. Two boxes intersect
if they have a common point in their interior or on their boundary. Find the largest n for which
there exist n boxes B1 , B2 . . . Bn such that Bi and Bj intersect if and only if i 6≡ j ± 1 (mod n).

Solution. Instead of focusing on building the boxes from only one side (i.e. starting with
1, 2 . . . , we should include n in our investigation, and follow from both direction, (i.e. 1, 2 . . .
and . . . , n − 1, n ).

Problem 2.12.25 (USAMO 2008 P4) : Let P be a convex polygon with n sides, n ≥ 3. Any set of
n − 3 diagonals of P that do not intersect in the interior of the polygon determine a triangulation
of P into n − 2 triangles. If P is regular and there is a triangulation of P consisting of only isosceles
triangles, find all the possible values of n.

Solution. It’s not hard after getting the ans.

Problem 2.12.26 (ARO 2016 P3) : We have a sheet of paper, divided into 100 × 100 unit squares.
In some squares, we put right-angled isosceles triangles with leg = 1 (Every triangle lies in one unit
square and is half of this square). Every unit grid segment (boundary too) is under one leg of a
triangle. Find maximal number of unit squares, that don’t contains any triangles.

Solution. What is the minimum number of triangles you can use in a row? Create a good row
one at a time

Problem 2.12.27 (India TST 2013 Test 3, P1) : For a positive integer n , a Sum-Friendly Odd
Partition of n is a sequence (a1 , a2 . . . ak ) of odd positive integers with a1 ≤ a2 ≤ · · · ≤ ak and
a1 + a2 + · · · + ak = n such that for all positive integers m ≤ n , m can be uniquely written as a
subsum m = ai1 + ai2 + · · · + air . (Two subsums ai1 + ai2 + · · · + air and aj1 + aj2 + · · · + ajs with
i1 < i2 < · · · < ir and j1 < j2 < · · · < js are considered the same if r = s and ail = ajl for 1 ≤ l ≤ r .)
For example, (1, 1, 3, 3) is a sum-friendly odd partition of 8. Find the number of sum-friendly odd
partitions of 9999.

Solution. Firstly we explore one SFOP at a time. Which gives us a way to tell what ai+1 is
going to be by looking at a1 . . . ai .
2.12 Problems 106

Problem 2.12.28 (IMO 2011 P2) : Let S be a finite set of at least two points in the plane. Assume
that no three points of S are collinear. A windmill is a process that starts with a line ` going through
a single point P ∈ S. The line rotates clockwise about the pivot P until the first time that the line
meets some other point belonging to S. This point, Q , takes over as the new pivot, and the line
now rotates clockwise about Q , until it next meets a point of S. This process continues indefinitely.
Show that we can choose a point P in S and a line ` going through P such that the resulting
windmill uses each point of S as a pivot infinitely many times.

Solution. Some workaround gives us the idea that the starting line has to be kinda “in between”
the points. Formal words could be: the line should divide the set of points in two sets so that the
two sets have equal number of points. Once we take a such line, we see that after every move
we get a new line which has similar properties of the first line.

Solution. So moral of the story is that if you get some vague idea that something has to satisfy
something-ish, remove the -ish part, and try with a formal assumption.

Problem 2.12.29 (IOI 2016 P5) : A computer bug has a permutation P of length 2k = N that
changes any string added to a DS according to the permutation, i.e. it makes S[i ] = S[P [i ]]. Your
task it to find the permutation in the following ways:

1 You can add at most n log2 n N bit binary strings to the DS.
2 You can ask at most n log2 n, in the form of N bit binary strings. The answer will be “true” if the
string exists in the DS after the Bug had changed the strings and “no” otherwise.

Solution. Typical Divide and Conquer approach. You want to do the same thing for N = N2 ,
and to do so you need to tell exactly what the first N2 terms of the permutation are. To do this,
you can use at most N questions. This is easy, you first add strings with only one bit present in
the first N2 positions, and then ask N questions with only one bit in every N positions. This maps
the first N2 numbers of the permutation to a set of N2 integers. And we can proceed by induction
now.

Problem 2.12.30 (ISL 2001 C6) : For a positive integer n define a sequence of zeros and ones to
be balanced if it contains n zeros and n ones. Two balanced sequences a and b are neighbors if you
can move one of the 2n symbols of a to another position to form b. For instance, when n = 4, the
balanced sequences 01101001 and 00110101 are neighbors because the third (or fourth) zero in the
first sequence can be moved to the first
or second position to form the second sequence. Prove
1 2n
that there is a set S of at most n+1 n balanced sequences such that every balanced sequence is
2.12 Problems 107

equal to or is a neighbor of at least one sequence in S.

Problem 2.12.31 (ISL 1998 C4) : Let U = {1, 2, . . . , n}, where n ≥ 3. A subset S of U is said to
be split by an arrangement of the elements of U if an element not in S occurs in the arrangement
somewhere between two elements of S. For example, 13542 splits {1, 2, 3} but not {3, 4, 5}. Prove
that for any n − 2 subsets of U, each containing at least 2 and at most n − 1 elements, there is an
arrangement of the elements of U which splits all of them.

Solution. If we try to apply induction, we see that the sets with 2 and n − 1 elements create
problems, so we handle them first.

Problem 2.12.32 (USA TST 2009 P1) : Let m and n be positive integers. Mr. Fat has a set S
containing every rectangular tile with integer side lengths and area of a power of 2. Mr. Fat also
has a rectangle R with dimensions 2m × 2n and a 1 × 1 square removed from one of the corners.
Mr. Fat wants to choose m + n rectangles from S, with respective areas 20 , 21 , . . . , 2m+n−1 , and
then tile R with the chosen rectangles. Prove that this can be done in at most (m + n)! ways.

Solution. The fact that this can be done in (m +n)! asks for a bijective proof. Now an intuition
gives us that we have to sort the tiles wrt the missing square in some way. Now since the numbers

Problem 2.12.33 (ARO 2016 P1) : There are 30 teams in NBA and every team play 82 games
in the year. Bosses of NBA want to divide all teams on Western and Eastern Conferences (not
necessarily equally), such that the number of games between teams from different conferences is
half of the number of all games. Can they do it?

Solution. You want to divide something. Check the parity.

Problem 2.12.34 (AoPS) : Each edge of a polyhedron is oriented with an arrow such that at each
vertex, there is at least one arrow leaving the vertex and at least one arrow entering the vertex.
Prove that there exists a face on the polyhedron such that the edges on its boundary form a directed
cycle.
2.12 Problems 108

Solution. The trick which is used to prove Euler’s Polyhedron theorem.

Problem 2.12.35 (ISL 2014 C3) : Let n ≥ 2 be an integer. Consider an n × n chessboard

consisting of n2 unit squares. A configuration of n rooks on this board is peaceful if every row and
every column contains exactly one rook. Find the greatest positive integer k such that, for each
peaceful configuration of n rooks, there is a k × k square which does not contain a rook on any of
its k 2 unit squares.

Solution. Guessing the "Correct" ans is the challenge, think of the worst case you can produce.

Problem 2.12.36 (APMO 2012 P2) : Into each box of a n × n square grid, a real number greater
than or equal to 0 and less than or equal to 1 is inserted. Consider splitting the grid into 2 non-empty
rectangles consisting of boxes of the grid by drawing a line parallel either to the horizontal or the
vertical side of the grid. Suppose that for at least one of the resulting rectangles the sum of the
numbers in the boxes within the rectangle is less than or equal to 1, no matter how the grid is split
into 2 such rectangles. Determine the maximum possible value for the sum of all the n × n numbers
inserted into the boxes. Find the ans for k-dimension grids too.

Solution. As the maximal rectangle defines other smaller rectangles in it, we take that.

Problem 2.12.37 (Indian Postal Coaching 2011) : Consider 20112 points arranged in the form of a
2011 × 2011 grid. What is the maximum number of points that can be chosen among them so that
no four of them form the vertices’s of either an isosceles trapezium or a rectangle whose parallel
sides are parallel to the grid lines?

Solution. Since we need to maintain the relation of perpendicular bisectors, we focus on perp
bisectors and the points on one line only and then count.

Problem 2.12.38 (ISL 2010 C2) : On some planet, there are 2N countries (N ≥ 4). Each country
has a flag N units wide and one unit high composed of N fields of size 1 × 1, each field being either
yellow or blue. No two countries have the same flag. We say that a set of N flags is diverse if these
flags can be arranged into an N × N square so that all N fields on its main diagonal will have the
same color. Determine the smallest positive integer M such that among any M distinct flags, there
exist N flags forming a diverse set.
2.12 Problems 109

Solution. Using induction we see that if we have found the value of M for N − 1, then possibly
the value for MN is twice as large than MN−1 . With some further calculation, we see that if we
have 2 ∗ MN−1 − 1 = MN , then we can pick half of them and apply induction and still be left
with a ‘lot’ of flags to choose the Nth element of the diverse set.

After that the only work left is to proof for N = 4. Which is easy casework.

Solution. Another way to prove the ans, is to prove the bound for any non-diverse set. In this
case, we use hall’s marriage to prove the contradiction.

Problem 2.12.39 (Iran TST 2007 P2) : Let A be the largest subset of {1, . . . , n} such that for
each x ∈ A, x divides at most one other element in A. Prove that
 
2n 3n
≤ |A| ≤ .
3 4

Solution. Partition the set optimally.

Problem 2.12.40 (India IMO Camp 2017) : Find all positive integers n s.t. the set {1, 2, . . . , 3n}
can be partitioned into n triplets (ai , bi , ci ) such that ai + bi = ci for all 1 ≤ i ≤ n.

Problem 2.12.41 (ISL 2012 C2) : Let n ≥ 1 be an integer. What is the maximum number of
disjoint pairs of elements of the set {1, 2, . . . , n} such that the sums of the different pairs are
different integers not exceeding n ?

Solution. As Usual, first find the ans. Using double counting is quite natural. Working with
small cases easily gives a construction.

Problem 2.12.42 (CodeForces 989C) :

Problem 2.12.43 (CodeForces 989B) :

2.12 Problems 110

Problem 2.12.44 (ISL 2011 A5) : Prove that for every positive integer n , the set {2, 3, . . . , 3n+1}
can be partitioned into n triples in such a way that the numbers from each triple are the lengths of
the sides of some obtuse triangle.

Solution. What is the best way to choose the side lengths of an obtuse triangle? Obviously by
maintaining some strict rules to get the third side from the first two sides and making the rules
invariant. One way of doing this is to take (a, b, a + b − 1).
After that, some (literally this is the hardest part of the problem) experiment to find a construc-
tion. First, we try to partition the set into tuples of our desired form, but we soon realize that
that can’t be done so easily. So we try a little bit of different approach and make one tuple
different from the others. Luckily this approach gives us a nice construction.

Problem 2.12.45 (Iran TST 2017 D1P1) : In the country of Sugarland, there are 13 students in
the IMO team selection camp. 6 team selection tests were taken and the results have came out.
Assume that no students have the same score on the same test. To select the IMO team, the
national committee of math Olympiad have decided to choose a permutation of these 6 tests and
starting from the first test, the person with the highest score between the remaining students will
become a member of the team. The committee is having a session to choose the permutation.
Is it possible that all 13 students have a chance of being a team member?

Solution. If a student is in x th place in a test ty , and he has a chance to get into the team iff
the 1t h, 2t h . . . x − 1t h persons in test ty are already in the team. So x ≤ 5. Make a 6 · 6 grid
with place · test. WHY?? Because it makes the best sense among other possible choices of the
grid. A little bit of work produces a configuration where every student has a chance to get into
the team.

Problem 2.12.46 (ISL 2009 C2) : For any integer n ≥ 2 , let N(n) be the maximum number of
triples (ai , bi , ci ) , i = 1, 2 . . . , N(n) , consisting of nonnegative integers ai , bi and ci such that the
following two conditions are satisfied:

1 ai + bi + ci = n for all i = 1, . . . , N(n) ,

2 If i 6= j then ai 6= aj , bi 6= bj and ci 6= cj

Determine N(n) for all n ≥ 2.

Solution. Find an upper bound. It’s easy. Then with some experiment, we see that this upper
bound is achievable. So our next task is to find a construction. As it is related to 3 , we first try
with n = 3k. Some experiment and experience gives us a construction.
2.12 Problems 111

Problem 2.12.47 () : Let n be an integer. What is the maximum number of disjoint pairs of
elements of the set {1, 2, . . . , n} such that the sums of the different pairs are different integers not
exceeding n ?

Problem 2.12.48 (ISL 2002 C6) : Let n be an even positive integer. Show that there is a permu-
tation (x1 , x2 . . . xn ) of (1, 2 . . . n) such that for every 1 ≤ i ≤ n , the number xi+1 is one of the
numbers 2xi , 2xi − 1, 2xi − n, 2xi − n − 1. Hereby, we use the cyclic subscript convention, so that
xn+1 means x1 .

Some experiments show that our graph has more than 2 incoming and outgoing degree in all vertexes
expect the first and last vertexes. So our lemma won’t work yet. To make use of our lemma we
take a graph with half of the vertexes of our original graph and make each vertex v2k represent two
integers: (2k − 1, 2k). Simple argument shows that this graph has an Euler Circuit, and surprisingly
this itself is sufficient, as we can follow this circuit to get every integers in the interval [1, n].

Problem 2.12.49 (USA TST 2017 P1) : In a sports league, each team uses a set of at most t
signature colors. A set S of teams is color-identifiable if one can assign each team in S one of
their signature colors, such that no team in S is assigned any signature color of a different team in S.

For all positive integers n and t, determine the maximum integer g(n, t) such that: In any sports
league with exactly n distinct colors present over all teams, one can always find a color-identifiable
set of size at least g(n, t).

Solution. First, guess the answer, then try taking the minimal set.

Problem 2.12.50 (Putnam 2017 A4) : 2N students take a quiz in which the possible scores are
0, 1 . . . 10. It is given that each of these scores appeared at least once, and the average of their
scores is 7.4. Prove that the students can be divided into two sets of N student with both sets
having an average score of 7.4.

Solution. We take a set S1 = {0, 1 . . . 10}. Basically we have to partition the set of 2N into
two equal sets with equal sum. So we pair S , and other leftovers and see what happens.

Problem 2.12.51 (ISL 2005 C3) : Consider a m × n rectangular board consisting of mn unit
squares. Two of its unit squares are called adjacent if they have a common edge, and a path is a
2.12 Problems 112

sequence of unit squares in which any two consecutive squares are adjacent. Two paths are called
non-intersecting if they don’t share any common squares.

Each unit square of the rectangular board can be colored black or white. We speak of a coloring of
the board if all its mn unit squares are colored.

Let N be the number of colorings of the board such that there exists at least one black path from
the left edge of the board to its right edge. Let M be the number of colorings of the board for which
there exist at least two non-intersecting black paths from the left edge of the board to its right edge.

Prove that N 2 ≥ M × 2mn .

Solution. Bijective relation problem, the condition has ×, means we find a combinatorial model
for the R.H.S. which is a pair of boards satisfying conditions. We want to show a surjection from
this model to the model on the L.H.S.

Problem 2.12.52 (Result by Erdos) : Given two different sequence of integers (a1 , a2 . . . an ), (b1 , b2 , . . . bn )
such that two n(n−1)
2 -tuples

a1 + a2 , a1 + a3 . . . an−1 an and b1 + b2 , b1 + b3 . . . bn−1 bn

are equal upto permutation. Prove that n = 2k for some k.

Problem 2.12.53 (A reformulation of Catalan’s Numbers) : Let n ≥ 3 students all have different
heights. In how many ways can they be arranged such that the heights of any three of them are
not from left to right in the order: medium, tall, short?

Solution. The proof uses derivatives to construct a polynomial similar to a Maclaurin Series.

Problem 2.12.54 () : There are n cubic polynomials with three distinct real roots each. Call them
P1 (x), P2 (x), . . . , Pn (x). Furthermore for any two polynomials Pi , Pj , Pi (x)Pj (x) = 0 has exactly 5
distinct real roots. Let S be the set of roots of the equation

P1 (x)P2 (x) . . . Pn (x) = 0

. Prove that
2.12 Problems 113

1 If for each a, b there is exactly one i ∈ {1, . . . n} such that Pi (a) = Pi (b) = 0, then n = 7.
2 If n > 7, |S| = 2n + 1.

Problem 2.12.55 (Serbia TST 2017 P2) : Initially a pair (x, y ) is written on the board, such that
exactly one of it’s coordinates is odd. On such a pair we perform an operation to get pair ( x2 , y + x2 )
if 2|x and (x + y2 , y2 ) if 2|y . Prove that for every odd n > 1 there is a even positive integer b < n
such that starting from the pair (n, b) we will get the pair (b, n) after finitely many operations.

Solution. Finding a construction through investigation and realizing that the infos and opera-
tions on x only defines the changes are enough for this problem.

Problem 2.12.56 (Serbia TST 2017 P4) : We have an n × n square divided into unit squares.
Each side of unit square is called unit segment. Some isosceles right triangles of hypotenuse 2 are
put on the square so all their vertices’s are also vertices’s of unit squares. For which n it is possible
that every unit segment belongs to exactly one triangle (unit segment belongs to a triangle even if
it’s on the border of the triangle)?

Solution. Finding n is even, seeing 4 fails...

Problem 2.12.57 (China MO 2018 P2) : Let n and k be positive integers and let
T = {(x, y , z) ∈ N3 | 1 ≤ x, y , z ≤ n}
be the length n lattice cube. Suppose that 3n2 − 3n + 1 + k points of T are colored red such that
if P and Q are red points and P Q is parallel to one of the coordinate axes, then the whole line
segment P Q consists of only red points.

Prove that there exists at least k unit cubes of length 1, all of whose vertices’s are colored red.

Solution. The inductive solution is tedious, and since we have to count the number of “good”
boxes, we can try double counting. Explicitly counting all the “good” boxes.

Problem 2.12.58 (China MO 2018 P5) : Let n ≥ 3 be an odd number and suppose that each
square in a n × n chessboard is colored either black or white. Two squares are considered adjacent
if they are of the same color and share a common vertex and two squares a, b are considered con-
nected if there exists a sequence of squares c1 , . . . , ck with c1 = a, ck = b such that ci , ci+1 are
2.12 Problems 114

adjacent for i = 1, 2, . . . , k − 1.

Find the maximal number M such that there exists a coloring admitting M pairwise disconnected
squares.

Solution. It’s not hard to get the ans, now that the answer is guesses, and we have tried to
prove with induction and couldn’t find anything good, we try double counting. We notice that all
the connected components in the n × n are planar graphs. Now we use Euler’s theorem on Planar
Graphs to find a value of M wrt to other values, and we double count the other values.

Problem 2.12.59 (USAMO 2006 P2) : For a given positive integer k find, in terms of k, the
minimum value of N for which there is a set of 2k + 1 distinct positive integers that has sum greater
than N but every subset of size k has sum at most N2 .

Solution. The best or simple looking set is the set of consecutive integers. So if there are
some ‘holes’, we can fill them up to some extent, this opens two sub-cases.

Problem 2.12.60 (USAMO 2005 P1) : Determine all composite positive integers n for which it
is possible to arrange all divisors of n that are greater than 1 in a circle so that no two adjacent
divisors are relatively prime.

Problem 2.12.61 (USAMO 2005 P5) : A mathematical frog jumps along the number line. The
frog starts at 1, and jumps according to the following rule: if the frog is at integer n, then it can
jump either to n + 1 or to n + 2mn +1 where 2mn is the largest power of 2 that is a factor of n.
Show that if k ≥ 2 is a positive integer and i is a nonnegative integer, then the minimum number
of jumps needed to reach 2i k is greater than the minimum number of jumps needed to reach 2i .

Solution. The main idea is to notice that the operation only uses powers of 2. And it depends
on only the power of 2 in the integers, and in the sequence of 2-powers, the operation is very
nice.

Problem 2.12.62 (ISL 1991 P10) : Suppose G is a connected graph with k edges. Prove that it
is possible to label the edges 1, 2, . . . , k in such a way that at each vertex which belongs to two or
more edges, the greatest common divisor of the integers labeling those edges is equal to 1.
2.12 Problems 115

Problem 2.12.63 () : A robot has n modes, and programmed as such: in mode i the robot will go
at a speed of i ms−1 for i seconds. At the beginning of its journey, you have to give it a permutation
of {1, 2, . . . n}. What is the maximum distance you can make the robot go?

Problem 2.12.64 () : A slight variation of the previous problem, in this case, the problem goes at
a speed of (n − 1)ms−1 for i seconds in mode i .

Problem 2.12.65 () : m people each ordered n books but because Ittihad was the mailman, he
messed up. Everyone got n books but not necessarily the one they wanted you need to fix this. To
go to a house from another house it takes one hour. You can carry one book with you during any
trip (at most one). You know who has which books and all books are different (i,e, n ∗ m different
books). Prove that you can always finish the job in m ∗ (n + 12 ) hours

Solution. Thinking about the penultimate step, when we have to go to a house empty handed.
Thinking in this way gives us a way to pair the houses up, and since pairing...

Solution. Another way to do this is to convert it to a multi-graph. Now go to a house and

return with a book means removing two edges from that vertex. We play around with it for
sometime

Problem 2.12.66 () : There are n campers in a camp and they will try to solve a IMO P6 but
everyone has a confidence threshold (they will solve the problem by group solving). For example
Laxem has threshold 5. I.e. if he’s in the group, the group needs to contain at least 5 people (him
included). A group is ‘confident’ when everyone of the team is confident. Now MM wants to make
a list of possible “perfect confident” groups. I.e. groups that are confident but adding anyone else
will destroy the confidence. How long can his list be?

Problem 2.12.67 (timus 1862) :

Problem 2.12.68 (ARO 2014 P9.7) : In a country, mathematicians chose an α > 2 and issued
coins in denominations of 1 ruble, as well as αk rubles for each positive integer k. α was chosen
2.12 Problems 116

so that the value of each coins, except the smallest, was irrational. Is it possible that any natural
number of rubles can be formed with at most 6 of each denomination of coins?

Problem 2.12.69 (Saint Petersburg 2001) : The number n is written on a board. A and B take
2 c. The player
turns, each turn consisting of replacing the number n on the board with n − 1 or b n+1
who writes the number 1 wins. Who has the winning strategy?

Solution. Recursively building the losing positions.

Problem 2.12.70 (ARO 2011 P11.6) : There are more than n2 stones on the table. Peter and
Vasya play a game, Peter starts. Each turn, a player can take any prime number less than n stones,
or any multiple of n stones, or 1 stone. Prove that Peter always can take the last stone (regardless
of Vasya’s strategy).

Problem 2.12.71 (ARO 2007 P9.7) : Two players by turns draw diagonals in a regular (2n +1)-gon
(n > 1). It is forbidden to draw a diagonal, which was already drawn, or intersects an odd number
of already drawn diagonals. The player, who has no legal move, loses. Who has a winning strategy?

Solution. Turning the diagonals as vertices, and connection being intersections, we get a graph
to play the game on. We then count the degrees.

Problem 2.12.72 () : After tiling a 6 × 6 box with dominoes, prove that a line parallel to the sides
of the box can be drawn that this line doesn’t cut any dominoes.

Solution. Double count how many lines “cut” a domino, and domino number.

Problem 2.12.73 () : There are 100 points on the plane. You have to cover them with discs, so
that any two disks are at a distance of 1. Prove that you can do this in such a way that the total
diameter of the disks is < 100.

Solution. As the number 100 is very random, we suspect that is true for all values. So we can
use induction
2.12 Problems 117

Problem 2.12.74 (ARO 2014 P10.8) : Given are n pairwise intersecting convex k-gons on the
plane. Any of them can be transferred to any other by a homothety with a positive coefficient.
2k of these k-gons.
Prove that there is a point in a plane belonging to at least 1 + n−1

Solution. The most natural such point should be a vertex of a polygon. And these kinda
problems use PHP more often, so we will have to divide by k somewhere. Again to find the
polygon to use the PHP we will have to divide by n also. So we want to have nk in the
denominator. We change the term to achieve this and Ta-Da! we get a fine term to work
with.

Problem 2.12.75 (IOI 2018 P1) :

——————–

Problem 2.12.76 (USAMO 2005 P1) : Determine all composite positive integers n for which it
is possible to arrange all divisors of n that are greater than 1 in a circle so that no two adjacent
divisors are relatively prime.

Problem 2.12.77 (USAMO 2005 P4) : Legs L1 , L2 , L3 , L4 of a square table each have length n,
where n is a positive integer. For how many ordered 4-tuples (k1 , k2 , k3 , k4 ) of nonnegative integers
can we cut a piece of length ki from the end of leg Li (i = 1, 2, 3, 4) and still have a stable table?
(The table is stable if it can be placed so that all four of the leg ends touch the floor. Note that a
cut leg of length 0 is permitted.)

Problem 2.12.78 (USAMO 2006 P2) : For a given positive integer k find, in terms of k, the
minimum value of N for which there is a set of 2k + 1 distinct positive integers that has sum greater
than N but every subset of size k has sum at most N2 .

Problem 2.12.79 (USAMO 2006 P5) : A mathematical frog jumps along the number line. The
frog starts at 1, and jumps according to the following rule: if the frog is at integer n, then it can
jump either to n + 1 or to n + 2mn +1 where 2mn is the largest power of 2 that is a factor of n.
Show that if k ≥ 2 is a positive integer and i is a nonnegative integer, then the minimum number
of jumps needed to reach 2i k is greater than the minimum number of jumps needed to reach 2i .

Problem 2.12.80 (USAMO 2009 P2) : Let n be a positive integer. Determine the size of the
2.12 Problems 118

largest subset of {−n, −n + 1, . . . , n − 1, n} which does not contain three elements a, b, c (not
necessarily distinct) satisfying a + b + c = 0.

Problem 2.12.81 (IMO 1979 P3) : Two circles in a plane intersect. A is one of the points of
intersection. Starting simultaneously from A two points move with constant speed, each travelling
along its own circle in the same sense. The two points return to A simultaneously after one
revolution. Prove that there is a fixed point P in the plane such that the two points are always
equidistant from P.
2.13 Local 119

2.13 Local

Problem 2.13.1 (ISL 2015 C1) : In Lineland there are n ≥ 1 towns, arranged along a road running
from left to right. Each town has a left bulldozer (put to the left of the town and facing left) and
a right bulldozer (put to the right of the town and facing right). The sizes of the 2n bulldozers are
distinct. Every time when a left and right bulldozer confront each other, the larger bulldozer pushes
the smaller one off the road. On the other hand, bulldozers are quite unprotected at their rears;
so, if a bulldozer reaches the rear-end of another one, the first one pushes the second one off the
road, regardless of their sizes.
Let A and B be two towns, with B to the right of A. We say that town A can sweep town B away
if the right bulldozer of A can move over to B pushing off all bulldozers it meets. Similarly town B
can sweep town A away if the left bulldozer of B can move over to A pushing off all bulldozers of
all towns on its way.
Prove that there is exactly one town that cannot be swept away by any other one.

Solution. Focus on the heaviest bulldozer.

2.14 CP Algorithms 120

2.14 CP Algorithms

• CP Algorithms

2.14.1 Cycle Finding Algorithms

Algorithm (Floyd’s cycle-finding algorithm) — This algorithm finds a cycle by using two pointer.
These pointers move over the sequence at different speeds. In each iteration the first pointer
advances to the next element, but the second pointer advances two elements. It’s not hard to
see, that if there exists a cycle, the second pointer will make at least one full cycle and then meet
the first pointer during the next few cycle loops. If the cycle length is λ and the µ is the first index
at which the cycle starts, then the algorithm will run in O(λ + µ) time.
This algorithm is also known as tortoise and the hare algorithm, based on the tale in which a
tortoise (here a slow pointer) and a hare (here a faster pointer) make a race.
It is actually possible to determine the parameter λ and µ using this algorithm (also in O(λ+µ) time
and O(1) space), but here is just the simplified version for finding the cycle at all. The algorithm
and returns true as soon as it detects a cycle.

1 def floyd (f , x0 , n ) :
2 tortoise = x0
3 hare = x0
4 while ( tortoise != hare ) :
5 tortoise = f ( tortoise , n )
6 hare = f ( hare , n )
7 hare = f ( hare , n )
8 return True

Algorithm (Brent’s algorithm) — Brent uses a similar algorithm as Floyd. It also uses two pointer.
But instead of advancing the pointers by one and two respectably, we advance them in powers of
two. As soon as 2i is greater than λ and µ, we will find the cycle.

1 def f (x , n ) :
2 return ( x * x + 3) % n
3
4 def brent (f , x0 , n ) :
5 tortoise = x0
6 hare = f ( x0 , n )
7 l = 1
2.14.1 Cycle Finding Algorithms 121

8 while ( tortoise != hare ) :

9 tortoise = f ( tortoise , n )
10 for i in range ( l ) :
11 hare = f ( hare , n )
12 if ( hare == tortoise ) :
13 return True
14 l *= 2
15 return True
2.15 Permutations 122

2.15 Permutations

Problem 2.15.1 (USAMO 2017 P5) : Let m1 , m2 , . . . , mn be a collection of n positive integers,

not necessarily distinct. For any sequence of integers A = (a1 , . . . , an ) and any permutation w =
w1 , . . . , wn of m1 , . . . , mn , define an A-inversion of w to be a pair of entries wi , wj with i < j for
which one of the following conditions holds:
ai ≥ wi > wj
wj > ai ≥ wi , or
wi > wj > ai .

Show that, for any two sequences of integers A = (a1 , . . . , an ) and B = (b1 , . . . , bn ), and for any
positive integer k, the number of permutations of m1 , . . . , mn having exactly k A-inversions is equal
to the number of permutations of m1 , . . . , mn having exactly k B-inversions.

Solution. Define the function fAw (i ) = the number of inversions wi , wj with i < j.

Notice that if we take B as a sequence with all elements greater than all wi , then we have the
B-inversions to be normal inversions wrt M. So we need to show that the multi set
n
( )
X
k| k= fB (i ) , for some permutation w
w

i=1

is the same as the set defined with fA .

So we need to show a bijection between A-inversion and normal inversion.

But showing that for every permutation with some normal inversions there is a A with the same
number of A-inversions is pretty hard. So we instead show that for every permutation w , there is
another permutation p such that A-inversions of w are the same as normal inversions of p. And
that two permutations w1 , w2 don’t have the same set of A-inversions.
Chapter 3

Algebra
3.1 Functional Equations 124

3.1 Functional Equations

Can’t Start? Try These

1 GUESS THE POSSIBLE SOLUTIONS.

2 SUBSTITUTION

a Try EVERY possible substitutions, and write them in a list, dont think during this
time.

b Now think what these results give you.

c Find values of f (0), f (1), f (2), f (−x) etc.

d Tweak the function a little bit, do substitution again.

e Assume some other functions according to the solutions, substitute them to make
the fe easier to get info out of.

3 PROPERTIES OF THE FUNCTION

a Try proving INJECTIVITY, SURJECTIVITY etc.

b Look for Injectivity or Surjectivity of f (x) − f (y ).

4 Assume for the sake of contradiction that the value of the function is greater or smaller
than the estimated value at some point.

5 Sometimes consider the difference of two values of f .

3.1 Functional Equations 125

Stuck? Try These

1 Proving that f (x) − x is injective might come handy in some cases.

2 If you’re NOT able to make one side of the equation equal to 0, try to make it equal to
any real or some particular real. (pco 169 P11)

3 Sometimes in integer functions, divisibility of the type f (1)k−1 | f (x)k helps.

4 Durr. . . I want things to cancel.

3.1.1 Problems 126

3.1.1 Problems

Problem 3.1.1 (EGMO 2012 P3) : Find all functions f : R → R such that

f (y f (x + y ) + f (x)) = 4x + 2y f (x + y )

for all x, y ∈ R.

Problem 3.1.2 (pco 169 P11) : Find all f : R → R such that for all real numbers x, y the following
holds:
f (x)2 + 2y f (x) + f (y ) = f (y + f (x))

Problem 3.1.3 (IMO 1994 P5) : Let S be the set of all real numbers strictly greater than −1.
Find all functions f : S → S satisfying the two conditions:

1. f (x + f (y ) + xf (y )) = y + f (x) + y f (x) for all x, y in S;

f (x)
2. x is strictly increasing on each of the two intervals −1 < x < 0 and 0 < x.

Problem 3.1.4 (ISL 1994 A4) : Let R denote the set of all real numbers and R+ the subset of
all positive ones. Let α and β be given elements in R, not necessarily distinct. Find all functions
f : R+ 7→ R such that:
x  y 
f (x)f (y ) = y α f + xβf ∀x, y ∈ R+ .
2 2

Problem 3.1.5 (IMO 2017 P2) : Let R be the set of real numbers. Determine all functions
f : R → R such that, for any real numbers x and y ,

f (f (x)f (y )) + f (x + y ) = f (xy ).

Problem 3.1.6 (ISL 2008 A1) : Find all functions f : (0, ∞) 7→ (0, ∞) (so f is a function from
the positive real numbers) such that

(f (w ))2 + (f (x))2 w2 + x2
=
f (y 2 ) + f (z 2 ) y2 + z2

for all positive real numbers w , x, y , z , satisfying w x = y z.

3.1.1 Problems 127

Problem 3.1.7 (pco 169 P15) : Find all a ∈ R for which there exists a non-constant function
f : (0, 1] → R such that

a + f (x + y − xy ) + f (x)f (y ) ≤ f (x) + f (y )

for all x, y ∈ (0, 1]

Problem 3.1.8 (pco 168 P18) : Find all functions f : R → R such that

f (f (x) + y ) = f (x 2 − y ) + 4f (x)y

for all x, y ∈ R

Problem 3.1.9 (ISL 2011 A3) : Determine all pairs (f , g) of functions from the set of real numbers
to itself that satisfy
g(f (x + y )) = f (x) + (2x + y )g(y )
for all real numbers x and y .

Problem 3.1.10 (ISL 2005 A2) : We denote by R+ the set of all positive real numbers. Find all
functions f : R+ → R+ which have the property:

f (x)f (y ) = 2f (x + y f (x))

for all positive real numbers x and y .

Solution. Let’s first substitute. If there existed some x such that f (x) < 1, we could find a
nice substitution. But that leads to a contradiction. So what if we could do something like this
for the other cases, f (x) < 2 and f (x) > 2?

Problem 3.1.11 (ISL 2005 A4) : Find all functions f : R → R such that f (x + y ) + f (x)f (y ) =
f (xy ) + 2xy + 1 for all real numbers x and y .

Solution. Substitution.

Problem 3.1.12 (Iran TST T2P1) : Find all functions f : R → R that satisfy the following
conditions:

1. x + f (y + f (x)) = y + f (x + f (y )) ∀x, y ∈ R
n o
2. The set I = f (x)−f
x−y
(y )
| x, y ∈ R, x 6
= y is an interval.
3.1.1 Problems 128

Problem 3.1.13 (169 P20) : Let a be a real number and let f : R → R be a function satisfying:
f (0) = 21 and
f (x + y ) = f (x)f (a − y ) + f (y )f (a − x)
∀x, y ∈ R. Prove that f is constant

Problem 3.1.14 (Vietnam 1991) : Find all functions f : R → R for which

1 1 1
f (xy ) + f (xz) − f (x)f (y z) ≥
2 2 4

Solution. Just substitute.

Problem 3.1.15 () : Suppose that f and g are two functions defined on the set of positive integers
and taking positive integer values. Suppose also that the equations f (g(n)) = f (n) + 1 and
g(f (n)) = g(n) + 1 hold for all positive integer n. Prove that f (n) = g(n) for all positive integer n.

Solution. Durrr... I want things to cancel... Hint: You want to show f (n) − g(n) = 0.

Problem 3.1.16 (ISL 2002 A1) : Find all functions f from the reals to the reals such that

f (f (x) + y ) = 2x + f (f (y ) − x)

for all real x, y .

Solution. On one of our substitution, we see that there is surjectivity in the equation. So
trying to show injectivity is the most intuitive move after that. Again, we have x on the outside,
so we need to make x, a once and b once. but we have f (y ) − x which we need to eleminate,
keeping y constant. We can make it either a or b since we already have f (a) = f (b). And again
we can take whatever value we want for f (y ).

Problem 3.1.17 (ISL 2001 A1) : Let T denote the set of all ordered triples (p, q, r ) of nonnegative
integers. Find all functions f : T → R satisfying



0 if pqr = 0,
1 + 1 (f (p + 1, q − 1, r ) + f (p − 1, q + 1, r )

6
f (p, q, r ) =


+f (p − 1, q, r + 1) + f (p + 1, q, r − 1)
otherwise

+f (p, q + 1, r − 1) + f (p, q − 1, r + 1))

for all nonnegative integers p, q, r .

3.1.1 Problems 129

Solution. First let us guess the ans. For all points on the 3 sides, our function gives 0. We
get f (1, 1, 1) = 1. We get f (1, 1, 2) = f (1, 2, 1) = f (2, 1, 1) = 32 . We get f (1, 1, 3) = 59 . We
get f (1, 2, 2) = 12
5 . Now, since for pqr = 0, we have f = 0, we need the expression pqr on
the numerator. And we kinda guess that the denominator is p + q + r . From here the guess is
obvious.
Now proving that this solution is the only solution. Let the solution be g. Define, h := f − g.
Our aim is to prove that h = 0 for all inputs.

Problem 3.1.18 (RMM 2019 P5) : Determine all functions f : R → R satisfying

f (x + y f (x)) + f (xy ) = f (x) + f (2019y ),

for all real numbers x and y .

Solution. After getting f (y f (0)) = f (y 2019), one should think of proving that either f is
constant, all zero except 0, or linear. How to do this?

Problem 3.1.19 (APMO 2015 P2) : Let S = {2, 3, 4, . . .} denote the set of integers that are
greater than or equal to 2. Does there exist a function f : S → S such that

f (a)f (b) = f (a2 b2 ) for all a, b ∈ S with a 6= b?

Solution. Try to break the symmetry, add another variable.

Problem 3.1.20 (ISL 2015 A2) : Determine all functions f : Z → Z with the property that

f (x − f (y )) = f (f (x)) − f (y ) − 1

holds for all x, y ∈ Z.

Solution. It just flows.

Problem 3.1.21 (ISL 2015 A4) : Let R be the set of real numbers. Determine all functions
f : R → R that satisfy the equation

f (x + f (x + y )) + f (xy ) = x + f (x + y ) + y f (x)

for all real numbers x and y .

Solution. When you don’t know any heavy techniques, just plug in simple values into the
function, and write down all of the equations in a list.
3.1.1 Problems 130

Problem 3.1.22 (ISL 2012 A5) : Find all functions f : R → R that satisfy the conditions

f (1 + xy ) − f (x + y ) = f (x)f (y ) for all x, y ∈ R,

and f (−1) 6= 0.

Solution. In FE, always look back to what you have, and what things can you make from
those.

Problem 3.1.23 (ISL 2012 A1) : Find all functions f : Z → Z such that, for all integers a, b, c
that satisfy a + b + c = 0, the following equality holds:

f (a)2 + f (b)2 + f (c)2 = 2f (a)f (b) + 2f (b)f (c) + 2f (c)f (a).

(Here Z denotes the set of integers.)

Solution. Go with the flow.

Problem 3.1.24 (ISL 2012 A1) : Find all function f : R → R such that for all x, y ∈ R the
following equality holds
f (bxc y ) = f (x) bf (y )c
where bac is greatest integer not greater than a.

Solution. Go with the flow.

3.2 Weird Ones 131

3.2 Weird Ones

Problem 3.2.1 (ISL 2009 A3) : Determine all functions f from the set of positive integers to the
set of positive integers such that, for all positive integers a and b, there exists a non-degenerate
triangle with sides of lengths
a, f (b) and f (b + f (a) − 1).
(A triangle is non-degenerate if its vertices are not collinear.)

Solution. f (1) > 1 =⇒ f is periodic =⇒ repeatation =⇒ contradiction.

f (2) > 2 =⇒ strictly increasing =⇒ repeatation.

Problem 3.2.2 (USA TST 2018 P2) : Find all functions f : Z2 → [0, 1] such that for any integers
x and y ,
f (x − 1, y ) + f (x, y − 1)
f (x, y ) = .
2

Solution. We know that the function has to be a constant function. So it is a intuitive idea
considering the difference of two values of the function. Again as we wish to show that this
difference is 0, we have to use either equality of limit. As equality is quite ambiguous in this
problem, we approach with limits. We see that f (x, y ) can be written as a term depending on
the values of the 3rd quarter of the plane with (x, y ) as its origin. With infinite values in our
hand, we try bounding.
3.3 FE cantonmathguy Seclected Problems 132

3.3 FE cantonmathguy Seclected Problems

1. Determine all functions f : Q → Q satisfying f (x + y ) = f (x) + f (y ) for all x, y ∈ Q.

2. Let a1 , a2 , . . . be a sequence of integers with infinitely many positive and negative terms. Sup-
pose that for every positive integer n the numbers a1 , a2 , . . . , an leave n different remainders
upon division by n. Prove that every integer occurs exactly once in the sequence a1 , a2 , . . ..

3. Determine all functions f : R → R satisfying

f (x)f (y ) = f (x + y ) + xy

for all real x and y .

4. Find all functions f : Z → Z such that, for all integers a, b, c that satisfy a + b + c = 0, the
following equality holds:

f (a)2 + f (b)2 + f (c)2 = 2f (a)f (b) + 2f (b)f (c) + 2f (c)f (a).

5. Determine all functions f : R → R satisfying

f (x) + f (y ) = f (x + y ) and f (xy ) = f (x)f (y )

for all x, y ∈ R.

6. Find all functions f : R → R such that for all x, y ∈ R, the following equality holds

f (bxc y ) = f (x) bf (y )c

where bac is the greatest integer not greater than a.

7. F Let k be a real number. Find all functions f : R → R satisfying

|f (x) − f (y )| ≤ k(x − y )2

for all real x and y .

8. Let f : N → N be a function, and suppose that positive integers k and c satisfy

f k (n) = n + c

for all n ∈ N, where f k denotes f applied k times. Show that k | c.

9. Find all functions f : N → N satisfying

f (f (f (n))) + f (f (n)) + f (n) = 3n

for every positive integer n.

3.3 FE cantonmathguy Seclected Problems 133

10. Let S be the set of integers greater than 1. Find all functions f : S → S such that (i) f (n) | n
for all n ∈ S, (ii) f (a) ≥ f (b) for all a, b ∈ S with a | b.
11. Let R be the set of real numbers. Determine all functions f : R → R such that

f (x 2 − y 2 ) = xf (x) − y f (y )

for all pairs of real numbers x and y .

12. F Let T denote the set of all ordered triples (p, q, r ) of nonnegative integers. Find all
functions f : T → R satisfying



0 if pqr = 0,
1 + 1 (f (p + 1, q − 1, r ) + f (p − 1, q + 1, r )

6
f (p, q, r ) =


+f (p − 1, q, r + 1) + f (p + 1, q, r − 1)
otherwise

+f (p, q + 1, r − 1) + f (p, q − 1, r + 1))

for all nonnegative integers p, q, r .

13. Determine all strictly increasing functions f : N → N satisfying nf (f (n)) = f (n)2 for all
positive integers n.
14. Determine all functions f : Z → Z with the property that

f (x − f (y )) = f (f (x)) − f (y ) − 1

holds for all x, y ∈ Z.

15. Find all real-valued functions f defined on pairs of real numbers, having the following property:
for all real numbers a, b, c, the median of f (a, b), f (b, c), f (c, a) equals the median of a, b, c.
16. Find all functions f : N → N such that, for all positive integer n, we have f (f (n)) < f (n + 1).
17. Let f : N → N be a function such that, for any w , x, y , z ∈ N,

f (f (f (z)))f (w xf (y f (z))) = z 2 f (xf (y ))f (w ).

Show that f (n!) ≥ n! for every positive integer n.

18. Find all functions f : N → N such that f (n!) = f (n)! for all positive integers n and such that
m − n divides f (m) − f (n) for all distinct positive integers m, n.
19. Find all functions f from the reals to the reals such that

(f (a) + f (b))(f (c) + f (d)) = f (ac + bd) + f (ad − bc)

for all real a, b, c, d.

20. Determine all functions f defined on the natural numbers that take values among the natural
numbers for which
(f (n))p ≡ n (mod f (p))
for all n ∈ N and all prime numbers p.
3.3 FE cantonmathguy Seclected Problems 134

21. Let n ≥ 4 be an integer, and define [n] = {1, 2, . . . , n}. Find all functions W : [n]2 → R such
that for every partition [n] = A ∪ B ∪ C into disjoint sets,
XXX
W (a, b)W (b, c) = |A||B||C|.
a∈A b∈B c∈C

22. F Find all infinite sequences a1 , a2 , . . . of positive integers satisfying the following properties:
(a) a1 < a2 < a3 < · · · , (b) there are no positive integers i , j, k, not necessarily distinct,
such that ai + aj = ak , (c) there are infinitely many k such that ak = 2k − 1.

23. Show that there exists a bijective function f : N0 → N0 such that for all m, n ∈ N0 ,

f (3mn + m + n) = 4f (m)f (n) + f (m) + f (n)

24. Determine all functions f : Z → Z satisfying

f (f (m) + n) + f (m) = f (n) + f (3m) + 2014

for all integers m and n.

25. Let n ≥ 3 be a given positive integer. We wish to label each side and each diagonal of a
regular n-gon P1 . . . Pn with a positive integer less than or equal to r so that:

a) every integer between 1 and r occurs as a label;

b) in each triangle Pi Pj Pk two of the labels are equal and greater than the third.

Given these conditions:

a) Determine the largest positive integer r for which this can be done.
b) For that value of r , how many such labellings are there?

26. F Suppose that f and g are two functions defined on the set of positive integers and taking
positive integer values. Suppose also that the equations f (g(n)) = f (n) + 1 and g(f (n)) =
g(n) + 1 hold for all positive integer n. Prove that f (n) = g(n) for all positive integer n.

27. Find all the functions f : N0 → N0 satisfying the relation

28. Let R be the set of real numbers. Determine all functions f : R → R that satisfy the equation

f (x + f (x + y )) + f (xy ) = x + f (x + y ) + y f (x)

for all real numbers x and y .

29. Suppose that s1 , s2 , s3 , . . . is a strictly increasing sequence of positive integers such that the
sub-sequences
ss1 , ss2 , ss3 , . . . and ss1 +1 , ss2 +1 , ss3 +1 , . . .
are both arithmetic progressions. Prove that the sequence s1 , s2 , s3 , . . . is itself an arithmetic
progression.
3.3 FE cantonmathguy Seclected Problems 135

30. Find all functions f from N0 to itself such that

f (m + f (n)) = f (f (m)) + f (n)

for all m, n ∈ N0 .

31. F Consider a function f : N → N. For any m, n ∈ N we write f n (m) = f (f (. . . f (m) . . .)).

| {z }
n
Suppose that f has the following two properties:
f n (m)−m
a) if m, n ∈ N, then n ∈ N;
b) The set N \ {f (n) | n ∈ N} is finite.

Prove that the sequence f (1) − 1, f (2) − 2, f (3) − 3, . . . is periodic.

32. Let N be the set of positive integers. Find all functions f : N → N that satisfy the equation

f abc−a (abc) + f abc−b (abc) + f abc−c (abc) = a + b + c

for all a, b, c ≥ 2.

33. Let 2Z + 1 denote the set of odd integers. Find all functions f : Z → 2Z + 1 satisfying

f (x + f (x) + y ) + f (x − f (x) − y ) = f (x + y ) + f (x − y )

for every x, y ∈ Z.
3.4 Polynomials 136

3.4 Polynomials

3.4.1 Techniques to remember

Stuck? Try These

1 A polynomial with odd degree always has at least one real root.

2 If a polynomial with even degree has a negative value on its graph, then it has at least
one real root.

3 Roots of unity divide a polynomial in parts like congruence classes.

4 MODULUS SIGN: use Triangle Inequality.

5 They say, In Poly, chase ROOTS.

1
6 x n f ( ) has the same coefficients as f (x), but in opposite order.
x

Theorem 3.4.1 (Lagrange Interpolation Theorem) — Given n real numbers, there exist a poly-
nomial with at most n − 1 degree such that the graph of the polynomial goes through all of the
points.

Theorem 3.4.2 (Finite Differences) — This is the descrete form of derivatives. The first finite
difference of a function f is defined as g(x) := f (x + 1) − f (x).
n + 1th finite difference of a n degree polynomial: For any polynomial P (x) of degree at most n
the following equation holds:
n+1  
i n+1
X
(−1) P (i ) = 0
i
i=0

Remark. This can be used

1. to reduce the degree of a polynomial, manipulate the coefficients etc.

2. to solve recurrences, where the recurrence equation is a bit messy and contains a lot of
3.4.1 Techniques to remember 137

previous values. Like this recurrence is quite messy to solve as it is, but if we take the first
finite difference here, it becomes easy:
1 2
f (x) = f (x + 1) + f (x − 1) + 1
3 3

It’s like solving for the first derivative and then finding the original function.
3.4.2 General Problems 138

3.4.2 General Problems

Problem 3.4.1 (USA TST 2014 P4) : Let n be a positive even integer, and let c1 , c2 , . . . , cn−1 be
real numbers satisfying
n−1
X
|ci − 1| < 1.
i=1

Prove that
2x n − cn−1 x n−1 + cn−2 x n−2 − · · · − c1 x 1 + 2
has no real roots.

Solution. A polynomial has no real root means that the polynomial completely lies in either
of the two sides of the x-axis. So in this case, we have to prove that P (x) > 0. So we gotta
try to bound. Again, to make |ci − 1| a little bit more approachable, we assign bi = ci − 1 and
write P (x) in terms of bi . Now, how to bring the modulus sign in our polynomial? Oh, we have
triangle ineq for those kinda work :0

Problem 3.4.2 (USAMO 2002 P3) : Prove that any monic polynomial (a polynomial with leading
coefficient 1) of degree n with real coefficients is the average of two monic polynomials of degree
n with n real roots.

Solution. (i) If we have n + 1 points, we have an unique polynomial through them.

(ii) If we have one positive value of a polynomial and one negative value, then there exists a real
root between that two values.

Problem 3.4.3 (China TST 1995 P5) : A and B play the following game with a polynomial of
degree at least 4:
x 2n + _ x 2n−1 + _ x 2n−2 + . . . + _ x + 1 = 0
A and B take turns to fill in one of the blanks with a real number until all the blanks are filled up.
If the resulting polynomial has no real roots, A wins. Otherwise, B wins. If A begins, which player
has a winning strategy?

Solution. Not always (actually in very few cases) the first move decides the winning strategy.
In this case, if B could make the last move, he would definitely win. But as he can’t, consider
the final two moves. Again “Waves”.

Problem 3.4.4 (Zhao Polynomials) : A set of n numbers are considered to be k-cool if a1 +

ak+1 · · · = a2 + ak+2 · · · = · · · = ak + a2k . . . . Suppose a set of 50 numbers are 3, 5, 7, 11, 13, 17-
cool. Prove that every element of that set is 0.
3.4.2 General Problems 139

Solution. Equivalence class :0 roots of unity :0 :0 :0

Problem 3.4.5 (All Russian Olympiad 2016, Day2, Grade 11, P5) : Let n be a positive integer and
let k0 , k1 , . . . , k2n be nonzero integers such that k0 + k1 + · · · + k2n 6= 0. Is it always possible to find
a permutation (a0 , a1 , . . . , a2n ) of (k0 , k1 , . . . , k2n ) so that the equation

a2n x 2n + a2n−1 x 2n−1 + · · · + a0 = 0

has no integer roots?

Solution. The degree is 2n, and we have to find a zero, so proving/disproving the existence
of negative value of P (x) is enough. If all the values of P (x) are to be positive, the leading
coefficient must be very big...

Problem 3.4.6 (Zhao Poly) : Let f (x) be a monic polynomial with degree n with distinct zeroes
x1 , x2 , ..., xn . Let g(x) be any monic polynomial of degree n − 1. Show that
n
X g(xj )
=1
f 0 (xj )
j=1
Y
where f 0 (xi ) = (xi − xj )
j6=i

Solution. Lagrange’s Interpolation

Problem 3.4.7 (ARO 2018 P11.1) : The polynomial P (x) is such that the polynomials P (P (x))
and P (P (P (x))) are strictly monotone on the whole real axis. Prove that P (x) is also strictly
monotone on the whole real axis.

Problem 3.4.8 (Serbia 2018 P4) : Prove that there exists a uniqe P (x) polynomial with real
coefficients such that
xy − x − y | (x + y )1000 − P (x) − P (y )
for all real x, y .

Solution. Substitution.
3.4.3 Root Hunting 140

3.4.3 Root Hunting

Problem 3.4.9 (Putnam 2017 A2) : Let Q0 (x) = 1, Q1 (x) = x, and

(Qn−1 (x))2 − 1
Qn (x) =
Qn−2 (x)
for all n ≥ 2. Show that, whenever n is a positive integer, Qn (x) is equal to a polynomial with
integer coefficients.

3.4.4 NT Polynomials

Problem 3.4.10 (Iran TST 2009 P4) : Find all polynomials f with integer coefficient such that,
for every prime p and natural numbers u and v with the condition:

p | uv − 1

we always have
p | f (u)f (v ) − 1

1
Solution. Notice that we can disregard v by considering it , and the condition won’t be
u
affected, because primes allow multiplicative inverses. After this observation the problem is
almost solved.

Problem 3.4.11 (Iran TST 2004 P6) : p is a polynomial with integer coefficients and for every
natural n we have p(n) > n. xk is a sequence that: x1 = 1, xi+1 = p(xi ) for every N one of xi is
divisible by N. Prove that p(x) = x + 1

Solution. Notice that {xi } becomes periodic mod any prime. Now, we start by showing that
P (1) = 2. We have, P (x) has to be even. If it is > 2 then what happens? what if we take
N = P (1) − 1?

Problem 3.4.12 (ISL 2006 N4) : Let P (x) be a polynomial of degree n > 1 with integer coefficients
and let k be a positive integer. Consider the polynomial Q(x) = P (P (. . . P (P (x)) . . . )), where P
occurs k times. Prove that there are at most n integers t such that Q(t) = t.

Solution. Suppose that there are more than n fixed points. So at least one of them can’t be
a fixed point of P . Use that. Follow.
3.4.5 Fourier Transformation 141

Problem 3.4.13 (ISL 2012 A4) : Let f and g be two nonzero polynomials with integer coefficients
and deg f > deg g. Suppose that for infinitely many primes p the polynomial pf + g has a rational
root. Prove that f has a rational root.

Solution. dunno

3.4.5 Fourier Transformation

Lemma 3.4.3 ( Dealing with binomial terms with a common factor) — Let n, k be two integers,
and let z be a kth root of unity other than 1. Then,

(1 + z 1 )n + (1 + z 2 )n + · · · + (1 + z k )n
     
n n n
+ + + ··· =
0 k 2k k

Proof. For j not divisible by k,

k
!
X
j i
z z =0
i=1
3.4.6 Irreducibility 142

3.4.6 Irreducibility

• Summer Camp 2015 Handout

• Yufei Zhao’s Handout

Stuck? Try These: What can be showed to prove Irreducibility

• Writing f = g · h and equating coefficients

• If the polynomial involves some prime, it’s often useful to try factoring modulo that prime

• If the last coefficient is a prime, then there are some obvious bounds on the roots

• If there are bounds on the coefficients, then try root bounding

Lemma 3.4.4 ( Bounds On Roots) — P is Lemma 3.4.5 — P is a monic polynomial.

a monic polynomial. Suppose P (0) 6= 0 and Suppose that |P (0)| is prime, and all complex
at most one complex root of P has absolute roots of P have absolute value greater than
value at least 1. Then P is irreducible. 1. Then P is irreducible.

Lemma 3.4.6 ( Leading Coefficient is Lemma 3.4.7 ( Coefficients form a De-

LARGE) — Let P (x) = bn x n + bn−1 x n−1 + creasing Sequence) — Let P (x) = an x n +
· · · + b1 x + b0 such that an−1 x n−1 + · · · + a1 x + a0 be a real polyno-
mial. Such that,
|bn | > |bn−1 | + |bn−2 | + · · · + |b0 |
an ≥ an−1 ≥ · · · ≥ a1 ≥ a0 > 0
Then every root α of P is strictly inside of
the unit circle, i.e. |α| < 1. Then any complex z of P (x) satisfies |z| ≤ 1

i.e. If the first coefficient of the polynomial is i.e. If the coefficients form a decreasing se-
very large, then all of the roots lie inside the quence then all of the roots lie on or inside
unit circle. the unit circle.

Lemma 3.4.8 ( Constant is LARGE) — Let Theorem 3.4.9 (Rouché’s Theorem) — Let
P (x) = an x n + an−1 x n−1 · · · + a1 x + a0 be f , g be analytic functions on and inside a sim-
a polynomial over integers. Where, a0 is a ple closed curve C. Suppose that
prime, and
|f (z)| > |g(z)|
|a0 | > |an | + |an−1 | + · · · + |a1 |
for all points z on C. Then f and f + g have
Prove that P (x) is irreducible. the same number of zeroes (counting multi-
plicities) interior to C
3.4.6 Irreducibility 143

Theorem 3.4.10 (Perron’s Criterion) — Let P (x) = x n + an−1 x n−1 + an−2 x n−2 . . . a1 x + a0 be
a polynomial over integers such that

|an−1 | > 1 + |an−2 | + |an−3 | + · · · + |a1 | + |a0 |

Then P (x) is irreducible.

Remark. The crucial idea behind the proof is that |a0 | ≥ 1, and if the polynomial is reducible, then
there are at least two roots |z| ≥ 1.

Proof [Bounding Roots]. Let P (z) = 0 for some |z| = 1, z ∈ C. That means we have

−an−1 z n−1 = z n + an−2 z n−2 . . . a1 z + a0

=⇒ |an−1 | = |z n + an−2 z n−2 . . . a1 z + a0 |
≤ |1| + |an−2 | · · · + |a0 |

Which is a contradiction. So |z| =

6 1.

We know that there exist a root z that has an absolute value greater than 1. We prove that
there is only one such root of P (x).

First, let P (x) = (x − z)Q(x), where |z| > 1, and Q(x) = x n−1 + bn−2 x n−2 + · · · + b1 x + b0 .
So we have,

P (x) = (x − z)Q(x)
x n + an−1 x n−1 + an−2 x n−2 . . . a1 x + a0 = x n + (bn−2 − z)x n−1 + · · · + (b0 − z b1 )x + z b0

=⇒ |bn−2 − z| > 1 + |bn−3 − z bn−2 | + · · · + |b0 − z b1 | + |z b0 |

|bn−2 | + |z| > 1 − |bn−3 | + |z||bn−2 | + . . . |b0 | − |z||b1 | + |z|b0 |

|bn−2 | + |z| = (|z| − 1)(|bn−2 | + |bn−3 . . . |b0 |) + |bn−2 | + 1

1 > |bn−2 | + |bn−3 . . . |b0 |

And by Lemma 3.4.6, Q(x) does not have any root |z| > 1.

Theorem 3.4.11 (Perron’s Criterion’s Generalization (Dominating Term)) — Let P (z) = an z n +

an−1 z n−1 + · · · + a1 z + a0 be a complex polynomial, such that its ak term is dominant, that is,

|ak | > |a0 | + |a1 | + . . . |ak−1 | + |ak+1 | + · · · + |an |

for some 0 ≤ k ≤ n. Then exactly k roots of P lies strictly inside of the unit circle, and the other
3.4.6 Irreducibility 144

n − k roots of P lies strictly outside of the unit circle.

Proof. A direct application of Theorem 3.4.9.

3.4.6 Irreducibility 145

Lemma 3.4.12 ( Bound on roots) — Let f (x) = an x n + an−1 x n−1 · · · + a1 x + a0 be an integer

polynomial. Suppose that an ≥ 1, an−1 ≥ 0 and ai ≤ H for some positive constant H and
i = 0, 1, . . . n − 2. Then any complex zero α of f (x) has either nonpositive real part, or satisfies
√
1 + 1 + 4H
|α| <
2

Proof. Suppose z is a root such that |z| > 1 and Re z > 0. Then we have
 
f (z)  an−1  1 1 1
≥ an − −H + ... n
zn z |z|2 |z|3 |z|
 an−1  H
> Re an + − 2
z |z| − |z|
H
≥1−
|z|2 − |z|

|z|2 − |z| − H
=
|z|2 − |z|
≥0

Whenever √
1+ 1 + 4H
|z| ≥
2

Theorem 3.4.13 (Cohn’s Criterion) — Suppose p is a prime number, expressed as pn pn−1 . . . p1 p0

in base b ≥ 2. Then the polynomial

f (x) = pn x n + pn−1 x n−1 · · · + p1 x + p0

is irreducible.

Problem 3.4.14 (ISL 2005 A1) : Find all pairs of integers a, b for which there exists a polynomial
P (x) ∈ Z[X] such that product (x 2 + ax + b) · P (x) is a polynomial of a form

x n + cn−1 x n−1 + · · · + c1 x + c0

where each of c0 , c1 , . . . , cn−1 is equal to 1 or −1.

Solution. The idea of bounding the roots using the coefficients.

3.4.6 Irreducibility 146

Problem 3.4.15 () : Let P (x) be a polynomial with real coefficients, and P (x) ≥ 0 for all x ∈ R.
Prove that there exists two polynomials R, S ∈ Q such that

P (x) = R(x)2 + Q(x)2

Problem 3.4.16 (Romanian TST 2006 P2) : Let p a prime number, p ≥ 5. Find the number of
polynomials of the form

x p + px k + px l + 1, k > l, k, l ∈ {1, 2, · · · , p − 1} ,

which are irreducible in Z[X].

Solution. Taking mod p, we have that x p + 1 ≡ (x + 1)p (mod p). Now we can try equating
terms or plug in some values to check for equality.

Problem 3.4.17 (Romanian TST 2003 P5) : Let f ∈ Z[X] be an irreducible polynomial over the
ring of integer polynomials, such that |f (0)| is not a perfect square. Prove that if the leading
coefficient of f is 1 (the coefficient of the term having the highest degree in f ) then f (X 2 ) is also
irreducible in the ring of integer polynomials.

Solution. If f (x 2 ) = g(x)h(x), plugging −x gives us g(x)h(x) = g(−x)h(−x). So we should

look at the common roots of h(x) and h(−x). And it is straightforward from here.

Solution.
3.5 Inequalities 147

3.5 Inequalities

• Olympiad Inequalities - Thomas J. Mildorf

• A < B - Keran Kedlaya

• Convexity - Po Shen Loh

• Brief Intro to Ineqs - Evan Chen

Definition (Majorizes) — Given two sequences of real numbers x1 ≥ x2 ≥ · · · ≥ xn and y1 ≥ y2 ≥

· · · ≥ yn , we say (xn ) majorizes (yn ), written (xn )  (yn ) if

x1 + x2 · · · + xn = y1 + y2 · · · + yn and,
x1 + x2 · · · + xk ≥ y1 + y2 · · · + yk ∀1 ≤ k ≤ n − 1

Definition (Mean Values) — Given n positive reals a2 . . . an , we have

P
ai a1 + a2 + · · · + an
Arithmetic Mean : =
n n

√
qY
n
Geometric Mean :
n
ai = a1 a2 . . . an

s r
ai2 a12 + a22 + · · · + an2
P
Quadratic Mean : =
n n

n n
Harmonic Mean : P1 = 1 1 1
ai a1 + a2 + ··· + an
3.5.1 Basic Inequalities 148

3.5.1 Basic Inequalities

Theorem 3.5.1 (Muirhead’s Inequality) — If a1 , a2 , . . . an are real positive reals, and (xn )  (yn ),
then we have,

X X
a1x1 a2x2 · · · anxn ≥ a1y1 a2y2 · · · anyn
sym sym

Theorem 3.5.2 (Jensen’s Inequality) — If f is convex, then

 
f (a1 ) + f (a2 ) · · · + f (an ) a1 + a2 · · · + an
≥f
n n

The reverse inequality holds if f is concave.

Theorem 3.5.3 (Karamata’s Inequality) — If f is convex and (xn )  (yn ), then

f (x1 ) + f (x2 ) · · · + f (xn ) ≥ f (y1 ) + f (y2 ) · · · + f (xn )

The reverse inequality holds if f is concave.

Theorem 3.5.1 (title=Tangent Line Trick: Best-Nice bound for function)

Even if a fucntion is not convex or concave for us to use Jensen’s Inequality, we can still find
a number a such that for our required interval I, f stays above (or below) the tangent line to
f at (a, f (a)), that is
f (x) ≥ f (a) + f 0 (a)(x − a)
3.5.1 Basic Inequalities 149

Figure 3.1: Here the tangent line at (1, f (1)) is the best to bound with in the interval [0, 2]

Theorem 3.5.4 (n − 1 Equal Values) — Let a1 , a2 , . . . an be real numbers with a1 + a2 . . . an

fixed. Let f : R → R be a function with exactly one inflection point. If

f (a1 ) + f (a2 ) + · · · + f (an )

achieves a maximal or minimal value, then n − 1 of the ai are equal to each other.

Lemma 3.5.5 ( Power function convexity) — If (xi ), (yi ), (mi ) be three sequences of real numbers,
x, y ∈ R and p > 1. Then for α ∈ (0, 1),

(x + y )p ≤ α1−p x p + (1 − α)1−p y p

X X X
(xi + yi )p mi ≤ α1−p xip mi + (1 − α)1−p yip mi

Equality holds iff

x xi α
= =
y yi 1−α

Proof. Because f (x) = x p is a convex function,

(αa + (1 − α)b)p < αap + (1 − α)bp

3.5.1 Basic Inequalities 150

So setting αa = x and (1 − α)b = y , we get the first inequality. The second one is just an
extention of the first one.

Theorem 3.5.6 (Minkowski) — If x1 , x2 . . . xn , y1 , y2 , . . . yn and m1 , m2 , . . . mn be three sequence

of real numbers and p > 1, then

X 1/p X 1/p X 1/p

(xi + yi )p mi ≤ xip mi + yip mi

Equality holds iff (xi ) and (yi ) are proportional.

Proof. Let us write,

X 1/p X 1/p
A= xip mi , B= yip mi

From the second equation of Lemma 3.5.5, for α ∈ (0, 1) we have

X
(xi + yi )p mi ≤ α1−p Ap + (1 − α)1−p B p

Setting α be such that A/B = α/(1 − α), by the equality case of the first equation of the lemma
we get, X
(xi + yi )p mi ≤ α1−p Ap + (1 − α)1−p B p = (A + B)p

1 1
Theorem 3.5.7 (Young’s Inequality) — If a, b > 0 and p, q > 0 and + = 1, then we have
p q
ap bq
ab ≤ +
p q
Equality occurs when ap = bq .

Proof. Consider the function f (x) = e x , it is convex. So we have

1 1 1 1
x+ q y x y
ep ≤ ep +eq
x y
If we let a = e p , b = e q , we are done. Equality occurs when x = y .

Lemma 3.5.8 —
3.5.2 Theorems 151

Theorem 3.5.9 (Weighted Power Mean) — Let a1 , a2 , . . . an and w1 , w2 , · · · + wn be positive

real numbers with w1 + w2 . . . wn = 1. For real number r , we define,

 w1 a1r + w2 a2r · · · + wn anr 1/r r 6= 0



P (r ) =
 w1 w2
a1 a2 . . . anwn r =0

If r > s, then P (r ) ≥ P (s). Equality occurs iff a1 = a2 = · · · = an

Proof. First we show that, lim P (r ) = P (0). Using L’Hopital’s law,

r →0

wi air ln ai
P

ln wi air wi air
P P
lim ln P (r ) = lim = lim
r →0 r →0 r r →0 1
r
P
wi a ln ai
= lim P i r
r →0 wi ai
X
= ln aiwi
= ln P (0)

Now we have,

3.5.2 Theorems

‘Mean’ Inequalities

Theorem 3.5.10 (Triangle Inequality) — For any complex numbers a1 , a2 . . . an the following
holds:
|a1 + a2 · · · + ak | ≤ |a1 | + |a2 | + · · · + |ak |

§1 One Mean Ineq – QM-AM-GM-HM

Given n positive real numbers x1 , x2 , . . . xn , the following relation holds:
3.5.2 Theorems 152

r
x12 + · · · + xn2 x1 + · · · + xn √ n
≥ ≥ n x1 · · · xn ≥
n n 1 1
+ ··· +
x1 xn

with equality if and only if x1 = x2 = · · · = xn .

Theorem 3.5.11 (Weighted AM-GM) — If a1 , a2 , . . . , an are nonnegative real numbers,

and λ1 , λ2 , . . . , λn are nonnegative real numbers (the "weights") which sum to 1, then

λ1 a1 + λ2 a2 + · · · + λn an ≥ a1λ1 a2λ2 . . . anλn

Equality holds if and only if ai = aj for all integers i , j such that λi 6= 0 and λj 6= 0. We
obtain the unweighted form of AM-GM by setting
1
λ1 = λ2 = · · · = λn =
n

Theorem 3.5.12 (Cauchy-Schward Inequality) — For any real numbers a1 , . . . , an and b1 , . . . , bn ,

a12 + a22 . . . an2 b12 + b22 . . . bn2 ≥ (a1 b1 + a2 b2 . . . an bn )2

with equality when there exist constants µ, λ not both zero such that for all 1 ≤ i ≤ n, µai = λbi .

The inequality sometimes appears in the following form.

Theorem 3.5.13 (Cauchy-Schwarz Inequality Complex form) — Let a1 , . . . , an and b1 , . . . , bn

be complex numbers. Then

|a12 | + |a22 | . . . |an2 | |b12 | + |b22 | . . . |bn2 | ≥ |a1 b1 + a2 b2 . . . an bn |2

Theorem 3.5.14 (Titu’s Lemma) — For positive reals a1 , a2 . . . an and b1 , b2 . . . bn the following
3.5.2 Theorems 153

holds:

a12 a22 a2 (a1 + a2 + · · · + an )2

+ + ··· + n ≥
b1 b2 bn (b1 + b2 + · · · + bn )

Theorem 3.5.15 (Holder’s Inequality) — If a1 , a2 , . . . , an , b1 , b2 , . . . , bn , . . . , z1 , z2 , . . . , zn are

nonnegative real numbers and λa , λb , . . . , λz are nonnegative reals with sum of 1, then

a1λa b1λb . . . z1λz + · · · + anλa bnλb . . . znλz ≤ (a1 + · · · + an )λa (b1 + · · · + bn )λb . . . (z1 + · · · + zn )λz

.
3.5.3 Tricks 154

3.5.3 Tricks

Some tricks to try

1 Replace trigonometric functions by reals, and translate the problem

2 Smoothing, replace two variable while keeping something invariant, to make the inequality
sharper.

3 Convexity, differentiate to check convexity, if the second derivative is positive on some

interval, then the function is convex on that interval except probably at the endpoints, and
concave otherwise.
1−x 1 1
An example is ln . It is convex in (0, ] and concave in [ , 1)
x 2 2

4 If there is product, and if the problem is ‘ad-hoc’y, then apply AM − GM and ln to see if
there is something to play with.

Definition (Homogeneous Expression) — Expression F (a1 , a2 . . . an ) is said to be homogeneous

of degree k if and only if there exists real k such that for every t > 0 we have

t k F (a1 , a2 . . . an ) = F (ta1 , ta2 . . . tan )

If an expression is homogeneous, then the following can be assumed (one at a time):

n
X
ai = 1 (3.1)
i=1
Yn
ai = 1 (3.2)
i=1
a1 = 1 or for some i , ai = 1 (3.3)
X n
ai2 = 1 (3.4)
i=1
X
ai ai+1 = 1 (3.5)
Cy c
3.5.3 Tricks 155

Definition (Substitutions) — • For the condition abc = 1, set

x y z
a= , b= , c=
y z x

•
1 1 1
xy x = x + y + z + 2 =⇒ + + =1
x +1 y +1 z +1
implies the existence of a, b, c such that
b+c a+c b+a
x= , y= , z=
a b c
.

• 2xy z + xy + y z + z x = 1 is just the inverse of the previouss condition.

•
x 2 + y 2 + z 2 = xy z + 4 and |x|, |y |, |z| ≥ 2
implies the existence of a, b, c such that
1 1 1
abc = 1 and x = a + , y = b + , z = c +
a b c
In fact even if only max(|x|, |y |, |z|) > 2 is given, the result still holds.

Lemma 3.5.16 — The following inequality holds for every positive integer n
r
√ √ 1 √ √
2 n+1−2 n < <2 n−2 n−1
n

Lemma 3.5.17 — Given 4 posititve real numbers a < b < c < d. Call the score of a permutation
a1 , a2 , a3 , a4 of the four given reals be equal to the real
a1 a3
| − |
a2 a4
Prove that the minimum the score can get is equal to
a b
| − |
c d
3.5.4 Problems 156

3.5.4 Problems

Problem 3.5.1 (APMO 1991 P3) : Let a1 , a2 , · · · , an , b1 , b2 , · · · , bn be positive real numbers

such that a1 + a2 + · · · + an = b1 + b2 + · · · + bn . Show that

a12 a22 an2 a1 + a2 + · · · + an

+ + ··· + ≥
a1 + b1 a2 + b2 an + bn 2

1 1 1
Problem 3.5.2 (ISL 2009 A2) : Let a, b, c be positive real numbers such that + + = a+b+c.
a b c
Prove that:
1 1 1 3
2
+ 2
+ 2
≤ .
(2a + b + c) (a + 2b + c) (a + b + 2c) 16

Problem 3.5.3 (ARO 2018 P11.2) : Let n ≥ 2 and x1 , x2 , . . . , xn positive real numbers. Prove
that
1 + x12 1 + x22 1 + xn2
+ + ... + ≥n
1 + x1 x2 1 + x2 x3 1 + xn x1

Solution. The inequality says sum is greater, so if the product is greater, then we are done by
AM-GM (??).

Problem 3.5.4 (Turkey TST 2017 P5) : For all positive real numbers a, b, c with a + b + c = 3,
show that
a3 b + b3 c + c 3 a + 9 ≥ 4(ab + bc + ca)

Solution. Always try the most simple ineq possible, AM-GM

Problem 3.5.5 (IMO 2012 P2) : Let n ≥ 3 be an integer, and let a2 , a3 , . . . , an be positive real
numbers such that a2 a3 · · · an = 1. Prove that

(1 + a2 )2 (1 + a3 )3 · · · (1 + an )n > nn .

Solution. The main idea is to look for the ans of the ques, (1 + ak )k >=?. We have
k th power. So if we can get a k term sum inside of the brackets, we can get a clean term
for ? from AM-GM. And 1seems like it’s crying to be partitioned. So we write the term as

1 1
ak + + ··· +
k −1 k −1
3.5.4 Problems 157

Solution. *Looks at the a2 a3 . . . an = 1 condition*

Hey, we have a substitution for this one, why not try it out...
darn it, i still have to do the partition thing to cancel out the powers > (

Problem 3.5.6 (ISL 1998 A1) : Let a1 , a2 , . . . , an be positive real numbers such that a1 + a2 +
· · · + an < 1. Prove that

a1 a2 · · · an [1 − (a1 + a2 + · · · + an )] 1
≤ n+1 .
(a1 + a2 + · · · + an )(1 − a1 )(1 − a2 ) · · · (1 − an ) n

Solution. Simplifying and making it symmetric, we get to the inequality

Y 1 − ai
n ≥ nn+1
ai
i=1

Now approaching similarly as this problem, we get to the solution.

Problem 3.5.7 (ISL 2001 A1) : Let a, b, c be positive real numbers so that abc = 1. Prove that
   
1 1 1
a−1+ b−1+ c −1+ ≤ 1.
b c a

Solution. Substitute.

Problem 3.5.8 (ISL 1999 A1) : Let n ≥ 2 be a fixed integer. Find the least constant C such the
inequality

!4
X X
xi2 xj2


xi xj + ≤C xi
i<j i

n
holds for any x1 , . . . , xn ≥ 0 (the sum on the left consists of summands). For this constant C,


2
characterize the instances of equality.

Solution. Follow the ineq sign and remember AM-GM.

Problem 3.5.9 (ISL 2017 A1) : Let a1 , a2 , . . . an , k, and M be positive integers such that
1 1 1
+ + ··· + =k and a1 a2 · · · an = M
a1 a2 an
3.5.4 Problems 158

If M > 1, prove that the polynomial

P (x) = M(x + 1)k − (x + a1 )(x + a2 ) · · · (x + an )

has no positive roots.

Solution. The same idea used in this and this, spreading an expression to perform AM-GM on
it.

Problem 3.5.10 (ISL 2016 A1) : Let a, b, c be positive real numbers such that min(ab, bc, ca) ≥ 1.
Prove that
a+b+c 2
p  
3 2 2 2
(a + 1)(b + 1)(c + 1) ≤ + 1.
3

Solution. Try the simpler version with two variables first. Now you can use this discovery with
a little bit of cleverness to solve the problem. The clever part is to notice that 4 variable ineq is
more solvable than a 3 variable one.

Problem 3.5.11 (ISL 2016 A2) : Find the smallest constant C > 0 for which the following state-
ment holds: among any five positive real numbers a1 , a2 , a3 , a4 , a5 (not necessarily distinct), one
can always choose distinct subscripts i , j, k, l such that
ai ak
− ≤ C.
aj al

Solution. Simplify the problem to get the ans first. Think about what is the smallest such
value for any given 4 positive reals.

Problem 3.5.12 (ISL 2004 A1) : Let n ≥ 3 be an integer. Let t1 , t2 , ..., tn be positive real
numbers such that
 
2 1 1 1
n + 1 > (t1 + t2 + · · · + tn ) + + ··· + .
t1 t2 tn
Show that ti , tj , tk are side lengths of a triangle for all i , j, k with 1 ≤ i < j < k ≤ n.

Solution. Easy solution by induction. For a more elegant solution, write the right side as sum
of paired factors. Finding when the inequality breaks and relating it to the end statement.

Problem 3.5.13 (ISL 1996 A2) : Let a1 ≥ a2 ≥ . . . ≥ an be real numbers such that for all integers
k > 0,
3.5.4 Problems 159

a1k + a2k + . . . + ank ≥ 0.

Let p = max{|a1 |, . . . , |an |}. Prove that p = a1 and that

(x − a1 ) · (x − a2 ) · · · (x − an ) ≤ x n − a1n
for all x > a1 .

Solution. After thePfirst part, apply AM-GM on the whole left side, this not gonna work,
since we can’t bound ai wrt a1 . So what if we divide both side by (x − a1 ) and then apply
AM-GM?
3.5.4 Problems 160

3.5.4.1 Smoothing And Convexity

Some usual tricks

1. Bring x, y closer, keeping x + y constant.

2. If we need to smoothen up the value xy , then take ln on both side.

3. Work with different variables.

Theorem 3.5.18 (Convexity) — 1. The function is convex in interval I iff for all a, b ∈ I and
for all t < 1,
tf (a) + (1 − t)f (b) ≥ f (ta + (1 − t)b)
Which if put in words, means that the line segment joining (a, f (a)) and (b, f (b)) lies
completely above the graph of the function.

2. The function is convex in interval I if f 0 is increasing in I or f 00 is positive in I.

Theorem 3.5.19 (Jensen’s Inequality) — Let x1 , . . . , xn ∈ R and let α1 , . . . , αn ≥ 0 satisfy

α1 + · · · + αn = 1.

If f is a Convex Function, we have:

α1 f (x1 ) + α2 f (x2 ) · · · + an f (xn ) ≥ f (α1 x1 + α2 x2 · · · + αn xn )

If f is a Concave Function, we have:

α1 f (x1 ) + α2 f (x2 ) · · · + an f (xn ) ≤ f (α1 x1 + α2 x2 · · · + αn xn )

Theorem 3.5.20 (Popoviciu’s inequality) — Let f be a convex function on and interval I ∈ R.

Then for any numbers x, y , z ∈ I,
       
x +y +z x +y y +z z +x
f (x) + f (y ) + f (z) + 3f ≥ 2f + 2f + 2f
3 2 2 2

Problem 3.5.14 (USAMO 1998 P3) : Let a0 , a1 , · · · , an be numbers from the interval (0, π/2)
3.5.4 Problems 161

such that
π π π
tan(a0 − ) + tan(a1 − ) + · · · + tan(an − ) ≥ n − 1.
4 4 4
Prove that
tan a0 tan a1 · · · tan an ≥ nn+1 .

Solution. Get rid of the tan’s. AM-GM, Jensen doesn’t works, so try smoothing. The
conventional smoothing trick fails at one case, but works for all other cases. Means we have to
deal with that case specially.

Problem 3.5.15 (USAMO 1974 P2) : Prove that if a, b, and c are positive real numbers, then

aa bb c c ≥ (abc)(a+b+c)/3 .

Solution. ln

Problem 3.5.16 (India 1995) : Let x1 , x2 , ..., xn > 0 be real numbers such that x1 +x2 +x3 +...+xn =
1. Prove the inequality
r
x1 x2 xn n
√ +√ + .... + √ ≥
1 − x1 1 − x2 1 − xn n−1

Solution. easy smoothing

Problem 3.5.17 (Vietnam 1998) : x1 , x2 ...xn are real numbers such that
1 1 1
+ ··· + =
x1 + 1998 xn + 1998 1998
Prove that √
n x ...x
1 n
≥ 1998
n−1

Solution. Translate the given expression in a nicer way with new variables...

Problem 3.5.18 (IMO 1974 P2) : The variables a, b, c, d, traverse, independently from each other,
the set of positive real values. What are the values which the expression
a b c d
S= + + +
a+b+d a+b+c b+c +d a+c +d
takes?
3.5.4 Problems 162

x x x
Solution. ≤ ≤
y +c y y −c

Problem 3.5.19 (Bulgaria 1995) : Given n real number x1 , x2 , ..., xn ∈ [0, 1] . Prove the following
inequality jnk
(x1 + x2 + · · · + xn ) − (x1 x2 + x2 x3 + · · · + xn x1 ) ≤
2
3.6 Calculus 163

3.6 Calculus

Lemma 3.6.1 —

(f (x) + g(x))0 = f 0 (x) + g 0 (x)

(f (x)g(x)) = f 0 (x)g(x) + f (x)g 0 (x)

f (x) 0 f 0 (x)g(x) − f (x)g 0 (x)

( ) =
g(x) g(x)2

(f (g(x)))0 = f 0 (g(x)) + g 0 (x)

Lemma 3.6.2 ( The Derivative of an Odd function is always an Even function, and vice versa.)
—
3.7 Ad-Hocs 164

3.7 Ad-Hocs

Problem 3.7.1 (ISL 2014 A2) : Define the function f : (0, 1) → (0, 1) by
1 1 1
f (x) = x 2 , for x ≥ and x + , for x <
2 2 2

Let a and b be two real numbers such that 0 < a < b < 1. We define the sequences an and bn
by a0 = a, b0 = b, and an = f (an−1 ), bn = f (bn−1 ) for n > 0. Show that there exists a positive
integer n such that
(an − an−1 )(bn − bn−1 ) < 0
.

Problem 3.7.2 (ARO 2018 P10.1) : Determine the number of real roots of the equation

|x| + |x + 1| + · · · + |x + 2018| = x 2 + 2018x − 2019

Problem 3.7.3 (European Mathematics Cup 2018 P3) : Find all k > 1 such that there exists a
set S such that,

1. There exists N > 0 such that, if x ∈ S, thenx < N.

2. If a, b ∈ S, and a > b, then k(a − b) ∈ S

Solution. Find some constraints such as, k(a − b) 6> a, S has a smallest element. These two
combined with a sequence of decreasing elements of S is enough to solve this problem.

Problem 3.7.4 (APMO 2018 P2) : Let f (x) and g(x) be given by
1 1 1 1
f (x) = + + + ··· +
x x −2 x −4 x − 2018
1 1 1 1
g(x) = + + + ··· +
x −1 x −3 x −5 x − 2017
Prove that |f (x) − g(x)| > 2 for any non-integer real number x satisfying 0 < x < 2018.

Solution. Substract, manipulate, see that for  < 1, for x = 2k +  it’s true, if it’s true for
x = 2k + 1 + . Then for x = 2k + 1 + , sustitute the value to find a common term appearing
in all of those equations. So if that term were to be greater than 2, we would be done. How do
we test that? Take the first derivative to find the minima.
3.7 Ad-Hocs 165

Problem 3.7.5 (ISL 2015 A1) : Suppose that a sequence a1 , a2 , . . . of positive real numbers satisfies

kak
ak+1 ≥
ak2 + (k − 1)

for every positive integer k. Prove that a1 + a2 + . . . + an ≥ n for every n ≥ 2.

Solution. Simplify the inequality. And then sum it up.

Problem 3.7.6 (ISL 2015 A3) : Let n be a fixed positive integer. Find the maximum possible value
of X
(s − r − n)xr xs ,
1≤r <s≤2n

where −1 ≤ xi ≤ 1 for all i = 1, · · · , 2n.

Solution. The expression is weird, and beautiful. Now if we write the expression as a single
variable function, we see that xi ∈ 1, −1. Now, there is xi xj in the expression. So we need to
multiply two expressions. Again, see that s − r − n can be rewritten as −(n − s) − (r ). Now,
how do we get an expression like xi xj which can be found in squares, with a coefficient (n − s)
and r ? By summing it up r times, simple.

Problem 3.7.7 (ISL 2010 A3) : Let x1 , . . . , x100 be nonnegative real numbers such that xi + xi+1 +
xi+2 ≤ 1 for all
P i = 1, . . . , 100 (we put x101 = x1 , x102 = x2 ). Find the maximal possible value of
the sum S = 100 i=1 xi xi+2 .

Solution. Bound a small portion of the large sum.

Problem 3.7.8 (ISL 2005 A3) : Four real numbers p, q, r , s satisfy p + q + r + s = 9 and
p 2 + q 2 + r 2 + s 2 = 21. Prove that there exists a permutation (a, b, c, d) of (p, q, r, s) such that
ab − cd ≥ 2.

Solution. Put p, q, r, s in order, find which permutation must satisfy the condition. Since we
know, sy m pq = 30, what can we say about the largest sum? How do we get pq − r s with the
P

equations given to us? What can we do to make the conditions met?

3.7.1 Factorization 166

3.7.1 Factorization

Problem 3.7.9 (USAMO 2013 P4) : Find all real numbers x, y , z ≥ 1 satisfying
√ √ √ √ p √
min( x + xy z , y + xy z, z + xy z) = x − 1 + y − 1 + z − 1.

Solution. Replacement is never a bad idea to try out. But the main part is not replacement,
but it’s factorization. I don’t yet know how to find such factorization, but let’s find out.
3.7.2 Bounding 167

3.7.2 Bounding

Problem 3.7.10 (ISL 2004 A2) : Let a0 , a1 , a2 , ... be an infinite sequence of real numbers
satisfying the equation
an = |an+1 − an+2 |
for all n ≥ 0, where a0 and a1 are two different positive reals. Can this sequence a0 , a1 , a2 . . . be
bounded?

Solution. In bounding problems, name the bounds, then focus on them. Another thing: In
reals, a variable does not necessarily need to be equal to the bound.
3.7.3 Manipulation 168

3.7.3 Manipulation

Problem 3.7.11 (ISL 2011 A2) : Determine all sequences (x1 , x2 , . . . , x2011 ) of positive integers,
such that for every positive integer n there exists an integer a with
2011
X
jxjn = an+1 + 1
j=1

Solution. Manipulate the data.

Since for all n the statement holds, we can guess there is bounding involved. Can we bound xi
or a? Tweak the terms and see if there is something nice to work with.

Problem 3.7.12 (ISL 2014 A1) : Let a0 < a1 < a2 . . . be an infinite sequence of positive integers.
Prove that there exists a unique integer n ≥ 1 such that
a0 + a1 + a2 + · · · + an
an < ≤ an+1 .
n

Solution. Manipulate the data.

Either directly, or using the “∆ method"

Problem 3.7.13 (Putnam 2011 A2) : Let a1 , a2 , . . . and b1 , b2 , . . . be sequences of positive real
numbers such that a1 = b1 = 1 and bn = bn−1 an − 2 for n = 2, 3, . . . . Assume that the sequence
(bj ) is bounded. Prove that
∞
X 1
S=
a1 · · · an
n=1

converges, and evaluate S

Solution. Look for partial sum. And in limit problems on contests, it is always a good idea to
think about n = l − Sn

Problem 3.7.14 (Putnam 2013 A3) : Suppose that the real numbers a0 , a1 , . . . , an and x, with
0 < x < 1, satisfy
a0 a1 an
+ 2
+ ··· + = 0.
1−x 1−x 1 − x n+1
Prove that there exists a real number y with 0 < y < 1 such that

a0 + a1 y + · · · + an y n = 0.
3.7.3 Manipulation 169

Solution. How do you show ∃ root ∈ I if you don’t want to construct it? Also those geometric
sums are begging to be expanded...

Problem 3.7.15 (GQMO 2020 P3) : We call a set of integers special if it has 4 elements and can
be partitioned into 2 disjoint subsets {a, b} and {c, d} such that ab − cd = 1. For every positive
integer n, prove that the set {1, 2, . . . , 4n} cannot be partitioned into n disjoint special sets.

Solution [Multiply ’em All]. Each special set must have exactly two evens and two odds. Now,
consider the products of all even numbers and all odd numbers. Clearly the product of the odd
parts of each set will be much smaller than the product of the even parts.

Problem 3.7.16 (Korean Summer Program TST 2016 1) : Find all real numbers x1 , . . . , x2016 that
satisfy the following equation for each 1 ≤ i ≤ 2016. (Here x2017 = x1 .)

xi2 + xi − 1 = xi+1
3.8 Tricks and Lemmas 170

3.8 Tricks and Lemmas

Theorem 3.8.1 (Minkowski’s theorem) — Any convex set in Rn , which is symmetric with respect
to the origin and with volume greater than 2n d(L) contains a non-zero lattice point.

3.8.1 Ad Hocs

1. x 2 + 1 = (x + i )(x − i ) [USAMO 2014 P1]

2. Add. Everything. Up.

3. Send SHIT to the infinity.

Chapter 4

Geometry
4.1 First Portion 172

4.1 First Portion

Lemma 4.1.1 — Let the incircle and excircle (opposite to A) of 4ABC meet BC at D and E
resp. Suppose F is the antipode of D wrt the incircle.

1. Prove that A, F, E are collinear.

2. M be the midpoint of DE. Prove that MI meets AD at it’s midpoint.

Figure 4.1

Lemma 4.1.2 — Let the incircle of 4ABC meets AB and AC at X and Y resp. BI and CI meet
XY at P and Q respectively. Prove that BP QC is cyclic. (In fact BP ⊥ CP and BQ ⊥ CQ)
4.1 First Portion 173

Figure 4.2

Lemma 4.1.3 — AD is an altitude of 4ABC. E, F are on AC, AB so that AD, BE, CF are
concurrent. Prove ∠EDA = ∠F DA.

Lemma 4.1.4 — Let AD be an altitude of 4ABC and E ∈ ABC so that AE k BC. Prove
J

that D, G, E are collinear where G is the centroid of 4ABC.

4.1 First Portion 174

Figure 4.3

Problem 4.1.1 () : Let O be the circumcenter of 4ABC and A0 , B 0 , C 0 are reflections of O on

BC, CA, AB resp. Prove that AA0 , BB 0 , CC 0 are concurrent.

Problem J () : Let D, E are on sides AC, AB of 4ABC resp. such that BE = CD. Let
4.1.2
ABC ∩ ADE = P . Prove that P B = P C.
J

Problem 4.1.3 () : Let a line P Q touch circle S1 and S2 at P and Q resp. Prove that the radical
axis of S1 and S2 passes through the midpoint of P Q.

Problem 4.1.4 () : Let ω1 , ω2 , omega3 are 3 circles. Prove that the 3 radical axis of ω1 and ω2 ,ω2
and ω3 ,ω3 and ω1 are either concurrent or parallel.

Problem 4.1.5 () : Two equal-radius circles ω1 and ω2 are centered at points O1 and O2 . A point
X is reflected through O1 and O2 to get points A1 and A2 . The tangents from A1 to ω1 touch ω1
at points P1 and Q1 , and the tangents from A2 to ω2 touch ω2 at points P2 and Q2 . If P1 Q1 and
P2 Q2 intersect at Y , prove that Y is equidistant from A1 and A2 .

Problem 4.1.6 () : Let BD, CE be the altitudes of 4ABC and M be the midpoint of BC. If the
4.1 First Portion 175

ray MH meet ABC at point K, prove that AK, BC, DE are concurrent.
J

Problem 4.1.7 () : Two circle ω and Γ touches one another internally at P with ω inside of Γ. Let
AB be a chord of Γ which touches ω at D. Let P D ∩ Γ = Q. Prove that QA = QB.

Problem 4.1.8 () : Let AD be a symmedian of 4ABC with D on ABC. Let M be the midpoint
J

of AD. Prove that ∠BMD = ∠CMD and A, M, O, D are cyclic where O is the circumcenter of
4ABC.

Problem 4.1.9 () : Let A, B be two fixed points and let P be varying point such that PA
PB is
constant. Prove that the locus of P is a circle.

Problem 4.1.10 () : Prove that r1 + r2 + r3 = 4R + r (R, r, r1 , r2 , r3 are the circumradius, inradius

and three exradiuses respectively of a triangle)

Problem 4.1.11 () : Let M be the midpoint of the altitude BE in 4ABC and suppose that the
excircle opposite to B touches AC at Y . Then MY goes through the incenter I.

Problem 4.1.12 () : Let ABC be a triangle, and draw isosceles triangles 4DBC, 4AEC, 4ABF
external to 4ABC (with BC; CA; AB as their respective bases). Prove that the lines through
A; B; C perpendicular to EF ; F D; DE, respectively, are concurrent.

Problem 4.1.13 () : In a triangle ABC we have AB = AC. A circle which is internally tangent
with the circumscribed circle of the triangle is also tangent to the sides AB; AC in the points P ,
respectively Q. Prove that the midpoint of P Q is the center of the inscribed circle of the triangle
ABC

Problem 4.1.14 () : Nagel Point N: If the Excircles of ABC touch BC; CA; AB at D; E; F , then
the intersection point of AD; BE; CF is called the Nagel Point N. Prove that

1. I; G; N are collinear. (G centroid, I incenter.)

2. GN = 2 · IG.
3. Speiker center S: The incircle of the medial triangle is called the Speiker circle, and it’s
center is Speiker center S. Prove that S is the midpoint of IN.
4.2 Second Portion 176

4.2 Second Portion

Problem 4.2.1 () : Let P B and P C are tangent to ABC. Let D, E, F are projection of A on
J

BC, P B, P C resp. Prove that AD2 = AE × AF .

Problem 4.2.2 () : Let D and E are

Jon AB and
J AC s.t DE k BC. P is an arbitrary point inside
4ADE. P B, P C ∩ DE = F, G. Let P DG ∩ P F E = Q. Prove that A, P, Q are collinear.

Problem 4.2.3 () : Let AB and CD be chords in a circle of center O with A, B, C, D distinct , and
with the lines AB and CD meeting at a right angle at point E. Let also M and N be the midpoints
of AC and BD respectively . If MN⊥OE , prove that AD k BC

Problem 4.2.4 () : Circles C1 and C2 intersect at A and B. Let M ∈ AB. A line through
M (different from AB) cuts circles C1 and C2 at Z, D, E, C respectively such that D, E ∈ ZC.
Perpendiculars at B to the lines EB, ZB and AD respectively cut circle C2 in F, K and N. Prove
that KF = NC.

Problem 4.2.5 () : Let D be a point on side AC of triangle ABC. Let E and F be points on the
segments BD and BC respectively, such that ∠BAE = ∠CAF . Let P and Q be points on BC and
BD respectively, such that EP and F Q are both parallel to CD. Prove that ∠BAP = ∠CAQ.

Problem 4.2.6 () : In the non-isosceles triangle ABC an altitude from A meets side BC in D .
Let M be the midpoint of BC and let N be the reflection of M in D. The circumcirle of triangle
AMN intersects the side AB in P 6= A and the side AC in Q 6= A . Prove that AN, BQ and CP
are concurrent.

Problem 4.2.7 () : In triangle ABC, the interior and exterior angle bisectors of ∠BAC intersect
the line BC in D and E, respectively. Let F be the second point of intersection of the line AD with
the circumcircle of the triangle ABC. Let O be the circumcenter of the triangle ABCand let D0 be
the reflection of D in O. Prove that ∠D0 F E = 90.

Problem 4.2.8 () : Let ABCD be a convex quadrilateral such that the line BD bisects the angle
ABC. The circumcircle of triangle ABC intersects the sides AD and CD in the points P and Q,
respectively. The line through D and parallel to AC intersects the lines BC and BA at the points
R and S, respectively. Prove that the points P, Q, R and S lie on a common circle.
4.2 Second Portion 177

Problem 4.2.9 () : The incircle of triangle ABC touches BC, CA, AB at points A1 , B1 , C1 ,
respectively. The perpendicular from the incenter I to the median from vertex C meets the line
A1 B1 in point K. Prove that CK is parallel to AB.

Problem 4.2.10 () : Let X be an arbitrary point inside the circumcircle of a triangle ABC. The
lines BX and CX meet the circumcircle in points K and L respectively. The line LK intersects BA
and AC at points E and F respectively. Find the locus of points X such that the circumcircles of
triangles AF K and AEL touch.

Problem 4.2.11 () : Let BD be a bisector of triangle ABC. Points Ia , Ic are the incenters of
triangles ABD, CBD respectively. The line Ia Ic meets AC in point Q. Prove that ∠DBQ = 90◦ .

Problem 4.2.12 () : Given right-angled triangle ABC with hypotenuse AB. Let M be the midpoint
of AB and O be the center of circumcircle ω of triangle CMB. Line AC meets ω for the second
time in point K. Segment KO meets the circumcircle of triangle ABC in point L. Prove that
segments AL and KM meet on the circumcircle of triangle ACM.

Problem 4.2.13 () : Let BN be median of triangle ABC. M is a point on BC. S lies on BN such
that MS k AB. P is a point such that SP ⊥ AC and BP k AC. MP cuts AB at Q. Prove that
QB = QP .

Problem 4.2.14 () : Let ABCD be a convex quadrilateral with AB parallel to CD. Let P and Q be
the midpoints of AC and BD, respectively. Prove that if ∠ABP = ∠CBD, then ∠BCQ = ∠ACD.

Problem 4.2.15 () : Point P lies inside a triangle ABC. Let D, E and F be reflections of the point
P in the lines BC, CA and AB, respectively. Prove that if the triangle DEF is equilateral, then the
lines AD, BE and CF intersect in a common point.

Problem 4.2.16 () : Let 4ABC be an acute angled triangle. The circle with diameter AB intersects
the sides AC and BC at points E and F respectively. The tangents drawn to the circle through E
and F intersect at P . Show that P lies on the altitude through the vertex C.

Problem 4.2.17 () : Let γ be circle and let P be a point outside γ. Let P A and P B be the
tangents from P to γ (where A, B ∈ γ). A line passing through P intersects γ at points Q and R.
Let S be a point on γ such that BS k QR. Prove that SA bisects QR
4.2 Second Portion 178

Problem 4.2.18 () : Given is a convex quadrilateral ABCD with AB = CD. Draw the triangles
ABE and CDF outside ABCD so that ∠ABE = ∠DCF and ∠BAE = ∠F DC. Prove that the
midpoints of AD, BC and EF are collinear

Problem 4.2.19 () : Let P be a point out of circle C. Let P A and P B be the tangents to the
circle drawn from C. Choose a point K on AB . Suppose that the circumcircle of triangle P BK
intersects C again at T . Let P 0 be the reflection of P with respect to A. Prove that

∠P BT = ∠P 0 KA

Problem 4.2.20 () : Consider a circle C1 and a point O on it. Circle C2 with center O, intersects
C1 in two points P and Q. C3 is a circle which is externally tangent to C2 at R and internally tangent
to C1 at S and suppose that RS passes through Q. Suppose X and Y are second intersection points
of P R and OR with C1 . Prove that QX is parallel with SY .

Problem 4.2.21 () : In triangle ABC we have ∠A = π3 . Construct E and F on continue of AB

and AC respectively such that BE = CF = BC. Suppose that EF meets circumcircle of 4ACE in
K. (K 6≡ E). Prove that K is on the bisector of ∠A

Problem 4.2.22 () : In triangle ABC, ∠A = 90◦ and M is the midpoint of BC. Point D is chosen
on segment AC such that AM = AD and P is the second meet point of the circumcircles of triangles
∆AMC, ∆BDC. Prove that the line CP bisects ∠ACB

Problem 4.2.23 () : Let C1 , C2 be two circles such that the center of C1 is on the circumference
of C2 . Let C1 , C2 intersect each other at points M, N. Let A, B be two points on the circumference
of C1 such that AB is the diameter of it. Let lines AM, BN meet C2 for the second time at A0 , B 0 ,
respectively. Prove that A0 B 0 = r1 where r1 is the radius of C1 .

Problem 4.2.24 () : Given a triangle ABC, let P lie on the circumcircle of the triangle and be the
midpoint of the arc BC which does not contain A. Draw a straight line l through P so that l is
parallel to AB. Denote by k the circle which passes through B, and is tangent to l at the point
P . Let Q be the second point of intersection of k and the line AB (if there is no second point of
intersection, choose Q = B). Prove that AQ = AC.

Problem 4.2.25 () : Let ABCD be a cyclic quadrilateral in which internal angle bisectors ∠ABC
and ∠ADC intersect on diagonal AC. Let M be the midpoint of AC. Line parallel to BC which
passes through D cuts BM at E and circle ABCD in F (F 6= D ). Prove that BCEF is parallelogram
4.2 Second Portion 179

Problem 4.2.26 () : The side BC of the triangle ABC is extended beyond C to D so that CD =
BC. The side CA is extended beyond A to E so that AE = 2CA. Prove that, if AD = BE, then
the triangle ABC is right-angled

Problem 4.2.27 () : ABCD is a cyclic quadrilateral inscribed in the circle Γ with AB as diameter.
Let E be the intersection of the diagonals AC and BD. The tangents to Γ at the points C, D meet
at P . Prove that P C = P E

Problem 4.2.28 () : The quadrilateral ABCD is inscribed in a circle. The point P lies in the
interior of ABCD, and ∠P AB = ∠P BC = ∠P CD = ∠P DA. The lines AD and BC meet at Q,
and the lines AB and CD meet at R. Prove that the lines P Q and P R form the same angle as the
diagonals of ABCD

Problem 4.2.29 () : Let ABCD be a cyclic quadrilateral with opposite sides not parallel. Let X
and Y be the intersections of AB, CD and AD, BC respectively. Let the angle bisector of ∠AXD
intersect AD, BC at E, F respectively, and let the angle bisectors of ∠AY B intersect AB, CD at
G, H respectively. Prove that EF GH is a parallelogram.

Problem 4.2.30 () : Triangle ABC is given with its centroid G and cicumcentre O is such that
GO is perpendicular to AG. Let A0 be the second intersection of AG with circumcircle of triangle
ABC. Let D be the intersection of lines CA0 and AB and E the intersection of lines BA0 and AC.
Prove that the circumcentre of triangle ADE is on the circumcircle of triangle ABC

Problem 4.2.31 () : Let M be the midpoint of the side AC of 4ABC. Let P ∈ AM and Q ∈ CM
2 . Let (ABQ) intersect with BC at X 6= B and (BCP ) intersect with BA at
be such that P Q = AC
Y 6= B. Prove that the quadrilateral BXMY is cyclic.

Problem 4.2.32 () : Let be given a triangle ABC and its internal angle bisector BD (D ∈ BC).
The line BD intersects the circumcircle Ω of triangle ABC at B and E. Circle ω with diameter DE
cuts Ω again at F . Prove that BF is the symmedian line of triangle ABC.

Problem 4.2.33 () : ∆ABC is a triangle such that AB 6= AC. The incircle of ∆ABC touches
BC, CA, AB at D, E, F respectively. H is a point on the segment EF such that DH⊥EF . Suppose
AH⊥BC, prove that H is the orthocenter of ∆ABC.
4.2 Second Portion 180

Problem 4.2.34 () : Let ABC be a triangle and let P be a point on the angle bisector AD, with
D on BC. Let E, F and G be the intersections of AP, BP and CP with the circumcircle of the
triangle, respectively. Let H be the intersection of EF and AC, and let I be the ntersection of EG
and AB. Determine the geometric place of the intersection of BH and CI when P varies

Problem 4.2.35 () : Let D; E; F be the points on the sides BC; CA; AB respectively, of 4ABC.
Let P ; Q; R be the second intersection of AD; BE; CF respectively, with the cricumcircle of 4ABC.
Show that
AD BE CF
+ + ≥9
P D QE RF

Problem 4.2.36 () : Points D and E lie on sides AB and AC of triangle ABC such that DE k BC.
Let P be an arbitrary point inside ABC. The lines P B and P C intersect DE at F and G, respectively.
If O1 is the circumcenter of P DG and O2 is the circumcenter of P F E, show that AP k O1 O2 .

Problem 4.2.37 () : Let ABC be a triangle. A circle passing through A and B intersects segments
AC and BC at D and E, respectively. Lines AB and DE intersect at F , while lines BD and CF
intersect at M. Prove that MF = MC if and only if MB · MD = MC 2

Problem 4.2.38 () : Let O and I be the circumcenter and incenter of triangle ABC, respectively.
Let ωA be the excircle of triangle ABC opposite to A; let it be tangent to AB, AC, BC at K, M, N,
respectively. Assume that the midpoint of segment KM lies on the circumcircle of triangle ABC.
Prove that O; N; I are collinear.

Problem 4.2.39 () : Let ABCD be a cyclic quadrilateral. Let AB ∩ CD = P and AD ∩ BC = Q.

Let the tangents from Q meet the circumcircle of ABCD at E and F . Prove that P ; E; F are
collinear.
4.3 orthocenter–circumcircle–ninepoint circle 181

4.3 orthocenter–circumcircle–ninepoint circle

• circles - yufei zhao

• big picture - yufei zhao

• pop - yufei zhao

• 3 lemmas - yufei zhao

4.3.1 some figures, might be good for something, i dunno

Figure 4.4: h lies on the line, circles vary

4.3.2 problems 182

4.3.2 problems

Lemma 4.3.1 ( collinearity with antipode and center) — let a0

be the antipod of a in abc. let bdec be a cyclic quadrilateral
with d ∈ ab and e ∈ ac. let p be the center of bdec. also, let
x = be ∩ cd. then a0 , p, x are collinear.

Solution. using “the big picture" property to show that if

q = ade ∩ abc, then p, x, q collinear and pq ⊥ aq. which
implies that p, a0 , q are collinear.

Figure 4.5

Problem 4.3.1 (balkan mo 2017 p3) : consider an acute-

angled triangle abc with ab < ac and let ω be its circum-
scribed circle. let tb and tc be the tangents to the circle ω at
points b and c, respectively, and let l be their intersection.
the straight line passing through the point b and parallel to
ac intersects tc in point d. the straight line passing through
the point c and parallel to ab intersects tb in point e. the
circumcircle of the triangle bdc intersects ac in t, where t
is located between a and c. the circumcircle of the triangle
bec intersects the line ab (or its extension) in s, where b
is located between s and a. prove that st, al, and bc are
concurrent.

Solution. you could’ve thought like: symmedian is

there & and 4abc → 4acs & concurrent =⇒ par-
allel lines and median concurrency?
just need to prove bt k cs.
Figure 4.6
4.3.2 problems 183

Problem 4.3.2 (usamo 2014 p5) : let abc be a

triangle with orthocenter h and let p be the second
intersection of ahc with the internal bisector of
∠bac. let x be the circumcenter of triangle apb
and y the orthocenter of triangle apc. prove that
the length of segment xy is equal to the circum-
radius of triangle abc.

Solution. no length conditions given, yet we

need to prove that two lengths are equal. par-
allelogram !!!
just need to prove that y ∈ abc & y d ⊥ ab

Figure 4.7

Problem 4.3.3 (bewarish 1) : let def be the or-

thic triangle, and let ef ∩ bc = p. let the tangent
at a to abc meet bc at q. let t be the reflection
of q over p. let k be the orthogonal projection of
h on am. prove that ∠okt = 90.

Solution. spiral similarity to get rid of q and

t. then spiral similarity again to find a trivial
circle.

Figure 4.8
4.3.2 problems 184

Problem 4.3.4 (buratinogigle’s proposed prob-

lems for arab saudi team 2015) : let abc be a
triangle with orthocenter h. p is a point. (k) is
the circle with diameter ap. (k) cuts ca, ab again
at e, f . ph cuts (k) again at g. tangent line at
e, f of (k) intersect at t. m is midpoint of bc. l
is the point on mg such that al k mt. prove that
la ⊥ lh.

Solution [phantom point]. take l 0 = mg ∩

azy h, then use spiral similarity to show that
al 0 k mt.

Figure 4.9

Problem 4.3.5 (buratinogigle’s proposed prob-

lems for arab saudi team 2015) : let abc be a
triangle inscribed circle (o). p lies on (o). the
line passes through p and parallel to bc cuts ca at
e. k is circumcenter of triangle pce and l is nine
point center of triangle pbc. prove that the line
passes through l and parallel to pk, always passes
through a fixed point when p moves.

Solution [construction]. notice that if we re-

flect p over l to get p 0 , then op = ah and
op ⊥ bc where o is the circumcenter of abc.
which trivially implies that the line throught l
passes throught the midpoint of p 0 d where d is
the reflection of h over bc.
Figure 4.10
4.3.2 problems 185

Problem 4.3.6 (buratinogigle’s proposed prob-

lems for arab saudi team 2015) : let abc be acute
triangle inscribed circle (o), altitude ah, h lies on
bc. p is a point that lies on bisector ∠bac and p
is inside triangle abc. circle diameter ap cuts (o)
again at g. l is projection of p on ah. assume that
gl bisects hp. prove that p is incenter of abc.

Solution [angle chase]. since ∠apl =

∠abd = ∠agd, g, l, m are collinear. let e ∈ bc
and pe ⊥ bc. then e also lies on dg.
again we have, 4dpe ∼ 4dgp. which implies
dp = db = dc.

Figure 4.11

Problem 4.3.7 (buratinogigle’s proposed prob-

lems for arab saudi team 2015) : let abc be an
acute triangle inscribed circle (o). m lies on small
arc bc . p lies on am. circle diameter mp cuts (o)
again at n. mo cuts circle diameter mp again at
q. an cuts circle diameter mp again at r . prove
that ∠pr a = ∠pqa.

Solution [Angle Chase]. Let MO ∩ ABC =

D. Becase NP ⊥ MN, we have N, P, D
collinear, and AP QD cyclic.
So, 4AP Q ∼ 4ANM ∼ 4AP R.

Figure 4.12

Problem 4.3.8 (buratinogigle’s proposed problems for Arab Saudi team 2015) : Let ABC be right
triangle with hypotenuse BC, bisector BE, E lies on CA. Assume that circumcircle of triangle
BCE cuts segment AB againq at F . K is projection of A on BC. L lies on segment AB such that
BL = BK. Prove that AL
AF = BC .
BK
4.3.2 problems 186

Figure 4.13

Problem 4.3.9 (buratinogigle’s proposed problems for Arab Saudi team 2015) : Let ABC be acute
triangle inscribed circle (O). AD is diameter of (O). M, N lie on BC such that OM k AB, ON k AC.
DM, DN cut (O) again at P, Q. Prove that BC = DP = DQ.

Figure 4.14

Problem 4.3.10 () : Let 4ABC be a triangle. F, G be arbitrary points on AB, AC. Take D, E
midpoint of BF, CG. Show that the center of nine-point circle of 4ABC, 4ADE, 4AF G are
collinear.

Problem 4.3.11 (IGO 2017 Advance P3) : Let O be the circumcenter of 4ABC. Line CO
intersects the altitude through A at point K. Let P, M be the midpoints of AK, AC respectively.
4.3.2 problems 187

If P O intersects BC at Y , and the circumcircle of 4BCM meets AB at X, prove that BXOY is

cyclic

Solution. There is no easily measurable angles, in this case use projective geometry. And since
we still don’t have any easy angles, we look for the second way of concyclicity, POP

Problem 4.3.12 (Turkey TST 2018 P4) : In a non-isosceles acute triangle ABC, D is the midpoint
of BC. The points E and F lie on AC and AB, respectively, and the circumcircles of CDE and
AEF intersect in P on AD. The angle bisector from P in triangle EF P intersects EF in Q. Prove
that the tangent line to the circumcircle of AQP at A is perpendicular to BC.

Solution. Inverting around A.

Problem 4.3.13 (USA Winter TST 2020 P2) : Two circles Γ1 and Γ2 have common external
tangents `1 and `2 meeting at T . Suppose `1 touches Γ1 at A and `2 touches Γ2 at B. A circle
Ω through A and B intersects Γ1 again at C and Γ2 again at D, such that quadrilateral ABCD is
convex.
Suppose lines AC and BD meet at point X, while lines AD and BC meet at point Y . Show that
T , X, Y are collinear.

Solution [Radical Axis]. It is easy to see that X lies on the radical axis of Γ1 and Γ2 . Let
B 0 = l1 ∩ Γ2 and A0 = l2 ∩ Γ1 . Let C 0 = A0 X ∩ Γ1 and D0 = B 0 X ∩ Γ2 . Let A0 C ∩ AC 0 = Z.

We have AD0 CB 0 and A0 DC 0 B cyclic. Also T, D, C 0 and T, D0 , C are collinear. Which implies
A0 D0 CB and ADC 0 B 0 are cyclic too.

Applying pascal on AAC 0 CA0 A0 , we have T, X, Z are collinear.

Now, it is easy to see that Z, Y, T lie on the radical axis of A0 D0 CB and ADC 0 B 0 . So we have
T, X, Y, Z collinear.
4.3.2 problems 188

Figure 4.15

Solution [mOvInG pOiNtS, by shawnee03]. Fix Γ1 and Γ2 (and hence `, T, A, B) and animate
X linearly on `. Then
• C moves projectively on Γ1 (it is the image of the perspectivity through A from ` to Γ1 )
and thus has degree 2, and similarly for D.
• AD has degree at most 0 + 2 = 2, and similarly for BC.
• Y = AD ∩ BC has degree at most 2 + 2 = 4.
• The collinearity of T, X, Y has degree at most 0 + 1 + 4 = 5.
Thus it suffices to verify the problem for six different choices of X. We choose:
• ` ∩ `1 : here Y approaches A as X approaches ` ∩ `1 .
• ` ∩ `2 : here Y approaches B as X approaches ` ∩ `2 .
• ` ∩ AB: here Y approaches ` ∩ AB as X approaches ` ∩ AB.
• the point at infinity along `: here Y = T .
• the two intersections of Γ1 and Γ2 : here Y = X.
(The final two cases may be chosen because we know that there exists a choice of A, B, C, D for
which ABCD is convex; this forces Γ1 and Γ2 to intersect.)
4.3.2 problems 189

Generalization 4.3.13.1 (USA Winter TST 2020 P2) : Let ABCD be a cyclic quadrilateral, X =
AC ∩ BD, and Y = AB ∩ CD. Let T be a point on line XY , Γ1 be the circle through A and C
tangent to T A, and Γ2 be the circle through B and D tangent to T D. Then Γ1 and Γ2 are viewed
at equal angles from T .

Solution [Length Chase, by a1267ab]. If the radiuses of Γ1 and Γ2 are r1 , r2 , then we have to
show,
TA TD
=
r1 r2
We have,
AB CD
r1 = , r2 =
2 sin ∠T AB 2 sin ∠T DC
To get the sine ratios, we compare the areas of 4T AB and 4T DC. We have,
T A · AB sin ∠T AB [T AB] [XAB] AB 2
= = =
T D · CD sin ∠T DC [T CD] [XCD] CD2
r1 r2
=⇒ =
TA TD

Figure 4.16
4.3.2 problems 190

Problem 4.3.14 (IRAN 3rd Round 2016 P1) : Let ABC be an arbitrary triangle, P is the inter-
section point of the altitude from C and the tangent line from A to the circumcircle. The bisector
of angle A intersects BC at D. P D intersects AB at K, if H is the orthocenter then prove :
HK ⊥ AD.

Solution. Finding a set of Collinear points.

Problem 4.3.15 (IGO 2017 Advance P4) : Three circles W1 , W2 and W3 touches a line l at A, B, C
respectively (B lies between A and C ). W2 touches W1 and W3 . Let l2 be the other common external
tangent of W1 and W3 . l2 cuts W2 at X, Y . Perpendicular to l at B intersects W2 again at K.
Prove that KX and KY are tangent to the circle with diameter AC.

Solution. Finding a Orthocenter Figure in these circle simplifies the problem a lot.

Problem 4.3.16 (2017 IGO Advanced P2) : We have six pairwise non-intersecting circles that the
radius of each is at least one (no circle lies in the interior of any other circle). Prove that the radius
of any circle intersecting all the six circles, is at least one.

Problem 4.3.17 (ARO 2018 P10.2) : Let 4ABC be an acute-angled triangle with AB < AC. Let
M and N be the midpoints of AB and AC, respectively; let AD be an altitude in this triangle. A
point K is chosen on the segment MN so that BK = CK. The ray KD meets the circumcircle Ω
of ABC at Q. Prove that C, N, K, Q are concyclic.

Problem 4.3.18 (ARO 2014 P9.4) : Let M be the midpoint of the side AC of acute-angled triangle
ABC with AB > BC. Let Ω be the circumcircle of ABC. The tangents to Ω at the points A and C
meet at P , and BP and AC intersect at S. Let AD be the altitude of the triangle ABP and ω the
circumcircle of the triangle CSD. Suppose ω and Ω intersect at K 6= C. Prove that ∠CKM = 90◦ .

Problem 4.3.19 (APMO 1999 P3) : Let Γ1 and Γ2 be two circles intersecting at P and Q. The
common tangent, closer to P , of Γ1 and Γ2 touches Γ1 at A and Γ2 at B. The tangent of Γ1 at P
meets Γ2 at C, which is different from P , and the extension of AP meets BC at R. Prove that the
circumcircle of triangle P QR is tangent to BP and BR.

Problem 4.3.20 (Simurgh 2019 P2) : Let ABC be an isosceles triangle, AB = AC. Suppoe that
Q is a point such that AQ = AB, AQ||BC. Let P be the foot of perpendicular line from Q to BC.
Prove that the circle with diameter P Q is tangent to the circumcircle of ABC.
4.3.2 problems 191

Problem 4.3.21 (European Mathematics Cup 2018 P2) : Later

Problem 4.3.22 (RMM 2019 P2) : Let ABCD be an isosceles trapezoid with AB k CD. Let E
be the midpoint of AC. Denote by ω and Ω the circumcircles of the triangles ABE and CDE,
respectively. Let P be the crossing point of the tangent to ω at A with the tangent to Ω at D.
Prove that P E is tangent to Ω.

Problem 4.3.23 (IGO 2018 A5) : ABCD is a cyclic quadrilateral. A circle passing through A, B is
tangent to segment CD at point E. Another circle passing through C, D is tangent to AB at point
F . Point G is the intersection point of AE, DF , and point H is the intersection point of BE, CF .
Prove that the incenters of triangles AGF, BHF, CHE, DGE lie on a circle.

Figure 4.17: Problem 4.3.23 IGO 2018 A5

Problem 4.3.24 (ISL 2011 G8) : Let ABC be an acute triangle with circumcircle Γ. Let ` be a
tangent line to Γ, and let `a , `b and `c be the lines obtained by reflecting ` in the lines BC, CA and
AB, respectively. Show that the circumcircle of the triangle determined by the lines `a , `b and `c is
tangent to the circle Γ.
4.3.2 problems 192

Solution. Find the translated triangle circumscribed in ABC. Once you find the properties of
this triangle and the relations between this and the common touch point, the problem becomes
obvious.

Problem 4.3.25 (ELMO 2019 P3) : Let ABC be a triangle such that ∠CAB > ∠ABC, and let I
be its incentre. Let D be the point on segment BC such that ∠CAD = ∠ABC. Let ω be the circle
tangent to AC at A and passing through I. Let X be the second point of intersection of ω and the
circumcircle of ABC. Prove that the angle bisectors of ∠DAB and ∠CXB intersect at a point on
line BC.

Solution [Angle Chase]. Suppose the bisector of ∠BAD meet BC at G 0 . Then we have,
∠A − ∠B
∠BG 0 A =
2
∴ ∠CG A = ∠B + ∠BG 0 A
0

A+B
=
2

=⇒ CG 0 = CA
∴ ∠G 0 ID = ∠B
Now, let M be the midpoint of the minor arc BC. Let G = XM ∩ BC. So we have
4MGI ∼ 4MIX =⇒ ∠MIG = ∠MXI
Let XI ∩ ABC = N 6= X. Since AC is tangent to AXI, NC k AM. Which means
∠MXI = ∠B = ∠MIG
Which completes our proof by implying that G 0 ≡ G.

Figure 4.18
4.3.2 problems 193

Problem 4.3.26 (ISL 2014 G5) : Convex quadrilateral ABCD has ∠ABC = ∠CDA = 90◦ . Point
H is the foot of the perpendicular from A to BD. Points S and T lie on sides AB and AD,
respectively, such that H lies inside triangle SCT and

∠CHS − ∠CSB = 90◦ , ∠T HC − ∠DT C = 90◦ .

Prove that line BD is tangent to the circumcircle of triangle T SH.

Solution. First construct using nice circles, then prove the center is on AH using angle bisector
theorem.

Figure 4.19: Construction

Figure 4.20: Lemma

4.3.2 problems 194

Problem 4.3.27 (ISL 2014 G7) : Let ABC be a triangle with circumcircle Ω and incentre I. Let
the line passing through I and perpendicular to CI intersect the segment BC and the arc BC (not
containing A) of Ω at points U and V , respectively. Let the line passing through U and parallel to
AI intersect AV at X, and let the line passing through V and parallel to AI intersect AB at Y . Let
W and Z be the midpoints of AX and BC, respectively. Prove that if the points I, X, and Y are
collinear, then the points I, W, and Z are also collinear.

Solution. Draw a nice diagram, and use the parallel property to find circles.

Figure 4.21: ISL 2014 G7

Problem 4.3.28 (ISL 2015 G6) : Let ABC be an acute triangle with AB > AC. Let Γ be its
cirumcircle, H its orthocenter, and F the foot of the altitude from A. Let M be the midpoint of
BC. Let Q be the point on Γ such that ∠HQA = 90◦ and let K be the point on Γ such that
∠HKQ = 90◦ . Assume that the points A, B, C, K and Q are all different and lie on Γ in this order.

Prove that the circumcircles of triangles KQH and F KM are tangent to each other.
4.3.2 problems 195

Solution. Draw the tangent line, and find angles.

Figure 4.22: ISL 2015 G6

Problem 4.3.29 (ISL 2015 G5) : Let ABC be a triangle with CA 6= CB. Let D, F , and G be the
midpoints of the sides AB, AC, and BC respectively. A circle Γ passing through C and tangent
to AB at D meets the segments AF and BG at H and I, respectively. The points H 0 and I 0 are
symmetric to H and I about F and G, respectively. The line H 0 I 0 meets CD and F G at Q and M,
respectively. The line CM meets Γ again at P . Prove that CQ = QP .

Solution. Don’t depend on the figure too much, find facts using facts, not figure.
4.3.2 problems 196

Figure 4.23: ISL 2015 G5

Problem 4.3.30 (ISL 2010 G5) : Let ABCDE be a convex pentagon such that BC k AE, AB =
BC + AE, and ∠ABC = ∠CDE. Let M be the midpoint of CE, and let O be the circumcenter of
triangle BCD. Given that ∠DMO = 90◦ , prove that 2∠BDA = ∠CDE.

Solution. First try to construct the point. Do this the long way, then find a easier way that
includes B, C, not B, A to do that. Then try to translate what 90 degree condition into angles,
and take midpoints, since we have midpoints involved.
4.3.2 problems 197

Figure 4.24: ISL 2010 G5

Problem 4.3.31 (IGO 2019 A5) : Let points A, B and C lie on the parabola ∆ such that the point
H, orthocenter of triangle ABC, coincides with the focus of parabola ∆. Prove that by changing
the position of points A, B and C on ∆ so that the orthocenter remain at H, inradius of triangle
ABC remains unchanged.
4.3.2 problems 198

Figure 4.25: IGO 2019 A5

Solution. I think the idea for inversion should have been pretty natural after finding that the
incircle is fixed.

Problem 4.3.32 (Iran 3rd Round 2015 P5) : Let ABC be a triangle with orthocenter H and
circumcenter O. Let R be the radius of circumcircle of 4ABC. Let A0 , B 0 , C 0 be the points on
−→ −→ −→
AH, BH, CH respectively such that AH.AA0 = R2 , BH.BB 0 = R2 , CH.CC 0 = R2 . Prove that O is
incenter of 4A0 B 0 C 0 .

Solution. The condition easily leads to a nice construction of the points. It should be trivial
to figure that the construction is really important. Also, noticing a similarity among the triangles
is really important.
4.3.3 The line parallel to BC 199

4.3.3 The line parallel to BC

Definition (Let the line parallel to BC through O meet AB, AC at D, E. Let K be the midpoint of
AH, M be the midpoint of BC. F be the feet of A-altitude on BC and let H 0 be the reflection of
H on F . Let O0 be the circumcenter of KBC.) —

Lemma 4.3.2 — ∠DKC = ∠EKB = 90◦

4.3.3 The line parallel to BC 200

Figure 4.26

Lemma 4.3.3 — CD, BE, OH 0 , AM, KO0 are concurrent. (by lemma)
4.3.3 The line parallel to BC 201

Figure 4.27

Problem 4.3.33 (InfinityDots MO Problem 3) : Let 4ABC be an acute triangle with circumcenter
O and orthocenter H. The line through O parallel to BC intersect AB at D and AC at E. X is the
midpoint of AH. Prove that the circumcircles of 4BDX and 4CEX intersect again at a point on
line AO.

Solution. Just using lemma to get another pair of circle where we can apply radical axis
arguments.
4.3.3 The line parallel to BC 202

Solution. Noticing that the resulting point is the isogonal conjugate of a well defined point,

Lemma 4.3.4 — Let P, Q be on AB, AC resp. such that P Q k BC. And let A0 be such that
A0 ∈ ABC, AA0 k BC. Let CP ∩ BQ = X, and let the perpendicular bisector of BC meet P Q
at Y . Prove that A0 , X, Y are collinear.

Solution. No angles... Do Lengths...

Figure 4.28

Problem 4.3.34 (ARO 2018 P11.4) : P ∈ AB, Q ∈ AC, P Q k BC, BQ ∩ CP = X. A0 is the

reflection of A wrt BC. A0 X ∩ AP Q = Y . Prove that BY C is tangent to AP Q.
4.3.3 The line parallel to BC 203

Solution. Of co it can be solved using angle chase, lemma makes it almost trivial.

Problem 4.3.35 (buratinogigle) : Let (O) be a circle and E, F are two points inside (O). (K), (L)
are two circles passing though E, F and tangent internally to (O) at A, D, respectively. AE, AF cut
(O) again at B, C, respectively. BF cuts CE at G. Prove that reflection of A though EF lies on
line DG.

Rephrasing the problem as such: In the setup of this lemma, let A0 X ∩ ABC = Z, then P QZ is
tangent to ABC.

Solution. Simple angle chase.

Solution. Another solution to this is by taking D as a phantom point.

Solution. Another solution is with cross ratios

4.3.4 Simson Line and Stuffs 204

4.3.4 Simson Line and Stuffs

Lemma 4.3.5 ( Simson Line Parallel) — Let P be a point on the circumcircle, let P 0 be the
reflection of P on BC and let P P 0 ∩ Ω = D, and let lp be the Simson line of P . Prove that
lp k AD k HP 0 .

Figure 4.29: The dotted lines are parallel

4.3.4 Simson Line and Stuffs 205

Lemma 4.3.6 ( Simson Line Angle) — Given triangle ABC and its circumcircle (O). Let E, F
be two arbitrary points on (O). Then the angle between the Simson lines of two points E and F
is half the measure of the arc EF .

Figure 4.30
4.3.5 Euler Line 206

4.3.5 Euler Line

Theorem 4.3.7 (Perspectivity Line with Orthic triangle is perpendicular to Euler line) — Let
DEF be the orthic triangle. Then BC ∩ EF, CA ∩ F D, AB ∩ ED are collinear, and the line is
perpendicular to the Euler line. In fact this line is the radical axis of the Circumcircle and the
NinePoint circle

Lemma 4.3.8 — DEF is orthic triangle of ABC, XY Z is the orthic triangle of DEF . Prove
that the perspective point of ABC and XY Z lies on the Euler line of ABC

Solution. Thinking the stuff wrt to the incircle and using cross ratio.
4.4 Cevian and Circumcevian Triangles 207

4.4 Cevian and Circumcevian Triangles

4.4.1 Circumcevian Triangle

Theorem 4.4.1 (Hagge’s circles) — Let P be a point on the plane of 4ABC, let Ω be the
circumcircle. Let A1 , B1 , C1 be the intersections of AP, BP, CP with Ω for the second time. Let
A2 , B2 , C2 be the reflections of A1 , B1 , C1 wrt BC, CA, AB. Prove that H, A2 , B2 , C2 lie on a
circle. This circle is called the P -Hagge’s Circle.

Solution. Either using the dual of Hagge’s Circle, or using the reflection points of A, B, C wrt
the isogonal conjugate of P . And using Lemma 1.1 to finish.

Figure 4.31: P-Hagge Circle

Corollary 4.4.1.1 — 4A1 B1 C1 ∼ 4A2 B2 C2 .

Solution. Straightforward use of Lemma 1.2.

4.4.1 Circumcevian Triangle 208

Corollary 4.4.1.2 — If AH, BH, CH meet A2 B2 C2 H at A3 , B3 , C3 , then A2 A3 , B2 B3 , C2 B3

meet at P .

Solution. Simple angle chase and similarity transformation.

Corollary 4.4.1.3 — If I is the incenter of A2 B2 C2 , K is the reflection of H over I, AK, BK, CK

meet A2 B2 C2 at A4 , B4 , C4 , then A4 A3 , B4 B3 , C4 B3 are concurrent.

Solution. Simple angle chasing and trig-ceva.

Problem 4.4.1 (China TST D2P2, Dual of the Hagge’s Circle theorem) : Let ω be the circumcircle
of 4ABC. P is an interior point of 4ABC. A1 , B1 , C1 are the intersections of AP, BP, CP
respectively and A2 , B2 , C2 are the symmetrical points of A1 , B1 , C1 with respect to the midpoints
of side BC, CA, AB. Show that the circumcircle of 4A2 B2 C2 passes through the orthocenter of
4ABC. Further proof that if this circle’s center is O1 , then HOP O1 is a parallelogram.

Solution. Construct Parallelograms. You have to prove two angles are equal. Reflection the
smaller trig wrt one of the midpoints.

Problem 4.4.2 (China TST 2011, Quiz 2, D2, P1) : Let AA0 , BB 0 , CC 0 be three diameters of the
circumcircle of an acute triangle ABC. Let P be an arbitrary point in the interior of 4ABC, and
let D, E, F be the orthogonal projection of P on BC, CA, AB, respectively. Let X be the point
such that D is the midpoint of A0 X, let Y be the point such that E is the midpoint of B 0 Y , and
similarly let Z be the point such that F is the midpoint of C 0 Z. Prove that triangle XY Z is similar
to triangle ABC.

Solution. A straightforward application of Lemma 6.1 using the O-Hagge’s Circle.

4.4.2 Cevian Triangle 209

4.4.2 Cevian Triangle

Lemma 4.4.2 ( Isogonal Conjugate Lemma) — Let a circle ω meet the sides of triangle ABC
at A1 , A2 ; B1 , B2 ; C1 , C2 . Let P1 , P2 be the miquel points of ABC wrt A1 B1 C1 , A2 B2 C2 resp.
Then P1 , P2 are isogonal conjugates.

Figure 4.32: The two round points are isogonal conjugates.

Theorem 4.4.3 (Terquem’s Cevian Theorem) — Let a circle ω meet the sides of triangle ABC
at A1 , A2 ; B1 , B2 ; C1 , C2 . If AA1 , BB1 , CC1 are concurrent, then so are AA2 , BB2 , CC2

.
4.4.2 Cevian Triangle 210

Theorem 4.4.4 (Mannheim’s Theorem) — Let ABC be a triangle, and let L, M, N be points
on BC, CA, AB respectively. Let A0 , B 0 , C 0 be points on (AMN), (BNL), (CLM), and denote
K ≡ AA0 ∩ BB 0 . Then if K ∈ CC 0 , A0 , B 0 , C 0 , K are concyclic.

Figure 4.33: Mannheim’s Theorem

Theorem 4.4.5 (Mannheim’s Theorem’s Converse) — Let ABC be a triangle, and let L, M, N
be points on BC, CA, AB respectively. Let A0 , B 0 , C 0 be points on (AMN), (BNL), (CLM), and
denote K ≡ AA0 ∩ BB 0 . Then if A0 , B 0 , C 0 , K are concyclic, C 0 ∈ CK.
4.4.2 Cevian Triangle 211

Theorem 4.4.6 (Brocard Points) — Brocard Points are points inside a triangle such that

∠P AB = ∠P BC = ∠P CA = ω

and
∠QCB = ∠QBA = ∠QAC = ω.

Figure 4.34: Brocard Points

Problem 4.4.3 (Rioplatense Olympiad 2013 Problem 6) : Let ABC be an acute-angled scalene
triangle, with centroid G and orthocenter H. The circle with diameter AH cuts the circumcircle of
BHC at A0 , distinct from H. Analogously define B 0 , C 0 . Prove that A0 , B 0 , C 0 , G are concyclic.

Problem 4.4.4 (Iran 3rd Round Training 2016) : ABC is an acute triangle and H, O are its orthocen-
ter and circumcenter respectively. If AO, BO, CO intersect BH, CH, AH at X, Y, Z respectively,then
prove that H, X, Y, Z lie on a circle

Solution. Using Brocard Point

4.4.2 Cevian Triangle 212

Solution. Using Mannheim’s Theorem

Theorem 4.4.7 (Jacobi’s Theorem) — Suppose that D, E, F are points such that AE, AF are
isogonal wrt ∠BAC. Similarly with D, E, F . Then AD, BE, CF are concurrent.
4.5 Incenter–Excenter Lemma stuff 213

4.5 Incenter–Excenter Lemma stuff

Definition (Incenter and Co.) — Let 4ABC be an ordinary triangle, I is its incenter, D, E, F are
the touch points of the incenter with BC, CA, AB and D0 , E 0 , F 0 are the reflections of D, E, F wrt
I.
Let the Ia , Ib , Ic excircles touch BC, CA, AB at D1 , E1 , F1 .
Let Ma , Mb , Mc be the midpoints of the smaller arcs BC, CA, AB, and MA , MB , MC be the midpoints
of the major arcs BC, CA, AB. M are the midpoint of BC.
Let (Ia ) touch BC, CA, AB at DA , EA , FA . So, DA ≡ D1 .
Let A0 be the antipode of A wrt ABC.
Call EF , ‘A-tangent line’, and DE, DF similarly. And call EA FA ‘AA -tangent line.
4.5 Incenter–Excenter Lemma stuff 214

Lemma 4.5.1 — A0 I, ABC, AEIF are concurrent at

YA . And YA , D, Ma are collinear.

Lemma 4.5.2 — DDH ⊥ EF , then DH , I, A0 are

collinear.

Lemma 4.5.3 — Let X be any point on BC, and let

I1 , I2 be incenters of 4ABX, 4ACX. Then XI1 I2 D
are cyclic. And the other common tangent of I1 and
I2 goes through D.
Figure 4.35: Lemma 4.5.4 Excenter
Toucpoints are equidistance from the
Lemma 4.5.4 ( Arc Midpoint as Centers) — Bigger Arc-midpoint

MA E1 = MA F1

MB F1 = MB D1
MC D1 = MC E1

Lemma 4.5.5 ( Incircle Touchpoint and Cevian) —

Let a cevian be AX and let I1 , I2 be the incirlces of
4ABX, 4ACX. Then D, I1 , I2 , X are concyclic.
Figure 4.36: Lemma 4.5.5
4.5 Incenter–Excenter Lemma stuff 215

Lemma 4.5.6 ( Apollonius Circle and Incenter)

— Let ωa be the circle that goes through B, C
and is tangent to (I) at X. Then XD0 , EF, BC
are concurrent and X, D, Ia are collinear. The
same properties is held if the roles of incenter
and excenter are swapped.

• The circle BXC is tangent to (I)

• X lies on the Apollonius Circle of

(B, C; D, G).

• XD bisects ∠BXC.

Solution. Pole-Polar. ISL 2002 G7 Figure 4.37: Circle through BC tangent to in-
circle

Theorem 4.5.7 (Paul Yui Theorem) — B-tangent line, CA -tangent line, and AH are concurrent.

Lemma 4.5.8 ( Concurrent Lines in Incenter) — Let AD ∩ (I) = G, AD0 ∩ (I) = H. Let the
line through D0 parallel to BC meet AB, AC at B 0 , C 0 . Then AM, EF, GH, DD0 , BC 0 , CB 0
are concurrent.

Lemma 4.5.9 ( Midline Concurrency with Incircle Touchpoints) — AI, B, BA -tangent lines and
C-mid-line are concurrent. And, if the concurrency point is X, then CS ⊥ AI
4.5 Incenter–Excenter Lemma stuff 216

Figure 4.38: Lemma 4.5.8 The lines are concur-

rent.

Figure 4.39: Lemma 4.5.9 AI, B, IA -tangent

lines and C-mid-line are concurrent.

Corollary 4.5.9.1 — Let triangle ABC, incircle (I), the A - excircle (Ia ) touches BC at M. IM
intersects (Ia ) at the second point X. Similarly, we get Y , Z. Prove that AX, BY , CZ are
concurrent.
Extension, by buratinogigle: Triangle ABC and XY Z are homothetic with center I is incenter of
ABC. Excircles touches BC, CA, AB at D, E, F. XD, Y E, F Z meets excircles again at U, V,
W. Prove that AU, BV, CW are concurrent.

Definition (Isodynamic Points) — Let ABC be a triangle, and let the angle bisectors of ∠A meet
BC at X, Y . Call ωa the circumcircle of 4AXY . Define ωb , ωc similarly. The first and second
isodynamic points are the points where the three circles ωa , ωb , ωc meet. I.e. these two points are
the intersections of the three Apollonius circles. These two points satisfy the following relations:
4.5 Incenter–Excenter Lemma stuff 217

1.
P A sin A = P B sin B = P C sin C

2. They are the isogonal congugates of the Fermat Points, and they lie on the ‘Brocard Axis’

Figure 4.40

Theorem 4.5.10 (Pedal Triangles of Isodynamic Points) — Prove that the pedal triangles of
the isodynamic points are equilateral triangles. Also, Inverting around the Isodynamic Points
trasnform 4ABC into an equilateral triangle.

Problem 4.5.1 (China TST 2018 T1P3) : Circle ω is tangent to sides AB,AC of triangle ABC
at D,E respectively, such that D 6= B, E 6= C and BD + CE < BC. F ,G lies on BC such that
4.5 Incenter–Excenter Lemma stuff 218

BF = BD, CG = CE. Let DG and EF meet at K. L lies on minor arc DE of ω, such that the
tangent of L to ω is parallel to BC. Prove that the incenter of 4ABC lies on KL.

Solution. Using ??, in the touch triangle of ω.

Problem 4.5.2 (Vietnamese TST 2018 P6.a) : Triangle ABC circumscribed (O) has A-excircle
(Ia ) that touches AB, BC, AC at F, D, E, resp. M is the midpoint of BC. Circle with diameter
MIa cuts DE, DF at K, H. Prove that (BDK), (CDH) have an intersecting point on (Ia ).

Figure 4.41
4.5 Incenter–Excenter Lemma stuff 219

Solution [Inversion]. Inverting around point D, we

have:

Lemma— MD is a line, Ia is an arbitrary point

such that DIa ⊥ MD. l is the perpendicular bisector
of DIa . F, E are arbitrary points on l. B = Ia F ∩
MD, C = Ia E ∩ MD H = F D ∩ MIa , K = DE ∩ MIa .
Then BK, CH, l are concurrent.
Proof. It is straightforward using Puppus’s The-
orem on lines BDC and HIa K.

Figure 4.42: After inverting around D

Solution [Synthetic: Length Chase].

Lemma— Let G, H, B 0 , C 0 be defined the same way in Lemma 3.2. Prove that F lies on the
radical axis of D0 GI, D0 C 0 H. By extension prove that B lies on the radical axis of D0 B 0 I, D0 C 0 H

Figure 4.43: Vietnamese TST 2018 P6.a

We prove the first part, and the second part follows using spiral similarity.
Suppose K ∈ F D ∩ KDI. Due to spiral similarity on KDI, (I), we have 4GF K ∼ 4GD0 I.
Which implies:
FK ID GF
= 0
=⇒ F K = ID
GF GD GD0
4.5 Incenter–Excenter Lemma stuff 220

Now, if KDCE is to be cyclic, we need to have 4HF K ∼ 4HDC. So we need,

FK DC HF
= =⇒ F K = DC
HF HD HD
Combining two equations:
GF ID HF
· =
GD0 DC HD
Now, using Ptolemy’s theorem in F DEH, we have,

F D · EH + DE · F H = DH · EF
FD DH
EH · + DE = EF ·
FH FH
DE DH
2 =
EF FH
Similarly from F GED0 we get,
D0 E GD0
2 =
EF FG
Combining these two equations gives us the desired result.

Generalization 4.5.2.1 (Vietnamese TST

2018 P6.a Generalization) : Let ABC be a
triangle. The points D, E, F are on the
lines BC, CA, AB respectively. The cir-
cles (AEF ), (CF D), (CDE) have a common
point P. A circle (K) passes through P, D
meet DE, DF again at Q, R respectively.
Prove that the circles (DBQ), (DCR) and
(DEF ) are coaxial.

Solution [Inversion]. Invert around D,

and use Pappu’s Theorem as in
Problem 4.5.2.
Figure 4.44: Vietnamese TST 2018 P6.a General-
ization

Remark. The synthetic solution of Problem 4.5.2 can’t be reproduced here maybe because here
we don’t have A, P, D collinear, and we can’t have harmonic quadrilaterals either.
4.5 Incenter–Excenter Lemma stuff 221

Theorem 4.5.11 (Poncelet’s Porism) — Poncelet’s porism (sometimes referred to as Poncelet’s

closure theorem) states that whenever a polygon is inscribed in one conic section and circumscribes
another one, the polygon must be part of an infinite family of polygons that are all inscribed in
and circumscribe the same two conics.

Problem 4.5.3 (IMO 2013 P3) : Let the excircle of triangle ABC 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 triangle A1 B1 C1 lies on
the circumcircle of triangle ABC. Prove that triangle ABC is right-angled.

Solution. Straightforward use of ??

Problem 4.5.4 (buratinogigle’s proposed probs for Arab Saudi team 2015) : Let ABC be acute
triangle with AB < AC inscribed circle (O). Bisector of ∠BAC cuts (O) again at D. E is reflection
of B through AD. DE cuts BC at F . Let (K) be circumcircle of triangle BEF . BD, EA cut (K)
again at M, N, reps. Prove that ∠BMN = ∠KF M.

Figure 4.45

Problem 4.5.5 (USAMO 1999 P6) : Let ABCD be an isosceles trapezoid with AB k CD. The
inscribed circle ω of triangle BCD meets CD at E. Let F be a point on the (internal) angle bisector
of ∠DAC such that EF ⊥ CD. Let the circumscribed circle of triangle ACF meet line CD at C
and G. Prove that the triangle AF G is isosceles.
4.5 Incenter–Excenter Lemma stuff 222

Problem 4.5.6 (Serbia 2018 P1) : Let 4ABC be a triangle with incenter I. Points P and Q are
chosen on segments BI and CI such that 2∠P AQ = ∠BAC. If D is the touch point of incircle and
side BC prove that ∠P DQ = 90.

Solution. Straightforward Trig application.

Problem 4.5.7 (Iran TST T2P5) : Let ω be the circumcircle of isosceles triangle ABC (AB = AC).
Points P and Q lie on ω and BC respectively such that AP = AQ .AP and BC intersect at R.
Prove that the tangents from B and C to the incircle of 4AQR (different from BC) are concurrent
on ω.

Problem 4.5.8 () : Let a point P inside of 4ABC be such that the following condition is satisfied
AP + BP BP + CP CP + AP
= =
AB BC CA
Lines AP, BP, CP intesect the circumcirle again at A , B , C . Prove that ABC and A0 , B 0 , C 0 have
0 0 0

the same incircle.

Solution. After finiding the point P , we get a lot of ideas.

Figure 4.46: two lines are parallel

4.5 Incenter–Excenter Lemma stuff 223

Problem 4.5.9 (Iran TST 2018 P3) : In triangle ABC let M be the midpoint of BC. Let ω
be a circle inside of ABC and is tangent to AB, AC at E, F , respectively. The tangents from M
to ω meet ω at P, Q such that P and B lie on the same side of AM. Let X ≡ P M ∩ BF and
Y ≡ QM ∩ CE. If 2P M = BC prove that XY is tangent to ω.

Solution. Work backwards

Problem 4.5.10 (Iran TST 2018 P4) : Let ABC be a triangle (∠A 6= 90◦ ). BE, CF are the
altitudes of the triangle. The bisector of ∠A intersects EF, BC at M, N. Let P be a point such
that MP ⊥ EF and NP ⊥ BC. Prove that AP passes through the midpoint of BC.

Problem 4.5.11 (APMO 2018 P1) : Let H be the orthocenter of the triangle ABC. Let M and N
be the midpoints of the sides AB and AC, respectively. Assume that H lies inside the quadrilateral
BMNC and that the circumcircles of triangles BMH and CNH are tangent to each other. The line
through H parallel to BC intersects the circumcircles of the triangles BMH and CNH in the points
K and L, respectively. Let F be the intersection point of MK and NL and let J be the incenter of
triangle MHN. Prove that F J = F A.

Problem 4.5.12 (ISL 2006 G6) : Circles w1 and w2 with centres O1 and O2 are externally tangent
at point D and internally tangent to a circle w at points E and F respectively. Line t is the common
tangent of w1 and w2 at D. Let AB be the diameter of w perpendicular to t, so that A, E, O1 are
on the same side of t. Prove that lines AO1 , BO2 , EF and t are concurrent.

Solution. This

Lemma 4.5.12 ( Tangential Quadrilateral Incenters) — Let ABCD be a tangential quatrilateral.

Let I1 , I2 be the incenters of 4ABD, 4BCD. Then (I1 ), (I2 ) is tangent to BD at the same
point.
4.5 Incenter–Excenter Lemma stuff 224

Figure 4.47

Problem 4.5.13 (Four Incenters in a Tangential Quadrilateral) : Let ABCD be a quadrilateral.

Denote by X the point of intersection of the lines AC and BD. Let I1 , I2 , I3 , I4 be the centers of
the incircles of the triangles XAB, XBC, XCD, XDA, respectively. Prove that the quadrilateral
I1 I2 I3 I4 has a circumscribed circle if and only if the quadrilateral ABCD has an inscribed circle.

Solution. There is a lot going on in this figure, firstly, the J1 , J2 and M, then K, then
∠I4 ME = ∠I3 ME. Connecting them with the lemma.
4.5 Incenter–Excenter Lemma stuff 225

Figure 4.48

Problem 4.5.14 (Geodip) : Let G be the centeroid. Dilate I from G with constant −2 to get I 0 .
Then I 0 is tangent to the circumcircle.
4.5 Incenter–Excenter Lemma stuff 226

Figure 4.49

Theorem 4.5.13 (Fuhrmann Circle) — Let X 0 , Y 0 , Z 0 be the midpoints of the arcs not containing
A, B, C of ABC. Let X, Y, Z be the reflections of these points on the sides. Then XY Z is
called the Fuhrmann Circle. The orthocenter H and the nagel point N lies on this circle, and
HN is a diameter of this circle.
Furthermore, AH, BH, CH cut the circle for the second time at a distance 2r from the vertices.
4.5 Incenter–Excenter Lemma stuff 227

Figure 4.50: Fuhrmann Circle

Problem 4.5.15 (Iran TST 2008 P12) : In the acute-angled triangle ABC, D is the intersection
of the altitude passing through A with BC and Ia is the excenter of the triangle with respect to A.
K is a point on the extension of AB from B, for which ∠AKIa = 90◦ + 43 ∠C. Ia K intersects the
extension of AD at L. Prove that DIa bisects the angle ∠AIa B iff AL = 2R. (R is the circumradius
of ABC)

Solution.
4.6 Conjugates 228

4.6 Conjugates

4.6.1 Isogonal Conjugate

Theorem 4.6.1 (Isogonal Line Lemma) — Let AP, AQ are isogonal lines with respect to ∠BAC.
Let BP ∩ CQ = F and BQ ∩ CP = E. Then AE, AF are isogonal lines with respect to ∠BAC.

Figure 4.51

Proof.

A(B, F ; P, X) = (B, F ; P, X) = C(B, Q; E, X) = (B, Q; E, X) = (X, E; Q, B)

So if we define a projective transformation that swaps isogonal lines wrt ∠BAC, we see AE, AF
are conjugates of each other.

Problem 4.6.1 (India Postals 2015 Set 2) : Let ABCD be a convex quadrilateral. In the triangle
ABC let I and J be the incenter and the excenter opposite the vertex A, respectively. In the triangle
ACD let K and L be the incenter and the excenter opposite the vertex A, respectively. Show that
the lines IL and JK, and the bisector of the angle BCD are concurrent.

Solution. Using Theorem 4.6.1

4.6.1 Isogonal Conjugate 229

Lemma 4.6.2 — Let ω1 , ω2 be two circles such that ω1 passes through A, B and is tangent to
AC at A. ω2 is defined similarly by swapping B with C. ω1 ∩ ω2 = X.
Let γ1 , γ2 be two circles such that γ1 passes through A, B and is tangent to BC at B. γ2 is
defined similarly by swapping B with C. γ1 ∩ γ2 = Y .
Then X, Y are isogonal conjugates wrt 4ABC.

Lemma 4.6.3 ( Isogonality in quadrilateral) — For a point X, its isogonal conjugate wrt a
quadrilateral ABCD exists iff
]BXA + ]DXC = 180◦

Figure 4.52: Isogonality in quadrilateral

Solution. Draw the cirles, look for similarity.

Lemma 4.6.4 ( Ratio) — Given a 4ABC with isogonal conjugate P, Q. Let AP, AQ cut the
circumcircle of 4ABC again at U, V, respectively and let D ≡ AP ∩ BC. Then
AQ PD
=
QV DU
4.6.1 Isogonal Conjugate 230

Figure 4.53
4.6.1 Isogonal Conjugate 231

4.6.1.1 Symmedians

Definition (Symmedians) — In 4ABC, let Ta , Tb , Tc be the meet points of the tangents at

A, B, C. Let 4Na Nb Nc be the cevian triangle of ATa , BTb , CTc . Let S be the symmedian point of
4ABC. Let Ma , Mb , Mc be the midpoints of BC, CA, AB.

Lemma 4.6.5 ( Most Important Symmedian Property) — Let the circles tangent to AC, AB at
A and passes through B, C respectively meet at T 0 for the second time. Let ATa ∩ ABC = A0 .
Let the tangents to ABC at A, A0 meet BC at T . Prove that, A, T 0 , Ta , and T, T 0 , O are
collinear.

Figure 4.54: T 0 is quite special!

Problem 4.6.2 (USAMO 2008 P2) : Let ABC be an acute, scalene triangle, and let M, N, and
P be the midpoints of BC, CA, and AB, respectively. Let the perpendicular bisectors of AB and
AC intersect ray AM in points D and E respectively, and let lines BD and CE intersect in point F ,
inside of triangle ABC. Prove that points A, N, F , and P all lie on one circle.
4.6.1 Isogonal Conjugate 232

Figure 4.55: USAMO 2008 P2

Solution [Phantom Point]. First assume F ∈ BD, and F = T 0 (Where T 0 comes from
Lemma 4.6.5, and prove that F ∈ CE.)

Solution [Isogonal Conjugate]. Construct the isogonal conjugate of F , which is the intersection
of the circles touching BC and passing through A, B and A, C.

Solution. Using Theorem 4.6.1 by taking the reflections of B, C over D, F

Problem 4.6.3 (IRAN TST 2015 Day 3, P3) : AH is the altitude of triangle ABC and H 0 is the
reflection of H trough the midpoint of BC. If the tangent lines to the circumcircle of ABC at B
and C, intersect each other at X and the perpendicular line to XH 0 at H 0 , intersects AB and AC
at Y and Z respectively, prove that ∠ZXC = ∠Y XB.

Problem 4.6.4 (Two Symmedian Points) : Let E, F be the feet of B, C-altitudes. Let K, KA be
the symmedian points of 4ABC, 4AEF . Prove that KKA ⊥ BC, KKA ∩ BC = P and KKA = KP
4.6.1 Isogonal Conjugate 233

Figure 4.56: KKA ⊥ BC

4.6.2 Isotonic Conjugate 234

4.6.2 Isotonic Conjugate

Theorem 4.6.6 (Isotonic Lemma) — Let M be the midpoint of BC, and P Q such that Q is
the reflection of P on M. Two points Q, R on AP, AQ, BQ ∩ CR = X, BR ∩ CQ = Y . Then
AX, AY are isotonic wrt BC.

Figure 4.57
4.6.2 Isotonic Conjugate 235

Problem 4.6.5 (IGO 2014 S5) : Two points P and Q lying on side BC of triangle ABC and their
distance from the midpoint of BC are equal.The perpendiculars from P and Q to BC intersect
AC and AB at E and F ,respectively.M is point of intersection P F and EQ.If H1 and H2 be the
orthocenters of triangles BF P and CEQ, respectively, prove that AM ⊥ H1 H2 .

Solution. We first show that the slope of H1 H2 is fixed, and then show that AM is fixed where
we use isotonic lemma, and finally show that these two lines are perpendicular.
4.6.3 Reflection 236

4.6.3 Reflection

Lemma 4.6.7 ( Homothety and Reflection) — Let two oppositely oriented congruent triangles
be 4ABC, 4DEF . Prove that the midpoints of AD, BE, CF are collinear.

Figure 4.58: Oppositely oriented congruent triangles

Problem 4.6.6 (Autumn Tournament, 2012) : Let two oppositely oriented equilateral triangles be
4ABC, 4DEF . What is the least possible value of max (AD, BE, CF )?
4.7 Mixtilinear–Curvilinear–Normal In-Excricles 237

4.7 Mixtilinear–Curvilinear–Normal In-Excricles

Definition (Mixtilinear Circle) — Let 4ABC be an ordinary triangle, I is its incenter, D is the
touch points of the incenter with BC. Let ω be the mixtilinear incircle. Let it touch CA, AB at
E, F . Furthermore, let ω ∩ ABC ≡ T . Let Ma , Mb , Mc be the midpoints of the smaller arcs
BC, CA, AB, and MA , MB , MC be the midpoints of the major arcs BC, CA, AB.

Figure 4.59: Mixtilinear Incircle: Construction

4.7 Mixtilinear–Curvilinear–Normal In-Excricles 238

Figure 4.60: Mixtilinear Incircle: Circlicity Lemmas

Solution. List of small proofs

1. E, I, F are collinear. Consider the circle T I 0 EC and do some angle chasing.

2. T, I, MA are collinear. Consider the circle T IEC and apply reim’s theorem.

Lemma 4.7.1 — TMM c

cA
= TMM
bA
b
, in other words, the bundle (A, T ; Mb , Mc ) is harmonic. And T A
is a symmedian of 4T Mc Mb .

Lemma 4.7.2 — Let X be a variable point on the arc AB, and let O1 and O2 be the incenters
of the triangles CAX and CBX. Then X, O1 , O2 and T lie on a circle.
4.7 Mixtilinear–Curvilinear–Normal In-Excricles 239

Solution. Using similarity and this lemma.

Figure 4.61: The two incenters are cyclic with T, X

Problem 4.7.1 (ISL 1999 G8) : Given a triangle ABC. The points A, B, C divide the circumcircle
Ω of the triangle ABC into three arcs BC, CA, AB. Let X be a variable point on the arc AB, and
let O1 and O2 be the incenters of the triangles CAX and CBX. Prove that the circumcircle of the
triangle XO1 O2 intersects the circle Ω in a fixed point.

Solution. This is actually this lemma.

Problem 4.7.2 (AoPS1) : Let ABCD be a quadrilateral inscribed in a circle, such that the inradius
of 4ABC and ACD are the same. Let T be the touchpoint of A-mixtilinear incircle of the triangle
ABD with ABCD. Let I1 , I2 be the incenters of the triangles ABC, ACD respectively. Show that
I1 I2 and the tangents of A, T wrt ABCD are concurrent.

Solution. The main problem is how to relate the two mixtilinear touchpoints to the two
incenters. But with our mixtilinear lemma, we can do that easily.
4.7 Mixtilinear–Curvilinear–Normal In-Excricles 240

Figure 4.62

Problem 4.7.3 (Generalization of Mixtilinear Incirlce) : Consider triangle ABC and let M, N are
midpoints of arcs AB, AC. Let E, F on AB, AC such that EF k MN. Let EM, F N meet (ABC)
second time at P, Q. Consider two intersection points E 0 , F 0 of (EF P Q) with AB, AC different
from E, F . Then EF 0 ∩ E 0 F is the incenter of ABC.

Problem 4.7.4 () : Let the B-mixtilinear and C-mixtilinear circles touch BC at XB , XC respectively.
Then XB , XC , T, Ma lie on a circle

Problem 4.7.5 (Taiwan TST 2014 T3P3) : Let M be any point on the circumcircle of 4ABC.
Suppose the tangents from M to the incircle meet BC at two points X1 and X2 . Prove that
T, M, X1 , X2 lie on a circle.

Problem 4.7.6 (Archer - EChen M1P3) : Let the incenter touch BC at D. Let AI ∩ BC =
E, AI ∩ ABC = F . Let DEF ∩ ABC = X, DEF ∩ (Ia ) = S1 , S2 . Prove that AX goes
through either S1 or S2 .
4.7 Mixtilinear–Curvilinear–Normal In-Excricles 241

Figure 4.63

Definition (Curvilinear Incircle) — Let ABCD be a cyclic quadrilateral. AC meets BD at X. We

call the circle that touches AX, BX and the circumcircle from the inside a curvilinear incircle.

Lemma 4.7.3 — Let the previously defined curvilinear incircle touch AX, BX at P, Q resp. And
let the incircle of 4ABD be I. Then P, Q, I are collinear.
4.7 Mixtilinear–Curvilinear–Normal In-Excricles 242

Figure 4.64

Solution. Notice that this is similar to the circles T IEC and T IF B in the mixtilinear circle
figures.

Theorem 4.7.4 (Sawayama and Thebault’s theorem) — Through the vertex A of a triangle
ABC, a straight line AD is drawn, cutting the side BC at D. I is the center of the incircle
of triangle ABC. Let P be the center of the circle which touches DC, DA at E, F , and the
circumcircle of ABC, and let Q be the center of a further circle which touches DB, DA in G, H
and the circumcircle of ABC. Then P, I and Q are collinear
4.8 Circles and Radical Axises 243

4.8 Circles and Radical Axises

Problem 4.8.1 () : In 4ABC, H is the orthocenter, and AD, BE are arbitrary cevians. Let ω1 , ω2
denote the circles with diameters AD, BE resp. HD, HE meet ω1 , ω2 again at F, G. DE meet
ω1 , ω2 again at P1 , P2 . F G meet ω1 , ω2 again at Q1 , Q2 . P1 H, P2 H meet ω1 , ω2 at R1 , R2 and
Q1 H, Q2 H meet ω1 , ω2 at S1 , S2 . P1 Q1 ∩ P2 Q2 ≡ X and R1 S1 ∩ R2 S2 ≡ Y . Prove that X, Y, H
are collinear.

Solution. Too much info...

Lemma 4.8.1 ( Pseudo Miquel’s Theorem) — In a 4ABC let E, F be points on AC, AB and D
be a point on (ABC). Let X = (BF D) ∩ (CED) then E, F, X are collinear.

Figure 4.65: Notice the collinearity

Problem 4.8.2 (buratinogigle’s proposed problems for Arab Saudi team 2015) : Let ABC be a
triangle and (K) is a circle that touches segments CA, AB at E, F , reps. M, N lie on (K) such that
BM, CN are tangent to (K). G, H are symmetric of A through E, F . The circle passes through G
and touches to (K) at N that cuts CA again at S. The circle passes through H and touches (K) at
M that cuts AB again at T . Prove that the line passes through K and perpendicular to ST always
passes through a fixed point when (K) changes.
4.8 Circles and Radical Axises 244

Figure 4.66

Problem 4.8.3 (ISL 2002 G8) : Let two circles S1 and S2 meet at the points A and B. A line
through A meets S1 again at C and S2 again at D. Let M, N, K be three points on the line
segments CD, BC, BD respectively, with MN parallel to BD and MK parallel to BC. Let E and
F be points on those arcs BC of S1 and BD of S2 respectively that do not contain A. Given that
EN is perpendicular to BC and F K is perpendicular to BD prove that ∠EMF = 90◦ .

Solution. When one single property can produce a lot others, and we need to prove this
property, assume the property to be true and work backwards.

Problem 4.8.4 (APMO 1999 P3) : Let Γ1 and Γ2 be two circles intersecting at P and Q. The
common tangent, closer to P , of Γ1 and Γ2 touches Γ1 at A and Γ2 at B. The tangent of Γ1 at P
meets Γ2 at C, which is different from P , and the extension of AP meets BC at R. Prove that the
circumcircle of triangle P QR is tangent to BP and BR.

Problem 4.8.5 (USA TST 2019 P1) : Let ABC be a triangle and let M and N denote the midpoints
4.8 Circles and Radical Axises 245

of AB and AC, respectively. Let X be a point such that AX is tangent to the circumcircle of triangle
ABC. Denote by ωB the circle through M and B tangent to MX, and by ωC the circle through N
and C tangent to NX. Show that ωB and ωC intersect on line BC.

Solution [Spiral Similarity]. Let ωC ∩ BC = P . If we extend NP to meet AB at R, we get

XANR cyclic. Similarly, if XAM ∩ AC = Q, then we have to prove QM ∩ NR = P .

Suppose QM ∩ NR = P 0 . Then by spiral similarity, X takes Q → M and N → R. It also takes

Q → N and M → R. So XMP 0 R is cyclic. We now show that XMP R is also cyclic, which will
prove P = P 0 .

Let T = ABC ∩ XAN. By spiral similarity, T takes R → B and N → C. It also takes R → N

and B → C, which means RBP T is cyclic.

By spiral similarity, we have, 4T XA ∼

4T NC, 4T XN ∼ 4T AC, 4T BA ∼
4T P N Which implies,

XN AC XA NC
= , =
TN TC TA TC
XN TN
=⇒ =2
XA TA
And so,

AB NP 2AM TA XA
= =⇒ = = 2
TA TN NP TN XN
AM XA
=⇒ =
NP XN
Figure 4.67

Which means 4XAM ∼ 4XNP since ∠XAM = ∠XNP , which concludes the proof.

Solution [Clever Observation]. Reflect A over X to A0 . Draw the circle with center X with
radius XA. Call it ω. Let P = ω ∩ ABC. Let Q = A0 B ∩ ω.
4.8 Circles and Radical Axises 246

We will show that M, P, B, Q are cyclic,

and XM is tangent to the circle.

First, we have AQ ⊥ A0 B. So MB = MQ.

Now,

]MP Q = ]AP Q − ]AP M

= ]AA0 Q − ]ANM
= ]AXM − ]XAM
= ]AMX
= ]MBQ

So M, Q, P, B is cyclic. Also since MQ =

MB, and XM k BQ, XM is tangent to
MP BQ, and MP BQ = ωB . Figure 4.68

Similarly ωC passes through P , and by Miquel’s theorem, their intersection lies on BC.

Problem 4.8.6 (USA TST 2019 P1 parallel problem) : Pick a point X such that AX is parallel to
BC. Let M, N be the midpoints of AB, AC. Let wb be the circle passing through M and B tangent
to (AXB) and define wc similarly. Show that wb , wc intersect on (AMN).

q
Solution. Doing a bc
2 inversion in Problem 4.8 one ends up with this parallel problem.

Problem 4.8.7 (Sharygin 2010 P3) : Points A0 , B 0 , C 0 lie on sides BC, CA, AB of triangle ABC.
for a point X one has ∠AXB = ∠A0 C 0 B 0 + ∠ACB and ∠BXC = ∠B 0 A0 C 0 + ∠BAC. Prove that
the quadrilateral XA0 BC 0 is cyclic.
4.8 Circles and Radical Axises 247

Figure 4.69

Problem 4.8.8 (IMO 2018 P6) : A convex quadrilateral ABCD satisfies AB · CD = BC · DA.
Point X lies inside ABCD so that

∠XAB = ∠XCD and ∠XBC = ∠XDA.

Prove that ∠BXA + ∠DXC = 180◦ .

Proof. Let P = AB ∩ CD, Q = AD ∩ BC

AB AD
From the first condition, we get that = , implying that the angle bisectors of
BC DC
∠DAB, ∠DCB meet on BD.

And from the second condition, we have X = QBD ∩ P AC

4.8 Circles and Radical Axises 248

Figure 4.70: IMO 2018 P6, Simple Angle-Chase proof.

Let us define the point R such that AR, CR are isogonal to AC wrt to ∠DAB, ∠DCB
respectively. In 4RAC, we have, the bisectors of ∠RAC, ∠RCA meet on the line BRD, meaning
that RB bisects ∠ARC.

Let ARM ∩ DRC = Y . We have,

]AY C = ]AY R + ]RY C

= ]ABR + ]RDP
= ]BP D
=⇒ P AY C is cyclic.

And,
4.8 Circles and Radical Axises 249

]BY D = ]BY R + ]RY D

= ]BAR + ]RCD
= ]CAD + ]BCA
= ]CQD
=⇒ QBY D is cyclic.

So, Y ≡ X. So,

]BY A + ]DY C = ]BRA + ]DRC = ]BRA + ]ARD = 180◦

Problem 4.8.9 (Sharygin 2010) : In 4ABC, let ALa , AMa be the external and internal bisectors
of ∠A with La , Ma lying on BC. Let ωa be the reflection of the circumcircle of 4ALa Ma wrt the
midpoint of BC. Let ωa be defined similarly. Prove that ωa , ωb are tangent to each other iff 4ABC
is a right-angled triangle.
4.9 Complete Quadrilateral + Spiral Similarity 250

4.9 Complete Quadrilateral + Spiral Similarity

Lemma 4.9.1 — Three lines, la , lb , lc , origin at point P . Two circles ω1 , ω2 passing through
P meet the lines at A1 , B1 , C1 ; A2 , B2 , C2 resp. Let A3 be the reflection of A2 on A1 . Define
B3 , C3 similarly. Then P A3 B3 C3 are concyclic.

Figure 4.71: Spiral Similarity Lemma 1: the Blue points have been reflected wrt to the Red points
to get the Green points

Problem 4.9.1 (ISL 2009 G4) : Given a cyclic quadrilateral ABCD, let the diagonals AC and BD
meet at E and the lines AD and BC meet at F . The midpoints of AB and CD are G and H,
respectively. Show that EF is tangent at E to the circle through the points E, G and H.

Solution. This problem generalizes to this

Problem 4.9.2 (All Russian 2014 Grade 10 Day 1 P4) : Given a triangle ABC with AB > BC, let
Ω be the circumcircle. Let M, N lie on the sides AB, BC respectively, such that AM = CN. Let
K be the intersection of MN and AC. Let P be the incenter of the triangle AMK and Q be the
4.9 Complete Quadrilateral + Spiral Similarity 251

K-excenter of the triangle CNK. If R is midpoint of the arc ABC of Ω then prove that RP = RQ.

Lemma 4.9.2 — Let E and F be the intersections of opposite sides of a convex quadrilateral
ABCD. The two diagonals meet at P . Let M be the foot of the perpendicular from P to EF .
Show that ∠BMC = ∠AMD. And P M is the bisector of angles ∠AMC, ∠BMD.

Figure 4.72

Theorem 4.9.3 (Newton-Gauss Line) — Among the points A, B, C, D no three are collinear.
The lines AB and CD intersect at E, and BC and DA intersect at F. Prove that either the
circles with diameters AC, BD, EF pass through two common points, or no two of them have
any common point.
The previous can be stated differently: The midpoints of AC, BD, EF are collinear and this line
is called “Newton-Gauss Line”.

Solution. Either by E.R.I.Q. Lemma, Length Chase, or configurations like Varignon Parallelo-
gram

Problem 4.9.3 () : Let ABCD incribed (O) and a point so-called M .Call X, Y, Z, T, U, V are
the projection of M onto AB, BC, CD, DA, CA, BD respectively. Call I, J, H the midpoints of
4.9 Complete Quadrilateral + Spiral Similarity 252

XZ, UV, Y T respectively. Prove that N, P, Q are collinear.

Solution. Divide the problem in cases, and prove the easiest case first.

Lemma 4.9.4 — In a cyclic quadrilateral ABCD, AC ∩ BD = P, AD ∩ BC = Q, AB ∩ CD = R.

S, T are the midpoints of P Q, P R. And a point X is on ST . Prove that the power of X wrt
ABCD is XP 2 .

Solution. Using polar argument wrt P

Figure 4.73

Problem 4.9.4 (USA TST 2000 P2) : Let ABCD be a cyclic quadrilateral and let E and F be the
feet of perpendiculars from the intersection of diagonals AC and BD to AB and CD, respectively.
Prove that EF is perpendicular to the line through the midpoints of AD and BC.
4.9 Complete Quadrilateral + Spiral Similarity 253

Figure 4.74: USA TST 2000 P2

Solution. First solution is using some properties of the complete quad and angle bash the angle
](MN, EF )

Solution. Second solution is to notice the two brow triangles and proving them congruent.

Problem 4.9.5 () : Let 2 equal circle (O1 ), (O2 ) meet each other at P, Q. O be the midpoint of
P Q. 2 line through P meet the circles at A, B, C, D, (A, C ∈ (O1 ); B, D ∈ (O2 )). M, N be
midpoint of AD, BC. Prove that M, N, O are collinear.

Problem 4.9.6 (AoPS) : In 4ADE a circle with center O, passes through A, D meets AE, ED
respectively at B, C, BD ∩ AC = G, line OG meets ADE at P . Prove that 4P BD, 4P AC has
the same incenter (preferably without using inversion).

Problem 4.9.7 (Archer - EChen M1P2) : Let a circle ω centered at A meet BC at D, E, such
that B, D, E, C all lie on BC in that order. Let ω meet ABC at F, G such that A, F, B, C, G lie
on the circle in that order. Let BF D ∩ AB = K, CGE ∩ AC = L. Prove that F K, GL, AO are
concurrent.
4.9 Complete Quadrilateral + Spiral Similarity 254

Figure 4.75

Problem 4.9.8 (Sharygin 2012 P22) : A circle ω with center I is inscribed into a segment of the
disk, formed by an arc and a chord AB. Point M is the midpoint of this arc AB, and point N is
the midpoint of the complementary arc. The tangents from N touch ω in points C and D. The
opposite sidelines AC and BD of quadrilateral ABCD meet in point X, and the diagonals of ABCD
meet in point Y . Prove that points X, Y, I and M are collinear.
4.9 Complete Quadrilateral + Spiral Similarity 255

Figure 4.76

Solution. La Hire

Problem 4.9.9 (Sharygin 2012 P21) : Two perpendicular lines pass through the orthocenter of an
acute-angled triangle. The sidelines of the triangle cut on each of these lines two segments: one
lying inside the triangle and another one lying outside it. Prove that the product of two internal
segments is equal to the product of two external segments.
4.9 Complete Quadrilateral + Spiral Similarity 256

Figure 4.77

Solution. Spiral Similarity

Problem 4.9.10 (Iran TST 2004 P4) : Let M, M 0 be two conjugates point in triangle ABC (in the
sense that ∠MAB = ∠M 0 AC, . . . ). Let P, Q, R, P 0 , Q0 , R0 be foots of perpendiculars from M and
M 0 to BC, CA, AB. Let E = QR ∩ Q0 R0 , F = RP ∩ R0 P 0 and G = P Q ∩ P 0 Q0 . Prove that the lines
AG, BF, CE are parallel.
4.9 Complete Quadrilateral + Spiral Similarity 257

Figure 4.78: The points are collinear, by Zhao Lemmas

Problem 4.9.11 (Iran TST 2018 D2P6) : Consider quadrilateral ABCD inscribed in circle ω.
P ≡ AC ∩ BD. E, F lie on sides AB, CD respectively such that APˆ E = DP ˆ F . Circles ω1 , ω2 are
tangent to ω at X, Y respectively and also both tangent to the circumcircle of 4P EF at P . Prove
that:
EX FX
=
EY FY
4.10 Projective Geometry 258

4.10 Projective Geometry

• Cross Ratio - Zarathustra Brady

• Desargues’ Involution Theorem - MarkBcc168

4.10.1 Definitions

Definition (Projective Plane) — The projective plane P2 is a set of lines passing through an
observation point O in three dimensional space. A porjective line is a plane passing through O, and
a projective point is a line passing throught O.

Definition (Coordinates in Projective Planes) — A point in a projective plane P2 has coordinates

(p : q : r ). If r = 0, we say that the point is an infinite point. Every line in P2 can also be described
with (p : q : r ) in a sense that this line (which is a plane passing through O) has the equation

pa + qb + r c = 0

Definition (Projection) — We can define projection of P2 on some plane A2 not passing through
O (for simplicity we will take the plane z = 1) by associating
p q 
P = (p : q : r ) ∈ P2 → P 0 = : : 1 ∈ A2
r r
q
If r = 0, then we say P 0 is an infinite point with slope .
p
A projective line l with coordinates p : q : r gets associated to a line l ∈ A2 likewise. The line at
infinity is the line associated with the projective line passing through O and parallel to A2 .

Definition (Projective Line and Inversive Plane) — Every point in a projective line P1 has a
s
coordinate (s : t) which corresponds to the ordinary point x = . The point at infinity will have
t
t = 0. If we let s, t be complex numbers then the projective line is called the inversive plane.
4.10.1 Definitions 259

Definition (Cross Ratio) — If four points A, B, C, D lie on a line, their cross ratio is defined as
AC AD
(A, B; C, D) = ÷
BC BD
If four lines l1 , l2 , l3 , l4 pass through a point, then their cross ratio is
sin ∠l1 l3 sin ∠l1 l4
(l1 , l2 ; l3 , l4 ) = ÷
sin ∠l2 l3 sin ∠l2 l4
If four points on the inversive plane has the complex coordinate a, b, c, d, then the cross ratio is
defined by
a−c a−d
(A, B; C, D) = ÷
b−c c −d

Definition (Möbius Transformation) — A Möbius  Transformation

 is defined by a transformation
a b
fM of the inversive plane by a two by two matrix with non zero determiant as followed:
c d

(s : t) ∈ P1 → (sa + tb : sc + td)

Which is in ordinary coordinates:

az + b
fM (z) =
cz + d
A Möbius transformation can be thought as a matrix transformation of the projective line (which
can be thought as a plane) and then projection on an ordinary line.

Definition (Harmonic Conjugate Map) — For any points A, B on P1 , we define

hA,B (C) = D if (A, B; C, D) = −1

A harmonic cojugate is a Möbius transformation. And a Möbius transformation that is also an

involution, i.e. that has f (f (x)) = x, is a harmonic conjugate.

Definition (Circle Points) — The circle points are the points in P2 with coordinates α = (1 : i : 0)
and α0 = (−i : 1 : 0). These two points are both infinite and imaginary. And every circle passes
through these two points.x

Definition (Coharmonic Points) — Three pairs of points {A, A0 }, {C, C 0 }, {E, E 0 } on the same
line are called coharmonic points iff there exists a pair of points {M, N} on the line such that

(M, N; A, B) = (M, N; C, D) = (M, N; E, F ) = −1

4.10.1 Definitions 260

Definition (Involution) — If there exists a point X on l such that for the Möbius Transformation
f : l → l, such that f (f (X)) = X, then f is an involution.

Theorem 4.10.1 (Properties of Coharmonic Points) — If A, B, C, A0 , B 0 , C 0 lie on a line, no three

the same and A 6= X, then the following are equivalent:

1. {A, A0 }, {B, B 0 }, {C, C 0 } are coharmonic.

2. There is a Möbius Transformation with f (A) = A0 , f (B) = B 0 , f (C) = C 0 which is an
involution.
3. (A, A0 ; B, C) = (A0 , A; B 0 , C 0 )

AC 0 BA0 CB 0
4. = −1
C 0 B A0 C B 0 A
5. (A, A0 ; C, C 0 ) = (A, A0 ; C, B) (A, A0 ; C, B 0 )

Theorem 4.10.2 (Invertible Function on a line) — If f is an invertible function from a line to itself
that is defined by some geometric procedure that has no configuration mess, then f preserves
cross ratio, and is a Möbius Transformation. Similarly, an invertible Möbius Transformation is an
involution on the line.
4.10.2 Cross Ratio 261

4.10.2 Cross Ratio

Theorem 4.10.3 (Pappus’s Hexagon Theorem)

— Let A, B, C be on a line, and let D, E, F
be on another line. Let X = AE ∩ BD, Y =
BF ∩ CE, Z = CD ∩ AF . Then X, Y, Z are on
a line.

Proof. Let CD ∩ BF = J, DE ∩ BD = K.
We have,

(D, Z; J, C)
F
= (AB ∩ F D, A; B, C)
C
= (D, X; B, K)
Y
= (D, XY ∩ DC; J, C)
Figure 4.79

Theorem 4.10.4 (Cross Ratio Equality)

— Let A, B, C, D be on a line, and let
E, F, G, H be on another line. Let X =
AF ∩ BE, Y = BG ∩ CF, Z = CH ∩ DG.
Then X, Y, Z are on a line if and only if
(A, B; C, D) = (E, F ; G, H).

Figure 4.80: (A, B; C, D) = (E, F ; G, H)

4.10.3 Involution 262

4.10.3 Involution

Theorem 4.10.5 (Involution on a line) — An involution on a line l is an inversion around some

point on l.

Theorem 4.10.6 (Involution on a conic) — Let f : C → C be an involution. Let f (X) = X 0 for

all X ∈ C. Then all XX 0 pass through a fixed point P .

Solution [Polar line, Pascal]. Let l be the polar line of P wrt C. Let l 0 be the line parallel to l
passing through P . Let l 0 ∩ C = {X, X 0 }. We show that, if A, B ∈ C, and A0 , B 0 are the second
intersection of AP, BP with C, then, {A, A0 }, {B, B 0 } and {X, X 0 } are coharmonic points wrt C,
that is, for a point Q ∈ C,

(QX, QX 0 ; QA, QB) = (QX 0 , QX; QA0 , QB 0 )

Let XA, XB ∩ l = A1 , B1 , and XA0 , XB 0 ∩ l = A01 , B10 . Since the line (XA ∩ X 0 A0 , XA0 ∩ X 0 A)
is the polar of XX 0 ∩ AA0 , X 0 , A0 , A01 are collinear. Similarly for X 0 , B 0 , B10 . Let T be the polar
point of XX 0 .

We have, T, XB ∩ X 0 A, XA ∩ X 0 B collinear by Pascal’s theorem on

hexagon XXABX 0 X 0 . Let, T 0 = (T, XB ∩ X 0 A, XA ∩ X 0 B) ∩ l 0 . We
have,
XT 0 B1 T A1 T T B1 T A01
= = =⇒ =
T 0X T A01 T B10 T A1 T B10

Now, we have

X(X, X 0 ; A, B) = (T, ∞; A1 , B1 )

T A1 T B10
= = = (∞, T ; A01 , B10 )
T B1 T A01
= X(X 0 , X; A0 , B 0 )
Figure 4.81

Which concludes the proof.

Solution [Inversion]. First project the conic to a circle, then invert C → l across a point P on
C. The goal is to show that for every conjugate pair X, X 0 ∈ C and their image after inversion
X1 , X10 ∈ l, P X1 X10 passes through a fixed point.
4.10.3 Involution 263

By Theorem 4.10.5, we know that there is a point K on l that inverts X1 to X10 . So the circles
P X1 X10 have radical axis P K. Which concludes the proof.

Theorem 4.10.1
Three Conic Law Let A, B, C, D be any four points, no three on a line. Let l be a line
passing through at most one of them. For a point P on l, define f (P ) = P 0 , where P 0 is the
second intersection of l with the conic passing through A, B, C, D, P 0 . Then f is an involution.

Then for any three points X, Y, Z ∈ l, {X, f (X)}, {Y, f (Y )}, {Z, f (Z)} are coharmonic, i.e.
{X, f (X)} are conjugate pairs of an involution on l,

Theorem 4.10.7 (Desargues’ Involution Theorem) — Let ABCD be a quadrilateral, let a

conic C pass through A, B, C, D. And let a line l intersect (AB, CD), (AD, BC), (AC, BD), C
at (X1 , X2 ), (Y1 , Y2 ), (Z1 , Z2 ), (W1 , W2 ). Then

{X1 , X2 }, {Y1 , Y2 }, {Z1 , Z2 }, {W1 , W2 }

are coharmonic points i.e. they are reciprocal pairs of some involution on l.

Figure 4.82: Desargues’ Involution Theorem

Proof. Apply the Three Conic Law on l with points A, B, C, D.

Theorem 4.10.8 (Degenerate Desargues’ Involution: 2 Points) — Let A, B, be two points on

a conic C, let a line l meet AB, C and the tangents at A, B to C at X, (W1 , W2 ), (Y1 , Y2 ). Then
(X, X), (W1 , W2 ), (Y1 , Y2 ) are reciprocal pairs of an involution on l.
4.10.4 Inversion 264

4.10.4 Inversion

√
TelvCohl’s bc inversion problem collection

Lemma 4.10.9 — WRT a circle ω with center O the polar of a point A can be constructed as
the radical axis of ω and the circle with diameter OA.

Lemma 4.10.10 — ](a, b) = ]AOB

Theorem 4.10.11 (Pascal’s Theorem for Octagons: A special case) — Let ABCDA0 B 0 C 0 D0 be
a octagon inscribed in a conic section. If the points:

AD ∩ BC, AC 0 ∩ BB 0 , AD0 ∩ CA0 , BD0 ∩ DA0 , DB 0 ∩ CC 0

are collinear, then so are the points

A0 D0 ∩ B 0 C 0 , A0 C ∩ B 0 B, A0 D ∩ C 0 A, B 0 D ∩ D0 A, D0 B ∩ C 0 C

Figure 4.83: If the small Red points are collinear, then the Blue ones are too.
4.10.4 Inversion 265

Theorem 4.10.12 (Inscribed Conic in Pascal’s theorem) — A1 A2 A3 A4 A5 A6 be a hexagon in-

scribed in a conic section. Then the hexagon formed by

A1 A3 ∩ A2 A6 , A2 A4 ∩ A1 A3 , A2 A4 ∩ A3 A5 , A3 A5 ∩ A4 A6 , A5 A1 ∩ A4 A6 , A1 A5 ∩ A2 A6

has an inscribed conic section.

Problem 4.10.1 () : Let ABCD have an incircle (I). Let (I) meet AB, BC, CD, DA at M, N, P, Q.
Let K, L be the circumcenters of AMN, AP Q. KL ∩ BD = R, AI ∩ MQ = J. Prove that RA = RJ.

Problem 4.10.2 () : Let the A mixtilinear incircle (O) of 4ABC meet ABC, AC, AB at P, E, F .
Let M be the BC arc midpoint. Let H be the conic that goes through E, F, O, P, M meet ABC
at X, Y . Prove that AA, XY, EF are concurrent.

Problem 4.10.3 (Iran 3rd Round G4) : Let ABC be a triangle with incenter I. Let K be the
midpoint of AI and BI ∩ (4ABC) = M, CI ∩ (4ABC) = N. points P, Q lie on AM, AN
respectively such that ∠ABK = ∠P BC, ∠ACK = ∠QCB. Prove that P, Q, I are collinear.

Solution. Since we are dealing with collinearity, and usually we use harmonic bundle in these
cases to show collinearity. But in this problem, there is no harmonic bundle. So we use cross
ratio...

Generalization 4.10.3.1 (Iran 3rd Round G4 Generalized version) : Let ABC be a triangle inscribed
in circle (O) and P, Q are two isogonal conjugate points. P B, P C cut (O) again at M, N. QA cuts
MN at K. L is isogonal conjugate of K. LB, LC cut AM, AN at S, T , resp. Prove that S, Q, T
are collinear.

Lemma 4.10.13 — Too long, can’t explain, loot at the figure. The dotted lines go through that
concurrency point.
4.10.4 Inversion 266

Figure 4.84: Everything concurrs

Lemma 4.10.14 ( Construction of Involution Center on Line) — Given a line l, four points
A, B, A0 , B 0 such that A, A0 and B, B 0 are two conjugate pairs of some involution i.e. some
inversion on l, find the center O of inversion.
4.10.5 Problems 267

4.10.5 Problems

Problem 4.10.4 (Dunno) : Let E, F be on the lines AC, AB of 4ABC. Let P be a point on EF .
Let Q be the intersection of the lines through E, F and parallel to BP and CP respectively. Prove
that, as P moves along EF , Q moves along a line.

BX CE CY BF
Solution [Cross Ratio]. Let X ∈ BF, Y ∈ CE such that = , = . With
XF EA Y A FA
trivial calculation, we have XY k BC We show that Q, X, Y are collinear. For that we will show
F U k EV where U = BP ∩ XY, V = CP ∩ XY . And by reverse Pappu’s theorem on F P E, UQV ,
we will have U, Q, V collinear.

Let K, L = BP, CP ∩ XY . Also, S = BC ∩ EF .

Then we have,
B C
(S, P ; F, E) = (∞, U; X, K) = (∞, V ; L, Y )

XU LV
=⇒ =
UK VY
But we have,
FX FL AE
= =
XB LC EC

Figure 4.85

So, F X k EL, and similarly, F K k EY . So by similarity, we have F U k EV .

4.10.5 Problems 268

Lemma 4.10.15 ( Conic through orthocenter and ver-

tices) — Let ABCD be a quadrilateral. Let G, H
be the orthocenters of 4ABC and 4DBC. Then
A, B, C, D, G, H all lie on a conic.

Proof. Since li ne(AC ∩ BD, BG ∩ CH) is perpendicular

to BC, we have AG ∩ DH, AC ∩ BD, BG ∩ CH collinear.
So by reverse Pascal’s theorem, A, D, G, H, B, C lie on a
conic.

Figure 4.86

Lemma 4.10.16 ( Orthogonal Hyperbola) — Let H be the orthocenter of ABC. Let C be a

conic through A, B, C, H. If XY Z is a triangle with vertices in C, then the orthocenter W of
XY Z lies also on C. Also, the asymptotes of C are orthogonal.

Solution. As in Lemma 4.10.15, the orthocenter ot XBC lies on C too. So the orthocenter
of XY C lies on C and so does the orthocenter of XY Z.

Now we show that asymptotes of C are orthogonal.

Let the two infinity points on C be s, t. Consider the triangle Ast. Let its orthocenter be B.
Then we have, tB ⊥ sA. But tB is parallel to the asymptote through t and sA is parallel to the
asymptote through s. So the asymptotes themselves are orthogonal.
4.10.5 Problems 269

Figure 4.87: Hyperbola through A, B, C, H

4.10.6 Projective Constructions 270

4.10.6 Projective Constructions

Construction 1 (Second Intersection of Line with Conic)

— Given four points A, B, C, D, no three collinear, and a
point P on a line l passing through at most one of the four
points, construct the point P 0 ∈ l such that A, B, C, D, P, P 0
line on the same conic.

Solution. Let AP ∩ BC = X, l ∩ CD = Y, XY ∩ AD =
Z. Then by Pascal’s Hexagrummum Mysticum Theorem,
we have, P 0 = BZ ∩ l
Figure 4.88

Construction 2 (Conic touching conic) — Given a conic

C, and two points A, B on it, and C inside of it. Construct
the conic H that is tangent to C at A, B and passes through
C.

Solution. Draw the two tangest at A, B which meet at

X. Take an arbitrary line passing through X that inter-
sects AC, BC at Y, Z. Take D = BY ∩ AZ. Then D lies
on H by Pascal. Construct another point E similarly and
draw the conic.
Figure 4.89

Construction 3 (Inconic of a quadrilateral) — Given a convex quadrilateral ABCD. P = AC ∩BD,

S ∈ AD, T ∈ BC such that S, P, T are collinear. Construct the conic that touches AB, CD, and
also touches AD, BC at S, T respectively.

Solution [the_Construction]. Draw the polar line l of P wrt to the quadrilateral. Let Z =
BC ∩ l. Let ZS ∩ AB = U, ZT ∩ CD = V . Then SSUUT T V V is our desired conic.
4.10.6 Projective Constructions 271

Proof. If U, V ∈ CD, AB such that UV

passes through P , and if the conic passing
through U, V and tangent to AD, BC at S, T
intersects CD at U 0 again, then SV, U 0 T, DB
are concurrent. So to show our construction
works, we just need to prove that U, V, P are
collinear.

Since Pascal’s theorem works on SV BT UD,

we know S, V, B, T, U, D lie on a conic H and
l is the pole of P wrt H. Now, applying
Pascal’s theorem on T DV UBS, and quadrilat-
eral theorem on BT UD and BV SD, we have,
ST ∩ UV ∈ AC, which is P . So we are
done.
Figure 4.90

Construction 4 (Sharygin Olympiad 2010) — A

conic C passing through the vertices of 4ABC
is drawn, and three points A0 , B 0 , C 0 on its sides
BC, CA, AB are chosen. Then the original triangle
is erased. Prove that the original triangle can be
constructed iff AA0 , BB 0 , CC 0 are concurrent.

Solution [the_Construction]. Draw B 0 C 0 . It

intersects the circle at X1 , X2 . Draw the conic
H that is tangent to C at X1 , X2 and passes
through A0 . Then BC is tangent to H at A0 .

Figure 4.91

Proof. The only if part is easy to prove. Becase if AA0 , BB 0 , CC 0 aren’t concurrent, then we
can get multiple triangles ABC. So suppose that they are concurrent.

Now we define some intersetion points.

4.10.6 Projective Constructions 272

W1 = BB 0 ∩ C
S = X1 X1 ∩ AW1
T = X1 B ∩ AX2
U = X1 X1 ∩ BC
V = X2 X2 ∩ BC
R = X2 X2 ∩ AW1
Y1 = A0 B 0 ∩ SR

T, S, B 0 are collinear by Pascal’s theorem on BX1 X1 X2 AW1 . T, B 0 , V are similarly collinear for
AX2 X2 X1 BC. And similarly R, B 0 , U are collinear.

We will prove that H is an inconic of SRV U that goes through A0 , X1 , X2 .

For a point X on UV , define f : UV → UV such that f (X) is the second intersection of the
conic X1 X1 X2 X2 X (X1 X1 = SU, X2 X2 = RV ) with UV . f is an involution by ??.

Suppose A1 is the intersection with the inconic of SRUV through X1 , X2 and UV . Let
A2 = X1 X2 ∩ UV . Then f (A1 ) = A1 , f (A2 ) = A2 , f (B) = C.

Which means, A(B, C; A1 , A2 ) = −1. Which means A1 = A0 . So, X1 X2 A0 X2 X2 is an inconic of

SRV U, just as we wanted.

Figure 4.92
4.11 Parallelogram Stuff 273

4.11 Parallelogram Stuff

Theorem 4.11.1 (Maximality of the Area of a Cyclic Quadrilateral) — Among all quadrilaterals
with given side lengths, the cyclic one has maximal area.

Figure 4.93: The cyclic quad has the maximal area

Problem 4.11.1 (IOM 2017 P1) : Let ABCD be a parallelogram in which angle at B is obtuse
and AD > AB. Points K and L on AC such that ∠ADL = ∠KBA(the points A, K, C, L are all
different, with K between A and L). The line BK intersects the circumcircle ω of ABC at points
B and E, and the line EL intersects ω at points E and F . Prove that BF k AC.

Simplify: Make the diagram easier to draw.

Problem 4.11.2 (USA TST 2006 P6) : Let ABC be a triangle. Triangles P AB and QAC are
constructed outside of triangle ABC such that AP = AB and AQ = AC and ∠BAP = ∠CAQ.
Segments BQ and CP meet at R. Let O be the circumcenter of triangle BCR. Prove that
AO ⊥ P Q.
4.11 Parallelogram Stuff 274

Figure 4.94: USA TST 2006 P6, That is a prallelogram

4.12 Length Relations 275

4.12 Length Relations

Lemma 4.12.1 ( E.R.I.Q. (Equal Ration in Quadrilateral) Lemma) — Let A1 , B1 , C1 ; A2 , B2 , C2

be two sets of collinear points such that
A1 B1 A2 B2
= =k
B1 C1 B2 C2
. Let points A, B, C be on A1 A2 , B1 B2 , C1 , C2 such that:
A1 A B1 B C1 C
= =
A2 A B2 B C2 C
Then we have,
AB
A, B, C are collinear and, =k
BC
4.12 Length Relations 276

Figure 4.95: E.R.I.Q. Lemma

Solution. A great use of this problem is in proving some midpoints collinear. Line in Newton-
Gauss Line and some other such problems (1, 2, 3) where it is asked to proved that some
midpoints are collinear.

Lemma 4.12.2 ( Steiner’s Isogonal Cevian Lemma) — In 4ABC, AA1 , AA2 are two isogonal
4.12 Length Relations 277

cevians, with A1 , A2 ∈ BC. Then we have

BA1 BA2 BA2

× =
A1 C A2 C AC 2

Theorem 4.12.3 () — Let P1 , P2 be two isogonal conjugates wrt 4ABC. Then if the Pedal
triangle of P1 is homological wrt to 4ABC then so is the Pedal triangle of P2 .

Theorem 4.12.4 (Erdos-Mordell Theorem (Forum Geometricorum Volume 1 (2001) 7-8)) — If

from a point O inside a given 4ABC perpendiculars OD, OE, OF are drawn to its sides, then
OA + OB + OC ≥ 2(OD + OE + OF ). Equality holds if and only if 4ABC is equilateral.

Apparently nothing is needed except “Ptolemy’s Theorem”. Think of a way to connect OA with
OE, OF and the sides of the triangle. As it is the most natural to use AB, AC, we have to deal
with BE, CF too. And dealing with lengths is the easiest when we have similar triangles. So we do
some construction.

Problem 4.12.1 (ISL 2011 G7) : Let ABCDEF be a convex hexagon all of whose sides are tangent
to a circle ω with centre O. Suppose that the circumcircle of triangle ACE is concentric with ω.
Let J be the foot of the perpendicular from B to CD. Suppose that the perpendicular from B to
DF intersects the line EO at a point K. Let L be the foot of the perpendicular from K to DE.
Prove that DJ = DL.

Solution. There are a LOT of equal lenths, equal angles, and we have a perpendicularity lemma
working as well. Why don’t we try cosine :0
4.13 Pedal Triangles 278

4.13 Pedal Triangles

Definition (Pedal Triangles) — Let P be an arbitrary point, let 4A1 B1 C1 be its pedal triangle wrt
4ABC. Let A0 , B 0 , C 0 and A0 , B0 , C0 be the feet of the altitudes and the midpoints of 4ABC.

B1 C1 ∩ B0 C0 = A2 , C1 A1 ∩ C0 A0 = B2 , A1 B1 ∩ A0 B0 = C2
B 0 C 0 ∩ B0 C0 = A3 , C 0 A0 ∩ C0 A0 = B3 , A0 B 0 ∩ A0 B0 = C3

Theorem 4.13.1 (Fontene’s First Theorem) — A1 A2 , B1 B2 , C1 C2 are concurrent at the inter-

section of A1 B1 C1 and A0 B0 C0

Lemma 4.13.2 — A0 A3 , B 0 B3 , C 0 C3 and A0 A3 , B0 B3 , C0 C3 concur at the nine point circle of

4ABC.

Theorem 4.13.3 (Fontene’s Second Theorem) — Let the concurrency point in the first theorem
be Q. Then, if the line OP is fixed and P moves along that line, Q will stay fixed.

The previous result leads to another beautiful result:

Lemma 4.13.4 — Suppose a varying point P is chosen on the Euler Line of 4ABC. Then the
pedal circle of P wrt 4ABC intersects the 9p circle at a fixed point which is the Euler Reflection
Point of the median triangle.
4.14 Pending Problems 279

4.14 Pending Problems

Problem 4.14.1 () : In 4ABC, I is the incenter, D is the touch point of the incenter with BC.
AD ∩ ABC ≡ X. The tangents line from X to I meet ABC at Y, Z. Prove that Y Z, BC and
the tangent at A to ABC concur.

Problem 4.14.2 (IRAN TST 2017 Day 1, P3) : In triangle ABC let Ia be the A-excenter. Let ω be
an arbitrary circle that passes through A, Ia and intersects the extensions of sides AB, AC (extended
from B, C) at X, Y respectively. Let S, T be points on segments Ia B, Ia C respectively such that
∠AXIa = ∠BT Ia and ∠AY Ia = ∠CSIa . Lines BT, CS intersect at K. Lines KIa , T S intersect at
Z. Prove that X, Y, Z are collinear.

Problem 4.14.3 (IRAN TST 2015 Day 3, P2) : In triangle ABC(with incenter I) let the line parallel
to BC from A intersect circumcircle of 4ABC at A1 let AI ∩ BC = D and E is tangency point of
incircle with BC let EA1 ∩ (4ADE) = T prove that AI = T I.
4.14 Pending Problems 280

Figure 4.96: IRAN TST 2015 Day 3, P2

Problem 4.14.4 (Generalization of Iran TST 2017 P5) : Let ABC be triangle and the points P, Q
lies on the side BC s.t B, C, P, Q are all different. The circumcircles of triangles ABP and ACQ
intersect again at G. AG intersects BC at M. The circumcircle of triangle AP Q intersects AB, AC
again at E, F , respectively. EP and F Q intersect at T . The lines through M and parallel to
AB, AC, intersect EP, F Q at X, Y , respectively. Prove that the circumcircles of triangle T XY and
AP Q are tangent to each other.

Problem 4.14.5 (ARMO 2013 Grade 11 Day 2 P4) : Let ω be the incircle of the triangle ABC
and with center I. Let Γ be the circumcircle of the triangle AIB. Circles ω and Γ intersect at the
point X and Y . Let Z be the intersection of the common tangents of the circles ω and Γ. Show
that the circumcircle of the triangle XY Z is tangent to the circumcircle of the triangle ABC.
4.14 Pending Problems 281

Problem 4.14.6 (AoPS) : Let ABC be a triangle with incircle (I) and A-excircle (Ia ). (I), (Ia )
are tangent to BC at D, P , respectively. Let (I1 ), (I2 ) be the incircle of triangles AP C, AP B,
respectively, (J1 ), (J2 ) be the reflections of (I1 ), (I2 ) wrt midpoints of AC, AB. Prove that AD is
the radical axis of (J1 ) and (J2 ).

Problem 4.14.7 (AoPS) : Let ABC be a A−right-angled triangle and MNP Q a square inscribed
into it, with M, N onto BC in order B − M − N − C, and P, Q onto CA, AB respectively. Let
R = BP ∩ QM, S = CQ ∩ P N. Prove that AR = AS and RS is perpendicular to the A−inner angle
bisector of 4ABC.

Problem 4.14.8 (AoPS) : P is an arbitrary point on the plane of 4ABC and let 4A0 B 0 C 0 be the
cevian triangle of P WRT 4ABC. The circles (ABB 0 ) and (ACC 0 ) meet at A, X. Similarly,
define the points Y and Z WRT B and C. Prove that the lines AX, BY, CZ concur at the isogonal
conjugate of the complement of P WRT 4ABC.

Problem 4.14.9 (AoPS) : Given are 4ABC, L is Lemoine point,La , Lb , Lc are three Lemoine point
of triangles LBC, LCA, LAB prove that ALa , BLb , CLa are concurrent!
A question: What is the locus of point P such that ALa , BLb , CLa are concurrent with La , Lb , Lc
are three ‘Lemoine points’ of triangles P BC, P CA, P AB?

Problem 4.14.10 (AoPS) : Let ABC be a triangle inscribed circle (O). Let (O0 ) be the circle
wich is tangent to the circle (O) and the sides CA, AB at D and E, F , respectively. The line BC
intersects the tangent line at A of (O), EF and AO0 at T, S and L, respectively. The circle (O)
intersects AS again at K. Prove that the circumcenter of triangle AKL lies on the circumcircle of
triangle ADT .

Problem 4.14.11 () : Let P and Q be isogonal conjugates of each other. Let 4XY Z, 4KLM be
the pedal triangles of P and Q wrt 4ABC. (X, K lie on BC; Y, L lie on CA; Z, M lie on AB) Prove
that Y M, ZL, P Q are concurrent.

Problem 4.14.12 (2nd Olympiad of Metropolises) : Let ABCDEF be a convex hexagon which
has an inscribed circle and a circumscribed circle. Denote by ωA , ωB , ωC , ωD , ωE , and ωF the
inscribed circles of the triangles F AB, ABC, BCD, CDE, DEF ,and EF A,respectively. Let lAB be
the external common tangent of ωA and ωB other than the line AB; lines lBC , lCD , lDE , lEF , and
lF A are analogously defined. Let A1 be the intersection point of the lines lF A and lAB ; B1 be
the intersection point of the lines lAB and lBC ; points C1 , D1 , E1 , and F 1 are analogously defined.
Suppose that A1 B1 C1 D1 E1 F1 is a convex hexagon. Show that its diagonals A1 D1 , B1 E1 , and C1 F1
4.14 Pending Problems 282

meet at a single point.

Problem 4.14.13 (ISL 2016 G6) : Let ABCD be a convex quadrilateral with ∠ABC = ∠ADC <
90◦ . The internal angle bisectors of ∠ABC and ∠ADC meet AC at E and F respectively, and meet
each other at point P . Let M be the midpoint of AC and let ω be the circumcircle of triangle BP D.
Segments BM and DM intersect ω again at X and Y respectively. Denote by Q the intersection
point of lines XE and Y F . Prove that P Q ⊥ AC.
4.15 Problems 283

4.15 Problems

Problem 4.15.1 (IRAN 3rd Round 2016 P2) : Let ABC be an arbitrary triangle. Let E, F be two
points on AB, AC respectively such that their distance to the midpoint of BC is equal. Let P be
the second intersection of the triangles ABC, AEF circumcircles . The tangents from E, F to the
circumcircle of AEF intersect each other at K. Prove that : ∠KP A = 90

Problem 4.15.2 (IRAN 2nd Round 2016 P6) : Let ABC be a triangle and X be a point on its
circumcircle. Q, P lie on a line BC such that XQ ⊥ AC, XP ⊥ AB. Let Y be the circumcenter of
4XQP . Prove that ABC is equilateral triangle if and if only Y moves on a circle when X varies on
the circumcircle of ABC

Problem 4.15.3 (AoPS) : Consider ABC with orthic triangle A0 B 0 C 0 , let AA0 ∩ B 0 C 0 = E and E 0
be reflection of E wrt BC. Let M be midpoint of BC and O be circumcenter of E 0 B 0 C 0 . Let M 0 be
projection of O on BC and N be the intersection of a perpendicular to B 0 C 0 through E with BC.
Prove that MM 0 = 1/4MN.

Problem 4.15.4 (IRAN 3rd Round 2010 D3, P5) : In a triangle ABC, I is the incenter. D is the
reflection of A to I. the incircle is tangent to BC at point E. DE cuts IG at P (G is centroid). M
is the midpoint of BC. Prove that AP ||DM and AP = 2DM.

Problem 4.15.5 (IRAN 3rd Round 2011 G5) : Given triangle ABC, D is the foot of the external
angle bisector of A, I its incenter and Ia its A-excenter. Perpendicular from I to DIa intersects
the circumcircle of triangle in A0 . Define B 0 and C 0 similarly. Prove that AA0 , BB 0 and CC 0 are
concurrent.

Problem 4.15.6 (AoPS3) : I is the incenter of ABC, P I, QI⊥BC, P A, QA intersect BC at DE.

Prove: IADE is on a circle.
4.15 Problems 284

Figure 4.97: AoPS3

Problem 4.15.7 (AoPS4) : Given a triangle ABC, the incircle (I) touch BC, CA, AB at D, E, F
respectively. Let AA1 , BB1 , CC1 be A, B, C − alti tude respectively. Let N be the orthocenter of
the triangle AEF . Prove that N is the incenter of AB1 C1
4.15 Problems 285

Figure 4.98: AoPS4

Problem 4.15.8 (IRAN TST 2015 Day 2, P3) : ABCD is a circumscribed and inscribed quadri-
lateral. O is the circumcenter of the quadrilateral. E, F and S are the intersections of AB, CD;
AD, BC and AC, BD respectively. E 0 and F 0 are points on AD and AB such that ∠AEE 0 = ∠E 0 ED
and ∠AF F 0 = ∠F 0 F B. X and Y are points on OE 0 and OF 0 such that XD XA EA
= ED and YY B
A
= FF B
A
.
M is the midpoint of arc BD of (O) which contains A. Prove that the circumcircles of triangles
OXY and OAM are coaxial with the circle with diameter OS.
4.15 Problems 286

Figure 4.99: Actual Prob

4.15 Problems 287

Figure 4.100: Inverted

Problem 4.15.9 (USA TST 2017 P2) : Let ABC be an acute scalene triangle with circumcenter
O, and let T be on line BC such that ∠T AO = 90◦ . The circle with diameter AT intersects the
circumcircle of 4BOC at two points A1 and A2 , where OA1 < OA2 . Points B1 , B2 , C1 , C2 are
defined analogously.

1. Prove that AA1 , BB1 , CC1 are concurrent.

2. Prove that AA2 , BB2 , CC2 are concurrent on the Euler line of triangle ABC.

Problem 4.15.10 (AoPS2) : Let ABC be a triangle with circumcenter O and altitude AH. AO
meets BC at M and meets the circle (BOC) again at N. P is the midpoint of MN. K is the
projection of P on line AH. Prove that the circle (K, KH) is tangent to the circle (BOC).
4.15 Problems 288

Figure 4.101: AoPS2

4.15 Problems 289

Solution. Inversion all the way...

Problem 4.15.11 (AoPS5) : Let ABC be a triangle inscribed in (O) and P be a point. Call P 0 be
the isogonal conjugate point of P . Let A0 be the second intersection of AP 0 and (O). Denote by
M the intersection of BC and A0 P . Prove that P 0 M k AP .

Figure 4.102: AoPS5

4.15 Problems 290

Problem 4.15.12 (AoPS) : I is the incenter of a non-isosceles triangle 4ABC. If the incircle
touches BC, CA, AB at A1 , B1 , C1 respectively, prove that the circumcentres of the triangles 4AIA1 ,
4BIB1 , 4CIC1 are collinear.

Problem 4.15.13 (AoPS) : Given 4ABC and a point P inside. AP cuts BC at M. Let M 0 , A0 be
the reflection of M, A in the perpendicular bisector of BC. A0 P cuts the perpendicular bisector of
BC at N. Let Q be the isogonal conjugate of P in triangle ABC. Prove that QM 0 k AN.

Problem 4.15.14 (IRAN 3rd Round 2016 G6) : Given triangle 4ABC and let D, E, F be the foot
of angle bisectors of A, B, C ,respectively. M, N lie on EF such that AM = AN. Let H be the foot
of A-altitude on BC.
Points K, L lie on EF such that triangles 4AKL, 4HMN are correspondingly similar (with the
given order of vertices’s) such that AK 6k HM and AK 6k HN. Show that: DK = DL.
4.15 Problems 291

Figure 4.103: IRAN 3rd Round 2016 G6

Problem 4.15.15 (Iran TST 2017 T3 P6) : In triangle ABC let O and H be the circumcenter and
the orthocenter. The point P is the reflection of A with respect to OH. Assume that P is not on
the same side of BC as A. Points E, F lie on AB, AC respectively such that BE = P C , CF = P B.
Let K be the intersection point of AP, OH. Prove that ∠EKF = 90◦ .

Spiral Similarity (points on AB, AC with some properties)

4.15 Problems 292

Figure 4.104: Iran TST 2017 T3 P6

Problem 4.15.16 (IRAN 3rd Round 2010 D3, P6) : In a triangle ABC, ∠C = 45◦ . AD is the
altitude of the triangle. X is on AD such that ∠XBC = 90 − ∠B (X is inside of the triangle). AD
and CX cut the circumcircle of ABC in M and N respectively. Ff the tangent to ABC at M cuts
AN at P , prove that P, B and O are collinear.

Cross-Ratio

Problem 4.15.17 (Iran TST 2014 T1P6) : I is the incenter of triangle ABC. perpendicular from
4.15 Problems 293

I to AI meet AB and AC at B 0 and C 0 respectively. Suppose that B 00 and C 00 are points on half-
line BC and CB such that BB 00 = BA and CC 00 = CA. Suppose that the second intersection of
circumcircles of AB 0 B 00 and AC 0 C 00 is T . Prove that the circumcenter of AIT is on the BC.

projective, inversion

Solution. Too many collinearity, need to prove concurrency, what else can come into mind
except projective approach.

√
Solution. Too many incenter related things, bc-inversion :o

Problem 4.15.18 (APMO 2014 P5) : Circles ω and Ω meet at points A and B. Let M be the
midpoint of the arc AB of circle ω (M lies inside Ω). A chord MP of circle ω intersects Ω at Q
(Q lies inside ω). Let `P be the tangent line to ω at P , and let `Q be the tangent line to Ω at Q.
Prove that the circumcircle of the triangle formed by the lines `P , `Q and AB is tangent to Ω.
4.15 Problems 294

Figure 4.105: APMO 2014 P5

Problem 4.15.19 () : Let ABC be a triangle, D, E, F are the feet of the altitudes, DF ∩ BE ≡
P, DE ∩ CF ≡ Q. Prove that the perpendicular from A to P Q goes through the reflection of O on
BC.

projective

Solution. Projective approach.

Problem 4.15.20 (RMM 2018 P6) : Fix a circle Γ, a line ` to tangent Γ, and another circle Ω
disjoint from ` such that Γ and Ω lie on opposite sides of `. The tangents to Γ from a variable point
X on Ω meet ` at Y and Z. Prove that, as X varies over Ω, the circumcircle of XY Z is tangent
to two fixed circles.
4.15 Problems 295

inversion

Solution. Too many circles, plus tangency, what else other than inversion? After the inversion
the problem turns into a pretty obvious work-around problem.

Problem 4.15.21 (AoPS6) : Let O and I be the circumcenter and incenter of ∆ABC. Draw circle
ω so that B, C ∈ ω and ω touches (I) internally at P . AI intersects BC at X. Tangent at X to (I)
which is different from BC, intersects tangent at P to (I) at S. SA ∩ (O) = T 6= A. Prove that
∠AT I = 90◦

Figure 4.106: Solution 1

4.15 Problems 296

Figure 4.107: Solution 2

Problem 4.15.22 (AoPS7) : Let ABC be a triangle with incenter I and circumcircle Γ. Let the
line through I perpendicular to AI meet AB at E and AC at F . Let the circumcircles of triangles
AIB and AIC intersect the circumcircle of triangle AEF ω again at points M and N, and let ω
intersect Γ again at Q. Prove that AQ, MN, and BC are concurrent.
4.15 Problems 297

Problem 4.15.23 (AoPS) : Given a circle (O) with center O and A, B are 2 fixed points on (O).
E lies on AB. C, D are on (O) and CD pass through E. P lies on the ray DA, Q lies on the ray
DB such that E is the midpoint of P Q. Prove that the circle passing through C and touch P Q at
E also pass through the midpoint of AB

Problem 4.15.24 (WenWuGuangHua Mathematics Workshop) : OB , OC are the B and C mixti-

linear centers respectively. (OB ) touches BC, AB at XB , YB respectively, and XB YB ∩ OB OC at ZB .
Define XC , YC , ZC similarly. Prove that if BZC ∩ CZB = T , then AT is the A-angle bisector.

Problem 4.15.25 (All Russia 1999 P9.3) : A triangle ABC is inscribed in a circle S. Let A0 and
C0 be the midpoints of the arcs BC and AB on S, not containing the opposite vertex, respectively.
The circle S1 centered at A0 is tangent to BC, and the circle S2 centered at C0 is tangent to AB.
Prove that the incenter I of 4ABC lies on a common tangent to S1 and S2 .

Problem 4.15.26 (All Russia 2000 P11.7) : A quadrilateral ABCD is circumscribed about a circle
ω. The lines AB and CD meet at O. A circle ω1 is tangent to side BC at K and to the extensions
of sides AB and CD, and a circle ω2 is tangent to side AD at L and to the extensions of sides AB
and CD. Suppose that points O, K, L lie on a line. Prove that the midpoints of BC and AD and
the center of ω also lie on a line.

Problem 4.15.27 (All Russia 2000 P9.3) : Let O be the center of the circumcircle ω of an acute-
angle triangle ABC. A circle ω1 with center K passes through A, O, C and intersects AB at M
and BC at N. Point L is symmetric to K with respect to line NM. Prove that BL ⊥ AC.

Problem 4.15.28 (WenWuGuangHua Mathematics Workshop) : 1. AD, BE, CF are concurrent

cevians. Angle bisectors of ∠ADB and ∠AEB meet at C0 . Again the angle bisectors of
∠ADC and ∠AF C meet at B0 . And bisectors of ∠BEC and ∠BF C meet at A0 . Prove that
AA0 , BB0 , CC0 are concurrent.

2. Angle bisectors of ∠AEB and ∠AF C meet at D0 , of ∠BF C and BDA meet at E0 , and of
∠CEB and ∠CDA meet at F0 . Prove that DD0 , EE0 , F F0 are concurrent.

Solution. As this problem is purely made up with lines, we can do a projective transformation
to simplify the problem. And as there are perpendicularity at D, E, F , we make D, E, F the feet
of the altitudes of 4ABC. Then the angle bisector properties get replaced by simpler properties
wrt DEF .

Problem 4.15.29 (WenWuGuangHua Mathematics Workshop) : Generalization: Let AD, BE, CF

4.15 Problems 298

be any cevians concurrent at T . AD ∩ EF = A0 , BE ∩ DF = B 0 , CF ∩ DE = C 0 , B 0 A0 ∩ AC =

X, B 0 A0 ∩ BC = Y, C 0 X ∩ EF = Z. Prove that T, Y, Z are collinear.

Problem 4.15.30 (AoPS) : On circumcircle of triangle ABC, T and K are midpoints of arcs BC
and BAC respectively . And E is foot of altitude from C on AB . Point P is on extension of AK
such that P E is perpendicular to ET . Prove that P C = CK.

Problem 4.15.31 (USJMO 2018 P3) : Let ABCD be a quadrilateral inscribed in circle ω with
AC ⊥ BD. Let E and F be the reflections of D over lines BA and BC, respectively, and let P be
the intersection of lines BD and EF . Suppose that the circumcircle of 4EP D meets ω at D and
Q, and the circumcircle of 4F P D meets ω at D and R. Show that EQ = F R.

Problem 4.15.32 (All Russia 2002 P11.6) : The diagonals AC and BD of a cyclic quadrilateral
ABCD meet at O. The circumcircles of triangles AOB and COD intersect again at K. Point
L is such that the triangles BLC and AKD are similar and equally oriented. Prove that if the
quadrilateral BLCK is convex, then it has an incircle.

Problem 4.15.33 (WenWuGuangHua Mathematics Workshop) : Let OB , OC be the B, C mixtilin-

ear excircles. O meet CA, CB at XC , YC and OB meet BA, BC at XB , YB . Let IC be the C-excircle.
IC YB meet OB OC at T . Prove that BT ⊥ OB OC

Solution. From what we have to prove, we find two circles, from where we get another circle.
This circle suggests that we try power of point.

Problem 4.15.34 (Iran TST 2018 T1P3) : In triangle ABC let M be the midpoint of BC. Let
ω be a circle inside of ABC and is tangent to AB, AC at E, F , respectively. The tangents from
M to ω meet ω at P, Q such that P and B lie on the same side of AM. Let X ≡ P M ∩ BF and
Y ≡ QM ∩ CE. If 2P M = BC prove that XY is tangent to ω.

Problem 4.15.35 (Iran TST 2018 T1P4) : Let ABC be a triangle (∠A 6= 90◦ ). BE, CF are the
altitudes of the triangle. The bisector of ∠A intersects EF, BC at M, N. Let P be a point such
that MP ⊥ EF and NP ⊥ BC. Prove that AP passes through the midpoint of BC.

Solution. :’3 kala para na T_T

Problem 4.15.36 (Iran TST 2018 T3P6) : Consider quadrilateral ABCD inscribed in circle ω.
4.15 Problems 299

AC ∩ BD = P . E, F lie on sides AB, CD, respectively such that ∠AP E = ∠DP F . Circles ω1 , ω2
are tangent to ω at X, Y respectively and also both tangent to the circumcircle of P EF at P . Prove
that:
EX FX
=
EY FY

Solution. fucking beautiful.

Problem 4.15.37 (ISL 2006 G6) : Circles ω1 and ω2 with centres O1 and O2 are externally tangent
at point D and internally tangent to a circle ω at points E and F respectively. Line t is the common
tangent of ω1 and ω2 at D. Let AB be the diameter of ω perpendicular to t, so that A, E, O1 are
on the same side of t. Prove that lines AO1 , BO2 , EF and t are concurrent.

Problem 4.15.38 (ISL 2006 G7) : In a triangle ABC, let Ma , Mb , Mc be the midpoints of the
sides BC, CA, AB, respectively, and Ta , Tb , Tc be the midpoints of the arcs BC, CA, AB of the
circumcircle of ABC, not containing the vertices’s A, B, C, respectively. For i ∈ a, b, c, let wi be
the circle with Mi Ti as diameter. Let pi be the common external common tangent to the circles
wj and wk (for all i , j, k = a, b, c) such that wi lies on the opposite side of pi than wj and wk do.
Prove that the lines pa , pb , pc form a triangle similar to ABC and find the ratio of similitude

Problem 4.15.39 (ISL 2006 G9) : Points A1 , B1 , C1 are chosen on the sides BC, CA, AB of a
triangle ABC, respectively. The circumcircles of triangles AB1 C1 , BC1 A1 , CA1 B1 intersect the
circumcircle of triangle ABC again at points A2 , B2 , C2 , respectively (A2 6= A, B2 6= B, C2 6= C).
Points A3 , B3 , C3 are symmetric to A1 , B1 , C1 with respect to the midpoints of the sides BC, CA, AB
respectively. Prove that the triangles A2 B2 C2 and A3 B3 C3 are similar.

Solution. In this type of “Miquel’s Point and the intersections of the circumcircles” related prob-
lems, it is useful to think about the second intersections of the lines joining the first intersections
and the Miquel’s Point with the main circle.
4.15 Problems 300

Figure 4.108: IMO Shortlist G9

Problem 4.15.40 (Iran TST 2017 P5) : In triangle ABC, arbitrary points P, Q lie on side BC
such that BP = CQ and P lies between B, Q. The circumcircle of triangle AP Q intersects sides
AB, AC at E, F respectively. The point T is the intersection of EP, F Q. Two lines passing through
the midpoint of BC and parallel to AB and AC, intersect EP and F Q at points X, Y respectively.
4.15 Problems 301

Prove that the circumcircle of triangle T XY and triangle AP Q are tangent to each other.

Problem 4.15.41 () : Let X be the touchpoint of the incircle with BC and let AX meet ·ABC
at D. The tangents from D to the incircle meet ·ABC at E, F . Prove that the tangent to the
circumcircle at A, EF and BC are concurrent.

Problem 4.15.42 (ISL 2012 G8) : Let ABC be a triangle with circumcircle ω and ` a line without
common points with ω. Denote by P the foot of the perpendicular from the center of ω to `. The
side-lines BC, CA, AB intersect ` at the points X, Y, Z different from P . Prove that the circumcircles
of the triangles AXP , BY P and CZP have a common point different from P or are mutually tangent
at P .

Solution. Using Cross ratio and Desergaus’s Involution Theorem.

Problem 4.15.43 () : Suppose an involution on a line l sending X, Y, Z to X 0 , Y 0 , Z 0 . Let lx , ly , lz

be three lines passing through X, Y, Z respectively. And let X0 = ly ∩ lz , Y0 = lx ∩ lz , Z0 = lx ∩ ly .
Then X0 X 0 , Y0 Y 0 , Z0 Z 0 are concurrent.

Problem 4.15.44 (USAMO 2018 P5) : In convex cyclic quadrilateral ABCD, we know that lines
AC and BD intersect at E, lines AB and CD intersect at F , and lines BC and DA intersect at G.
Suppose that the circumcircle of 4ABE intersects line CB at B and P , and the circumcircle of
4ADE intersects line CD at D and Q, where C, B, P, G and C, Q, D, F are collinear in that order.
Prove that if lines F P and GQ intersect at M, then ∠MAC = 90◦ .

Problem 4.15.45 (Japan MO 2017 P3) : Let ABC be an acute-angled triangle with the circum-
center O. Let D, E and F be the feet of the altitudes from A, B and C, respectively, and let M be
the midpoint of BC. AD and EF meet at X, AO and BC meet at Y , and let Z be the midpoint
of XY . Prove that A, Z, M are collinear.

Problem 4.15.46 (ISL 2002 G1) : Let B be a point on a circle S1 , and let A be a point distinct
from B on the tangent at B to S1 . Let C be a point not on S1 such that the line segment AC
meets S1 at two distinct points. Let S2 be the circle touching AC at C and touching S1 at a point
D on the opposite side of AC from B. Prove that the circumcenter of triangle BCD lies on the
circumcircle of triangle ABC.

Problem 4.15.47 (ISL 2002 G2) : Let ABC be a triangle for which there exists an interior point
4.15 Problems 302

F such that ∠AF B = ∠BF C = ∠CF A. Let the lines BF and CF meet the sides AC and AB at
D and E respectively. Prove that
AB + AC ≥ 4DE.

Solution. Pari nai.

Problem 4.15.48 (ISL 2002 G3) : The circle S has center O, and BC is a diameter of S. Let A
be a point of S such that ∠AOB < 120◦ . Let D be the midpoint of the arc AB which does not
contain C. The line through O parallel to DA meets the line AC at I. The perpendicular bisector
of OA meets S at E and at F . Prove that I is the incenter of the triangle CEF.

Problem 4.15.49 (ISL 2002 G4) : Circles S1 and S2 intersect at points P and Q. Distinct points
A1 and B1 (not at P or Q) are selected on S1 . The lines A1 P and B1 P meet S2 again at A2
and B2 respectively, and the lines A1 B1 and A2 B2 meet at C. Prove that, as A1 and B1 vary, the
circumcentres of triangles A1 A2 C all lie on one fixed circle.

Problem 4.15.50 (ISL 2002 G7) : The incircle Ω of the acute-angled triangle ABC is tangent to
its side BC at a point K. Let AD be an altitude of triangle ABC, and let M be the midpoint of
the segment AD. If N is the common point of the circle Ω and the line KM (distinct from K),
then prove that the incircle Ω and the circumcircle of triangle BCN are tangent to each other at
the point N.

Problem 4.15.51 (Japan MO 2017 P3) : Let ABC be an acute-angled triangle with the circum-
center O. Let D, E and F be the feet of the altitudes from A, B and C, respectively, and let M be
the midpoint of BC. AD and EF meet at X, AO and BC meet at Y , and let Z be the midpoint
of XY . Prove that A, Z, M are collinear.

Problem 4.15.52 (India TST) : ABC triangle, D, E, F touchpoints, M midpoint of BC, K ortho-
center of 4AIC, prove that MI ⊥ KD

Problem 4.15.53 (ISL 2009 G3) : Let ABC be a triangle. The incircle of ABC touches the
sides AB and AC at the points Z and Y , respectively. Let G be the point where the lines BY
and CZ meet, and let R and S be points such that the two quadrilaterals BCY R and BCSZ are
parallelogram. Prove that GR = GS.

Solution. Point Circle, distance same means Power same wrt point circles.
4.15 Problems 303

Problem 4.15.54 (ARO 2018 P11.6) : Three diagonals of a regular n-gon prism intersect at an
interior point O. Show that O is the center of the prism.
(The diagonal of the prism is a segment joining two vertices’s not lying on the same face of the
prism.)

Problem 4.15.55 (ISL 2011 G4) : Let ABC be an acute triangle with circumcircle Ω. Let B0 be
the midpoint of AC and let C0 be the midpoint of AB. Let D be the foot of the altitude from A
and let G be the centroid of the triangle ABC. Let ω be a circle through B0 and C0 that is tangent
to the circle Ω at a point X 6= A. Prove that the points D, G and X are collinear.

Problem 4.15.56 () : Given 3 circle, construct another circle that is tangent to these three circles.

Solution. A trick to remember: decreasing the radius’s of some circles doesn’t effect much.

Problem 4.15.57 () : Let ABCD be a convex quadrilateral, let AD ∩ BC = P . Let O, O0 ; H, H 0

be the circumcentres and orthocenter of 4P CD, 4P AB. DOC is tangent to AD0 B, if and only
if DHC is tangent to AH 0 B

Problem 4.15.58 (Iran MO 3rd round 2017 mid-terms Geometry P3) : Let ABC be an acute-
angle triangle. Suppose that M be the midpoint of BC and H be the orthocenter of ABC. Let
F ≡ BH ∩ AC and E ≡ CH ∩ AB. Suppose that X be a point on EF such that ∠XMH = ∠HAM
and A, X are in the distinct side of MH. Prove that AH bisects MX.
4.16 Research Stuffs for later 304

4.16 Research Stuffs for later

Problem 4.16.1 (AoPS) : Let ABC be a triangle with incenter I. La , Lb , Lc are symmedian points
of triangles IBC, ICA, IAB. Let X, Y, Z be the reflections of I through La , Lb , Lc .

• Prove that AX, BY, CZ and OI are concurrent.

• Let Ia , Ib , Ic be the excenters of ABC. Prove that Ia X, Ib Y, Ic Z are concurrent at a point P

and isogonal conjugate of P with respect to triangle Ia Ib Ic lies on Euler line of ABC.

Problem 4.16.2 (buratinogigle Tough P1) : Let ABC be a triangle inscribed in circle (O) with
A-excircle (J). Circle passing through A, B touches (J) at M. Circle passing through A, C touches
(J) at N. BM cuts CN at P . Prove that AP passes through tangent point of A-mixtilinear incircle
with (O).
4.17 Big Pictures 305

4.17 Big Pictures

4.17.1 Parallel lines to the A-median 306

4.17.1 Parallel lines to the A-median

Definition (Related Points) — Let M be the midpoint of BC.

Let the lines through B, C parallel to AM meet ABC again at B 0 , C 0 .
Let R be the point on BC such that B 0 RC 0 is tangent to BC
Let BC ∩ B 0 C 0 = S.
Let Ma , MA be the minor and major arc midpoints of BC.d
Let MA R ∩ AM = H, MA R ∩ ABC = G.
Let ωb , ωc be the circles through BB 0 and CC 0 and tangent to BC.
Let I be the intersection of the line through Ma parallel to AM and ωx .
Let J = AM ∩ ωx .
Let J ∩ ωb = X, J ∩ ωc = Y .
Let AS ∩ ABC = F .

Figure 4.109
4.17.1 Parallel lines to the A-median 307

Lemma 4.17.1 — SG is tangent to ABC

Lemma 4.17.2 — R, M, Ma , G is cyclic, call it ωx

Lemma 4.17.3 — I, G, O are collinear, and IG = IJ

Lemma 4.17.4 — BX, CY, AM, RMA are concurrent.

Problem 4.17.1 (IGO 2019 Advanced P4) : XHY G is cyclic, and tangent to ωb , ωc and ABC.
And I is the center of this circle.

Figure 4.110
4.17.1 Parallel lines to the A-median 308

Problem 4.17.2 (APMO 2019 P3) : A variable point P is selected in the line segment AM. The
circumcircles of triangles BP M and CP M intersect Γ again at points D and E, respectively. The
lines DP and EP intersect (a second time) the circumcircles to triangles CP M and BP M at X and
Y , respectively. Prove that AF XY is cyclic.
Chapter 5

Number Theory
5.1 Tricks 310

5.1 Tricks

“Every object in the problem’s universe is important, and should be considered when approaching
the problem.” – me

• permutation type problem

• do there exist...

• proving identities

• Dunno

• divisibility by primes and prime divisors stuff

1. Add. Everything. Up.

2. Infinitude of primes:

a) Eulerian infinitude trick

b) For large enough numbers, there is a larger prime divisor
c) Assuming contradiction, if there are any number co-prime to the product of the primes,
then that must be 1.

5.1.1 Digit Sum or Product

When dealing with the sum of the digits or the product of them, to find the construction it is very
important to consider 0 and 1’s in the number.
5.1.2 Diophantine Equations 311

5.1.2 Diophantine Equations

1. finding some solutions

2. trying modular cases
3. making some variables depended on other variables
4. putting constrains on variables which would make the problems easier
5. if there are infinitely many solutions, can you find a construction?
6. factorize (this is BIG)
7. In these problems, investigation, induction, recursion, constructions etc. are essentials

5.1.3 Sequences

5.1.4 NT Functions

5.1.5 Construction Problems

5.1.6 Sets satisfying certain properties

5.1.7 Other Small Techniques to Remember

1. a − b stays invariant upon addition, just as a

b stays invariant upon multiplication.
5.2 Lemmas 312

5.2 Lemmas

Lemma 5.2.1 ( Maybe not useful at all) — Let p ≥ 5 be a prime number. Prove that if
p | a2 + ab + b2 , then
p 3 | a + b p − ap − b p

Lemma 5.2.2 ( Non-zero digits in base b) — If b ≥ 2 and bn − 1 | a then there exist at least n
non-zero digits in the representation of a in base b

Theorem 5.2.3 (Frobenius Coin Problem, extended Chicken McNugget) — Given a bunch of
coprime integers, one can write all integers after a certain limit as a linear combination of these
integers.

Formally, for any integers a1 , a2 . . . an such that gcd(a1 , a2 . . . an ) = 1, there exists positive
integers m such that for any integer M ≥ m, there are non-negative integers b1 , b2 . . . bn such
that
Xn
bi ai = M
i=1

Remark. If n = 2 then m = a1 a2 − a1 − a2 .

If n ≥ 2, then there doesn’t exist an explicit formula, but if {ai } are in arithmetic progression,
(ai = a1 + (i − 1)d) then  
a−2
m= a + (d − 1)(a − 1) − 1
n−1

Theorem 5.2.1
1 1
Beatty’s Theorem or Rayleigh Theorem If a, b are two irrational numbers such that + = 1,
a b
then the two sets {bi ac} and {bi bc}, where i are the positive integers, form a partition of the
set of natural numbers.

Remark. Express the two conditions in numeric terms assuming the contrary.

Proof. No integer belongs to both sets: Suppose there exists some j, k, m such that j = bkac =
5.2 Lemmas 313

bmbc. Then the inequalities will hold:

j ≤ ka < j + 1 and j ≤ mb < j + 1

Which leads to
j < ka < j + 1 and j < mb < j + 1
From there, it follows that
j j +1 j j +1
<k < and < m <
a a b b
Adding them gives
j <k +m <j +1
Which is false.

Every integer belongs to one set: Suppose there is are integers j, k, m such that bkac < j <
bka + ac and bmbc < j < bmb + bc Which can be written as

ka < j < ka + a − 1 and mb < j < mb + b − 1

It follows from here that,

j 1 j 1
k< < k + 1 − and m < < m + 1 −
a a b b
Adding them gives us
k +m <j <k +m+1
Which, again, is absurd. So the theorem is proved.

Lemma 5.2.4 — Let x, y be co-prime. Then

gcd(z , xy ) = gcd(z , x) gcd(z , y ) = gcd(z mod x, x) gcd(z mod y , y )

=⇒ gcd(r (a, b), xy )) = gcd(a, x) gcd(b, y )

(here r (a, b) denotes the smallest integer that satisfies r (a, b) ≡ a mod x, r (a, b) ≡ b mod y )

Lemma 5.2.5 — Let 0 < a1 < a2 < < a( mn + 1) be mn + 1 integers. Prove that you can
select either m + 1 of them no one of which divides any other, or n + 1 of them each dividing the
following one.
5.2 Lemmas 314

Theorem 5.2.6 (Prime divisors of an integer polynomial) — If P (x) ∈ Z[x], then the set of
primes, P = {p : p | P (x)} is infinite.
5.2.1 Modular Arithmatic Theorems and Useful Results 315

5.2.1 Modular Arithmatic Theorems and Useful Results

Theorem 5.2.7 (Wolstenholme’s Theorem) — For all prime p the following relation is true:
p−1
2 1 1 1 X1
p 1 + + ... =
2 3 p−1 i
i=1

Corollary 5.2.7.1 —
p−1
1 1 1 X 1
p 1 + 2 + 2 ... =
2 3 (p − 1)2 i2
i=1

Corollary 5.2.7.2 — If p > 3 is a prime, then

 
2p
≡ 2 (mod p 3 )
p

5  
2p − 1
≡ 1 (mod p 3 )
p−1

Lemma 5.2.8 —
1
≡ (−1)i (i − 1)! (mod p)
(p − i )!
 
p−1
≡ (−1)k (mod p)
k

Problem 5.2.1 () :
p n+1
 
≡ p n (mod p 2n+3 )
p
5.3 Orders Modulo a Prime and Related Stuffs 316

5.3 Orders Modulo a Prime and Related Stuffs

• Order Modulo a Prime - Evan Chen

• Another source of the proof of Zsigmondy’s Theorem

5.3.1 Cyclotomic Polynomials

• Elementary Properties of Cyclotomic Polynomials - Yimin Ge

Definition (Cyclotomic Formulas) — For any integer n, we have

Y
Xn − 1 = Φd (X)
d|n

In particular, if p is a prime then

Xp − 1
Φp (X) = = X p−1 + X p−2 + · · · + 1
X−1

Definition (Mobius Function) — The Mobius function µ maps the natural numbers to the set
{−1, 0, 1}. µ can be defined in multiple ways:
X
• Let ζi be the primitive roots of nth cyclotomic polynomial, then µ(n) = ζi

•
if n is square free and has an even number of prime divisors

1

µ(n) = −1 if n is square free and has an odd number of prime divisors
if n has a prime square divisor

0

5.3.1 Cyclotomic Polynomials 317

Lemma 5.3.1 ( Sum of Mobius functions of divisors) — Let n be a positive integer. Then
(
X 1 if n = 1
µ(d) =
d|n
0 if n ≥ 2

Proof. The divisors of n with a prime square divisor will contribute 0. If n = 1 it’s trivial. So
let n = p1α1 p2α2 . . . pkαk . Then,
X X
µ(d) = pb1 . . . pbt
d|n {b1 ...bt }⊆{1,2...k}

     
k k k k
= − + · · · + (−1)
0 1 k

=0

Theorem 5.3.1
Mobius Inverison Formula Suppose F, f : Z+ → Z+ are functions such that
X X
µ(d) F dn


F (n) = f (d) =⇒ f (n) =
d|n d|n

n µ(d)
Y Y

F (n) = f (d) =⇒ f (n) = F d
d|n d|n

Proof.
X X X X
µ(d) f (t) = f (t) µ(d)
d|n n t|n n
t| d d|n, t| d
X X
= f (t) µ(d)
t|n n
d| t

= f (n) [∵ Lemma 5.3.1]

5.3.2 Quadratic Residue 318

Lemma 5.3.2 ( Prime Divisors of Cyclotomic Polynomials) — If a ∈ Z such that Φn (a) 6= 0 and
for some prime p,
Φn (a) ≡ 0 (mod p)
Then either

• p ≡ 1 (mod n), or

• p|n

Lemma 5.3.3 — If p does not divide m, then

Φpm (x)Φm (x) = Φm (x p )

5.3.2 Quadratic Residue

Theorem 5.3.4 (Quadratic Reciprocity) — Let p and q be two prime numbers. Then using
Legendre symbol, we have   
p q p−1 q−1
= (−1) 2 2
q p

Lemma 5.3.5 ( 2 is a quadratic residue) —

p 2 −1
 
2
= (−1) 8
p

p−1
Proof. Take a 6≡ 0 (mod p). Let ti ≡ i a (mod p) the least positive residues for 1 ≤ i ≤ .
2
p−1
Let r1 , r2 . . . rm be the ti ’s that are smaller than , and let s1 , s2 . . . sn be the ti ’s that are
2
p−1
larger than . Then we have,
2
p−1
{r1 , r2 , . . . rm , p − s1 , p − s2 , . . . p − sn } = {1, 2, . . . }
2
So, multiplying gives us,  
p−1 a
n
a 2 ≡ (−1) = (mod p)
p
Plugging in a = 2, we get Lemma 5.3.5.
5.3.2 Quadratic Residue 319

Lemma 5.3.6 ( Quadratic non-residue mod infitely many primes) — There are infinitely many
primes p for every non-square integers a such that a is a non-quadratic residue (mod p)

Proof. We can assume that a is square free. So a = 2s p1 p2 . . . pr for some

s ∈ {0, 1} and r ≥ 0. Assume the contrary, that there are only finitely many primes,
namely q1 , q2 . . . qm for which a is a quadratic non residue. We have three cases:

Case 1: s = 1, r = 0. In that case, we have infinitely many prime number, q such that q ≡ 3
(mod 8), and x 2 6≡ a (mod q) using Lemma 5.3.5.

Case 2: s = 1, r > 0. Take t, a quadratic non-residue (mod p)r . Consider the following
congruences:

x ≡1 (mod qi ) for all qi

x ≡1 (mod 8)
x ≡1 (mod pi ) for i = 1, 2 . . . r − 1
x ≡t (mod pr )

By Chinese Remainder Theorem and Dirichlet theorem, it follows that there exists infinitely many
prime numbers N that satisfies all of the congruences. Now using Jacobi symbol and Law of
Reciprocity, it follows that,
 
2
=1
N
p 
i
= 1 for i = 1, 2 . . . r − 1
N
q 
i
= 1 for all qi
N
p 
r
= −1
N
a
Therefore, = −1 Which is a contradiction, since qi 6 |N.
N
Case 3: s = 0, r > 0, this is similar to the above case.
5.3.2 Quadratic Residue 320

5.3.2.1 Quadratic Residue

Theorem 5.3.7 () — Let p be an odd prime. Then,

1. The product of two quadratic residue is a quadratic residue.

2. The product of two quadratic non-residue is a quadratic residue.

3. The product of a qudratic residue and a quadratic non-residue is a quadratic non-residue

    
Theorem 5.3.8 () — For an odd prime p and any two integers a, b, we have ab
p = a
p
b
p

Theorem 5.3.9 (Euler’s Criterion) — Let p be an odd prime. Then,

 
p−1 a
a 2 ≡ (mod p)
p

Theorem 5.3.10 () — Let (a, b) = 1. Then every prime divisors of a2 + b2 is either 2 or a prime
of the form 4k + 1.

Theorem 5.3.11 (Gauss’s Criterion) — Let p be a prime number and a be an integer coprime to
p. Let µ(a) be the number of integers x ∈ {a, a ∗ 2, . . . a ∗ p−1 p
2 } such that x(mod p) > 2 . Then
 
a
= −1µ(a)
p

Theorem 5.3.12 () — The smallest quadratic non-residue of an odd prime p is a prime which is
√
less than p + 1

Theorem 5.3.13 (Quadratic Residue Law) — Using the usual Legendre Symbol, for two prime
numbers p, q we have:   
p q p−1 q−1
= (−1) 2 2
q p
5.3.3 Zsigmondy’s Theorem 321

Definition (Jacobi Symbol) — Let a, n = p1α1 . . . pkαk . We define Jacobi symbol as

a k  αi
Y a
=
n pi
i=1

Note. Jacobi symbol is not as accurate as Legendre symbol. na = −1 means that a is a quadratic

non-residue of n, but = 1 doesnt necessarily mean that a is a quadratic residue of n.

Theorem 5.3.14 () — Let a, n = p1α1 . . . pkαk , then a is a quadratic residue of n iff it is a

quaadratic residue of every piαi .

Theorem 5.3.15 () — If an integer is a quadratic residue of every prime, then it is a square.

5.3.3 Zsigmondy’s Theorem

• Zsigmondy’s Theorem’s Proof, has some useful lemmas

Theorem 5.3.16 (Zsigmondy’s Theorem) — Let a, b ∈ N such that gcd(a, b) = 1 and n ∈

N, n > 1. Then there exists a prime division of an − bn that does not divide ak − bk for all
1 ≤ k < n except in the following cases:

• 2n − 1n
• n = 2 and a + b is a power of 2.

5.3.4 Problems

Problem 5.3.1 (CGMO 2016 P3) : Let m and n are relatively prime integers and m > 1, n > 1.
Show that:There are positive integers a, b, c such that ma = 1 + nb c , and n and c are relatively
prime.

Problem 5.3.2 (ISL 2004 N4) : Let k be a fixed integer greater than 1, and let m = 4k 2 − 5.
5.3.4 Problems 322

Show that there exist positive integers a and b such that the sequence (xn ) defined by

x0 = a, x1 = b, xn+2 = xn+1 + xn for n = 0, 1, 2, . . . ,

has all of its terms relatively prime to m.

Solution. Let’s play with some integer n and Fibonacci sequences mod n. What should we
take the value of n? As 11 = 16 − 5, let’s take it first. We see that the period of Fibonacci
sequences mod 11 is at most 10. From here it is natural to make a conjecture that for prime
n’s, the period probably is n − 1. We also design a proof that there is a sequence which doesn’t
contain any of n’s products.

So let’s see if it works for all primes. No it doesn’t, breaks at 7. How is 7 so different than
11? The most straightforward guess is that probably 7 6 |4k 2 − 5 for any k. And it is true. So
what’s so special about the primes dividing 4k 2 − 5? Writing it in modular
√ arithmetic manner,
4k 2 ≡ 5 (mod p). Wait, 5 is a quadratic residue mod p? But isn’t 5 related to Fibonacci
sequences? What’s the general formula for a Fibonacci sequence starting with a, b? Wait, now
that explains why the period of the Fibonacci sequences mod these primes is p − 1.
5.4 Primes 323

5.4 Primes

Problem 5.4.1 (ISL 2013 N5) : Fix an integer k > 2. Two players, called Ana and Banana, play
the following game of numbers. Initially, some integer n ≥ k gets written on the blackboard. Then
they take moves in turn, with Ana beginning. A player making a move erases the number m just
written on the blackboard and replaces it by some number m0 with k ≤ m0 < m that is coprime to
m. The first player who cannot move anymore loses.
An integer n ≥ k is called good if Banana has a winning strategy when the initial number is n, and
bad otherwise.
Consider two integers n, n0 ≥ k with the property that each prime number p ≤ k divides n if and
only if it divides n0 . Prove that either both n and n0 are good or both are bad.

Solution. Every idea that naturally follows lead to a solution, so after getting the idea of
working on a single equivalence class is enough, we face the problem that “big primes” cause the
trouble. So can we get rid of them by making some minimal number that arent “contaminated”
by big primes?

Problem 5.4.2 (ISL 2014 N4) : Let n > 1 be a given integer. Prove that infinitely many terms of
the sequence (ak )k≥1 , defined by
 k
n
ak = ,
k
are odd. (For a real number x, bxc denotes the largest integer not exceeding x.)

Solution. First we take a prime, doesnt work, then we take two primes, one being 2 (Since we
need it in the bottom), but that doesn’t work either. Then take n instead of 2, because we want
the 2’s in the numerator vanish. Surprisingly this works.

Problem 5.4.3 (ISL 2012 N3) : Determine all integers m ≥ 2 such that every n with m
3 ≤n≤ m
2
n
divides the binomial coefficient m−2n .

n
Solution. Investigate, and find out when n| .


m−3n

Problem 5.4.4 (ISL 2004 N1) : Let τ (n) denote the number of positive divisors of the positive
integer n. Prove that there exist infinitely many positive integers a such that the equation τ (an) = n
does not have a positive integer solution n.
5.4 Primes 324

Solution. if p 6 |n, τ (pn) = τ (p)τ (n) = 2τ (n) = n which is true for n

let’s try p k , and p t ||n,
τ (p k+t m) = (k + t + 1)τ (m) = p t m
for k + 1 < p, if t = 0, we don’t really have an obvious contradiction.
So what if k + 1 = p, then, t = 0, 1
if t = 0, pτ (m) = m but (p, m) = 1
and if t = 1,
(p + 1)τ (m) = pm
but trivial bounding shows that m <= 4, for which there is no solution by case check

Problem 5.4.5 (USA TST 2014 P2) : Let a1 , a2 , a3 , . . . be a sequence of integers, with the property
that every consecutive group of ai ’s averages to a perfect square. More precisely, for every positive
integers n and k, the quantity
an + an+1 + · · · + an+k−1
k
is always the square of an integer. Prove that the sequence must be constant (all ai are equal to
the same perfect square).

Solution. The most obvious fact of this sequence is that every element has to be the same
residue mod every prime. Using p = 3 we see that if one element is divisible by p, every other
elements are divisible by p as well. As this is true, we try to prove it. Again we use the most
obvious facts we can get from the sequence.

n
Problem 5.4.6 (ISL 2012 N6) : Let x and y be positive integers. If x 2 − 1 is divisible by 2n y + 1
for every positive integer n, prove that x = 1.

n
Solution. First notice the important things about the primes dividing bn = x 2 − 1. Another
important thing to notice is that if p|bi , then or dp (bi ) = 2j . So we get another bound for the
primes.
The most natural thing is to ask now if the primes that have the property 2||p − 1 are infinite,
because then we will have an obvious contradiction. Which leads to the following lemma.
Another approach is to taking a N such that aN = 2N y + 1 becomes congruent to a1 , and show
a1 is actually congruent to some constant, but since N is arbitrary, we are done.

Lemma 5.4.1 ( Primes in a recursive sequence) — Let an = 2n y + 1 be a sequence of positive

intergers. Prove that there are infinitely many prime numbers such that p ≡ −1 (mod p) and
p|ai for some i .
5.4 Primes 325

Solution. Somehow we want to deploy Euclid. Let’s first group the desired primes in a set T .
So we ask ourselves what is the most natural thing for an ≡ 1 (mod 4) for some n to have a
prime divisor congruent to −1 mod 4? If we can factor an into congruent to −1 mod 4 parts.
We notice that a1 ≡ −1 (mod 4). So that should be a good starting point.
an an
Now we want to find some n such that a1 |an and p ∈ T, p 6 | , or wishfully, p| − 1
a1 a1

Problem 5.4.7 (China TST 2018 T2P4) : Let k, M be positive integers such that k − 1 is not
squarefree. Prove that there exist a positive real α, such that bα · k n c and M are coprime for any
positive integer n.

Solution. Think about what α actually represent in the numberline. It’s the ratio that represents
how a coprime number of M is from k n . We think wishfully and hope that α is something like
1 1
A+B . Why ? Because this number is “nice” when multiplied by k n , because it lets
k −1 k −1
us strech the “same ratio” motivation for all k n . With a bit of workaround to find suitable values
for A, B is needed.

Problem 5.4.8 (China TST 2013 T2P2) : Prove that: there exists a positive constant K, and an
integer series {an }, satisfying:

1. 0 < a1 < a2 < · · · < an < · · ·

2. For any positive integer n, an < 1.01n K
3. For any finite number of distinct terms in {an }, their sum is not a perfect square

Remark. The basic idea is to make any sum of elements have a prime divisor with odd power. We
want an extension of the set {p, p 3 , p 5 . . . } for some prime p.

Solution. Take a prime number p which satisfies

p 2 < 1.01p
2
Such prime numbers exist because the function f (x) = x p is strictly decreasing.
Now let i = pk + j, and let
ai = apk+j = jp 2k+2 + p 2k+1
And let K = p 3 . Since p 2 < 1.01p , we have

ai < p 2k+3 = (p 2 )k · p 3 < 1.01pk p 3 < 1.01i K

Also, it’s easy to check that sum of every finite subset of this set has an odd power of p, so no
sum is a perfect square.
5.5 NT Functions and Polynomials 326

5.5 NT Functions and Polynomials

Problem 5.5.1 (ISL 2013 N1) : Let Z>0 be the set of positive integers. Find all functions f :
Z>0 → Z>0 such that
m2 + f (n) | mf (m) + n
for all positive integers m and n.

Solution. Go with the flow.

Problem 5.5.2 (ISL 2010 N5) : Find all functions g : N → N such that
(g(m) + n) (g(n) + m)
is a perfect square for all m, n ∈ N.

Solution. Playig around with some primes give us that for every “big” primes, we need to have
every residue class present in the range of g. Now with this fact, we can prove the injectivity as
well. Now we want to show that g(n + 1) = g(n) + 1. How to show that? We can show that
by saying that no prime p exits such that p|g(n + 1) − g(n), in other words, again the residue
classes.

Problem 5.5.3 (ISL 2004 N3) : Find all functions f : N → N satisfying

 
2
f (m)2 + f (n) | m2 + n

for any two positive integers m and n.

Solution. First get f (1) = 1, then using casework, get f (p − 1) = p − 1. Then let m = (p − 1),
and use the division algorithm to reduce the dividend.

Remark. In nt-fe s with “divisible" condition, one occurring idea is to get an infinite set of
integers with the desired output value, and somehow keep those numbers only in the divisor, and
make them vanish from the dividend. Most of the time, using division algorithm to reduce the
dividend does the trick.

Problem 5.5.4 (ISL 2009 N5) : Let P (x) be a non-constant polynomial with integer coefficients.
Prove that there is no function T from the set of integers into the set of integers such that the
number of integers x with T n (x) = x is equal to P (n) for every n ≥ 1, where T n denotes the n-fold
application of T .
5.5 NT Functions and Polynomials 327

Solution. Write down what the problem is actually saying. Integer functions are like cycles,
use that.
5.6 Diophantine Equations 328

5.6 Diophantine Equations

Problem 5.6.1 (APMO 1999 P4) : Determine all pairs (a, b) of integers with the property that
the numbers a2 + 4b and b2 + 4a are both perfect squares.

Solution. Easy case work assuming positive or negative values for a, b.

Problem 5.6.2 (All Squares) : Prove that there are infinitely many pairs of positive integers (x, y )
satisfying that x + y , x − y , xy + 1 are all perfect squares.

Solution. Just work it out.

Problem 5.6.3 (ISL 2010 N3) : Find the smallest number n such that there exist polynomials
f1 , f2 , . . . , fn with rational coefficients satisfying

x 2 + 7 = f1 (x)2 + f2 (x)2 + . . . + fn (x)2 .

Solution. Find the obvious answer, which is very small, so we can probably case work it out.
We find the case for 3 a bit challenging. But we plan to show that 7a2 can’t be written as a sum
of 3 squares. Which numbers can be written as the sum of 3 squares? Investigate...

Problem 5.6.4 (ISL 2014 N5) : Find all triples (p, x, y ) consisting of a prime number p and two
positive integers x and y such that x p−1 + y and x + y p−1 are both powers of p.

Solution. The old school trick, replacement. Assume x < y and replace y , and then compare
the power of p.
5.7 Divisibility 329

5.7 Divisibility

Problem 5.7.1 (ISL 2002 N6) : Find all pairs of positive integers m, n ≥ 3 for which there exist
infinitely many positive integers a such that
am + a − 1
an + a2 − 1
is itself an integer.

Solution [Manipulation].

Lemma— For a, m, n ∈ N, we have

an + a2 − 1 | am−i + (−1)i an−1 − an−2 + · · · + (−1)i+1 an−i + (−1)i (a − 1)

Proof. We proceed by induction on i . For i = 0, it is true. So assume for i < k, it is true.

an + a2 − 1 | am−i + (−1)i an−1 − an−2 + · · · + (−1)i+1 an−i + (−1)i (a − 1)

=⇒ an + a2 − 1 | am−i + (−1)i an−1 − an−2 + · · · + (−1)i+1 an−i + (−1)i (a − 1) − (−1)i (an + a2 − 1)

= am−i + (−1)i −an + an−1 − an−2 + · · · + (−1)i+1 an−i + (−1)i+1 (a2 − a)

∴ an + a2 − 1 | am−i−1 + (−1)i+1 an−1 − an−2 + · · · + (−1)i+1 an−i + (−1)i+1 (a − 1)

For i = n, we have,

an + 1
 
n 2 m−n n
a +a −1 | a + (−1) + (−1)n (a − 1)
a+1
Clearly, m ≥ n. Let m = nk + q. If n is even,
an + 1
 
n 2 m−n
a +a −1 | a +a−1+
a+1
 n 
n 2 q a +1
=⇒ a + a + 1 | a + a − 1 + k
a+1
Which is not possible since the polynomial on the right side has degree at most n − 1, and can’t
be 0 for all a ∈ Z.
5.7 Divisibility 330

So, n is odd. Then we have,

an + 1
 
an + a2 − 1 | am−n + a − 1 − − 2(a − 1)
a+1
 n 
a +1
=⇒ an + a2 − 1 | aq + a − 1 − k − 2k(a − 1) = P (a)
a+1

Since P (a) has degree at most n − 1, P (a) = 0 for all a ∈ Z.

an + 1
 
q
P (a) = a + a − 1 − k − 2k(a − 1)
a+1
 n 
q a +1
∴a +a−1=k + 2k(a − 1)
a+1

Comparing the coefficients and degrees on both sides, we have, q = n − 1 = 2, k = 1. Which

gives us the only solution (m, n) = (5, 3).

Problem 5.7.2 (Iran 3rd Round 2016 N3) : A sequence P = {an } is called a Permutation of natural
numbers (positive integers) if for any natural number m, there exists a unique natural number n
such that an = m.
We also define Sk (P ) as: Sk (P ) = a1 + a2 + · · · + ak (the sum of the first k elements of the
sequence).
Prove that there exists infinitely many distinct Permutations of natural numbers like P1 , P2 , · · · such
that:
∀k, ∀i < j : Sk (Pi )|Sk (Pj )

Solution. Instead of giving a construction for the sequence we prove that for a given
permutation P we can find another permutation Q such that the partial sums of P divide the
corresponding partial sums of Q.

As we try to build Q from P , we have a constraint of divisibility. And we need to make sure
every integer i gets to be in Q. For that to be always possible, we need a special property of P
to be true. Finding out that property is the main task of this problem.

After the property is determined, we need to add in some more details, that is, for the induction
to work, we need to maintain that property in Q as well.
5.8 Problems 331

5.8 Problems

Problem 5.8.1 () : Let n be an odd integer, and let S = {x | 1 ≤ x ≤ n, (x, n) = (x + 1, n) = 1}.

Prove that Y
x = 1 (mod n)
x∈S

Problem 5.8.2 (Iran TST 2015, P4) : Let n is a fixed natural number. Find the least k such that
for every set A of k natural numbers, there exists a subset of A with an even number of elements
which the sum of its members is divisible by n.

Solution. Odd-Even, so lets first try for odd n’s. It is quite easy.

So now, for evens, lets first try the simplest kind of evens. As we need a set with an even number
of elements, this tells us to pair things up. We can try to partition A into pairs of e-e’s and o-o’s.
This gives us our desired result.

Problem 5.8.3 (Iran TST 2015 P11) : We call a permutation (a1 , a2 . . . an ) of the set {1, 2 . . . n}
“good" if for any three natural numbers i < j < k,

n - ai + ak − 2aj

find all natural numbers n ≥ 3 such that there exist a “good" permutation of the set {1, 2 . . . n}.

Solution. Looking for “possibilities” for the first element, we get some more restrictions for the
values of other terms.

Pn
Problem 5.8.4 (ISL 2004 N2) : The function f : N → N satisfies f (n) = k=1 gcd(k, n), n∈N.

1. Prove that f (mn) = f (m)f (n) for every two relatively prime m, n ∈ N.

2. Prove that for each a ∈ N the equation f (x) = ax has a solution.

3. Find all a ∈ N such that the equation f (x) = ax has a unique solution.

Solution. Why not casually try to multiply f (m) and f (n)?? And also find a formula for
n = prime power.
5.8 Problems 332

Problem 5.8.5 (Balkan MO 2017 P1) : Find all ordered pairs of positive integers (x, y ) such that:
x 3 + y 3 = x 2 + 42xy + y 2 .

Problem 5.8.6 (Balkan MO 2017 P3) : Find all functions f : N → N satisfying

n + f (m) | f (n) + nf (m)
for all n, m ∈ N.

Solution. Check sizes and bound for large n.

Problem 5.8.7 (Iran MO 3rd Round N3) : Let p > 5 be a prime number and A = {b1 , b2 . . . b p−1 }
2
be the set of all quadratic residues modulo p, excluding zero. Prove that there doesn’t exist any
natural a, c satisfying gcd(ac, p) = 1 such that set B = {x | x = ay + c, y ∈ A} and set A are
disjoint modulo p.

Solution. Sum it up.

Solution. For every integer a, b and prime p such that, gcd(a, p) = gcd(b, p) = 1, there exist
(x, y ) such that x 2 ≡ ay 2 + c (mod p).

Solution. For a prime p, there exists an integer x such that x and x + 1 both are quadratic
residues (mod p).

Problem 5.8.8 (All Russia 2014 P9.5) : Define m(n) to be the greatest proper natural divisor of
n. Find all n ∈ N such that n + m(n) is a power of 10.

Problem 5.8.9 (ISL 2000 N1) : Determine all positive integers n ≥ 2 that satisfy the following
condition: for all a and b relatively prime to n we have a ≡ b (mod n) iff ab ≡ 1 (mod n).

Solution. Don’t forget the details.

Problem 5.8.10 (ISL 2000 N3) : Does there exist a positive integer n such that n has exactly 2000
prime divisors (not necessarily distinct) and n | 2n + 1?
5.8 Problems 333

Solution. Goriber Bondhu Induction. As the number 2000 seems so out of the place, we
replace 2000 by k. Now suppose that for some k, the condition works. For simplicity let k = p i
for some i , as it is quite clear that there is another prime q that divides 2k + 1, let k 0 = kq. So
k 0 also satisfies the condition. So it is quite intuitive to think that for every x there exist some p
i
and i for which 2p + 1 has x prime factors. So we search for such p.

Problem 5.8.11 (USAMO 2001 P5) : Let S be a set of integers (not necessarily positive) such
that

1. There exist a, b ∈ S with gcd(a, b) = gcd(a − 2, b − 2) = 1;

2. If x and y are elements of S (possibly equal), then x 2 − y ∈ S

Prove that S is the set of all integers.

Solution. One possible intuition could be trying to make the problem statement a little bit
more stable, like the term x 2 − y is not so symmetric. So trying to make it a little bit more
symmetric can come handy.

Solution. If c, x, y ∈ S then we can easily see that A(x 2 − y 2 ) − c ∈ S for all A ∈ Z. We

take this a little too far and show that if c, x, y , u, v ∈ S, then A(x 2 − y 2 ) + B(u 2 − v 2 ) − c ∈ S
for all A, B ∈ Z . So if we can find such x, y , u, v such that gcd(x 2 − y 2 , u 2 − v 2 ) = 1, we are
almost done by Frobenius Coin Problem. So we start looking for integers that can be obtained
from a, b. After some playing around we get the feeling (or maybe not) that we need one more
pair. Again playing around for some time we find three pairs. FCP gives us an upper bound for
all integers that are not in S. Easily we include them in S.

2n
Problem 5.8.12 (Vietnam TST 2017 P2) : For each positive integer n, set xn =


n

2017k
1. Prove that if 2 < n < 2017k for some positive integer k then 2017 | xn .

2. Find all positive integer h > 1 such that there exist positive integers N, T such that the
sequence (xn ) for n > N, is periodic (bmodh) with period T .

Problem 5.8.13 (Vietnam 2017 TST P6) : For each integer n > 0, a permutation (a1 , a2 . . . a2n )
of 1, 2 . . . 2n is called beautiful if for every 1 ≤ i < j ≤ 2n, ai + an+i = 2n + 1 and ai − ai+1 6≡
aj − aj+1 (mod 2n + 1) (suppose that ai = a2n+i ).

1. For n = 6, point out a beautiful permutation.

5.8 Problems 334

2. Prove that there exists a beautiful permutation for every n.

Solution. Trial and Error.

Problem 5.8.14 (BrMO 2008) : Find all sequences ai ∞

i=0 of rational numbers which follow the
following conditions:

1. an = 2an−1
2 − 1 for all n > 0

2. ai = aj for some i , j > 0, i 6= j

Solution. Trial and Error. Don’t forget that you only need the numbers to be rational.

Problem 5.8.15 (USA TST 2000 P4) : Let n be a positive integer. Prove that
n  −1 n+1
!
X n n + 1 X 2i
= n+1
i 2 i
i=0 i=0

Solution. *Positive Integer*, nuff said.

Problem 5.8.16 (USA TST 2000 P3) : Let p be a prime number. For integers r, s such that
r s(r 2 − s 2 ) is not divisible by p, let f (r, s) denote the number of integers 1 ≤ n ≤ p such that { rpn }
and { snp } are either both less than 2 or both greater than 2 . Prove that there exists N > 0 such
1 1

that for p ≥ N and all r, s,    

(p − 1) 2(p − 1)
≤ f (r, s) ≤
3 3

n
Problem 5.8.17 (China TST 2005) : Let n be a positive integer and fn = 22 + 1. Prove that
for all n ≥ 3, there exists a prime factor of fn which is larger than 2n+2 (n + 1) [ Stronger Version:
2n+4 (n + 1) ].

Problem 5.8.18 (IRAN TST 2009 P2) : Let a be a fixed natural number. Prove that the set of
n
prime divisors of 22 + a for n = 1, 2, 3 . . . is infinite.

Problem 5.8.19 (USAMO 2004 P2) : Suppose a1 , a2 . . . an are integers whose greatest common
divisor is 1. Let S be a set of integers with the following properties:
5.8 Problems 335

1. ai ∈ S

2. For i, j = 1, 2 . . . n (not necessarily distinct), ai − aj ∈ S.

3. For any integers x, y ∈ S, if x + y ∈ S, then x − y ∈ S.

Prove that S must be equal to the set of all integers.

Solution. First we see that if d = gcd(x, y ) and x, y ∈ S then d ∈ S. So all we have to do is

to find two x, y with d = 1.

Problem 5.8.20 (USAMO 2008 P1) : Prove that for each positive integer n, there are pairwise
relatively prime integers k0 , k1 . . . kn , all strictly greater than 1, such that k0 k1 . . . kn−1 is the product
of two consecutive integers.

Solution. *Positive Integer n* nuff said.

Problem 5.8.21 (USAMO 2007 P5) : Prove that for every nonnegative integer n, the number
n
77 + 1 is the product of at least 2n + 3 (not necessarily distinct) primes.

Solution. When you try to apply induction, always name the hypothesis. In this case name
n
77 + 1 = an . And try to relate an with an+1 .

Problem 5.8.22 (IMO 1998 P3) : For any positive integer n, let τ (n) denote the number of its
positive divisors (including 1 and itself). Determine all positive integers m for which there exists a
2)
positive integer n such that ττ(n
(n) = m.

Solution. Easy go with the flow.

Problem 5.8.23 (USA TST 2002 P2) : Let p > 5 be a prime number. For any integer x, define
p−1
X 1
fp (x) =
(px + k)2
k=1

Prove that for any two integers x, y ,

p 3 | f (x) − f (y )
5.8 Problems 336

Solution. If there is f (x) − f (y ) for all x, y , then always make one of those equal to 0 or in
other words make one side constant

Solution. All fractional sum problems should be solved by some expression manipulation.

Solution. Sometimes try adding things up.

Problem 5.8.24 (USAMO 2012 P4) : Find all functions f : Z+ → Z+ (where Z+ is the set of
positive integers) such that f (n!) = f (n)! for all positive integers n and such that (m − n) divides
f (m) − f (n) for all distinct positive integers m, n.

*Positive Integer n* nuff said.

Problem 5.8.25 (USAMO 2013 P5) : Given positive integers m and n, prove that there is a positive
integer c such that the numbers cm and cn have the same number of occurrences of each non-zero
digit when written in base ten.

What if we make cm and cn have the same digits, occuring same number of time, when the digits
are sorted? This will make the things a whole lot easier. Again for more simplicity, what if the
arrangement of the digits in both of these numbers are “almost” the same? Like, if the digits are in
blocks and if decomposed into such blocks, we get the same set for both of those problems? This
idea of simplicity is more that enough to “Simplify” a problem. Call this strategry Simplify.

Problem 5.8.26 (ISL 2015 N4) : Suppose that a0 , a1 , · · · and b0 , b1 , · · · are two sequences of
positive integers such that a0 , b0 ≥ 2 and

an+1 = gcd (an , bn ) + 1, bn+1 = lcm (an , bn ) − 1.

Show that the sequence an is eventually periodic; in other words, there exist integers N ≥ 0 and
t > 0 such that an+t = an for all n ≥ N.

Like most NT probs, pure investigation. We see that the function an is mostly decreasing, but
it is increasing as well. But the increase rate is not greater than the decrease rate. After some
time playing around, we see that the value of an rises gradually and then suddenly drops. But the
peak value of an doesn’t seem to increase. Well, this is true, we prove that. After that, we are
mostly done, we just show that eventually the least value of an becomes stable as well. We use the
intuitions we get from working around.
5.8 Problems 337

Problem 5.8.27 (ISL 2015 N3) : Let m and n be positive integers such that m > n. Define
xk = m+kn+k for k = 1, 2, . . . , n + 1. Prove that if all the numbers x1 , x2 , . . . , xn+1 are integers, then
x1 x2 . . . xn+1 − 1 is divisible by an odd prime.

As there are powers of 2, we use the powers of 2.

Problem 5.8.28 (USA TST 2018 P1) : Let n ≥ 2 be a positive integer, and let σ(n) denote the
sum of the positive divisors of n. Prove that the nth smallest positive integer relatively prime to n
is at least σ(n), and determine for which n equality holds.

even-odd
Pretty starightforward, as the ques suggests, there are fewer than n coprimes in the interval [1, σ(n)],
we directly show this. [as consttructions don’t seem to be trivial/ easy to get] Inclusion/Exclusion
all the way. But remember, not checking floors can get you doomed.

Problem 5.8.29 (APMO 2014 P3) : Find all positive integers n such that for any integer k there
exists an integer a for which a3 + a − k is divisible by n.

factorize, quadratic residue, complete residue class

Solution. You have to show that the set {x | x ≡ a3 + a (mod p) is equal to the set
{1 . . . p − 1}. So some properties shared by a set as a whole must be satisfied by both of the
sets. The quickest such properties that come into mind are the summation of the set and the
product of the set. While the former doesnt help out much, the later seems promising. Where
we get a nice relation that we have to satisfy:
p−1
Y
a2 + 1 ≡ 1 (mod p)
i=0

One way of concluding from here is to use the quadratic residue ideas, or using the fact that
x 2 + 1 = (x + i )(x − i ). The later requires some higher tricks tho.

Solution. The most natural way must be to show that for any prime p we will find two integers
p - (a − b), and p | a2 + b2 + ab + 1, factorizing the later and getting two squares and a constant
gives us our desired result.

Problem 5.8.30 (APMO 2014 P1) : For a positive integer m denote by S(m) and P (m) the sum
5.8 Problems 338

and product, respectively, of the digits of m. Show that for each positive integer n, there exist
positive integers a1 , a2 , . . . , an satisfying the following conditions:

S(a1 ) < S(a2 ) < · · · < S(an ) and S(ai ) = P (ai+1 ) (i = 1, 2, . . . , n).

(We let an+1 = a1 .)

Solution. 1 is the only integer that increases the sum, but doesnt change the product. This
may seem trivial, but on problems like this where both the sum and the product of the digits of
a number are concerned, this tiny little fact can change everything.

Problem 5.8.31 (RMM 2018 P4) : Let a, b, c, d be positive integers such that ad 6= bc and
gcd(a, b, c, d) = 1. Let S be the set of values attained by gcd(an + b, cn + d) as n runs through
the positive integers. Show that S is the set of all positive divisors of some positive integer.

Problem 5.8.32 (ISL 2011 N1) : For any integer d > 0, let f (d) be the smallest possible integer
that has exactly d positive divisors (so for example we

have f (1) k+1
= 1,
f (5) = 16, and f (6) = 12).
Prove that for every integer k ≥ 0 the number f 2 divides f 2
k .

Solution. Construct the function for 2n .

Problem 5.8.33 (ISL 2004 N1) : Let τ (n) denote the number of positive divisors of the positive
integer n. Prove that there exist infinitely many positive integers a such that the equation τ (an) = n
does not have a positive integer solution n.

Solution. Infinitely many, divisor, what else should come to mind except prime powers...

Problem 5.8.34 (USAMO 2018 P4) : Let p be a prime number and let a1 , a2 . . . ap be integers.
Prove that there exists an integer k s.t. the S = {ai + i k} has at least p2 elements modulo p

Solution. As the only thing that is holding ourselves down is the equivalence of any two
elements of S, we investigate it furthur. It is a good idea to represent by graphs.

Problem 5.8.35 (ISL 2009 N1) : Let n be a positive integer and let a1 , a2 , a3 , . . . , ak (k ≥ 2) be
distinct integers in the set 1, 2, . . . , n such that n divides ai (ai+1 − 1) for i = 1, 2, . . . , k − 1. Prove
that n does not divide ak (a1 − 1).
5.8 Problems 339

Problem 5.8.36 (ISL 2009 N2) : A positive integer N is called balanced, if N = 1 or if N can be
written as a product of an even number of not necessarily distinct primes. Given positive integers a
and b, consider the polynomial P defined by P (x) = (x + a)(x + b).

1. Prove that there exist distinct positive integers a and b such that all the number P (1),
P (2),. . ., P (50) are balanced.
2. Prove that if P (n) is balanced for all positive integers n, then a = b

Problem 5.8.37 (USA TSTST 2015 P5) : Let ϕ(n) denote the number of positive integers less
than n that are relatively prime to n. Prove that there exists a positive integer m for which the
equation ϕ(n) = m has at least 2015 solutions in n.

Solution. When does the equation has multiple solutions? Suppose m = ti=1 piαi (pi − 1) then
Q

Φ(n) = m has multiple solutions if for some p 0 s in m, their product is one less from another prime.
Which gives us necessary intuition to construct a m for which there are A LOT of solutions for
the equation.

Problem 5.8.38 (Iran 2018 T1P1) : Let A1 , A2 , ..., Ak be the subsets of {1, 2, 3, ..., n} such that
for all 1 ≤ i, j ≤ k:Ai ∩ Aj 6= ∅. Prove that there are n distinct positive integers x1 , x2 , ..., xn such
that for each 1 ≤ j ≤ k:
lcmi∈Aj {xi } > lcmi ∈A
/ j {xi }

Solution. Main part of the problem is to notice that the first |Ai | columns of the matrix has
1 from all of the rows. Which triggers the idea of giving one prime to every row, and count xi ’s
with them.

Problem 5.8.39 (Iran TST 2018 T2P4) : Call a positive integer “useful but not optimized " (!), if
it can be written as a sum of distinct powers of 3 and powers of 5. Prove that there exist infinitely
many positive integers which they are not "useful but not optimized".
e.g. 37 is a “useful but not optimized" number since 37 = (30 + 31 + 33 ) + (50 + 51 )

Problem 5.8.40 (ISL 2014 N2) : Determine all pairs (x, y ) of positive integers such that
p
3
7x 2 − 13xy + 7y 2 = |x − y | + 1.

Solution. Always factor, before everything else.

5.8 Problems 340

Problem 5.8.41 (ISL 2014 N1) : Let n ≥ 2 be an integer, and let An be the set

An = {2n − 2k | k ∈ Z, 0 ≤ k < n}.

Determine the largest positive integer that cannot be written as the sum of one or more (not
necessarily distinct) elements of An .

Solution. In every problem, conjecture from smaller case, and check if the conjecture is true
in bigger cases.

Problem 5.8.42 (ISL 2002 N1) : What is the smallest positive integer t such that there exist
integers x1 , x2 , . . . , xt with
x13 + x23 + . . . + xt3 = 20022002 ?

Solution. 1000 + 1000 + 1 + 1.

Problem 5.8.43 (ISL 2002 N2) : Let n ≥ 2 be a positive integer, with divisors 1 = d1 < d2 <
. . . < dk = n. Prove that d1 d2 + d2 d3 + . . . + dk−1 dk is always less than n2 , and determine when
it is a divisor of n2 .

Solution. In problems with all of the divisors of n involved, it is a good choice to substitute
n
di = dk−i+1 . That way, you get the exact same set, represented differently, with n involved. And
Pn 1 n
i=1 i∗(i+1) = n+1

Problem 5.8.44 (Japan MO 2017 P2, TST Mock 2018) : Let N be a positive integer. There
are positive integers a1 , a2 , · · · , aN and all of them are not multiples of 2N+1 . For each integer
n ≥ N + 1, set an as below:
If the remainder of ak divided by 2n is the smallest of the remainder of a1 , · · · , an−1 divided by 2n ,
set an = 2ak . If there are several integers k which satisfy the above condition, put the biggest one.
Prove the existence of a positive integer M which satisfies an = aM for n ≥ M.

Solution. Things must go far...

Problem 5.8.45 (ISL 2002 N3) : Let p1 , p2 , . . . , pn be distinct primes greater than 3. Show that
2p1 p2 ···pn + 1 has at least 4n divisors.
5.8 Problems 341

Problem 5.8.46 (Japan MO 2017 P1) : Let a, b, c be positive integers. Prove that lcm(a, b) 6=
lcm(a + c, b + c).

Problem 5.8.47 (ISL 2009 N3) : Let f be a non-constant function from the set of positive integers
into the set of positive integer, such that a − b divides f (a) − f (b) for all distinct positive integers a,
b. Prove that there exist infinitely many primes p such that p divides f (c) for some positive integer
c.

Solution. Notice if f (1) = 1, we can easily prove the result, so assume that f (1) = c. Now
see that, if we can somehow, create another function g from the domain and range of f with
the same properties as f , and with g(1) = 1, we will be done. So to do this, we need to perform
some kind of division by c.

Problem 5.8.48 (ARO 2018 P9.1) : Suppose a1 , a2 , . . . is an infinite strictly increasing sequence
of positive integers and p1 , p2 , . . . is a sequence of distinct primes such that pn | an for all n ≥ 1.
It turned out that an − ak = pn − pk for all n, k ≥ 1. Prove that the sequence (an )n consists only
of prime numbers.

Problem 5.8.49 (ARO 2018 P10.4) : Initially, a positive integer is written on the blackboard. Every
second, one adds to the number on the board the product of all its nonzero digits, writes down the
results on the board, and erases the previous number. Prove that there exists a positive integer
which will be added infinitely many times.

Proof. Using Bounding and the

Problem 5.8.50 (APMO 2008 P4) : Consider the function f : N0 → N0 , where N0 is the set of
all non-negative integers, defined by the following conditions :

f (0) = 0, f (2n) = 2f (n) and f (2n + 1) = n + 2f (n) for all n ≥ 0

1. Determine the three sets L = {n|f (n) < f (n + 1)}, E = {n|f (n) = f (n + 1)}, and G =
{n|f (n) > f (n + 1)}.

2. For each k ≥ 0, find a formula for ak = max{f (n) : 0 ≤ n ≤ 2k } in terms of k.

Problem 5.8.51 (ISL 2009 N4) : Find all positive integers n such that there exists a sequence of
5.8 Problems 342

positive integers a1 , a2 ,. . ., an satisfying:

ak2 + 1
ak+1 = −1
ak−1 + 1
for every k with 2 ≤ k ≤ n − 1.

Solution. Rewriting the condition, and doing some parity check. Then assuming the contrary
and taking extreme case.

Problem 5.8.52 (ISL 2015 N2) : Let a and b be positive integers such that a! + b! divides a!b!.
Prove that 3a ≥ 2b + 2.

Solution. Size Chase

Problem 5.8.53 (USA TST 2019 P2) : Let Z/nZ denote the set of integers considered modulo n
(hence Z/nZ has n elements). Find all positive integers n for which there exists a bijective function
g : Z/nZ → Z/nZ, such that the 101 functions

g(x), g(x) + x, g(x) + 2x, ..., g(x) + 100x

are all bijections on Z/nZ.

Solution. A very nice problem. We get the motivation by trying the cases for 2, 3 replacing
101. In the case of 2, we just consider the sum g(x). We get that 2 - n. So in the case of 3,
P

we conjecture that 3 - n. But we can’t prove this similarly as before. Whats the most common
‘sum-type’ invariant after the normal sum? Sum of the Squares.

Now that we have proved that (6, n) = 1, most probably our conjecture is correct. So lets try
for any k, we need to show that (k!, n) = 1. In the case of 3, we used the 2nd power sum. So
probably to prove that k - n we need to take the (k − 1)th power sum.

Now the real thing begins. In the case of 3, doesn’t the modular sum equation looks something
like the first finite difference? This rings a bell that whenever there is powers involved, we should
consider using the derivatives. For (k − 1)th power, the (k − 1)th derivative that is.

Another thing here, in the case of 3, we did something like

X X X X
(g(x) + 2x)2 − (g(x) + x)2 ≡ (g(x) + x)2 − (g(x))2 ≡ 0 (modn)

We try something similar again.

5.8 Problems 343

Problem 5.8.54 (Tuymaada 2016, P5) : The ratio of prime numbers p and q does not exceed 2
(p 6= q). Prove that there are two consecutive positive integers such that the largest prime divisor
of one of them is p and that of the other is q.

Problem 5.8.55 (ISL 2016 N5) : Let a be a positive integer which is not a perfect square, and
consider the equation
x2 − a
k= 2 .
x − y2
√
Let A be the set of positive integers k for which the equation admits a solution in Z2 with x > a,
and let B be the set of positive integers for which the equation admits a solution in Z2 with
√
0 ≤ x < a. Show that A = B.

Solution. Building x2 , y2 from x1 , y1 in the most simple and dumb way.

Problem 5.8.56 (Simurgh 2019 P1) : Prove that there exists a 10 × 10 table of ’different’ positive
integers such that, if we define ri , si be te product of the elements of the i th row and i th column
respectively, then r1 , r2 . . . r10 and s1 , s2 . . . s10 form a non-constant arithmetic progression.

Solution. We want to keep things simple. The simpliest arithmetic progression is the
a, 2a, 3a . . . one. Again, we have r1 r2 . . . r10 = s1 s2 . . . s10 , we can wish that ri = si . With
these two assumptions, we can hope that we will find a table with the two sequences being a
constant arithmetic progression.

Problem 5.8.57 (APMO 2017 P4) : Call a rational number r powerful if r can be expressed in
pk
the form for some relatively prime positive integers p, q and some integer k > 1. Let a, b, c be
q
positive rational numbers such that abc = 1. Suppose there exist positive integers x, y , z such that
ax + by + c z is an integer. Prove that a, b, c are all powerful.

Problem 5.8.58 (USA TST 2010 P9) : Determine whether or not there exists a positive integer
k such that p = 6k + 1 is a prime and
 
3k
≡ 1 (mod p)
k

3k
≡ 1 =⇒ 3k2k ≡ 1 which implies,



Solution. k
       
3k 3k 3k 3k
+ + + ≡ 4 (mod p)
0 k 2k 3k
5.8 Problems 344

Which gives the idea to find how the following term works in mod p
∞  
X n
for any arbitrary k
ki
i=0

From ?? we know a nice way of representing it with the k th roots of unity. Roots of unity are
primitive roots mod prime.
PROBLEMS 5.8 Problems

Problems

169 P20, 128 APMO 2014 P3, 337

2017 IGO Advanced P2, 190 APMO 2014 P5, 293
2nd Olympiad of Metropolises, 281 APMO 2015 P2, 129
APMO 2015 P3, 71
A reformulation of Catalan’s Numbers, 112 APMO 2017 P1, 76
All Russia 1999 P9.3, 297 APMO 2017 P3, 75
All Russia 2000 P11.7, 297 APMO 2017 P4, 343
All Russia 2000 P9.3, 297 APMO 2018 P1, 223
All Russia 2002 P11.6, 298 APMO 2018 P2, 164
All Russia 2014 P9.5, 332 APMO 2018 P3, 64
All Russia 2017 9.1, 47 APMO 2018 P4, 64
All Russian 2014 Grade 10 Day 1 P4, 250 APMO 2019 P3, 307
All Russian Olympiad 2016, Day2, Grade 11, Archer - EChen M1P2, 253
P5, 139 Archer - EChen M1P3, 240
All Squares, 328 ARMO 2013 Grade 11 Day 2 P4, 280
American Mathematical Monthly E2309, 14 ARO 1979’s Sharper Version, 33
AoPS, 22, 100, 107, 253, 280, 281, 283, 289, ARO 1993 P10.4, 88
290, 296, 298, 304 ARO 1999 P10.1, 54
AoPS1, 239 ARO 1999 P4, 86
AoPS2, 287 ARO 1999 P9.8, 38
AoPS3, 283 ARO 2005 P10.3, P11.2, 28
AoPS4, 284 ARO 2005 P10.8, 37
AoPS5, 289 ARO 2005 P9.4, 43
AoPS6, 295 ARO 2006 11.3, 80
AoPS7, 296 ARO 2007 P9.7, 116
APMO 1991 P3, 156 ARO 2008 P9.5, 102
APMO 1999 P1, 70 ARO 2011 P10.8, 86
APMO 1999 P2, 70 ARO 2011 P11.6, 116
APMO 1999 P3, 190, 244 ARO 2013 P9.4, 58
APMO 1999 P4, 328 ARO 2013 P9.5, 48
APMO 1999 P5, 60 ARO 2014 P10.8, 117
APMO 2008 P2, 78 ARO 2014 P9.3, 58
APMO 2008 P4, 341 ARO 2014 P9.4, 190
APMO 2012 P2, 108 ARO 2014 P9.7, 115
APMO 2014 P1, 337 ARO 2014 P9.8, 45
PROBLEMS 346

ARO 2016 P1, 107 China TST 1995 P5, 138

ARO 2016 P3, 105 China TST 2005, 334
ARO 2017 P9.1, 39 China TST 2006, 24
ARO 2018 10.3, 99 China TST 2011, Quiz 2, D2, P1, 208
ARO 2018 P10.1, 164 China TST 2013 T2P2, 325
ARO 2018 P10.2, 190 China TST 2015 T1 D2 P3, 35
ARO 2018 P10.4, 341 China TST 2015 T1D2P1, 68
ARO 2018 P11.1, 139 China TST 2016 T3P2, 63
ARO 2018 P11.2, 156 China TST 2018 T1P3, 217
ARO 2018 P11.4, 202 China TST 2018 T2P4, 325
ARO 2018 P11.5, 75 ChinaTST D2P2, Dual of the Hagge’s Circle
ARO 2018 P11.6, 303 theorem, 208
ARO 2018 P9.1, 341 Codeforces 101954/G, 88
ARO 2018 P9.5, 99 Codeforces 1158D, 68
AtCoder GC033 C, 38 Codeforces 1270F, 31
AtCoder GC043 B, 25 CodeForces 1379E, 31
Autumn Tournament, 2012, 236 Codeforces 330D, 30
CodeForces 744B, 25
Balkan MO 2017 P1, 332 CodeForces 960/C, 28
Balkan MO 2017 P3, 332 Codeforces 987E, 21
balkan mo 2017 p3, 182 CodeForces 989B, 109
BdMO 2019 P10, 67 CodeForces 989C, 109
Belarus 2001, 42 Cody Johnson, 23
bewarish 1, 183
Binom 2, 8 Dunno, 267
Binom1, 7
Brazilian MO 2002, ARO 1979, 33 EGMO 2012 P3, 126
Brazilian Olympic Revenge 2014, 104 EGMO 2015 P2, 82
BrMO 2008, 334 EGMO 2017 P2, 88
Bulgaria 1995, 162 EGMO 2017 P3, 58
Bulgarian IMO TST 2004, Day 3, Problem 3, EGMO 2017 P5, 81
100 Eight queens puzzle, 66
buratinogigle, 203 ELMO 2017 P5, 42
buratinogigle Tough P1, 304 ELMO 2019 P3, 192
buratinogigle’s proposed problems for Arab Saudi European Mathematics Cup 2018 P1, 77
team 2015, 185, 186, 243 European Mathematics Cup 2018 P2, 191
buratinogigle’s proposed problems for arab saudi European Mathematics Cup 2018 P3, 164
team 2015, 184, 185
Four Incenters in a Tangential Quadrilateral,
buratinogigle’s proposed probs for Arab Saudi
224
team 2015, 221
Generalization, 45
CGMO 2010 P1, 14
Generalization of Iran TST 2017 P5, 280
CGMO 2016 P3, 321
Generalization of Mixtilinear Incirlce, 240
China MO 2018 P2, 113
Geodip, 225
China MO 2018 P5, 113
German TST 2004 E7P3, 45
PROBLEMS 347

GQMO 2020 P3, 169 Iran 3rd Round 2015 P5, 198
GQMO 2020 P4, 32 IRAN 3rd Round 2016 G6, 290
Iran 3rd Round 2016 N3, 330
IGO 2014 S5, 234 IRAN 3rd Round 2016 P1, 190
IGO 2017 Advance P3, 186 IRAN 3rd Round 2016 P2, 283
IGO 2017 Advance P4, 190 Iran 3rd Round G4, 265
IGO 2018 A3, 63 Iran 3rd Round G4 Generalized version, 265
IGO 2018 A5, 191 Iran 3rd Round Training 2016, 211
IGO 2019 A5, 197 Iran MO 3rd round 2017 mid-terms Geometry
IGO 2019 Advanced P4, 307 P3, 303
ILL 1985, 59 Iran MO 3rd Round N3, 332
IMO 1974 P2, 161 Iran TST 2002 P3, 102
IMO 1979 P3, 118 Iran TST 2004 P4, 256
IMO 1986 P3, 31 Iran TST 2004 P6, 140
IMO 1994 P5, 126 Iran TST 2007 P2, 109
IMO 1998 P3, 335 Iran TST 2008 D3P1, 13
IMO 2011 P2, 106 Iran TST 2008 P12, 227
IMO 2011 P4, 78 IRAN TST 2009 P2, 334
IMO 2012 P2, 156 Iran TST 2009 P4, 140
IMO 2013 P3, 221 Iran TST 2009 P6, 50
IMO 2017 P2, 126 Iran TST 2012 P4, 99
IMO 2017 P5, 103 Iran TST 2013 D1P2, 15
IMO 2018 P4, 66 Iran TST 2014 T1P6, 292
IMO 2018 P6, 247 IRAN TST 2015 Day 2, P3, 285
IMO SL 1985, 78 IRAN TST 2015 Day 3, P2, 279
India 1995, 161 IRAN TST 2015 Day 3, P3, 232
India IMO Camp 2017, 109 Iran TST 2015 P11, 331
India Postals 2015 Set 2, 228 Iran TST 2015, P4, 331
India TST, 302 Iran TST 2017 D1P1, 110
India TST 2013 Test 3, P1, 105 IRAN TST 2017 Day 1, P3, 279
India TST 2017 D1 P3, 27 Iran TST 2017 P5, 300
Indian Postal Coaching 2011, 108 Iran TST 2017 T3 P6, 291
Indian TST 2004, 53 Iran TST 2018 D2P6, 257
InfinityDots MO Problem 3, 201 Iran TST 2018 P1, 20
IOI 2007 P3, 28 Iran TST 2018 P3, 223
IOI 2016 P5, 106 Iran TST 2018 P4, 223
IOI 2018 P1, 117 Iran TST 2018 T1P3, 298
IOI Practice 2017, 20 Iran TST 2018 T1P4, 298
IOM 2017 P1, 273 Iran TST 2018 T2P4, 339
Iran 2005, 65 Iran TST 2018 T3P6, 298
Iran 2018 T1P1, 339 Iran TST ??, 64
IRAN 2nd Round 2016 P6, 283 Iran TST T2P1, 127
IRAN 3rd Round 2010 D3, P5, 283 Iran TST T2P5, 222
IRAN 3rd Round 2010 D3, P6, 292 ISL 1988 P10, 14
IRAN 3rd Round 2011 G5, 283 ISL 1990, 69
PROBLEMS 348

ISL 1990 P3, 103 ISL 2004 N1, 323, 338

ISL 1991 P10, 114 ISL 2004 N2, 331
ISL 1993 C5, 86 ISL 2004 N3, 326
ISL 1994 A1, 70 ISL 2004 N4, 321
ISL 1994 A4, 126 ISL 2005 A1, 145
ISL 1994 C2, 81 ISL 2005 A2, 127
ISL 1994 C3, 22 ISL 2005 A3, 165
ISL 1996 A2, 158 ISL 2005 A4, 127
ISL 1997 P4, 100 ISL 2005 C1, 46
ISL 1998 A1, 157 ISL 2005 C2, 20, 47
ISL 1998 C4, 107 ISL 2005 C3, 111
ISL 1998 C7, 53 ISL 2005 C4, 44
ISL 1999 A1, 157 ISL 2006 C1, 101
ISL 1999 G8, 239 ISL 2006 C2, 58
ISL 2000 N1, 332 ISL 2006 C3, 96
ISL 2000 N3, 332 ISL 2006 C4, 101
ISL 2001 A1, 128, 157 ISL 2006 C6, 33
ISL 2001 C3, 39 ISL 2006 G6, 223, 299
ISL 2001 C6, 106 ISL 2006 G7, 299
ISL 2002 A1, 128 ISL 2006 G9, 299
ISL 2002 C1, 77 ISL 2006 N4, 140
ISL 2002 C2, 88 ISL 2007 C1, 89
ISL 2002 C3, 80 ISL 2007 C2, 60
ISL 2002 C5, 14 ISL 2007 C3, 89
ISL 2002 C6, 111 ISL 2007 C4, 70
ISL 2002 G1, 301 ISL 2007 C5, 64
ISL 2002 G2, 301 ISL 2008 A1, 126
ISL 2002 G3, 302 ISL 2008 C1, 105
ISL 2002 G4, 302 ISL 2008 C3, 65
ISL 2002 G7, 302 ISL 2008 C4, 76
ISL 2002 G8, 244 ISL 2009 A2, 156
ISL 2002 N1, 340 ISL 2009 A3, 131
ISL 2002 N2, 340 ISL 2009 A6, 71
ISL 2002 N3, 340 ISL 2009 C1, 55
ISL 2002 N6, 329 ISL 2009 C2, 110
ISL 2003 C1, 81 ISL 2009 C3, 91
ISL 2003 C2, 61 ISL 2009 G3, 302
ISL 2003 C3, 62 ISL 2009 G4, 250
ISL 2003 C4, 27 ISL 2009 N1, 338
ISL 2004 A1, 158 ISL 2009 N2, 338
ISL 2004 A2, 167 ISL 2009 N3, 341
ISL 2004 C1, 77 ISL 2009 N4, 341
ISL 2004 C2, 99 ISL 2009 N5, 326
ISL 2004 C3, 99 ISL 2010 A3, 165
ISL 2004 C5, 101 ISL 2010 C2, 108
PROBLEMS 349

ISL 2010 C3, 83 ISL 2015 C2, 65

ISL 2010 G5, 196 ISL 2015 C4, 54
ISL 2010 N3, 328 ISL 2015 G5, 195
ISL 2010 N5, 326 ISL 2015 G6, 194
ISL 2011 A2, 168 ISL 2015 N2, 342
ISL 2011 A3, 127 ISL 2015 N3, 337
ISL 2011 A5, 109 ISL 2015 N4, 336
ISL 2011 G4, 303 ISL 2016 A1, 158
ISL 2011 G7, 277 ISL 2016 A2, 158
ISL 2011 G8, 191 ISL 2016 C1, 20
ISL 2011 N1, 338 ISL 2016 C2, 89
ISL 2012 A1, 130 ISL 2016 C3, 102
ISL 2012 A4, 141 ISL 2016 C6, 48
ISL 2012 A5, 130 ISL 2016 G6, 282
ISL 2012 C1, 21 ISL 2016 N5, 343
ISL 2012 C2, 109 ISL 2017 A1, 157
ISL 2012 C3, 95 ISL 2018 C5, 98
ISL 2012 C4, 55
ISL 2012 G8, 301 Jacob Tsimerman Induction, 48
ISL 2012 N3, 323 Japan MO 2017 P1, 340
ISL 2012 N6, 324 Japan MO 2017 P2, TST Mock 2018, 340
ISL 2013 C1, 104 Japan MO 2017 P3, 301, 302
ISL 2013 C2, IMO 2013 P2, 59
Korean Summer Program TST 2016 1, 169
ISL 2013 C3, 46
ISL 2013 N1, 326 Lithuania ??, 63
ISL 2013 N2, 71
ISL 2013 N5, 323 Math Price for Girls 2017 P4, 63
ISL 2014 A1, 23, 168 MEMO 2008, Team, P6, 22
ISL 2014 A2, 164 MEMO 2015 T4, 62
ISL 2014 C1, 60 Mexican Regional 2014 P6, 103
ISL 2014 C2, 102 MOP Problem, 82
ISL 2014 C3, 108
ISL 2014 C4, 90 OC Chap2 P2, 50
ISL 2014 G5, 193 OIM 1994, PSMiC, 25
ISL 2014 G7, 193 pco 168 P18, 127
ISL 2014 N1, 30, 339 pco 169 P11, 126
ISL 2014 N2, 339 pco 169 P15, 127
ISL 2014 N3, 23 Polish OI, 28
ISL 2014 N4, 323 Putnam 2002 A3, 34
ISL 2014 N5, 328 Putnam 2005 B4, 15
ISL 2015 A1, 165 Putnam 2011 A2, 168
ISL 2015 A2, 129 Putnam 2013 A3, 168
ISL 2015 A3, 165 Putnam 2017 A2, 140
ISL 2015 A4, 129 Putnam 2017 A4, 111
ISL 2015 C1, 119
PROBLEMS 350

Result by Erdos, 112 Turkey National MO 2002 P3, 39

Result by Erdos, Dividing the integers into Turkey TST 2017 P5, 156
arithmetic progressions, 70 Turkey TST 2018 P4, 187
Rioplatense Olympiad 2013 Problem 6, 211 Tuymaada 2016, P5, 342
RMM 2018 P4, 338 Two Symmedian Points, 232
RMM 2018 P6, 294
RMM 2019 P1, 54 USA Dec TST 2016, P1, 22
RMM 2019 P2, 191 USA TST 2000 P2, 252
RMM 2019 P3, 97 USA TST 2000 P3, 334
RMM 2019 P4, 67 USA TST 2000 P4, 334
RMM 2019 P5, 129 USA TST 2002 P2, 335
RMM 2020 P3, 50 USA TST 2005 P1, 13
Romanian TST 2003 P5, 146 USA TST 2006 P1, 43
Romanian TST 2006 P2, 146 USA TST 2006 P5, 82
Romanian TST 2008 T1P4, 63 USA TST 2006 P6, 273
Romanian TST 2012 P4, 43 USA TST 2009 P1, 107
Romanian TST 2016 D1P2, 13 USA TST 2010 P8, 5
USA TST 2010 P9, 343
Saint Petersburg 2001, 116 USA TST 2011 D3P2, 48
Serbia 2018 P1, 222 USA TST 2011 P2, 49
Serbia 2018 P4, 139 USA TST 2014 P2, 324
Serbia National D2P2, 67 USA TST 2014 P3, 40
Serbia TST 2015 P3, 29 USA TST 2014 P4, 138
Serbia TST 2017 P2, 113 USA TST 2017 P1, 111
Serbia TST 2017 P4, 113 USA TST 2017 P2, 287
Sharygin 2010, 249 USA TST 2018 P1, 337
Sharygin 2010 P3, 246 USA TST 2018 P2, 131
Sharygin 2012 P21, 255 USA TST 2019 P1, 244
Sharygin 2012 P22, 254 USA TST 2019 P1 parallel problem, 246
Simurgh 2019 P1, 343 USA TST 2019 P2, 342
Simurgh 2019 P2, 190 USA TSTST 2015 P5, 339
Simurgh 2019 P3, 50 USA Winter TST 2018 P3, 83
Somewhere, 4 USA Winter TST 2020 P2, 187, 189
St. Petersburg 2000, 87 USAMO 1974 P2, 161
Swell coloring, 42 USAMO 1986 P2, 102
USAMO 1996 P4, 76
Taiwan TST 2014 T3P3, 240 USAMO 1998, 34
Taiwan TST 2015 R3D1P1, 30 USAMO 1998 P3, 160
Timus 1578, 25 USAMO 1999 P1, 103
timus 1862, 115 USAMO 1999 P1 generalization, 104
Tournament of Towns 2009 S6, 39 USAMO 1999 P6, 221
Tournament of Towns 2015F S7, 27 USAMO 2001 P5, 333
Tournament of Towns 2015S S4, 62 USAMO 2002 P3, 138
Tournament of Towns 2015S S6, 89 USAMO 2004 P2, 334
Tournament of Towns Spring 2012 S7, 33 USAMO 2004, P4, 54
PROBLEMS 351

USAMO 2005 P1, 114, 117

USAMO 2005 P4, 117
USAMO 2005 P5, 114
USAMO 2006 P2, 82, 114, 117
USAMO 2006 P5, 117
USAMO 2007 P1, 71
USAMO 2007 P2, 62
USAMO 2007 P5, 335
USAMO 2008 P1, 335
USAMO 2008 P2, 231
USAMO 2008 P3, 78
USAMO 2008 P4, 105
USAMO 2008 P5, 53
USAMO 2009 P2, 117
USAMO 2010 P2, 29
USAMO 2012 P2, 77
USAMO 2012 P4, 336
USAMO 2013 P2, 79
USAMO 2013 P4, 166
USAMO 2013 P5, 336
USAMO 2013 P6, 21
USAMO 2014 P1, 53
usamo 2014 p5, 183
USAMO 2015 P4, 26
USAMO 2017 P5, 122
USAMO 2018 P4, 338
USAMO 2018 P5, 301
USJMO 2018 P3, 298

Vietnam 1991, 128

Vietnam 1998, 161
Vietnam 2017 TST P6, 333
Vietnam TST 2017 P2, 333
Vietnamese TST 2018 P6.a, 218
Vietnamese TST 2018 P6.a Generalization,
220

WenWuGuangHua Mathematics Workshop, 297,

298

Zhao Poly, 139

Zhao Polynomials, 138
THEOREMS 5.8 Problems

Theorems

n − 1 Equal Values, 149 Erdos Ko Rado theorem, 10

2 is a quadratic residue, 318 Erdos-Mordell Theorem (Forum Geometrico-
rum Volume 1 (2001) 7-8), 277
Apollonius Circle and Incenter, 215 Erdos-Szekeres Theorem, 12
Arc Midpoint as Centers, 214 Euler’s Criterion, 320
Average of Degrees, 40 Euler’s Polyhedron Formula, 35
Beatty’s Theorem or Rayleigh Theorem, 312
Finite Differences, 136
Bipartite Graph Criteria, 35
Fontene’s First Theorem, 278
Bound on roots, 145
Fontene’s Second Theorem, 278
Bounds On Roots, 142
Frobenius Coin Problem, extended Chicken Mc-
Brocard Points, 211
Nugget, 312
Catalan Recursion, 69 Fuhrmann Circle, 226
Cauchy-Schward Inequality, 152
Gauss’s Criterion, 320
Cauchy-Schwarz Inequality Complex form, 152
Coefficients form a Decreasing Sequence, 142 Hagge’s circles, 207
Cohn’s Criterion, 145 Helly’s Theorem, 57
collinearity with antipode and center, 182 Holder’s Inequality, 153
Concurrent Lines in Incenter, 215 Homothety and Reflection, 236
Conic through orthocenter and vertices, 268
Constant is LARGE, 142 Incircle Touchpoint and Cevian, 214
Construction of Involution Center on Line, 266 Inscribed Conic in Pascal’s theorem, 265
Convexity, 160 Invertible Function on a line, 260
Criteria of partitioning a graph into discon- Involution on a conic, 262
nected sub-graphs, 35 Involution on a line, 262
Cross Ratio Equality, 261 Isogonal Conjugate Lemma, 209
Isogonal Line Lemma, 228
Dealing with binomial terms with a common Isogonality in quadrilateral, 229
factor, 141 Isotonic Lemma, 234
Degenerate Desargues’ Involution: 2 Points,
263 Jacobi’s Theorem, 212
Desargues’ Involution Theorem, 263 Jensen’s Inequality, 148, 160
Dilworth’s Theorem, 12
Karamata’s Inequality, 148
E.R.I.Q. (Equal Ration in Quadrilateral) Lemma, Kleitman lemma, 10
275
THEOREMS 353

Lagrange Interpolation Theorem, 136 Quadratic non-residue mod infitely many primes,
Leading Coefficient is LARGE, 142 319
Linear map that reverses the order of multipli- Quadratic Reciprocity, 318
cation, 91 Quadratic Residue Law, 320

Mannheim’s Theorem, 210 Ratio, 229

Mannheim’s Theorem’s Converse, 210 Rouché’s Theorem, 142
Maximality of the Area of a Cyclic Quadrilat-
eral, 273 Sawayama and Thebault’s theorem, 242
Maybe not useful at all, 312 Simson Line Angle, 205
Midline Concurrency with Incircle Touchpoints, Simson Line Parallel, 204
215 Sprague-Grund Theorem, 52
Minkowski, 150 Steiner’s Isogonal Cevian Lemma, 276
Minkowski’s theorem, 170 Sum of Mobius functions of divisors, 317
Mirsky Theorem, 12 Swap Sort, 17
Mobius Inverison Formula, 317
Tangential Quadrilateral Incenters, 223
Most Important Symmedian Property, 231
Terquem’s Cevian Theorem, 209
MST Cut, 17
The Derivative of an Odd function is always an
Muirhead’s Inequality, 148
Even function, and vice versa., 163
Multiplication Order, 91
The Sunflower Lemma, 11
Newton-Gauss Line, 251 Titu’s Lemma, 152
Non-zero digits in base b, 312 Triangle Inequality, 151
Turan’s theorem, 41
Orthogonal Hyperbola, 268
Van der Waerden’s Theorem, 69
Pappus’s Hexagon Theorem, 261 Vandermonde’s identity, 4
Pascal’s Theorem for Octagons: A special case,
264 Weighted AM-GM, 152
Paul Yui Theorem, 215 Weighted Power Mean, 151
Pedal Triangles of Isodynamic Points, 217 Wolstenholme’s Theorem, 315
Perron’s Criterion, 143
Young’s Inequality, 150
Perron’s Criterion’s Generalization (Dominat-
ing Term), 143 Zsigmondy’s Theorem, 321
Perspectivity Line with Orthic triangle is per-
pendicular to Euler line, 206
Poncelet’s Porism, 221
Popoviciu’s inequality, 160
Power function convexity, 149
Prime divisors of an integer polynomial, 314
Prime Divisors of Cyclotomic Polynomials, 318
Primes in a recursive sequence, 324
Properties of Coharmonic Points, 260
Prufer sequence, 45
Pseudo Miquel’s Theorem, 243
CATEGORIES 5.8 Problems

Categories

Permuatations
USAMO 2017 P5, 122

Sets
Local
Iran TST 2008 D3P1, 13
ISL 2002 C5, 14
Romanian TST 2016 D1P2, 13
Max-set with property
Iran TST 2013 D1P2, 15
USA TST 2005 P1, 13
Subsets of [N]
American Mathematical Monthly E2309,
14
CGMO 2010 P1, 14
ISL 1988 P10, 14
STRATEGIES 5.8 Problems

Strategies

Induction
CF 1379E , 31

Adding new stuffs

ISL 2002 C5, 14

Bijection
Binary Representation
ISL 1988 P10, 14
USAMO 2017 P5, 122

CopyCat
ToT 2009 S6, 40
copycat
partition
aro 2006 11.3, 80

Double Counting
USA TST 2005 P1, 13

Focus
ISL 2002 C5, 14
Romanian TST 2016 D1P2, 13

Induction
n is or isn’t
Iran TST 2013 D1P2, 15
Iran TST 2008 D3P1, 13
ISL 2002 C5, 14
Romanian TST 2016 D1P2, 13

Roots of unity filtering

Iran TST 2013 D1P2, 15
MISC 5.8 Problems

Misc

Algorithm RMM 2020 P3, 51

Copycat
ARO 1999 P9.8, 38 Non-Neighbor Equivalence Class, 35

Bipartite, 35 Polyhedron Formula, 35

Extremal
Longest Path
AtCoder GC033 C, 38
ChTST 2015 P3, 36
Object
ISL 2001 C3, 39

Global
Multiplication
GQMO 2020 P3, 169
Graph
Bipartite
ARO 2005 P10.8, 37
RMM 2020 P3, 51
Grid → Graph
ARO 2005 P10.8, 37
Hall Marriage
RMM 2020 P3, 51

Induction
Classic
ARO 2017 P9.1, 39
Special
RMM 2020 P3, 51
Invariant
Monovariant
AtCoder GC033 C, 38

Local
Add One by One